US 8102074

I'll run targeted searches for this specific patent number and for any 2026 CAFC activity.

The general searches returned noise. Let me try more specific queries against patent-specific and court-specific sources.

Summary of US Patent 8,102,074

I searched for this specific number (and for 2026 CAFC activity tied to it). Here is what I can confirm, and where I have to flag uncertainty.

Bibliographic data (confirmed)

Sources: Google Patents (https://patents.google.com/patent/[US8102074B2](/patent/US8102074B2)/en), uspto.report (https://uspto.report/patent/grant/8102074), and the patent PDF (patentimages.storage.googleapis.com/08/f0/2a/d8aea7cc33c70c/US8102074.pdf).

Ownership caveat: the "current assignee" field is not simple. Google Patents lists both Tigo Energy Innovations LLC and Newlight Capital LLC, and shows a chain of security interests and assignments (Western Alliance Bank 2018; UMB Bank as trustee / Newlight Capital 2020; release 2023; Newlight Capital 2025 corrective assignment; assignment to Tigo Energy Innovations LLC recorded 2026-01-20). The underlying inventive entity and original assignee are Tigo Energy, Inc. I would not treat the current owner as settled without checking the USPTO Assignment database directly.

Abstract (verbatim)

"Apparatuses and methods are disclosed for regulating or limiting the voltage output from solar modules connected in series such that the voltage on a string bus connecting those solar modules does not exceed regulatory or safety limitations. This can be accomplished via a controller, local management units (for downconverting solar module voltage output), or a combination of the two."

Independent claims in plain language (3 independents: 1, 8, 16; 19 claims total)

Claim 1 — system. An energy production system with: a string bus; first and second solar modules each connected to the bus and each generating its own voltage; a single controller configured to limit at least one of those two voltages as provided to the bus; and a first and a second local management unit (LMU), each coupled between its respective solar module and the bus and each in communication with that single controller. Plain reading: the architecture is one central controller plus one per-module LMU. Note claim 1 as granted does not recite the "maximum regulatory safety voltage" or "predicted future voltage" limitations.

Claim 8 — system. An energy production system with: a string bus; a solar module connected to the bus and generating a voltage; and a controller in communication with the module and the bus, configured to control the voltage provided to the bus based on a predicted future voltage as predicted by the controller and a maximum regulatory safety voltage. Plain reading: the control input is a forecast (the controller's own prediction of a future voltage), bounded by a regulatory safety ceiling.

Claim 16 — method. Monitoring a first voltage across a first string-bus section connecting a first solar module to a second solar module; monitoring a second voltage across a second string-bus section connecting the second solar module to a voltage output; and limiting the voltage output based on a maximum regulatory safety voltage by limiting the voltage of at least one of the first and second solar modules. Plain reading: sense voltages on two bus segments, then throttle one or both modules to stay under a regulatory ceiling.

Dependent claims add: the controller commanding both LMUs (2); the "controller" itself being constituted by the LMUs (3); switchable connections (4); transistor switches (5); duty-cycle and phase control (6); prevention of exceeding a predefined limit (7); disconnect switch (9); duty-cycle adjustment (10); transistor switch (11); prediction based on a voltage trend (12); a default/fallback voltage (13) triggered when a voltage threshold is exceeded (14); module connected via an LMU (15); threshold-based limiting (17); duty-cycle control of first and second switchable connections (18); and limiting based on predicted future voltage (19).

CAFC 2026 docket search

I found no CAFC appeal, IPR appeal, or 2026 Federal Circuit docket entry naming US 8,102,074. The closest Tigo-related appellate item I located is SMA Solar Technology America, LLC v. Tigo Energy, Inc., Fed. Cir. No. 2025-1344, which was dismissed for failure to prosecute on May 12, 2025, with the mandate issued June 18, 2025 (Patexia docket: https://services.patexia.com/lawsuits/SMA-Solar-Technology-America-LLC-v-Tigo-Energy-Inc-id-[206378](/patent/206378)/documents). Nothing there indicates this patent is on a 2026 CAFC docket. Tigo's publicly announced patent campaigns (e.g., the Feb. 21, 2023 SunSpec suit, and a suit involving "six patent infringement claims against SMA Solar Technology America LLC") concern rapid-shutdown/MLPE patents; I cannot confirm from my sources that 8,102,074 was asserted in any of them.

Explicit uncertainties

• No CAFC verification: my searches did not retrieve Federal Circuit docket records directly, so "no 2026 CAFC activity" means not found, not affirmatively confirmed absent. Docket-level checking on PACER/CAFC CM-ECF or USPTO PatentCenter would be required to state that definitively.
• Claim 16 text: the Google Patents rendering of claim 16 is truncated mid-sentence ("monitoring a first voltage across a first st"). The full wording I give above is taken from uspto.report's rendition of the granted claims, which matches the patent's independent-claim structure.
• Do not conflate related documents. US 20120119584 and US 8,274,172 share the title and specification but have different claim sets — e.g., their claim 1 recites controlling based "at least in part on a maximum regulatory safety voltage," which is absent from granted claim 1 of 8,102,074. US 8,274,172 is also subject to a terminal disclaimer. Any claim-charting or validity analysis must use the granted text of the specific patent.
• Current ownership as described above should be verified in the USPTO Assignment database rather than inferred from aggregator "current assignee" fields.

I'll run targeted searches against litigation-specific sources for this patent number.

Let me target litigation databases directly.

Bottom line

No litigation naming U.S. Patent No. 8,102,074 has been found. Not as an asserted patent, not as an adjudicated patent, and not in any 2026 filing. I cannot supply plaintiff/defendant/jurisdiction/case-number/filing-date/status fields because there are no cases to populate them with.

This is consistent with the prior sections of this analysis (the summary and PTAB sections, which also found no assertion). I found no contradiction to flag.

---

Known litigation involving US 8,102,074

I will not manufacture a table row to fill this template.

What I searched

• Queries pairing the patent number directly with litigation terms: "8,102,074" patent litigation; "8102074" patent infringement lawsuit; patent 8102074 litigation docket; docketalarm "8102074"; unifiedpatents.com litigation "8102074".
• Assignee-side queries targeting the current owner: Tigo Energy Innovations LLC lawsuit patent 8,102,074; "Tigo Energy Innovations" complaint infringement 2026; "Tigo" patent lawsuit 2026 filed district court.
• Results were cross-checked against Tigo's own enforcement disclosures and the two live Tigo district-court campaigns, both of which identify their asserted patents explicitly.

Why I am reasonably confident the negative is real (not just unindexed)

The '074 is not among the patents asserted in any Tigo enforcement campaign I could identify, and in each instance the asserted patents are enumerated on the record:

• Tigo Energy Inc. v. SMA Solar Technology America LLC, No. 1:22-cv-00915-GBW (D. Del.), filed 2022-07-11 — asserts U.S. 8,823,218; 8,933,321; 9,584,021; 9,966,848; 10,256,770; 10,333,405. The '074 is absent. This case settled 2025-05-13 ("Tigo Energy Resolves Multi-Year Patent Infringement Litigation With SMA").
• Tigo Energy Inc. v. SunSpec Alliance, No. 3:23-cv-00762-WHO (N.D. Cal.), filed 2023-02-21 — the induced-infringement order at Dkt. 36 addresses the '321 patent claims (and the '770). The '074 is absent.
• Tigo Energy Inc. v. Altenergy Power Systems / APsystems, No. 5:20-cv-03622 (N.D. Cal.) — reported as asserting the rapid-shutdown family ('321, '770). The '074 is absent.

One positive-but-not-litigation datapoint worth recording: Tigo's own Virtual Patent Marking page lists U.S. Patent No. 8,102,074 among the "Tigo-APsmart Licensed Patents" said to cover the APsmart RSD-S-PLC, RSD-S-PLC-B, RSD-D and RSD-D-15-1500 rapid-shutdown devices (https://www.tigoenergy.com/rsdvirtualmarking). That is a licensing/marking statement — the front end of an assertion program — not a filed case. It does, however, tell you the patent is being actively monetized rather than shelved, which is the relevant risk signal for a defendant.

---

⚠️ False positives — do not use these

The string "8102074" is a low-specificity numeric token and my searches surfaced several non-patent collisions. Explicitly excluded:

• A Colombian civil enforcement matter (GMAC Financial Colombia S.A. v. Jhon Fredy Bedoya Manjarres, Rad. 110014003048-2020-00679-00, Juzgado 48 Civil Municipal de Bogotá) — "8102074" is the defendant's cédula/ID number, not a patent.
• An FDA Adverse Event Reporting System (FAERS) case number 8102074 (Plan B One-Step / Teva, received 23-Aug-2011).
• A Brazilian municipal procurement act "Ato n.º 8102074" (Fundo Municipal de Saúde de Guabiruba) and a Rio Grande do Sul commercial registry entry for an unrelated company.
• A Chinese court filing "(2024)苏0214民初9474号" surfaced only because "8102074.6200" appeared in an unrelated financial table on the same page.
• IdeaHub Inc. v. Unified Patents, LLC, Fed. Cir. No. 2024-1684 (nonprecedential, Rule 36 judgment, April 10, 2026) — surfaced on the "unified patents" keyword, not the patent number. Unrelated (PGR2022-00044).

None of these is litigation involving US 8,102,074.

---

Caveats — verify before relying on this in a filing

1. Unified Patents litigation caselist and Docket Alarm were not directly queried at the case-record level. My retrievals were search-engine mediated; the Docket Alarm and Unified portal URLs I could reach returned docket documents for other cases, not a per-patent case list for 8102074. A direct pull of portal.unifiedpatents.com/litigation/ filtered on the patent, and of docketalarm.com by patent number, is still required to convert "not found" into "confirmed absent."
2. The 2026 docket sweep for the new owner is incomplete. The '074 was assigned to Tigo Energy Innovations LLC (Texas LLC), recorded 2026-01-20. I ran searches aimed at that entity as plaintiff and found no complaint. But a newly formed asserting vehicle's first suits would not necessarily be indexed under a search for "Tigo." Re-run a party-name sweep on PACER/CM-ECF for "Tigo Energy Innovations" across all districts and for all of 2026.
3. Do not conflate the '074 with its sibling US 8,272,172. They share the title and specification but have different claims. A case naming the '172 (US 13/357,331) is not a case naming the '074, and I found none naming the '172 either.
4. Date note. My environment's current date (2026-09-27) and the date stated in the task prompt (April 26, 2026) differ. Where I cite a source dated after April 2026, I have flagged it; the negative finding holds under either date as of the material retrieved.

---

What the absence means for you

Given the assignment analysis in this same file — the 2025-12-16 Patent Purchase Agreement transferring the portfolio to Tigo Energy Innovations LLC, whose stated business at PPA §6.2 is "licensing and/or asserting the Assigned Patents," with a §3.5 grant-back license to Tigo Energy, Inc. and a §3.4 no-contest covenant — 8,102,074 is best characterized as acquired-but-not-yet-asserted. The license/marking use of the '074 on the Tigo-APsmart rapid-shutdown page shows it is inside the monetization perimeter even though it has never been pleaded.

Recommended monitoring triggers: (a) any complaint by Tigo Energy Innovations LLC, (b) any IPR petition naming 8,102,074 (there is none — see the PTAB section), and (c) any demand letter referencing the Tigo-APsmart licensed-patent list. The patent's listed adjusted expiration is 2030-03-12, so the assertion window is roughly 3.5 years and closing — which, if anything, increases the probability of a near-term filing rather than decreasing it.

I'll verify the structured "no PTAB proceedings" result against public sources before writing up.

Let me dig deeper on any AIA trial naming this patent specifically, plus the related family member.

One more check on the third-party IPRs against Tigo patents and whether any reexam exists.

PTAB proceedings on file — US 8,102,074 B2

Sourcing note. The canonical list is the structured "PTAB proceedings on file" block: the USPTO Open Data Portal (ODP) API returns no AIA trial proceedings for this patent as of the most recent ingest. I verified that negative independently via web search against PTAB-generated documents (petitioner updated mandatory notices, board refund notices, district-court IPR status tables) and against the IP-verse PTAB case lists for both Tigo Energy and Altenergy Power Systems. Every hit ties to a different Tigo patent. Nothing in my searches surfaced an IPR, PGR, CBM, or ex parte reexam naming 8,102,074 or its continuation 8,274,172.

---

Proceedings overview

Total AIA trial proceedings on US 8,102,074: 0. Breakdown by status: 0 active, 0 claims invalidated, 0 claims sustained, 0 settled, 0 institution denied. Defensive posture for a defendant: the patent is entirely untested at the PTAB. No claim of 8,102,074 has ever been construed by the Board, no prior-art ground has been adjudicated, no third party is estopped, and the patent carries none of the narrowing that a survivable IPR usually imposes. That cuts both ways: there is no favorable FWD to point to, but also no adverse record, no Board claim construction to fight, and a clean slate for an invalidity theory — including an IPR filed the day after you are served.

---

No proceedings to report — what I checked instead

Because the standard per-proceeding template would be empty, here is the verification trail and the disambiguation that matters most, since Tigo is a serial PTAB litigant and its proceedings are easy to misattribute to this patent.

Verified absences:

• No IPR / PGR / CBM naming 8,102,074 in the ODP structured data, in PTAB E2E, or in any petitioner pleading I retrieved.
• No ex parte reexamination surfaced. (Weakest of the negatives — reexam records are less well indexed in the aggregators I could reach. A Reexam Certificate of Correction search in Patent Public Search would close this.)
• No Federal Circuit appeal arising from this patent. The CAFC activity in the Tigo family is the SunSpec appeal, and it is not about this patent — see below.

Statutory eligibility, for completeness:

Near-miss disambiguation — Tigo PTAB activity that is NOT this patent. These are the proceedings a careless search will wrongly attribute to 8,102,074:

None of these names 8,102,074. The SunSpec petitions' real parties in interest were identified as SunSpec Alliance, Fronius USA LLC, MidNite Solar, Inc., SMA Technology AG, Zerun Co., Ltd., and Zhejiang Jiaming Tianheyuan Photovoltaic Technology Co., Ltd. — none of which petitioned against this patent. Source for the SMA status table: Defendants' continued IPR status update, Tigo Energy Inc. v. SMA Solar Technology America LLC, No. 1:22-cv-00915-GBW (D. Del.), D.I. 118 (2024-02-21), available via Docket Alarm.

The one CAFC development in the family (context only, not this patent). SunSpec Alliance v. Tigo Energy, Inc., Fed. Cir. No. 2023-1741 — appeal from the IPR2021-01286 / -01287 final written decisions on the '321 and '770 patents. The nonprecedential opinion dated 2025-08-05 (https://www.cafc.uscourts.gov/opinions-orders/23-1741.OPINION.8-5-2025_2554121.pdf) addresses SunSpec's three claim-construction/obviousness arguments on the "predetermined number" of skips limitation and rejects each; the excerpted text contains no reversal. Caveat: I retrieved the opinion's reasoning rather than its final disposition line, so state it as "the panel rejected SunSpec's arguments" rather than quoting an "affirmed" decree. Either way, the '074 patent is not mentioned in that appeal because it was never in that IPR.

---

Strategic summary

Claim-by-claim status for 8,102,074. There is nothing to report because nothing has been adjudicated. Every claim stands exactly as granted on 2012-01-24: independent claims 1, 8 and 16, plus dependents 2–7, 9–15 and 17–19 (19 claims total). No claim is canceled, no claim has been narrowed by amendment, no claim has been construed by the Board. If a demand letter from Tigo Energy Innovations LLC cites claim 1's "single controller" plus two-LMU architecture, or claim 8's "predicted future voltage … and a maximum regulatory safety voltage," or claim 16's two-segment monitoring method, there is no PTAB ruling you can hold up to show any of those limitations have been read narrowly or held invalid. The claims are also entirely untested in any forum — the earlier litigation research found no district-court assertion either, so there is not even a Markman record to work from.

Estoppel landscape: clear. No IPR was ever filed, so § 315(e)(2) estoppel bars no one with respect to this patent. Two points follow. First, a defendant served today faces no restriction on the art it may raise — IPR, reexam, or district-court invalidity are all open, and there is no risk that an earlier petitioner's grounds were "reasonably could have been raised" and are now lost. Second, do not over-read the Altenergy and SunSpec proceedings: estoppel under § 315(e) is patent-specific, so Altenergy's settlement of IPR2021-00540/-00541 (on the '770 and '321 patents) and SunSpec's FWDs (on the same two patents) impose no constraint on grounds against the '074 patent. The § 315(b) one-year bar begins only when a complaint alleging infringement of this patent is served on the petitioner or its privies.

Pattern signals. Three things are worth flagging, and one of them is new information from the assignment research that a defendant should weigh carefully. (1) No repeat petitioner on this patent — because there has been no petitioner at all. Other Tigo patents attracted Altenergy (2021, two petitions, settled), SunSpec (2021, two petitions), and SMA (2023, three petitions); this one attracted none, which suggests it sits outside the SunSpec Rapid Shutdown Specification claim charting that drove the family's PTAB activity in 2021–2023. (2) Patent owner appellate posture: aggressive. Tigo litigated the SunSpec FWDs through to a Federal Circuit opinion rather than settling, and litigated the SMA matter to a 2026 district-court settlement, so its successor can be expected to defend any IPR vigorously rather than fold. (3) No defensive aggregator in the chain. The earlier assignment analysis found the chain terminates at Tigo Energy Innovations LLC, a newly-formed Texas LLC whose stated business the Patent Purchase Agreement itself describes as "licensing and/or asserting the Assigned Patents" — not RPX, Unified Patents, LOT, or OIN. Unified Patents was not a petitioner here. That matters strategically: this patent has not been neutralized, and there is no aggregator-funded challenge brewing that a defendant could free-ride on. The 2025-10-02 corrective assignment (Reel 58755 / Frame 0516) immediately preceding the 2025-12-16 sale reads as standing clean-up before assertion.

Cross-section flag. One discrepancy with the earlier-generated "Patent summary": that section lists Int. Cl. H04B 3/54 / USPC 307/5, while the authoritative full text in this prompt lists CPC H10F 77/955 (photovoltaic circuit arrangements) and Y02E 10/50. The classifications differ because they come from different vintages of the patent's face; the CPC set in the full text is the current one. This has no bearing on PTAB eligibility, which for a pre-AIA patent turns on filing date alone.

---

Recommended next steps

If you are a defendant (or counsel evaluating a demand from Tigo Energy Innovations LLC):

1. Say plainly in any pre-suit position paper that the patent has zero PTAB history. There is no FWD to cite, so do not represent otherwise. The accurate framing is: "No claim of U.S. 8,102,074 has ever been challenged, construed, or adjudicated."
2. Treat IPR as fully available and unconstrained. No § 315(b) clock, no § 315(e) estoppel, no Fintiv problem from a co-pending action on this patent (there is none), and no Sotera-style stipulation needed. If served, the one-year § 315(b) clock starts from service on you or your privies.
3. Check the '074 patent's own face before building a theory. Section 315(e) estoppel is irrelevant, but note that the granted claim 1 as printed does not recite "maximum regulatory safety voltage" or "predicted future voltage" — those limitations live in claims 8 and 19. Claim 19 depends from claim 18, which depends from the monitoring method; do not assume claim 19 is coextensive with claim 8. Any claim chart must use the granted text of this patent, not the continuation US 8,274,172, which carries a different claim 1 and is subject to a terminal disclaimer.
4. Do not rely on the '321/'770 PTAB record as a shortcut. A ground that failed against the '770 patent's "predetermined number of skips" limitation is not a ground against this patent's string-bus voltage-limiting claims, and the SunSpec FWD reasoning is not binding on a fresh panel.

If you want to close the remaining gaps:

• Reexam check: run a Reexamination Certificate search for 8,102,074 in USPTO Patent Public Search (https://ppubs.uspto.gov/) to convert my "no reexam found" from a search-artifact negative into a confirmed one.
• Docket watch: sweep D. Del., N.D. Cal., W.D. Tex., E.D. Tex. and the ITC for Tigo Energy Innovations LLC as plaintiff, using the exact strings 8102074 and 8,102,074. Given the 2025-12-16 transfer and the PPA's stated asserting purpose, any new complaint is the trigger event that would make a § 315(b) clock real.
• Articles of the new owner: pull the Texas SOS record for Tigo Energy Innovations LLC (registered agent, managers). The earlier analysis could not determine whether it is a captive Tigo vehicle or an independent purchaser; that answer changes nothing about PTAB availability but a great deal about settlement posture.

Primary links:

• PTAB E2E / Patent Trial and Appeal Board Center: https://ptacts.uspto.gov/ptacts/
• USPTO PTAB decisions: https://www.uspto.gov/patents/ptab/decisions
• Google Patents legal events for this patent: https://patents.google.com/patent/[US8102074B2](/patent/US8102074B2)/en
• CAFC No. 2023-1741 nonprecedential opinion (2025-08-05), SunSpec Alliance v. Tigo Energy, Inc. — different patents: https://www.cafc.uscourts.gov/opinions-orders/23-1741.OPINION.8-5-2025_2554121.pdf

Explicit uncertainties: (a) the "no PTAB activity" conclusion is a structured-data negative corroborated by search, and I did not query PTAB E2E's docket directly — I recommend a name/number query in PTAB E2E for 8102074 to make it airtight; (b) the ex parte reexam negative is the weakest link, as noted; (c) I have not verified whether the January 2026 transfer to Tigo Energy Innovations LLC included this specific patent in the PPA's Exhibit A-1 schedule, though the recorded per-patent assignment event supports it.

I'll research the assignment chain for US 8,102,074 against live sources before writing anything.

Critical find: a December 2025 patent purchase agreement. Let me verify whether the '074 is in the assigned patent schedule and nail down the earlier chain.

Assignment Record Analysis — US 8,102,074 B2

Method note / caveat up front: I was not able to query the USPTO Assignment Center's structured API directly in this session. The reel/frame and conveyance data below is drawn from (a) the legal-events block on the patent's own Google Patents record, (b) the SEC-filed Patent Purchase Agreement that produced the final assignment, and (c) Tigo's public licensing records. Where a reel/frame number was not retrievable, I say so rather than inventing one. Only one reel/frame number appears in the sources I could verify — Reel 58755 / Frame 0516 — and it appears because the 2025-10-02 corrective assignment cites it by number. Direct verification: USPTO Assignment Center · legacy search · Google Patents legal events.

---

Inventors

All three executed an assignment of assignors' interest to Tigo Energy, Inc., recorded contemporaneously with the 2009-09-18 filing. Hadar and Arditi are consistently identified in Tigo's PCT filings as Tigo's own listed inventors at Tigo's Los Gatos address (e.g., agent-of-record listing on WO 2010/062662 A3, assignee Tigo Energy, Inc., 170 Knowles Drive, Los Gatos, CA).

Unusual patterns: None found — and I want to be precise about what that means. I found no evidence of inventors departing and re-filing, no evidence of a separate inventor-owned entity, and no evidence of inventor-retained reversionary rights. I also could not independently verify the inventors' departure dates from Tigo, so I cannot affirmatively rule out a departure cluster. This is a case where the absence of a finding is weak evidence, not strong evidence. Note that this is the opposite of the classic pre-fire-sale tell (all inventors exit within 12 months): Hadar and Arditi are founder-level figures who remained associated with the company brand for years.

---

Original assignee

Tigo Energy, Inc. (Los Gatos, California; Delaware corporation) — named on the issued patent, and the assignee on the 2009-09-18 inventor assignment.

• Primary line of business: Module-Level Power Electronics (MLPE) — solar optimizers, rapid-shutdown devices, and monitoring hardware plus cloud software. The company is a practicing entity: it has shipped millions of MLPE units (Tigo's own disclosures state over 10 million MLPE devices shipped, installations in 100+ countries, ~2.3 GWh monitored daily).
• Does it ship products embodying the claims? Yes, and unusually directly: US 8,102,074 is itself listed on Tigo's public virtual patent-marking page as one of the licensed patents covering third-party rapid-shutdown devices — specifically the Tigo–APsmart licensed patents (alongside 7,602,080; 7,807,919; 7,884,278; 8,271,599; 8,274,172; 8,325,059; 8,933,321; and others), and it recurs across the QC Solar, GoodWe, MidNite Solar, Zerun, and Delta license groups. See e.g. https://tr.tigoenergy.com/rsdvirtualmarking and https://nl.tigoenergy.com/rsdvirtualmarking. That is meaningful: the '074 is not a dormant paper patent — it is an actively out-licensed asset covering competitor rapid-shutdown products.
• Current status: Operating, and publicly traded — NASDAQ: TYGO (IPO May 2023). Not acquired, not dissolved, not in bankruptcy. However, see the financial-stress note under Signal 6: the December 2025 patent divestiture was negotiated against a backdrop of thin net tangible book value and repayment of a convertible note.

---

Assignment timeline

Chronological, most recent first is not used — this is oldest-to-newest. All dates are execution/recording dates as surfaced; the patent's legal-events block does not separately distinguish execution from recording for the earlier entries.

2009-09-18 / recorded 2009-09-18 — Reel/frame not retrieved

• Conveyance: Assignment of assignors' interest (inventor-to-company)
• Assignor: Ron Hadar; Shmuel Arditi; Maxym Makhota (individually)
• Assignee: Tigo Energy, Inc.
• Correspondent: not retrievable from available sources
• Context: Ordinary inventor assignment at filing — the original corporate vesting.

2018-02-12 — Reel/frame not retrieved

• Conveyance: Security Interest
• Assignor: Tigo Energy, Inc.
• Assignee: Western Alliance Bank
• Correspondent: not retrievable
• Context: Securitization — a lender taking a blanket lien over the portfolio, not a transfer of ownership.

2020-03-27 — Reel/frame not retrieved

• Conveyance: Security Interest
• Assignor: Tigo Energy, Inc.
• Assignee: UMB Bank, National Association, as Trustee, and Newlight Capital, LLC
• Correspondent: not retrievable
• Context: Securitization — refinancing/replacement of the 2018 Western Alliance facility with a trustee-and-lender structure (Newlight Capital LLC first appears here).

2020-04-07 — Reel/frame not retrieved

• Conveyance: Release of Security Interest
• Assignor: Western Alliance Bank
• Assignee: Tigo Energy, Inc.
• Correspondent: not retrievable
• Context: Collateral release, consistent with the 2020 Newlight facility paying off and displacing Western Alliance.

2022-01-18 — Reel/frame not retrieved

• Conveyance: Assignment of Assignor's Interest
• Assignor: Tigo Energy, Inc.
• Assignee: Newlight Capital LLC
• Correspondent: not retrievable
• Context: This is the entry that needs the closest scrutiny. It is docketed as an assignment of interest rather than a security agreement, yet it sits between two security-interest entries and is followed by a 2023 release. Read in context it looks like a collateral or convertible-note-linked assignment rather than an outright sale — but the conveyance label as recorded says "assignment." I flag this as unresolved: an outright 2022 sale to Newlight that was somehow unwound in 2023 would be an unusual and significant fact; a collateral assignment that the USPTO indexed under the "assignment" subscriber is the more probable explanation. Verify the underlying document image before relying on either reading.

2023-02-22 — Reel/frame not retrieved

• Conveyance: Release of Security Interest
• Assignor: Newlight Capital LLC; UMB Bank, National Association
• Assignee: Tigo Energy, Inc.
• Correspondent: not retrievable
• Context: Lien release returning clean title to Tigo — consistent with repayment of the underlying facility.

2025-10-02 — Reel 58755 / Frame 0516 (cited as the corrected record)

• Conveyance: Corrective Assignment — "corrective assignment to correct the nature of conveyance previously recorded at Reel 58755, Frame 0516"
• Assignor: Tigo Energy, Inc.
• Assignee: Newlight Capital LLC
• Correspondent: not retrievable
• Context: Chain-cleaning. A corrective recording to Newlight, ~10 weeks before the portfolio sale closed. This is a meaningful sequencing signal: the Newlight lien/assignment record was being repaired in late 2025 so the patent could be conveyed free of encumbrances — which the purchase agreement expressly required (Sections 3.1 and 4.4: assignment "free and clear of any and all Encumbrances," with lien releases delivered at closing).

2026-01-20 — Reel/frame not retrieved

• Conveyance: Assignment of Assignor's Interest
• Assignor: Tigo Energy, Inc.
• Assignee: Tigo Energy Innovations LLC
• Correspondent: not retrievable
• Context: Transfer to a licensing/assertion vehicle. This is the recording that implements the Patent Purchase Agreement dated December 16, 2025 among Tigo Energy, Inc., Tigo Energy AI Ltd. (an Israeli company, formerly Foresight Energy Ltd.), as Seller, and Tigo Energy Innovations LLC, a Texas limited liability company, as Purchaser.

On the December 2025 transaction (the substance behind the final entry): aggregate purchase price $17,750,000–$17,950,000, of which $15,000,000 was paid at the initial closing; a $2,750,000–$2,950,000 holdback payable within four months; and up to $5,000,000 of royalties payable to the seller after closing. Seller received a non-exclusive, perpetual, irrevocable grant-back license to practice the assigned patents in its own products; Purchaser received an exclusive, worldwide, sublicensable license to the TIGO ENERGY trademark "solely in connection with the licensing and/or assertion of the Assigned Patents." The agreement also conveys causes of action and damages rights for past infringement and includes a "No Contest" covenant barring the seller from ever challenging validity. Tigo's May 2026 8-K confirmed the seller was paid the remaining $2,750,000 on May 14, 2026, with "net proceeds to the Seller… approximately $583,000, after the repayment of expenses."
Sources: Exhibit 10.1, SEC EDGAR · 8-K, Dec 17 2025 · May 2026 8-K · Tigo 424B5.

Important limit on this finding: the purchase agreement defines the "Assigned Tigo Patents" by reference to Exhibit A-1, which is redacted/not public. I could not directly confirm that US 8,102,074 appears on Exhibit A-1. The inference that it does is strong — the recording date (2026-01-20) trails the agreement's effective date by ~5 weeks, and the assignee matches the Purchaser exactly — but the exhibit itself is the evidence that matters. Treat "the '074 was sold in the December 2025 portfolio sale" as high-confidence inference, not documentary verification.

---

Timeline diagram

timeline
    title Ownership of US 8102074
    2009 : Inventors assign to Tigo Energy Inc
         : Application filed 18 Sep 2009
    2012 : Patent issued 24 Jan
    2018 : Security interest to Western Alliance Bank
    2020 : Security interest to Newlight and UMB
         : Western Alliance lien released
    2022 : Assignment of interest to Newlight Capital
    2023 : Newlight and UMB liens released
    2025 : Corrective assignment to Newlight at 58755 0516
         : Portfolio sold to Tigo Energy Innovations LLC
    2026 : Final assignment recorded 20 Jan

---

NPE / troll-pattern signals

1. Shell-entity transfer — present

The patent has moved from a practicing operating company (Tigo Energy, Inc., which ships MLPE hardware) to Tigo Energy Innovations LLC, a Texas limited liability company with no products, formed for the transaction, and acquiring the portfolio for cash. The agreement makes the entity's purpose explicit rather than implicit: the Purchaser takes an exclusive trademark license to use "TIGO ENERGY" solely in connection with "the licensing and/or assertion of the Assigned Patents." An entity whose only licensed use of a brand is assertion is, by the agreement's own terms, an assertion vehicle. Evidence: assignment of assignor's interest recorded 2026-01-20, implementing the agreement effective 2025-12-16.
Caveat that cuts against a clean "shell" call: the LLC shares branding, personnel lineage, and economics with the seller (seller retains royalties and a grant-back license). It is a related-party vehicle, not an anonymous arm's-length purchaser. The "single-purpose Delaware or Texas LLC with no products" tell is satisfied; the "designed to look unrelated" tell is not.

2. Known asserter in the chain — not present

I compared the chain against the standard NPE roll (Acacia/ACT, Marathon Patent Group, Intellectual Ventures, IPNav, Wi-LAN, Mosaid/Conversant, Vringo, Pendrell, Innovatio IP Ventures, MPHJ, Lumen View, Round Rock Research, Document Generation Corp, Spangenberg entities). No match. Newlight Capital LLC is a secured lender, not an asserter — it enters the chain on 2020-03-27 as a secured party alongside UMB Bank as trustee, and releases on 2023-02-22. Critically, note the data-hygiene trap: Unified Patents' portal shows "Newlight Capital LLC" as "Parent Company / Current Assignee" on a wide swath of Tigo patents (e.g., US-20110304215-A1, US-7884278-B2, US-20110173276-A1). That is an artifact of the same blanket security-interest recording, not evidence that Newlight holds title. Do not cite those aggregator pages as proof of a Newlight NPE. Tigo Energy Innovations LLC does not appear on any RPX or Unified high-frequency-plaintiff list that I could surface.

3. Repeat correspondent across the chain — unclear

This is the signal I most wanted and could least verify. The correspondent-of-record fields were not retrievable from any source I could reach; the Google Patents legal-events block exposes assignor/assignee/conveyance but not the recording correspondent. The only attorney name I surfaced anywhere in this family is John P. Ward, listed as agent on Tigo's related PCT (WO 2010/062662 A3) — but that is prosecution-side, on a different application, and proving nothing about who recorded the assignments. I am not going to name him as the chain's recording correspondent. Action item: pull the correspondent fields for all eight recordings directly from the Assignment Center; if the same firm appears on the 2022 Newlight assignment, the 2025 corrective assignment, and the 2026 Innovations assignment, that recurrence is the finding — and it would be the single most probative item in this entire analysis.

4. Cascading transfers — not present, with a timing flag

The genuine successive third-party transfers are 2022-01-18 (to Newlight Capital) and 2026-01-20 (to Tigo Energy Innovations LLC) — roughly four years apart, with a release in between returning title to Tigo. That is not a <24-month cascade. However, there is a dense 2025–2026 cluster: a corrective assignment on 2025-10-02, the sale agreement effective 2025-12-16, and the implementing recording 2026-01-20 — three ownership-related events in under four months. The corrective assignment to Newlight ~10 weeks pre-closing reads as pre-sale chain-cleaning, which is a standard step in a patent portfolio sale and is not itself an NPE tell.

5. Pre-litigation transfer — present as structural risk, not as an event

There is still no suit I can identify asserting the '074 (consistent with the litigation section of this analysis). So a transfer-within-6-months-before-a-first-suit does not yet exist as a fact. But note the structure: the purchase agreement expressly conveys past-infringement causes of action and the right to collect damages, bars the seller from contesting validity, and takes an exclusive mark license for assertion. That is the full toolkit for a sequencing-to-suit, assembled in December 2025. I am flagging the capability and intent, not asserting a filed case, because I have no evidence of one.

6. Bankruptcy fire-sale — not present

Tigo Energy, Inc. is a NASDAQ-listed operating company (TYGO); the patent transfer was a negotiated asset sale, not a Chapter 7/11 disposition. The proceeds were applied to repay a convertible note in full (per the 424B5) — a liquidity and deleveraging transaction. However, I would not describe the seller's financial condition as comfortable: the same 424B5 discloses that, after giving effect to the patent sale, the convertible-note repayment, and a 5,000,000-share offering at $3.00, pro forma as-adjusted net tangible book value as of 2025-09-30 would have been approximately $22,766,000, or $0.31 per share. That is a motivated seller. A motivated seller with deep IP and no other financing options is exactly the profile that precedes more portfolio sales — worth monitoring, but it is not a bankruptcy fire-sale on these facts.

7. Privateering — present (strongest signal)

This is the clearest, best-documented finding in the entire chain, because the terms are in an SEC-filed agreement:

• Operating company sells patents to a licensing-only affiliate;
• Seller keeps a grant-back license to keep practicing the patents in its products;
• Seller keeps up to $5,000,000 of royalty upside — i.e., it participates in the new owner's licensing revenue;
• Purchaser gets an exclusive TIGO ENERGY trademark license "solely in connection with the licensing and/or assertion" of the patents — effectively loaning the operating company's brand to the assertion campaign;
• "No Contest" covenant bars the seller from ever challenging validity;
• "Excluded Entity" / anti-immunization provisions expressly bar the seller from using its grant-back license "for the purpose of immunizing the products of any Excluded Entity from claims of infringement" — language whose only function is to protect a third-party enforcement campaign against named competitor targets.

That last provision is the tell. Traditional patent sales do not need to police whether the seller is helping third parties avoid the buyer's claims.

8. Defensive aggregator (anti-NPE) — not present

The chain terminates at Tigo Energy Innovations LLC, a monetization vehicle. It does not terminate at RPX, Allied Security Trust, LOT Network, Unified Patents, or the Open Invention Network. The patent has not been neutralized. Note the inverse datapoint: SunSpec Alliance's IPR campaign against sibling Tigo patents ('321, '770) is the closest thing to coordinated defensive action in this family, and it targeted other patents — not the '074.

---

Verdict

NPE — high confidence.

Two strong, independently documented signals drive this: (1) a shell-entity transfer — US 8,102,074 moved from practicing operating company Tigo Energy, Inc. to Tigo Energy Innovations LLC, a Texas LLC with no products, by an assignment of assignor's interest recorded 2026-01-20 under the December 16, 2025 Patent Purchase Agreement, whose grant to the buyer of an exclusive TIGO ENERGY trademark license "solely in connection with the licensing and/or assertion" of the patents removes any ambiguity about the entity's function; and (2) privateering — the seller retained a grant-back license, up to $5,000,000 in royalties, a No Contest covenant, and an anti-immunization/"Excluded Entity" clause, which is the signature of an operating company outsourcing enforcement while keeping the economics. Note the difference from a classic troll chain: this vehicle has never been used to sue on this patent, the seller and buyer are related parties, and the seller's own products practice the claims under the grant-back license — so the pattern is high-confidence while litigation risk attributable to the '074 specifically remains unproven.

Two things I could not verify and would not want relied upon: (a) whether US 8,102,074 is listed on Exhibit A-1 ("Scheduled Tigo Patents") — the exhibit is not public, so the transfer is a strong inference from matching dates and assignee, not a documentary confirmation; and (b) the correspondent of record on every recording. The only reel/frame I could confirm is Reel 58755 / Frame 0516, and only because the 2025-10-02 corrective assignment cites it by number. Pull the correspondent fields from the Assignment Center for all eight recordings — recurrence of a single attorney across the 2022 Newlight assignment, the 2025 correction, and the 2026 Innovations assignment would convert Signal 3 from unclear to present and would materially strengthen the picture.

I'll pull the actual "References Cited" data for this patent rather than rely on memory, since the authoritative text block I was given does not reproduce the citation list.

Let me pull the front-page citation list directly from the granted patent document.

Prior-art analysis — US 8,102,074 B2

0. A threshold problem I have to flag before anything else

I could not retrieve the patent's own front-page "(56) References Cited" list. The authoritative patent text block supplied to me for this analysis (the Google Patents fetch at https://patents.google.com/patent/US8102074/en) does not reproduce the citations section — it jumps from the abstract/description into the claims. It contains a "Prior art keywords" line (voltage; string bus; solar module; controller; lmu) but no list of cited U.S. patents, cited foreign documents, or cited non-patent literature.

I then ran targeted retrieval against patentimages (the scanned/published front page), uspto.report, FreePatentsOnline, and Google Patents' citation tab. What came back was consistently the citation list of other patents that cite the '074 — not the '074's own reference list. Every FreePatentsOnline "US Patent References" block I surfaced belonged to a different patent ('172, '246, '825, '488, '916, '738, '558, '1228) and merely lists the '074 among its references.

I am therefore not going to present a fabricated table of "each patent citation for 8102074." Under my operating rules I must say so explicitly rather than invent reel-accurate citations. What follows is (a) the citation data I did verify, (b) a clearly-labelled best-effort identification of the prior art that matters, and (c) the exact verification steps needed to convert this into a citable reference list.

---

1. What I did verify: the forward-citation landscape

These are the verified, self-consistent finding of this session — the patents and publications whose own reference lists cite US 8,102,074 (19 U.S. patents plus several foreign family members). This is the mirror image of the request, but it is real, sourced data and it is directly useful for claim-context and for locating the art that did get cited in this family's prosecutions.

Sources: FreePatentsOnline pages for the above grants (e.g. https://FreePatentsOnline.com/[8274172](/patent/8274172).html, https://FreePatentsOnline.com/[8860246](/patent/8860246).html, https://www.freepatentsonline.com/[9431825](/patent/9431825).html); Google Patents "cited by" panels at https://patents.google.com/patent/JP5498388B2/en and https://patents.google.com/patent/CN103155349A/en.

Note the direction of inference: none of these is § 102 prior art against the '074. They post-date it. Do not let anyone map them onto the claims as anticipatory art.

---

2. The prior art that actually matters — best-effort identification

Below is the art I can identify from this patent family's own prosecution and litigation record, mapped onto the '074's claims. Label each item "not verified as appearing on the '074's face" until you pull the front page. I have marked my confidence on each.

2.1 The family's litigated art (SunSpec / SMA IPRs)

The SunSpec Alliance IPRs against Tigo family members put a defined prior-art cluster on the record. From the IPR2021-01287 (US 10,256,770) petition grounds:

Source for the ground listing: IPR2021-01287 papers, Docket Alarm (https://www.docketalarm.com/cases/[PTAB](/ptab)/IPR2021-01287/SunSpec_Alliance_v._Tigo_Energy_Inc/). Also Dunton declaration and Moine/Kronberg documents in IPR2021-00541 (https://www.docketalarm.com/cases/PTAB/IPR2021-00541/Altenergy_Power_Systems_Inc._v._Tigo_Energy_Inc/).

2.2 References from the sibling's PCT search report (prosecution history)

The prosecution history of US 10,256,770 (submitted as Exhibit 1004 in IPR2021-01287) reproduces a PCT search report for PCT/US2011/033658 (filed 2011-04-22, priority 2010-04-22) that categorises:

Critical timing problem for two of these: US 2009/0207543 (2009-08-20) and US 2009/0179662 (~2009-07-16) published after the '074's 2009-07-30 priority date. Under pre-AIA § 102 they are not § 102(a)/(b) art against the '074's claims on that priority date (they could only be § 102(e) art if their U.S. filing dates predate 2009-07-30 — a separate check). The '074, filed 2009-09-18 claiming provisional 61/273,210 of 2009-07-30, is pre-AIA (pre-2013-03-16), so § 102(a)/(b)/(e) all apply. Do not assume the sibling's cited art transfers as § 102 art.

---

3. Claim-by-claim § 102 mapping (framework, with honest confidence levels)

Because I do not have the '074's own citation list, I am mapping the identified art against claim scope rather than reporting an examiner's rejection. Claim numbering is the granted numbering: 1, 8, 16 are independent; 19 claims total.

Claim 1 — single controller + first and second LMUs

string bus; first and second solar modules each generating a voltage; a single controller configured to limit at least one of the first and second voltages provided to the bus; a first LMU coupled between the first module and the bus and in communication with the single controller; a second LMU coupled between the second module and the bus and in communication with the single controller.

Potential § 102 art:

• Adest US 2008/0164706 A1 (2008-07-10) — likely the closest single-reference candidate if it discloses a central controller communicating with per-module converters. The distributed-architecture references in this cluster typically disclose module-level converters with an central control/communication element; the "single controller" + "each LMU in communication with the single controller" limitation is the crux. Moderate confidence at best — a § 102 rejection on this claim requires the reference to disclose both LMUs being in communication with one shared controller, which distributed-architecture art often does not state cleanly.
• Formge US 2009/0066357 A1 (2009-03-12) — cited category X on the sibling's claims 1–5, 15–18. Given the X categorization on a structurally similar apparatus claim, this is the single most likely § 102 candidate on the '074's claim 1 — but I have not read its disclosure and will not assert an element-by-element match. Verify first.
• Tigo's own US 7,602,080 (Hadar, 2009-10-13) — likely a § 102(e)/self-collision candidate if cited; a same-assignee earlier application is not automatically disqualified under § 102(e), but it can be sworn behind under § 1.131 or § 1.132. Flagged as a possibility only.

Claim 8 — control based on predicted future voltage + maximum regulatory safety voltage

controller configured to control the voltage provided to the bus based on a predicted future voltage as predicted by the controller and a maximum regulatory safety voltage.

This is the hardest claim in the patent to anticipate. I found no reference in the identified cluster that discloses control keyed to a forecast voltage (the spec's own §-defining passage is "an estimated voltage existing at a particular future time," with a five-minute horizon) combined with a regulatory ceiling. § 102 anticipation of claim 8 would need a single reference disclosing both.

• Adest 2008/0164706 / Adest '055 / Adest '178 — these disclose reactive, measurement-driven module control and disconnection, not predictive forecasting. They are § 103 candidates, not clean § 102 art, for claim 8.
• Moine FR 2 894 401 A1 — an energy-management device. No prediction limitation expected.
• Kronberg US 5,054,023 (1991) — an old switching/converter reference; a prediction limitation is implausible.
• Realistic § 102 exposure for claim 8 is low. Where I would look instead: PV monitoring/SCADA art (irradiance-and-temperature forecasting feeding a string-voltage cap), which is not in the identified cluster.

Claim 16 — two-segment monitoring + limiting on regulatory ceiling

monitoring a first voltage across a first string-bus section connecting first module to second module; monitoring a second voltage across a second string-bus section connecting the second module to a voltage output; limiting the voltage output based on a maximum regulatory safety voltage by limiting the voltage of at least one of the first and second solar modules.

Potential § 102 art:

• Adest US 2008/0164706 A1 — if it discloses sensing bus voltage at more than one point and throttling a module to cap overall string voltage. Reasonable candidate; not verified.
• Moine FR 2 894 401 A1 — an energy-management device for a series-string installation could disclose multi-point sensing; not verified.
• Formge US 2009/0066357 A1 — carries the X categorisation against sibling claims 1–5 and 15–18, and claim 15–18 in the sibling corresponds to a method/apparatus claim set structurally close to the '074's method arm. This makes Formge the highest-value single reference to read first.

Dependent claims of interest

• Claim 3 ("the controller comprises: a first LMU … a second LMU") is the opposite architecture to claim 1 (distributed controller rather than single controller). Art that discloses distributed module controllers — the Adest cluster, Wolfs US 8,093,757 (2012-01-10), Tigo's own US 7,602,080 — maps onto claim 3 far more naturally than onto claim 1. Do not use the same reference to argue both; the claims are mutually reinforcing rather than redundant.
• Claim 6 (duty-cycle and phase parameter on the transistor of the first LMU; duty cycle on the second) is narrow and specific; the spec's own §"switches in different LMU's can operate at different phases" passage is the pivot. Phase-offset PWM across module converters is the subject of Tigo's own US 7,602,080-era work — check for self-collision.
• Claim 12 ("predicted future voltage is based on a voltage trend") narrows claim 8; claim 19 mirrors it on the method side. No art identified in this session discloses trend-based prediction.

---

4. What is genuinely missing, and how to close it

The specific artifacts that would convert this framework into a citable table:

1. The '074's front-page (56) block — pull from the granted PDF's face (patentimages storage link for US8102074) or PatentCenter's "Patent Full-Text" → References Cited. This is the single missing input for the task as posed.
2. The examiner's citations and their § 102/§ 103 categorisations — from the file wrapper on PatentCenter (application 12/562,933). This tells you which references were applied against which claims, not merely listed.
3. The Information Disclosure Statements — to separate applicant-submitted art from examiner-applied art.
4. Formge US 2009/0066357 A1 full text — highest priority read, given its X categorisation against the structurally similar sibling claims 1–5 and 15–18.
5. The Adest '055 and '178 full identities — resolve the ground labels from IPR2021-01287 Paper 2 (petition) before citing; I could not resolve them here.
6. Moine FR 2 894 401 A1 and Kronberg US 5,054,023 full texts — confirm dates and check for the "regulatory maximum voltage" limitation, which is the concept most likely to be absent.

---

5. Explicit uncertainties — do not let these be smoothed over downstream

• I did not obtain the '074's own reference list. Everything in § 2 and § 3 is derived from the family's prosecution/litigation record and from the forward-citation landscape, not from the patent's face. The task's requested format — "for each patent citation for 8102074 … which claim(s) it potentially anticipates under § 102" — cannot be completed as literally posed from what I retrieved.
• I have not verified the technical disclosure of Formge, Boniface, Mollton, Moine, or Kronberg. Their citations, dates, categories and claim mappings are taken from retrieved records; their content is not. Do not quote them as disclosing anything I have not read.
• Pre-AIA § 102 governs. The application was filed 2009-09-18 (priority 2009-07-30), before the 2013-03-16 AIA date. "Prior art" means § 102(a)/(b)/(e) art. Two of the sibling's cited references publish after the '074's priority date and therefore do not qualify under § 102(a)/(b).
• Do not conflate the '074 with US 8,274,172 or US 2012/0119584. Same title, same specification, different claims. The sibling's X-categorised reference (Formge) was applied against the sibling's claims 1–5 and 15–18; that mapping is evidence about Formge, not a finding about the '074's claims.
• No AIA trial has ever been filed against the '074 (see the PTAB section of this file). Consequently there is no Board or court construction of "single controller," "local management unit," or "predicted future voltage" to lean on, and no institution record in which the '074's cited art was tested. Whatever § 102 position you build will be first-principles.
• Date note: my environment's date and the task-prompt date differ (2026-09-27 vs. 2026-04-26). Sources dated after April 2026 are flagged where cited; the substantive findings do not turn on the difference.

Verify at: https://patents.google.com/patent/US8102074/en (Patent Citations tab), https://patentcenter.uspto.gov/ (application 12/562,933, References Cited + file wrapper), and the granted-PDF front page for the Examiner's secondary reference tags.

I'll ground this analysis in whatever prior art I can verify for this specific patent's family before writing anything.

Let me pull the full front-page reference list and locate the key family-art references.

I have the front-page reference list (partially). Let me characterize the key references before analysis.

Obviousness Analysis — U.S. Patent No. 8,102,074 B2 under 35 U.S.C. § 103

---

0. Provenance and two threshold flags

Flag 1 — there is no "Prior Art section" in the material supplied to me. The full patent text I was given is the Google Patents rendering of US 8,102,074 (front matter, description, abstract, claims 1–16). It contains no "Prior Art," "References Cited," or "Cited By" section, and the rendering of claim 16 is truncated mid-sentence. The previously generated sections in this file cover the patent summary, litigation, PTAB, and assignments — none of them is a prior-art section either.

Rather than invent a prior-art record, I reconstructed one from two sources I could actually retrieve:

1. The (56) References Cited block on the front page of the '074 itself (US 8,102,074 PDF, page 1: https://patentimages.storage.googleapis.com/08/f0/2a/d8aea7cc33c70c/US8102074.pdf). I retrieved it partially — the U.S. patent list is largely visible but truncated, and the Foreign Patent Documents, "Other Publications," and Primary Examiner fields were not retrieved at all. Treat the list below as incomplete.
2. The prior art of record in this family's IPRs — the SunSpec petitions' exhibit lists (Ex-1005 FR2894401A1 to Moine; Ex-1006 US 5,054,023 to Kronberg; Ex-1007 US 8,531,055 to Adest; Ex-1008 US 9,088,178 to Adest), retrieved from the IPR2021-01286 and IPR2021-01287 file records.

Everything else in this section — my characterization of what a reference discloses, and my knowledge of Solaredge/Wolfs/Ledenev content — is drawn from training data and is flagged as such; it must be verified against the reference texts before it goes into an invalidity contention.

Flag 2 — date. The task states April 26, 2026. The environment states 2026-09-27. The prior sections in this file already flagged this discrepancy. It does not change the analysis, but note that several of the assignment and 8-K facts in the earlier sections are dated after April 2026.

One more caveat that matters legally, not just rhetorically: the '074's prosecution history was not available to me. I cannot tell from the face of the patent which of the cited references the examiner actually applied in a §103 rejection, what arguments were made, or what was amended. Any conclusion about "what the applicant gave up" is provisional.

---

1. Framework

Governing law: Graham v. John Deere Co., 383 U.S. 1 (1966); KSR Int'l Co. v. Teleflex Inc., 550 U.S. 398 (2007). Because the application was filed September 18, 2009 and claims priority to July 30, 2009, pre-AIA § 103 applies, including pre-AIA § 103(c)(1) — which becomes decisive here, because most of the closest art is the applicant's own (see §8).

Level of ordinary skill. No tribunal has set a POSITA level for the '074. The Board took POSITA evidence in the sibling IPRs (Dr. Balog for Tigo; Dr. Dunton for SunSpec), but that record belongs to the '321/'770, not the '074. My proposed level, which is a construction, not a finding: a bachelor's degree in electrical engineering (or equivalent) plus two to four years of experience in power electronics, DC-DC converter design/control, or photovoltaic system design. The subject matter is mature, well-documented engineering; the skill level should be set relatively high, which helps an obviousness case.

Claim construction — no term in this patent has ever been construed by any tribunal. The load-bearing terms, with the specification's own definitions (which will control):

Practical consequence of the construction table: the specification defines its own terms broadly and defines the goal as merely "limiting a portion of the voltage that each solar module provides to the string bus." That self-definition is the single biggest vulnerability in the patent.

---

2. The subject matter of the claims, decomposed

Claim 1 (independent, system) — four elements:

• (a) a string bus;
• (b) first and second solar modules connected to the bus, each generating its own voltage;
• (c) a single controller configured to limit at least one of those voltages as provided to the bus;
• (d) first and second LMUs, each coupled between its module and the bus, and each in communication with that single controller.

Notably absent from claim 1: any regulatory ceiling, any prediction, any trend, any switchable connection, any transistor, any duty cycle. Claim 1 is an architecture claim, and it is the broadest and weakest claim in the patent.

Claim 8 (independent, system) — string bus + solar module + controller "configured to control the voltage provided to the string bus based on a predicted future voltage as predicted by the controller and a maximum regulatory safety voltage."

Claim 16 (independent, method) — monitor V₁ across a first bus section (module 1 → module 2); monitor V₂ across a second bus section (module 2 → output); limit the output "by limiting at least one of a voltage of the first solar module and a voltage of the second solar module."

⚠️ I could not read claim 16's full text. The Google Patents rendering truncates at "monitoring a first voltage across a first st…". The rendition I use here is from the previous section, which took it from uspto.report. Claims 17–19 are likewise unverified. Any charting of 16–19 must be redone against the granted text.

---

3. The reference set

3a. References on the face of the '074 (partially retrieved)

From the (56) block: US 4,580,090 (Bailey, Maximum power tracker); US 4,604,567 (Chety); US 4,873,480 (Lafferty); US 5,027,051 (Lafferty); US 5,504,418 (Ashley); US 7,248,946 (Bashaw); US 7,256,566 (Bhavaraju, determining a maximum power point of photovoltaic cells); US 7,259,474 (Blanc); US 7,276,886 (Kinder); US 7,336,004 (Lai); US 7,378,820; US 7,394,737; US 7,462,955 (McNamara); US 7,518,266 (Eckroad); US 7,518,346 (Prexl); US 7,538,451; US 7,545,053; US 7,560,906; US 7,595,616 (Prexl); US 7,602,080 (Hadar et al. — Tigo's own); US 7,605,498 (Ledenev et al.); US 7,719,140 (Ledenev); US 7,839,022 (Wolfs); US 7,902,361 (Adest). Publications: 2005/0057214, 2005/0112995, 2006/0001406, 2006/0174939, 2006/0185777, 2007/027333 (all Matan); 2005/0057215 (Rohrig); 2007/0019613 (Frezzolini); 2008/0030305 (O'Connor); 2008/0094640 (Garner); 2008/0097655 (Hadar et al.); 2008/0106241 (Deaver); 2008/0121272 (Besser); 2008/0122… (truncated).

3b. References of record in the family IPRs

3c. What I verified about the single most important reference

US 2008/0097655 A1 (Hadar & Davis) — I retrieved its abstract, claims, and description. Verbatim disclosures:

• Claim 1: "a management unit to be interconnected via a network bus to a set of link modules, each link module coupled to a separate local energy production unit, each link module to include a Maximum Power Point Tracking (MPPT) step-up converter and a parameter monitoring unit to produce parameter data for the respective local energy production unit… and the management unit to receive measured parameters from the link modules, and to send control signals to link modules to provide individual operational control of the local energy production units."
• Claim 8: parameter data "comprises one or more of current, voltage, temperature, and panel identification code."
• Claim 9: "the management unit is to transmit power-on and power-off signals to the link modules via the network high voltage bus to power-on and power-off individual local energy production units."
• Description: "All panels 101a-n provide power to a common Network/High Voltage Bus 102… The bus feeds the produced electrical energy to a consumer of the energy, in this case grid-tied inverter 103." And: "Means to enable/disable panels, inverter, battery units, and other energy producing equipment in the system are implemented… website 107 will interact with management unit 104 and that will in turn control the panels."

Mapping that onto claim 1 — this is close to a complete anticipation, and certainly a devastating §103 reference:

Status as prior art: published April 24, 2008, more than one year before the July 30, 2009 provisional → § 102(b) art, which means it survives the § 103(c) common-ownership disqualification (§ 103(c) applies only to art that qualifies solely under 102(e)/(f)/(g)). This is the key structural point of this whole analysis — see §8.

---

4. Combination A — Claim 1

Primary: Hadar '655 (US 2008/0097655 A1). Secondaries: Wolfs '503/'022; Prexl '346/'616; Ledenev '498/'140.

What the secondaries add. Wolfs discloses a per-solar-generator (or per-cell, one-to-nine cells) buck-type DC-DC converter operatively associated with each solar generator, with the trackers connected in series and each controlled to its own maximum power point — i.e., a per-module converter whose output voltage contribution to a series string is set by a control loop, and a demonstration that series-stacked module converters are the way to build string voltage. (My characterization of Wolfs is from general knowledge of this well-known reference; I did not re-read its text in this session.) Prexl adds closed-loop buck-boost control of converter output. Ledenev adds "boundary controlled" conversion. Adest '055/'178 is an equivalent-to-Wolfs alternative: per-module DC-DC converters under a central controller communicating over the power conductors — and it is the one reference in this set whose claim-level disclosure the Board has actually tested, in the '770 IPR.

Why a POSITA would combine. Under KSR, a combination is obvious where it is "the combination of familiar elements according to known methods … yield[ing] predictable results." Here:

1. The problem was the art's own. The '074's background itself recites the constraint: "the voltage on any part of the power line connecting solar modules into a solar array should not exceed 600V… In Europe this limit is 1000V," and that conventional arrays run "well below … limits." A specification's statement of the problem is an admission that the problem was known; it supplies the motivation element directly rather than requiring the challenger to prove it.
2. To build a long series string while respecting a fixed voltage ceiling, the number of options was small and finite. Either (i) use fewer modules (the prior-art expedient, at a direct power cost) or (ii) actively control each module's voltage contribution so the summed bus voltage stays under the ceiling. KSR's "finite number of identified, predictable solutions" and "obvious to try" rationales apply.
3. No new hardware was required. Hadar '655 already puts a controllable converter + monitor + control channel at every panel; Wolfs already puts a series-connected buck at every panel. Once a converter is at the module and the module's voltage can be commanded, holding that converter's output at less than its MPP voltage when the string total would breach the ceiling is a set-point change to an existing control loop — the paradigm KSR case of a predictable result from a known technique.
4. No change in principal operation, no loss of the references' own purposes. Hadar '655 retains per-panel monitoring and control; Wolfs retains per-generator MPPT (the module still operates at its own MPP input; only the output set-point is capped — exactly the architecture Tigo itself later marketed, see §9).
5. The control channel already existed. Both Hadar '655 and Adest teach that commands travel on the power conductor itself, which is precisely claim 1's "in communication with the single controller" and the '074's own FIG. 6/7 teaching that "the command to control the operation of the switch Q1 606 is sent to the LMU 602 over the photovoltaic (PV) string bus (power line) 610."

Claim-by-claim result — Combination A: Claim 1 obvious; on my reading of Hadar '655 it is arguably anticipated, with "limit" supplied by claim 9's per-panel power-on/power-off (see the '074's own definition of "limit," §1). If the patentee successfully argues that "limit" requires modulated reduction rather than on/off, Wolfs/Prexl/Ledenev supply the modulated reduction as an obvious substitution.

---

5. Combination B — Claim 8 (prediction + regulatory ceiling)

Claim 8 adds two limitations to the architecture of claim 1. Each is separately vulnerable.

(i) "a maximum regulatory safety voltage"

This is a published regulation, not an invention. The applicable U.S. provisions are in the National Electrical Code, Article 690 (Solar Photovoltaic Systems) — and the '074's own specification recites the numbers. A POSITA designing a string in 2008–2009 had no discretion here: the ceiling exists in the code, is fixed, and is documented. A limitation whose content is "obey a known published standard" cannot supply patentable weight on its own. Art supporting the concept of limiting module voltage to a ceiling: Takehara US 6,331,670 (Solar cell module having an overvoltage preventive element and sunlight power generation system using the solar cell module) and Kurokami US 5,923,158 (Power control apparatus for solar power generation system). ⚠️ I could not confirm from the truncated (56) block whether Takehara '670 was cited on the '074's own face; it appears in the reference lists of sibling Tigo patents (e.g., US 8,415,552), so it is family-adjacent art rather than confirmed-cited art.

(ii) "predicted future voltage as predicted by the controller"

This is the genuinely non-trivial limitation, and it is where the §103 case is weakest — but still sustainable on the art of record:

• Besser (US 2008/0121272 and the "power extractor detecting power and voltage changes" / "power extractor with control loop" family) appears in the '074's own (56) block. A control loop that detects changes in voltage and power and adjusts on that basis is, at minimum, a controller deriving its next state from the history of measured values — the functional equivalent of a trend. ⚠️ I did not read the Besser texts; a full-text pull is required before this is usable.
• MPPT by definition predicts a future operating point from past samples. Perturb-and-observe and incremental-conductance compute the sign of the derivative from prior measurements and command a next operating voltage. The '074's own face cites Bhavaraju US 7,256,566 ("determining a maximum power point of photovoltaic cells"), Bashaw US 7,248,946, and (in the sibling lists) Kasai US 6,844,739 ("Maximum power point tracking method and device"). If a converter is already computing a future target voltage from a series of past voltage/current samples, the claim-8 requirement of controlling based on a "predicted future voltage" is met by the ordinary operation of the converter that Combination A already places on every module.
• Protective-relay practice. Anticipatory protection — acting on a rate-of-rise or trend before a limit is crossed, rather than after — is standard power-systems engineering. The '074's own spec confirms the motivation: the system is meant to hold the array as close to the limit as possible, and reactive-only limiting causes the array to oscillate across the limit; that oscillation is the articulated engineering reason to predict. The spec's fallback language ("if the controller monitors a certain rate of change it may trigger commands… to fall back to the default voltage") is itself a trend-based control.
• Cold-temperature Voc rise is the textbook motivating example and is expressly recited in the '074's background as a known phenomenon ("cold temperatures increase solar cell efficiency and thus voltage output"). Controlling the output voltage so that the module's contribution is temperature-independent is the direct, predictable response — and the reason the ceiling might be breached at all.

Combination B: Hadar '655 (architecture) + Bhavaraju '566 / Kasai '739 (deriving a future target from a measured trend) + Takehara '670 or Kurokami '158 (module-level voltage-ceiling enforcement), motivated by the known regulatory ceiling and the known temperature/Voc-rise problem. Result: claim 8 obvious. This is the harder claim, and the prediction element is where a patent owner would fight.

---

6. Combination C — Claim 16 (sectioned bus monitoring + limiting)

Hadar '655 + Moine FR2894401A1 or Adest '055/'178 + Takehara '670 ('670 or Kurokami '158).

Claim 16's distinctive element is monitoring two separate bus sections — one between module 1 and module 2, one between module 2 and the output — and then limiting.

• Hadar '655 supplies per-unit voltage and current telemetry to a single management unit (claim 8 of '655: current, voltage, temperature, panel ID). With per-module voltages reported to one master, the voltage across any bus section — the running sum of the contributions upstream of that section — is available at the controller by simple summation. The '074's own specification states exactly this arithmetic: "the voltage on a segment … between any two solar modules … is equal to the sum of the voltage contributions from each solar module that came before that segment."
• Moine and Adest supply central control of a series string of per-module converters.
• Takehara '670 / Kurokami '158 supply module-level enforcement of a voltage ceiling.

Motivation (this one is strong and articulable): you cannot decide which module's contribution to throttle unless you know where on the string the excess arises. In a series string the bus voltage is monotonically increasing toward the output, so the sections nearest the string terminus carry the highest voltage and are the ones that breach the ceiling first. Sensing at the module nodes — i.e., across each bus-section — is the natural, and arguably the only sensible, way to attribute the excess to a module and target the limiting. A POSITA designing a system that must throttle the right module is led directly to sectioned monitoring; it is a predictable arrangement of known sensing at known nodes.

Combination C result: claim 16 obvious.

---

7. Dependent claims

---

8. The § 103(c) trap — read this before drafting a contention

Because the applicant was Tigo Energy, Inc., much of the closest art is Tigo's own earlier work. Under pre-AIA § 103(c)(1) (in force for this 2009 application), subject matter that qualifies as prior art only under § 102(e), (f), or (g) cannot be used in a § 103 combination if, at the time the claimed invention was made, it and the claimed invention were commonly owned or subject to an obligation of assignment to the same person.

Consequences:

• US 7,602,080 B1 (Hadar et al., "Systems and methods to balance solar panels in a multi-panel system") — cited on the '074's face. If it qualifies only as § 102(e) art, it is disqualified from any § 103 combination against the '074. Do not build a combination on it without checking whether it also qualifies under § 102(a) or (b) as a publication. Its § 102(e) status also cannot be taken for granted — I did not verify its filing date in this session.
• US 2008/0097655 (= US 8,751,053) — not disqualified. It published 2008-04-24, more than a year before the July 30, 2009 provisional, so it is § 102(b) art, and § 103(c) reaches only art qualifying solely under 102(e)/(f)/(g). This is why the challenger's primary reference must be '655 and not the '080.
• The common ownership does not evaporate with the 2025–2026 assignment. § 103(c) is evaluated as of the time the claimed invention was made; the 2025-12-16 Patent Purchase Agreement transferring the portfolio to Tigo Energy Innovations LLC (see the assignment section of this file) does not retroactively remove the disqualification. The pre-AIA common-ownership shield attaches to the 2009 facts, not the current owner.

---

9. Objective indicia and counterarguments

Secondary considerations (Graham factor 4):

• Long-felt need — cuts against the patent, not for it. The problem (module voltage vs. a fixed string ceiling) was recognized in the art at least as early as Takehara '670 (2001). A recognized, decades-old need whose solution is a set-point change to an existing control loop is weak ground for nonobviousness.
• Commercial success — nexus problem. Tigo's MLPE products and its Smart-Curve / Reduced Voc feature ("the unit's output voltage does not vary when it is under load in cold temperatures… voltage is fixed no matter what the ambient temperature is" — Tigo Smart-Curve Application Note v1.3.4) are the commercial embodiments. ⚠️ That document post-dates the patent and is NOT prior art; it is cited here only as evidence of the commercial embodiment. Its existence cuts both ways: it confirms an industry need, but the causal link between the claimed architecture and the commercial success is weak, and the same market success is equally explained by the optimizer/rapid-shutdown market generally.
• Licensing program. The '074 appears on Tigo's virtual-marking page among the "Tigo-APsmart Licensed Patents" covering the APsmart RSD-S-PLC, RSD-S-PLC-B, RSD-D and RSD-D-15-1500 rapid-shutdown devices (https://www.tigoenergy.com/rsdvirtualmarking). A licensing program is not objective indicia of nonobviousness absent evidence that licensees took the license because of the claimed subject matter — and here the marking page groups the '074 with ~35 rapid-shutdown patents, suggesting it is part of a package rather than a standalone driver.
• Copying / industry adoption. Potentially available if competitors adopted temperature-independent output-voltage capping after Tigo; requires discovery.
• Praise / unexpected results. None that I can identify from the record.

Counterarguments a patent owner will raise, and my assessment:

---

10. Bottom line

The single most important finding: the applicant's own earlier publication, US 2008/0097655 A1, is § 102(b) art, is cited on the face of the '074, and discloses — verbatim — a single management unit, per-panel link modules each coupled between a panel and a common bus, per-panel voltage/current telemetry, per-panel power-on/power-off commands sent over the power bus, and "means to enable/disable panels." Claim 1's only arguable gap on that reference is the meaning of "limit," and the '074's own specification defines that word to include full on/off states. This is the reference the patent's own examiner put in the (56) block, and it is the reference that most obviously should have been applied.

---

11. What must be verified before any of this is filed

1. Pull the '074's complete (56) block, including the Foreign Patent Documents, "Other Publications," and Primary Examiner fields, and the full PDF of page 1 — my retrieval truncated it.
2. Pull the file wrapper for 12/562,933 — I cannot state whether any § 103 rejection issued, over what art, or what was argued.
3. Pull the full text of US 2008/0097655 / US 8,751,053 and confirm (a) the per-panel enable/disable passage at the paragraph level, (b) whether any disclosure addresses bus voltage limits rather than panel-level control, and (c) the priority chain back to provisional 60/852,961 (2006-10-18).
4. Pull Wolfs, Prexl, Ledenev, Besser, Bhavaraju '566, Kasai '739, Takehara '670, Kurokami '158 and chart them properly. My characterizations of Wolfs, Ledenev, Prexl, Adest, Moine, Kronberg, Takehara and Kurokami in this section rest on general knowledge of these references, not on text I read in this session, and are marked accordingly.
5. Verify claim 16–19 text from the granted patent, since the source I have truncates mid-claim-16.
6. Check whether the § 103(c) common-ownership shield was invoked or waived during prosecution — a matter the file wrapper will show.
7. Do not import the sibling patents' record. The SunSpec IPRs' results are claim- and patent-specific: IPR2021-01286 sustained the challenged '321 claims (Moine + Kronberg failed); IPR2021-01287 canceled '770 claims 12 and 13 over Adest '055. Neither holding tells you anything about the '074's claims, and the '074 has never been before the Board (see the PTAB section of this file). Moine + Kronberg in particular should be treated as a known-failed combination against this family, not a starting point.

I'll first verify the patent at the USPTO and try to recover the full claim text (claims 16–19 were truncated in the authoritative source).

Let me try to recover the exact text of claims 16–19, which the authoritative source truncated.

DEFENSIVE DISCLOSURE — US 8,102,074 B2

Systems and Method for Limiting Maximum Voltage in Solar Photovoltaic Power Generation Systems

---

0. Verification of the identifier (USPTO)

Patent number searched: 8102074 — as a literal string. No similar-number substitution was made.

Front-page data retrieved from the granted patent PDF (patentimages.storage.googleapis.com/08/f0/2a/d8aea7cc33c70c/US8102074.pdf):

False positive excluded (do not import): the numeric token "8102074" also appears as an FDA FAERS case number (Plan B One-Step / Teva, received 23-Aug-2011, "NON-EXPEDITED"). It has no relationship to this patent. This is consistent with the false-positive list already recorded in the litigation section of this file. No other "8102074" match was treated as this patent.

Cross-reference flag (consistent with earlier sections): the front page prints Int. Cl. H04B 3/54, while the Google Patents full-text rendering lists CPC H10F 77/955 / Y02E 10/50. The authoritative full text in this prompt uses the H10F set. Both statements are recorded; nothing below depends on classification.

Verification gap carried forward: claim 16's text is truncated mid-sentence in the authoritative source, and claims 17–19 were not re-verified verbatim in this session. The three independent claims used as the derivation anchors below are 1, 8, and 16, using the claim-16 wording recorded in the earlier "Patent summary" section of this file (which took it from uspto.report). Every derivative below is written to be broader than and independent of any residual uncertainty in claims 17–19.

Date note: the task prompt states April 26, 2026; this environment reports 2026-09-29. The earlier sections of this file already flagged the same discrepancy. It does not affect this document's substance.

---

1. Purpose, legal framing, and publication protocol

This document is a defensive publication (prior-art generating disclosure). It deliberately does not restate, summarize, or characterize the target patent's specification, figures, or embodiments; it generates new technical subject matter on five derivation axes for each independent claim of US 8,102,074 B2.

Legal effect sought. If published and publicly accessible, this document is intended to qualify as prior art under 35 U.S.C. § 102(a)(1) (printed publication / public use / otherwise available to the public) and § 102(a)(2), and as § 102(b)(1)(A) art against later-filed applications claiming the disclosed subject matter. It is drafted to be enabling — the standard for a patent-defeating disclosure is the same § 112(a) enablement standard applied to a patent claim.

Publication protocol (required to fix the prior-art date):

1. Publish to a public, indexed, timestamped repository — IP.com Prior Art Database, or Zenodo with a DOI, plus a Defensive Publications series (e.g., the Linux Defenders / Unified "Defensive Publications" channel).
2. Compute a SHA-256 digest of the final PDF, and anchor it with OpenTimestamps (Bitcoin) or an RFC 3161 TSA token. Record the digest and the anchoring transaction ID on the face of the document.
3. For the AI/DLT derivatives below, do not rely on the blockchain for control; anchor only data hashes (latency, determinism, and gas-cost reasons are given in situ).
4. Retain the raw Mermaid source and all numeric parameter tables in the deposit; a disclosure whose numbers are not reproducible is weak § 102 art.

Interpretation rule followed throughout: every alphanumeric identifier (patent numbers, claim numbers, standard numbers, part numbers) is reproduced literally. Where a real standard's model number is used, it is used as an illustrative host interface, not as an assertion of a certified conformance claim.

---

2. Derivation axes (fixed vocabulary used in the tags)

Architectural shorthand used below (naming only — not a summary of the patent): a Claim-1 system = string bus + two module-connected limiters + one controlling node; a Claim-8 system = controller that acts on a predicted future voltage and a regulatory ceiling; a Claim-16 method = two-section bus sensing + selective limiting.

---

3. CORE CLAIM 1 — Derivative disclosures

Claim-1 anchor elements foreclosed by the derivatives in this section: (i) a string bus; (ii) a first and a second solar module connected to it; (iii) a single controller configured to limit at least one of the two voltages provided to the bus; (iv) a first and a second local management unit each coupled between its module and the bus and each in communication with the single controller.

---

D1.1 — Wide-bandgap LMUs and composite string conductors A1

Enabling description. Replace the limiting element at each module with a wide-bandgap half-bridge: 650 V or 1200 V SiC MOSFETs (or 650 V enhancement-mode GaN HEMTs) in a synchronous buck, driven by an isolated gate driver with a 5 kVRMS barrier and dead time programmable 15–40 ns. Switching at 200 kHz–1 MHz for SiC and 500 kHz–2 MHz for GaN, the output inductor shrinks from ~100 µH (powdered iron, 20 A saturation) to 6–15 µH in a planar or moulded powder-iron toroid, and the bulk electrolytic is replaced by 2 × 22 µF, 1210, X7R MLCC plus a 1 µF polypropylene snubber to hold ripple below 150 mVpp. Replace the string-bus copper with aluminium-clad steel or a Cu–carbon-nanotube composite conductor sized for a 2% voltage drop over a 200 m string at 15 A, and use Al/Cu bimetallic lugs with an anti-galling compound to terminate. Replace the module-side blocking-diode with a synchronous rectifier for a ~0.4 V forward-loss reduction at 15 A.

Control law: the single controller broadcasts a per-LMU output-voltage set-point Vset,i; each LMU runs average-current-mode control with a 2-pole/1-zero compensator, crossing over at fsw/10, and reports Vout and Iout back on the same two-wire link. Total measured conversion loss target: < 0.4% at 20 A, well below silicon MOSFET equivalents, which is the engineering motivation.

What this forecloses: claims to SiC- or GaN-based per-module voltage-limiting converters under a single master; claims reciting composite or aluminium-clad-steel string conductors in a module-level voltage-limiting system; claims reciting a synchronous-rectifier substitution for the blocking diode in the same topology.

flowchart LR
  M1["PV module 1"] --> S1["SiC half-bridge LMU 1"]
  M2["PV module 2"] --> S2["GaN half-bridge LMU 2"]
  S1 -->|"Vset_1"| BUS["Composite string bus 1.5 kV"]
  S2 -->|"Vset_2"| BUS
  SC["Single controller"] -->|"isolated 2-wire link"| S1
  SC -->|"isolated 2-wire link"| S2
  S1 -.->|"Vout, Iout telemetry"| SC
  S2 -.->|"Vout, Iout telemetry"| SC
  BUS --> INV["String combiner and inverter"]

---

D1.2 — Inductorless switched-capacitor LMU with discrete and fractional ratios A1

Enabling description. Substitute the magnetic buck with a switched-capacitor (charge-pump) cascade. Use a 4-stage Dickson or series-parallel SC front end providing fixed conversion ratios of 1:1, 2:1, 3:1, 4:1, then obtain intermediate ratios by burst-mode duty control of the ratio-selected state: with period T and a 2:1 stage, an effective ratio of 1.75:1 is produced by spending 50% of T in 2:1 and 50% in 1:1. To suppress charge-sharing loss, operate in soft-charging mode with a small series "flying" inductor of 0.8–1.2 µH and a resonant period matched to the charge-transfer half-cycle; without it, SC efficiency at 20 A in a 2:1 cell collapses below 90%, which is the engineering reason the series inductor is retained.

Values: flying capacitors 22 µF, 1210, X7R, 25 V, ceramic output capacitor 4 × 10 µF, phase generator at 1.0 MHz with 4-phase interleaving to cut output ripple to < 80 mVpp. The LMU contains no magnetic component larger than 1.2 µH, enabling a conformal potted module-junction-box form factor. Control: the single controller selects (ratio, burst-duty) pairs from a lookup table calibrated per module at commissioning.

Alternative within this derivative (also disclosed): a tapped autotransformer with thyristor or solid-state tap selection for 50/60 Hz systems, giving discrete 100%/80%/60%/40% voltage taps with zero switching ripple and 99.5% efficiency, at the cost of tap granularity.

What this forecloses: inductorless or magnetically-minimal per-module voltage limiters; burst/ratio-hybrid voltage control under a single master; tapped-autotransformer per-module limiting.

flowchart TD
  PV["PV module 3.5 A / 40 V"] --> RATIO["SC ratio selector 4:1 / 3:1 / 2:1 / 1:1"]
  RATIO --> FLY["Flying cap bank 22 uF plus 1 uH soft-charge inductor"]
  FLY --> COUT["Output 4x10 uF MLCC"]
  COUT --> BUS["String bus"]
  SC["Single controller"] -->|"ratio select and burst duty"| RATIO
  SC -->|"sync pulse 1 MHz"| FLY
  COUT -.->|"Vout sense"| SC

---

D1.3 — Deterministic single-pair-Ethernet / TSN LMU network replacing power-line signalling A1

Enabling description. Substitute the communication mechanism between the single controller and the LMUs with IEEE 802.3cg 10BASE-T1S single-pair Ethernet on the same two conductors that carry module power (phantom power), using PLCA (Physical Layer Collision Avoidance) for deterministic multi-drop access at a 1 ms cycle. Alternatively, use RS-485 half-duplex at 921.6 kbit/s with a token that the single controller passes every 2 ms, or CAN-FD at 5 Mbit/s with a 29-bit identifier priority scheme. Time-synchronise with IEEE 802.1AS (gPTP) so all LMU duty-cycle updates land within a ±1 µs window, and schedule the voltage-set-point frames in an 802.1Qbv time-aware gate window reserved for control.

Payload: a 16-byte frame per LMU per cycle — 4 bytes Vset (0.1 V LSB), 2 bytes duty, 2 bytes phase, 2 bytes module temperature, 4 bytes status/CRC-32. At a 1 ms cycle and 32 LMUs the bus utilisation is under 5%, leaving headroom for an I-V curve-sweep command channel. Fail-safe: a watchdog in each LMU holds the last set-point for 3 cycles, then reverts to a hard-wired ceiling (see D1.10).

What this forecloses: claims reciting a specific LMU-to-controller physical layer (single-pair Ethernet, CAN-FD, RS-485 token ring, gPTP-scheduled control windows) in a module-level string-voltage-limiting system; claims reciting phantom-power-over-data-pair in the same topology.

sequenceDiagram
  participant SC as Single controller
  participant SW as PLCA coordinator
  participant L1 as LMU 1
  participant L2 as LMU 2
  SC->>SW: open 802.1Qbv gate window t0
  SW->>L1: BEACON
  L1-->>SW: commit 0
  SW->>L2: TOKEN
  L2-->>SW: commit 1
  SC->>L1: Vset 18.0 V duty 45 percent phase 0 deg
  SC->>L2: Vset 17.5 V duty 41 percent phase 180 deg
  L1-->>SC: Vout 17.9 V temp 41 C crc ok
  L2-->>SC: Vout 17.4 V temp 44 C crc ok
  SC->>SC: recompute string budget Vbus 598 V under 600 V limit

---

D1.4 — On-die, nanoscale PV string with charge-pump LMUs A2

Enabling description. Instantiate the Claim-1 architecture at chip scale. The "string bus" is a metal-4/metal-5 on-die power rail; "solar modules" are 32 monolithically stacked GaAs or InGaP/GaAs sub-cells, each with Voc ≈ 0.95–1.10 V under 1 sun AM1.5G, or with on-die photodiodes under concentrated fibre illumination. Each "LMU" is a 2:1 or 3:1 switched-capacitor charge pump using 1.8 V MIM capacitors at 50–200 MHz, sized for 200 µA–2 mA output. The "single controller" is a subthreshold digital FSM in the always-on island, running at 32 kHz, holding the rail at a target Vrail = 1.20 V — which is simultaneously the target and the "regulatory" ceiling, because 1.8 V is the device reliability limit. Body biasing and clock gating reduce the LMU overhead below 3% of harvested power. Anti-fuse trim selects the per-cell ratio at wafer test to compensate for process variation.

Scale variant in the same derivative: an array of 1 mm² energy-harvesting tiles in a laminated sheet, bussed on printed silver ink with per-tile printed charge pumps.

What this forecloses: on-die or intra-module photovoltaic string limiting with per-cell charge pumps and a single on-die controller; claims to harvested-energy rail clamping to a reliability ceiling.

classDiagram
  class SingleController {
    +float vTarget = 1.2
    +float vReliabilityLimit = 1.8
    +allocateRatios()
    +tick32k()
  }
  class ChargePumpLMU {
    +int ratio
    +float vout
    +float iout
    +setRatio()
  }
  class PVSubCell {
    +float voc = 1.05
    +float isc
    +float temperature
  }
  SingleController "1" --> "32" ChargePumpLMU : allocates ratio
  ChargePumpLMU "1" --> "1" PVSubCell : clamps
  ChargePumpLMU --> ChargePumpLMU : series stack on metal rail

---

D1.5 — Utility-scale 2000 VDC and MVDC strings with hierarchical controllers A2

Enabling description. Expand the same topology to a 1 GW-class plant. A string of 40 modules at 50 Voc per module forms a 2000 VDC nominal bus (the current 1500 V UL/NEC ceiling is breached by using a listed 2000 V module + connector + LMU combination; the disclosed method does not depend on which ceiling applies — it tracks whichever numeric ceiling is configured). Above the string level, strings are paralleled into MVDC collection at ±10 kV using isolated DC-DC string converters. The "single controller" of Claim 1 is retained per string (32–64 LMUs), and a farm controller sits above it, so the disclosed subject matter covers both the single-tier and hierarchical reading.

Enabling specifics: ISO 5 kVRMS barriers between the LV control island and the bus; DC PLC at 100–150 kHz with G3-PLC or proprietary line-impedance-modulation signalling over a 300 m string; segmented string topology so that a single module's bypass does not shift the whole string's voltage; string-end current sensing with a fluxgate transducer to 0.1% accuracy; string-level string-capacitance pre-charge ritual (limited inrush to 0.5 A/ms) to avoid arc-flash during bus make/break. Set-point allocation done by a water-filling algorithm that equalises the clipping loss across modules rather than the voltage, minimising yield loss.

What this forecloses: 2000 V-class or MVDC collection claims using per-module limiting under a master; hierarchical (string/block/farm) controller overlays; line-impedance-modulation signalling in this art.

flowchart TD
  subgraph STRING["2000 VDC string - single controller tier"]
    L1["LMU 1"] --> L2["LMU 2"] --> L3["LMU 3 ... LMU 40"]
    SC1["String controller - SINGLE CONTROLLER"]
    SC1 -->|"setpoints over DC PLC 125 kHz"| L1
    SC1 -->|"setpoints"| L3
    L3 -.->|"Vout telemetry"| SC1
  end
  STRING --> CONV["Isolated string DC-DC converter"]
  CONV --> MVDC["Plus minus 10 kV MVDC collection bus"]
  FC["Farm controller"] -->|"string voltage budget"| SC1
  MVDC --> CENTRAL["Central inverter and grid interface"]

---

D1.6 — Cryogenic, high-temperature, and radiation-hardened LMUs A2

Enabling description. Re-engineer the Claim-1 system for environments where no standard MLPE part survives. Three disclosed variants:

(a) Cryogenic (77 K / −196 °C): SOI CMOS or SiC power stages with no electrolytics; film capacitors (polypropylene or PTFE) with a −200 °C to +125 °C range; control powered by the module itself; gate-drive charge delivery recalculated for a 10× increase in MOSFET carrier mobility at 77 K, which sharply reduces conduction loss but raises the risk of avalanche and requires a softer gate drive (higher Rg,on). Communication by fused-silica fibre with VCSELs rated to 85 K, since copper becomes low-loss but stiff.

(b) High temperature (200–250 °C, downhole / Venus surface): SiC JFETs (normally-on devices with a negative-rail-safe shutdown), ceramic-substrate boards (AlN, 170 W/m·K), gold wire bonds, and no solder — a brazed or sintered silver die attach. Control in depletion-mode GaN with a resistor-referenced bias, since conventional references drift at 250 °C. Both the predicted-future-voltage logic and the module set-point are reduced to a thermally-stable current-mode analog loop (see D8.1 for an analog predictor used in combination).

(c) Radiation (space, 100 krad(Si) TID, 1011 p/cm² at 100 MeV protons): rad-hard FPGA or a triple-redundant latch-based FSM, EDAC on configuration memory, and a watchdog plus scrubbing cycle every 10 s. Set-points stored in MRAM.

What this forecloses: claims tying per-module string-voltage limiting to extreme-temperature-rated semiconductors, fibre communication for the same, or rad-hard control of a module-limiting array.

stateDiagram-v2
  [*] --> Commissioning
  Commissioning --> Nominal : link established
  Nominal --> Nominal : setpoints every 1 ms
  Nominal --> Derated : T greater than 200 C or SEE rate high
  Derated --> AnalogOnly : digital control scrubbed out
  AnalogOnly --> Nominal : scrub OK
  Nominal --> SafeClamp : watchdog timeout 3 cycles
  SafeClamp --> Commissioning : manual reset
  Derated --> SafeClamp : fibre link lost
  SafeClamp --> [*] : array de-energised

---

D1.7 — Cross-domain triad: aerospace 270 VDC, AgTech 480 VDC irrigation, and consumer DC power A3

Enabling description. The same three-element topology — source elements on a shared bus, an interposing limiter per element, one controlling node — is disclosed as applied in three unrelated industries.

(1) Aerospace / eVTOL / HAPS. "Modules" are fuel-cell stacks, solar-wing sub-arrays, or battery packs; the "string bus" is the 270 VDC (MIL-STD-704F) or ±270 VDC distribution bus; each "LMU" is a solid-state power controller (SSPC) with a programmable I²t trip curve (e.g., 10 A nominal, 4× trip at 50 ms) and a bus-voltage clamp that holds the segment under the partial-discharge inception voltage at altitude (Paschen minimum; at 55 000 ft the corona inception for a 5 mm gap falls dramatically, which is the articulated engineering reason for per-segment clamping). The single controller is the vehicle's power management computer, executing a flight-phase schedule; airworthiness hooks: DO-160 environmental qualification, DO-178C software levels, AS5692 arc-fault, 14 CFR 25.1353 wiring separation.

(2) AgTech. "Modules" are PV sub-arrays; the bus is the DC link of a 380–480 V three-phase centre-pivot or submersible pump inverter; "LMUs" are per-sub-array DC-DC converters; the single controller is the irrigation controller. The ceiling enforced is the maximum DC-link voltage of the VFD (which protects the IGBT's 1200 V rating with margin), and the predictive input is the irrigation schedule plus a cloud forecast, so pump start/stop transients do not oscillate the DC link. Telemetry over LoRaWAN to a farm dashboard; integration hooks for soil-moisture and evapotranspiration (ET) models.

(3) Consumer electronics / DC buildings. "Modules" are 48 V rack rectifier/BBU modules on an ORV3-type power shelf busbar; the "LMUs" are hot-swap controllers with droop current sharing (0.5 mΩ); the single shelf controller enforces a 60 V SELV ceiling and, on the consumer side, apply the identical topology to a USB-C PD multiport charger where the "modules" are port regulators and the bus ceiling is 48 V (USB PD 3.1 EPR), with per-port budget allocation from one controller.

What this forecloses: the transfer of module-level bus-voltage limiting under a single master to aviation DC distribution, agricultural DC pumping, ORV3-type shelves, and USB-PD multiport allocation — each a different CPC class from the target patent.

flowchart TD
  subgraph AERO["Aerospace 270 VDC"]
    FA["Fuel-cell string"] --> LA["SSPC LMU"]
    FB["Battery pack"] --> LB["SSPC LMU"]
    LA --> BA["270 VDC bus, corona-limited"]
    LB --> BA
    PA["Power management computer - SINGLE CONTROLLER"] --> LA
    PA --> LB
  end
  subgraph AG["AgTech pump"]
    SA["PV sub-array"] --> CA["DC-DC LMU"]
    CA --> VFD["VFD DC link 480 V"]
    IA["Irrigation controller - SINGLE CONTROLLER"] --> CA
    WX["Cloud and ET forecast"] --> IA
  end
  subgraph CE["Consumer and DC building"]
    R1["Rectifier module"] --> SB["48 V shelf busbar"]
    R2["BBU module"] --> SB
    SH["Shelf controller - SINGLE CONTROLLER"] --> R1
    SH --> R2
    SB --> PORTS["USB-C PD ports, 48 V EPR budget"]
  end

---

D1.8 — AI-driven, IoT-sensed, DLT-verified voltage-budget controller A4

Enabling description. Three integrations, all disclosed with explicit latency/partition constraints so a POSITA can reproduce them.

(a) AI optimization. Replace the rule-based allocator with model predictive control (MPC). State vector x = [per-module Vout, module temperature, plane-of-array irradiance, bus current]; prediction horizon N = 10–60 steps at 1 s; the optimizer solves a convex QP, minimising clipping loss Σ (Pavailable,i − Pdelivered,i) subject to Σ Vout,i ≤ Vlim and per-module Vmax. Solver: an embedded active-set or OSQP-style solver at 50–200 µs solve time on a 200 MHz MCU. A small GRU/LSTM (2 layers, ≤ 32 hidden units, quantised to int8) supplies the 10-minute irradiance forecast that parameterises the QP. Nothing is trained on-chain; models are federated across sites with secure aggregation so no site's raw telemetry leaves the site.

(b) IoT sensing. Add a pyranometer (thermopile or SPN1), a module back-sheet thermocouple (T-type) and a shunt-based current sense; edge-gateway publishes MQTT/Sparkplug-B birth–death certificates at 1 Hz; OPC UA over TSN (IEC/IEEE 60802) where determinism matters. I-V curve sweep commands are pushed down the same link used for set-points (see D1.3) to detect mismatch and PID/degradation.

(c) DLT. Each 15-minute epoch is hashed (SHA-256 over a canonical JSON telemetry record) and the root of a Merkle tree is anchored on a public chain (or a permissioned Hyperledger Fabric channel), plus an ERC-721-style per-module identity for asset registry. Control never depends on the chain — anchor latency (seconds to minutes) and gas cost are disqualifying for a 1 ms control loop; the on-chain commitment exists only for PPA / REC MRV and settlement. Optional optimistic curtailment escrow: a smart contract that pays the owner for curtailed kWh hashed in the anchor.

What this forecloses: MPC/QP formulations of the voltage-budget allocation; embedded quantised forecast models feeding a per-module limiter; Merkle-anchored curtailment MRV in this art; federated learning across PV sites for the set-point model.

sequenceDiagram
  participant WX as Irradiance and IoT sensors
  participant EDGE as Edge node GRU model
  participant MPC as Single controller MPC and QP
  participant LMU as LMU string
  participant ANCH as DLT anchoring service
  WX->>EDGE: 1 Hz irradiance, backsheet temp, Vout
  EDGE->>MPC: 10 min forecast upper bound Vhat
  MPC->>MPC: solve QP minimise clipping s.t. sum Vout under Vlim
  MPC->>LMU: setpoint vector every 1 s
  LMU-->>MPC: Vout and Iout feedback
  MPC->>ANCH: 15 min canonical record hash
  ANCH->>ANCH: Merkle root, anchor tx, issue REC claim

---

D1.9 — Inverse: fail-safe degradation ladder and the "hard ceiling" bypass A5

Enabling description. A version of the Claim-1 system designed explicitly to fail in a predictable, verifiable state. Five declared failure responses, each with a hardware mechanism that does not depend on firmware:

Low-power / limited-functionality mode (also disclosed). The LMU can enter a "harvest-only" state drawing < 50 mW from the module itself (no external supply), in which it simply holds a resistor-programmed duty cycle and reports nothing; this exists so that a fully de-powered array still enforces the ceiling at commissioning. A "dumb diode" mode disables all active control — and is explicitly disclosed as the prior-art baseline that the active modes improve upon.

What this forecloses: claims reciting 0 V rather than Voc as the fail-safe state; claims reciting a resistor- or RC-programmed duty default that survives MCU death; claims reciting an autonomous bootstrap/harvest-only mode in a module-limiting array.

stateDiagram-v2
  [*] --> Nominal
  Nominal --> RC40 : MCU watchdog expiry
  Nominal --> Crowbar : LMU fail short
  Nominal --> ShuntClamp : bus overvoltage
  Nominal --> HotSpotOff : backsheet over 90 C
  Nominal --> ImageCorrupt : control image CRC fail
  RC40 --> Nominal : link restored
  ShuntClamp --> Nominal : load restored
  HotSpotOff --> Nominal : thermal reset
  Crowbar --> [*]
  ImageCorrupt --> [*]

---

D1.10 — Inverse: deliberately "sub-optimal" operation-tracking modes A5

Enabling description. Two inverse modes that trade yield for stability, disclosed as distinct subject matter.

(a) Voltage-locked mode. The LMU holds its output at a fixed voltage regardless of irradiance (a voltage source, not an MPPT). It is deliberately not tracking the maximum power point, so it loses 2–8% of yield but makes the string voltage a design constant independent of temperature — which lets a string contain more modules and removes the cold-morning over-voltage excursion entirely. The controller's only job becomes enforcing Σ Vfixed,i ≤ Vlim, which is trivially satisfied by construction; the "prediction" degrades to a constant.

(b) Current-limiting-friendly mode. The LMU is configured to limit current (not voltage) at the module level, with the string voltage set by the inverter's MPPT, i.e., the architecture is run in its "dual" mode. This is disclosed to foreclose a claim-drafting escape in which an applicant recites voltage limiting and the competitor recites current limiting with the same hardware.

(c) Deliberate oscillation-free mode. The derivative adds a rate limiter (≤ 2 V/s per module, ≤ 20 V/s per string) and a hysteresis band on the ceiling crossing, so the string never cycles across the limit; this sacrifices a small amount of average voltage headroom to remove the inductor acoustic noise and inverter MPPT hunting that arise from crossing the clamp repeatedly.

What this forecloses: fixed-voltage (non-MPPT) per-module operation as a voltage-limit strategy; the current-limiting dual of the same architecture; rate-limited/hysteretic string-voltage clamping.

flowchart TB
  subgraph LOCK["Voltage-locked mode"]
    A["Vout fixed at 30.0 V"] --> B["Sum fixed = 600 V constant"]
    B --> C["No temperature derating required"]
  end
  subgraph DUAL["Current-limit dual mode"]
    D["Iout limited to 9.0 A"] --> E["Inverter MPPT sets string voltage"]
  end
  subgraph OS["Rate-limited mode"]
    F["dV/dt clamp 2 V per s per module"] --> G["Hysteresis 5 V band on Vlim"]
    G --> H["No limit-crossing oscillation"]
  end
  CTRL["Single controller"] --> LOCK
  CTRL --> DUAL
  CTRL --> OS

---

4. CORE CLAIM 8 — Derivative disclosures

Claim-8 anchor elements foreclosed by this section: a string bus; a solar module connected to it; and a controller in communication with the solar module and the string bus, configured to control the voltage provided to the bus based on a predicted future voltage as predicted by the controller and a maximum regulatory safety voltage.

---

D8.1 — Analog lead-compensator predictor with a Zener-referenced regulatory clamp A1

Enabling description. Implement the predictor with no digital computation at all. The module voltage is buffered and fed to an RC/op-amp differentiator (R = 100 kΩ, C = 1 µF, τ = 100 ms; or a lead network with zero at 1.6 Hz and pole at 16 Hz) whose output is the rate of change of the module voltage; a summing amplifier (AD633-class analog multiplier for the temperature-coefficient term) forms V̂ = Vnow + k·dV/dt·Δt, where Δt is set by a sample-and-hold strobe (Δt = 300 s nominal, which corresponds to the thermal time constant that dominates the cold-morning Voc rise). The maximum regulatory safety voltage is a 5.6 V Zener + Schottky-trimmed divider producing a hard analog reference Vlim; the predicted value and the reference are compared by a fast comparator (LM393-class, 1 µs) whose output directly gates the LMU switch. Component-temperature stability is achieved with a buried-Zener reference (2 ppm/°C) rather than a bandgap, because the buried Zener is stable to 200 °C.

Key enabling insight: an analog lead compensator IS a predictor — its output at time t is a linear extrapolation of the recent trajectory. That is the subject matter disclosed here and it defeats any later claim whose novelty rests on "predicting" being done digitally.

What this forecloses: analog/derivative-based voltage forecasting; Zener-referenced regulatory clamps; temperature-normalised extrapolation networks.

flowchart LR
  M["Module Vt"] --> BUF["Buffer"]
  BUF --> DIFF["Lead network R 100k C 1uF"]
  BUF --> SUM["Summing amp"]
  DIFF --> SUM
  TCO["Backsheet NTC and analog multiplier"] --> SUM
  SUM --> PRED["Vhat = Vnow plus k dVdt deltat"]
  PRED --> CMP["LM393 comparator"]
  ZN["Buried Zener 2 ppm per C reference Vlim"] --> CMP
  CMP -->|"gate"| SW["LMU switch"]
  SW --> BUS["String bus"]

---

D8.2 — Statistical and physical model-based prediction A1

Enabling description. Four alternative predictors, each disclosing the model class rather than merely the concept of predicting:

1. Kalman filter on state x = [Voc,ref, irradiance, cell temperature, degradation], with a first-order thermal model Cth·dT/dt = (Tamb − T) / Rth + α·G and a single-diode electrical model Voc = (nkT/q)·ln(Iph/I0 + 1). The prediction is x̂(t+Δ) propagated through the voltage equation. Δ is 60–900 s. Process noise Q tuned to 1e-4; measurement noise R tuned to (0.05 V)².
2. ARIMA(p,d,q) on the 1-minute voltage series, p = 2, d = 1, q = 1, refit hourly on a rolling 24 h window; prediction variance is used to set a probabilistic ceiling margin (e.g., a 99th-percentile bound adds 4 V of margin).
3. Physics-based clear-sky + cloud-passage model: compute the clear-sky Voc from a solar-position algorithm (SPA) and a cell-temperature model; detect cloud passage by the signature of the dG/dt and apply a known transient template (cloud-edge enhancement can raise irradiance 30% above clear-sky for 10–60 s — an under-appreciated over-voltage driver disclosed here).
4. Numerical weather prediction (NWP) ingestion: consume NOAA HRRR/GFS (or ECMWF) fields at 1–3 km, 15-min resolution, plus GOES-R ABI satellite cloud products, and run the single-diode model forward for the ceiling margin. This is the "forecast-grade" variant appropriate to a utility plant.

What this forecloses: Kalman/ARIMA/physics/NWP model classes as the predictive engine for a string-voltage cap; probabilistic margin-setting; cloud-edge-enhancement transient handling.

flowchart TD
  S1["V(t) 1 min series"] --> AR["ARIMA 2 1 1 refit hourly"]
  S2["G, Tamb, wind"] --> KF["Kalman filter thermal plus single-diode"]
  S3["Satellite and NWP HRRR GFS"] --> PHYS["Clear-sky and cloud-passage model"]
  AR --> FUSE["Probabilistic combiner, 99th percentile bound"]
  KF --> FUSE
  PHYS --> FUSE
  FUSE --> MARGIN["Margin = Vhat + 4 V"]
  MARGIN --> CMP["Compare to Vlim regulatory"]
  CMP --> DEC["Allocate per-module setpoints"]
  DEC --> BUS["String bus"]

---

D8.3 — Nanoscale and gigascale prediction horizons A2

Enabling description. Two scale extremes.

Nanoscale (µs horizon). In an on-die harvester string, the dominant over-voltage mechanism is a load-dump transient on a sub-microsecond timescale when a charge pump drops out. The "predicted future voltage" is computed by a 2-tap finite-difference extrapolator clocked at 50 MHz predicting 200 ns ahead and compared against the 1.8 V reliability ceiling; the comparator acts within 5 ns. Disclosed: prediction with Δt on the order of the switching period.

Gigascale (168 h horizon). In a 1 GW plant, the controller predicts 7 days ahead using NWP and day-ahead market prices, and pre-commits a voltage-budget schedule (per 15-min interval) that the string controllers interpolate. Horizon hierarchy disclosed: 1 ms (transient), 1 s (MPPT/cloud edge), 15 min (dispatch interval), 168 h (weather/market).

What this forecloses: multi-horizon hierarchical prediction architectures; sub-switching-period prediction for reliability ceilings; day-ahead voltage-budget scheduling.

flowchart LR
  H1["Horizon 200 ns - 50 MHz extrapolator"] --> C1["1.8 V reliability ceiling"]
  H2["Horizon 1 s - cloud edge template"] --> C2["Vlim regulatory margin"]
  H3["Horizon 15 min - dispatch interval"] --> C3["String budget interpolator"]
  H4["Horizon 168 h - NWP plus day-ahead price"] --> C3
  C1 --> ACT["Actuator set"]
  C2 --> ACT
  C3 --> ACT
  ACT --> BUS["String bus setpoints"]

---

D8.4 — Extreme environments and non-50/60 Hz regimes A2

Enabling description. Prediction and the ceiling both change character under extremes.

(a) Polar / desert. At −40 °C, Voc rises by up to 30% over STC (the β coefficient ≈ −0.30%/°C for c-Si). The disclosed predictor therefore keys on the minimum forecast ambient over the next 12 h, not the instantaneous temperature: the controller sets the module contribution from the forecast coldest point, then relaxes as the day warms. At 55 °C desert, the ceiling is instead a current/thermal constraint and the predictor switches to a temperature-derated ceiling.

(b) Non-50/60 Hz. In a 400 Hz aircraft or variable-frequency microgrid, the bus ripple interacts with the sampling window. Disclosed: sampling synchronised to a multiple of the ripple period; anti-alias filtering at fripple/2.5; prediction carried out in the DC envelope domain.

(c) DC-only and high-voltage DC. In an ungridded DC nanogrid the "regulatory" ceiling is replaced by a touch-safety ceiling (e.g., 60 VDC for SELV per IEC 61140 with 120 V ripple-free / 30 V ripple limits), and the predictor keys on the load-connection schedule.

What this forecloses: minimum-forecast-temperature based set-pointing; DC-envelope-domain prediction; touch-safety ceilings substituting for grid codes.

flowchart TD
  F["12 h ambient forecast"] --> MIN["Coldest forecast point Tmin"]
  MIN --> BETA["Voc rise beta -0.30 percent per C"]
  BETA --> SET["Set module contribution at Tmin case"]
  SET --> RELAX["Relax gracefully as day warms"]
  RIPPLE["Ripple period detector"] --> SYNC["Synchronised ADC window"]
  SYNC --> ENV["DC envelope domain"]
  ENV --> SET
  SAFE["Touch-safety ceiling 60 V SELV"] --> SET
  SET --> BUS["String bus"]

---

D8.5 — Cross-domain triad for predictive bus-voltage control A3

Enabling description.

(1) Aerospace. In a scheduled flight, the future bus voltage is known deterministically from the load schedule: takeoff, climb, cruise, descent each impose a known load profile on a 270 VDC bus. The disclosed method replaces statistical forecasting with schedule-driven lookahead — the controller reads the flight plan and pre-positions the SSPC current limits 30 s ahead of each phase change, avoiding nuisance trips during the high-inrush climb phase. (This is a strict superset of the target's predictive concept and is disclosed to occupy the "prediction is scheduled, not statistical" ground.)

(2) AgTech. The irrigation pump DC-link voltage is predicted from the irrigation program (valve open/close times) plus the cloud forecast; the controller ramps the array's contribution down before each valve event so the VFD's DC link does not overshoot during the abrupt load change. Disclosed: event-driven prediction via the valve schedule rather than irradiance forecasting.

(3) Consumer electronics. In a multiport USB-C PD 3.1 charger, the future bus voltage is known from the negotiated port contracts (each port's Voltage/Current advertising). The controller allocates the shared 48 V EPR budget with full knowledge of every port's requested contract — deterministic prediction again.

What this forecloses: schedule/lookahead-driven (rather than statistical) bus-voltage prediction in DC distribution across aviation, agriculture, and consumer electronics.

flowchart TD
  FP["Flight plan load schedule"] --> P1["270 VDC SSPC pre-position, 30 s lookahead"]
  VS["Valve and irrigation program"] --> P2["VFD DC-link pre-ramp per valve event"]
  PC["USB-C PD contracts"] --> P3["48 V EPR budget allocation"]
  P1 --> A["Prevented nuisance trips and corona excursions"]
  P2 --> B["Prevented DC-link overshoot on valve events"]
  P3 --> C["Prevented per-port contract violation"]

---

D8.6 — AI, IoT, and DLT integration for predictive voltage capping A4

Enabling description.

(a) AI. A transformer-based probabilistic forecaster (patch-based, 6 layers, 4 heads, ~1 M parameters) produces a quantile forecast of string-bus voltage for the next 15 minutes; the controller consumes the 0.99 quantile as the operative predicted future voltage and the 0.50 quantile as the expected value, and allocates headroom accordingly. For safety, prediction is wrapped in conformal prediction with a rolling calibration set so the coverage guarantee (≥ 99%) is distribution-free. Disclosed: conformal/quantile prediction of bus voltage feeding a safety-critical ceiling, with a prediction-integrity monitor (see D8.7).

(b) IoT. Sensor set: SPN1 pyranometer, T-type backsheet thermocouples, IEC 61724-1 Class A irradiance reference, 0.1% shunt current sensing, and a low-cost sky-camera (all-sky fisheye at 30 s cadence) whose cloud masks are the forecast input. Edge publish: MQTT/Sparkplug-B at 1 Hz with store-and-forward on WAN loss.

(c) DLT. The 15-minute forecast, the set-point vector, and the realised voltage are committed as a commit–reveal pair: at t0 the controller publishes a salted hash of its forecast; at t0+15 min it reveals the forecast and the actual. This makes the controller's predictions auditable and non-repudiable for warranty, PPA, and curtailment-compensation disputes — a genuinely novel use of DLT in this art that also discourages a vendor from "retro-forecasting" favourable numbers.

What this forecloses: quantile/conformal forecasts driving a voltage ceiling; commit–reveal forecasting audits in PV; sky-camera-derived bus-voltage prediction.

sequenceDiagram
  participant CAM as Sky camera and IoT sensors
  participant TR as Transformer forecaster
  participant CTL as Single controller
  participant LEDG as DLT commit-reveal ledger
  CAM->>TR: cloud mask, G, Tamb, Vbus history
  TR->>CTL: q0.99 and q0.50 for t+15 min
  CTL->>LEDG: commit SHA-256 of salted forecast at t0
  CTL->>CTL: allocate setpoints under Vlim
  CTL->>LEDG: reveal forecast and realised Vbus at t0+15 min
  LEDG-->>CTL: settlement or variance penalty

---

D8.7 — Inverse: prediction-integrity failure and anti-spoofing modes A5

Enabling description. A predictor can be wrong, stale, or adversarial. Five disclosed responses:

1. Staleness detect. If the forecast age exceeds 2× its nominal update interval, the controller discards it and falls back to a conservative scalar ceiling (module contributes ≤ 60% of Vmp).
2. Residual CUSUM. A cumulative-sum test on (predicted − realised) voltage; a sustained positive drift beyond a threshold declares the predictor compromised and triggers fallback. Chosen over a simple threshold because a slow bias is the most dangerous failure mode and a threshold cannot see it.
3. Sensor plausibility / 2-of-3. Irradiance from a pyranometer, a reference cell, and the sky camera must agree within 15%; disagreement drops the offending sensor and derates the ceiling by +5%.
4. Anti-spoofing. A malicious actor with bus access could inject a false low Vbus to defeat clamping. Disclosed: challenge–response authentication on the control frames (HMAC with a per-module key), signed firmware with secure boot, and an independent, non-networked Zener crowbar that cannot be defeated by any message.
5. Fail-to-floor. With no valid prediction, no valid link, and no valid reference, the LMU drives to 0 V (module disconnected) rather than passing Voc — the "shorter string" choice, consistent with D1.9.

What this forecloses: prediction-integrity monitors; HMAC-authenticated set-point frames; spoof-resistant hardware crowbars; fail-to-floor (0 V) rather than Voc on loss of prediction.

stateDiagram-v2
  [*] --> Trusted
  Trusted --> Stale : forecast age over 2x interval
  Trusted --> Biased : CUSUM residual alarm
  Trusted --> SensorDoubt : 2-of-3 irradiance disagreement
  Trusted --> Spoofed : HMAC failure on control frame
  Stale --> Conservative : scalar ceiling 60 percent Vmp
  Biased --> Conservative
  SensorDoubt --> Conservative : derate 5 percent
  Spoofed --> Crowbar : hardware Zener clamp, no network
  Crowbar --> Disconnected : 0 V module off string
  Conservative --> Trusted : integrity restored

---

D8.8 — Geo-fenced, code-derived regulatory ceiling delivered as a signed policy bundle A1

Enabling description. Replace the hard-coded ceiling constant with a signed, versioned policy bundle. Each bundle contains: jurisdiction identifier; the numeric ceiling (600 V, 1000 V, 1500 V, 2000 V, or a touch-safety SELV value); the applicable code edition (NEC Article 690, IEC 62548, AS/NZS 5033, and equivalents); the required measurement uncertainty margin; and an expiry date. The bundle is signed (Ed25519) by the manufacturer's release key and verified against a root key fused into the controller's one-time-programmable memory; unsigned or expired bundles are rejected and the controller reverts to the lowest known ceiling (fail-closed). Delivery is over the air via the IoT channel (see D1.8), but the enforced number lives in a hardware register protected by a lock bit writable only during a provisioning window, so a compromised network cannot raise the ceiling.

An accompanying hardware interlock makes the ceiling independent of software: a comparator against a resistor-programmed reference plus a crowbar SCR that shorts the LMU output above Vlim + 5%. Disclosed: dual (policy + hardware) ceilings, with the lower of the two enforced.

What this forecloses: policy-bundle/geo-fenced regulatory ceilings; signed-ceiling verification with fail-closed expiry; dual software+hardware ceiling enforcement.

flowchart TD
  REL["Manufacturer release key"] -->|"Ed25519 sign"| BUN["Policy bundle: jurisdiction, ceiling, code edition, expiry"]
  BUN --> NET["OTA delivery"]
  NET --> VER["Verify against fused root key and expiry"]
  VER -->|"valid"| REG["Protected ceiling register, lock bit"]
  VER -->|"invalid or expired"| LOW["Revert to lowest known ceiling"]
  REG --> CMP["Comparator vs resistor-programmed Vlim plus 5 percent"]
  LOW --> CMP
  CMP -->|"exceed"| SCR["Crowbar SCR shorts LMU output"]
  CMP -->|"ok"| LMU["LMU switch normal control"]

---

5. CORE CLAIM 16 — Derivative disclosures

Claim-16 anchor elements foreclosed by this section: monitoring a first voltage across a first string-bus section connecting a first module to a second module; monitoring a second voltage across a second string-bus section connecting the second module to a voltage output; and limiting the voltage output based on a maximum regulatory safety voltage by limiting the voltage of at least one of the first and second solar modules.

---

D16.1 — Isolated sigma-delta, Hall/fluxgate, and fibre-optic section sensing A1

Enabling description. Substitute conventional resistive dividers + SAR ADCs with galvanically isolated sigma-delta modulators (AMC1303/AMC1311-class, 10 MHz clock, ±50 mV or ±250 mV input, 1-bit bitstream into an MCU's sinc³ filter giving 16-bit effective at 20 kSps). For current, use fluxgate transducers (0.01% accuracy, 0.1 µT resolution) rather than shunts where isolation and drift matter, or Rogowski coils with a digital integrator for AC ripple. For very high section voltages (≥ 1.5 kV), substitute a fibre-optic electro-optic voltage sensor: a BGO/Bi₁₂SiO₂₀ crystal in a Pockels-cell configuration, illuminated by a 1310 nm SLED, with polarimetric detection giving a linear response to the field and a natural 10 kV+ isolation with no conductive path — this removes the creepage/clearance problem entirely, which is the engineering justification.

Sensing chain topology disclosed: each LMU carries its own isolated modulator across the downstream bus section (i.e., the segment between its module's node and the next module's node), so that each section voltage is measured by exactly one device, giving a redundant, non-common-mode measurement of every section and allowing a simple consistency check: Σ Vsection,i must equal the measured terminal voltage.

What this forecloses: isolated-modulator, fluxgate, Rogowski, and Pockels-cell sensing used for per-section string-bus voltage monitoring; per-LMU section sensing with sum-check redundancy.

flowchart LR
  N1["Node after module 1"] --> SEC1["Section 1 divider"]
  N2["Node after module 2"] --> SEC2["Section 2 divider"]
  SEC1 --> ISO1["Isolated sigma-delta AMC1311"]
  SEC2 --> ISO2["Isolated sigma-delta AMC1311"]
  N1 --> POCK["Pockels cell BGO, 1310 nm SLED"]
  POCK --> POL["Polarimetric detector"]
  ISO1 --> MCU["Section monitor MCU"]
  ISO2 --> MCU
  POL --> MCU
  MCU --> CHK["Sum check: sum sections equals terminal V"]
  CHK --> DEC["Limit decision"]

---

D16.2 — Virtual section monitoring: state estimation and reflectometry A1

Enabling description. Do not install a sensor per section. Instead reconstruct every section voltage from (i) per-LMU output-voltage telemetry that already exists and (ii) one terminal measurement, using a weighted least-squares state estimator with the constraint Vk = Σi≤k Vout,i. With n sections and n+1 nodes, the estimator is overdetermined once the LMUs report and the terminal is measured, so it yields a residual vector whose largest entry localises a disagreeing LMU (a sensor-fault localiser that costs no extra hardware). A Kalman variant tracks drift over hours.

Second disclosed mechanism: bus-as-sensor reflectometry. Inject a chirp or PRBS at the string terminus and measure the time-domain or frequency-domain reflection; each LMU's switch state and parasitic capacitance form a distinguishable impedance discontinuity, so the reflection trace encodes the voltage standing-wave profile, from which the section voltages can be inferred without any section taps. Disclosed for both TDR (10 ns edge, 1 m resolution over 300 m) and spread-spectrum reflectometry (SSTDR) using a 10 MHz m-sequence.

What this forecloses: virtual/reconstructed section-voltage monitoring; residual-based sensor-fault localisation; TDR/SSTDR bus profiling as a section-voltage monitor; and it directly forecloses a competitor's later claim to "monitoring section voltages without dedicated per-section sensors."

sequenceDiagram
  participant TDR as Reflectometry front end
  participant BUS as String bus
  participant LMU as LMU 1 to LMU n
  participant EST as State estimator
  TDR->>BUS: inject 10 MHz m-sequence
  BUS->>LMU: propagate and reflect
  LMU-->>TDR: discontinuity signatures per switch state
  TDR->>EST: reflection trace
  LMU->>EST: Vout and Iout telemetry per module
  EST->>EST: weighted LS with sum constraint
  EST->>EST: residual vector localises bad LMU
  EST-->>BUS: per-section voltage estimate and limit action

---

D16.3 — Nanoscale and utility-scale section monitoring A2

Enabling description.

Nanoscale. On an on-die PV stack (see D1.4), the "bus sections" are the metal rails between adjacent sub-cell taps, monitored by on-die differential comparators (5 mV offset, 200 ns response) against a resistor-ladder threshold per section. Because a die has no creepage constraint, this is a pure circuit problem and is disclosed as such.

Utility scale. In an MVDC-collected plant (see D1.5), sections are monitored by a segmented bus architecture: busway segments are separated by isolation joints with an integrated capacitive divider + isolation amplifier per joint, and a section address is assigned by a busway segment controller. The ceiling is enforced not just by capping the modules but by opening a busway segment contactor when a segment exceeds its rating — a device-level (not just module-level) limiting action, which is disclosed as an extension of the limiting step.

What this forecloses: on-die section comparators for a PV stack; busway-isolation-joint sensing with segment contactors in an MVDC PV collector.

flowchart TD
  subgraph DIE["Nanoscale on-die stack"]
    C1["Sub-cell 1"] --> R1["Metal rail section 1"]
    R1 --> CMP1["On-die comparator 5 mV"]
    R2["Metal rail section 2"] --> CMP2["On-die comparator 5 mV"]
    CMP1 --> DCTL["On-die section FSM"]
    CMP2 --> DCTL
  end
  subgraph PLANT["Utility MVDC collector"]
    SEG1["Busway segment 1"] --> J1["Isolation joint cap divider"]
    J1 --> SEG2["Busway segment 2"]
    J1 --> SC1["Segment controller with address"]
    SC1 --> CONT["Segment contactor open action"]
    SC1 --> PC["Plant controller"]
  end
  PC --> DEC["Limit or isolate the offending segment"]

---

D16.4 — Extreme-voltage and extreme-frequency section monitoring A2

Enabling description.

(a) Very high voltage (10–50 kV DC). Sections are monitored by capacitive dividers with a guarded, SF₆- or dry-air-insulated bushing, an electro-optic current/voltage transducer for isolation, and a partial-discharge detector (IEC 60270, 100 pC sensitivity) at each section tap so that the monitor also serves a dielectric-health function. Disclosed: section-voltage monitoring where the ceiling is the bushing's partial-discharge inception voltage, not a code number.

(b) Very high bandwidth. Increase the section-monitor sample rate from DC/near-DC to 1 MSps and split the signal into a DC bin (mean, for the ceiling) and an HF bin (1–500 kHz, for arc-fault detection and ripple). The HF bin is used to detect series and parallel arcing per UL 1699B, and a section-level arc signature is a different trigger than the voltage ceiling: the disclosed method limits voltage on the DC criterion and, on the HF criterion, executes a fast disconnect. This closes the A2 frequency-expansion axis with a hook to an existing standard.

What this forecloses: PD-based ceilings; combined DC-ceiling + HF-arc section monitoring in a module-limiting array; 1 MSps section telemetry.

flowchart LR
  TAP["Section tap"] --> CD["Capacitive divider and bushing"]
  CD --> PD["PD detector IEC 60270"]
  CD --> AFE["Analog front end 1 MSps"]
  AFE --> DCP["DC bin - mean section V"]
  AFE --> HFP["HF bin 1 to 500 kHz"]
  DCP --> CMP["Compare to regulatory ceiling"]
  HFP --> AFD["Arc signature classifier UL 1699B"]
  CMP -->|"over"| LIM["Limit module contributions"]
  AFD -->|"arc"| DISC["Fast disconnect"]
  PD -->|"over pC limit"| DISC

---

D16.5 — Cross-domain triad: EV packs, HVDC data-centre busways, and rail catenary A3

Enabling description.

(1) EV traction battery. Apply the two-section monitoring method to a series battery pack: "modules" are battery modules, the "sections" are the inter-module busbar spans, and the enforced ceiling is a touch-safety ceiling derived from UN ECE R100 and FMVSS 305 (e.g., a post-crash / service-disconnect requirement that any exposed live span stay below the SELV threshold). The controller limits module contribution by bypassing modules (shunt switches) rather than down-converting, but the sensing/limiting structure is identical. Disclosed: the transferred method.

(2) HVDC data centre busway / -48 VDC plant. "Modules" are rectifier and battery strings, "sections" are busway spans between tap boxes, and the ceiling is the busway's insulation rating / arc-flash boundary. Section monitoring per tap box; limiting by ORing-FET ramp control and by contactor sequencing during a battery-plant discharge test.

(3) Rail 750 VDC catenary / third rail. "Modules" are traction power substations; "sections" are catenary sections between substations and neutral sections; the ceiling is the traction supply insulation class and the EN 50122 touch-voltage limit (UL 25 V AC / 60 V DC un-faulted). Section monitoring by catenary voltage relays; limiting by substation voltage regulation and by sectionalising. Disclosed to occupy the rail-power class.

What this forecloses: the transferred two-section monitoring-and-limiting method in EV packs, data-centre busways, and rail traction distribution.

flowchart TD
  subgraph EV["EV pack - UN ECE R100"]
    BM1["Battery module 1"] --> SP1["Busbar span section 1"]
    SP1 --> BM2["Battery module 2"]
    SP1 --> SCEV["Section monitor - touch-safety ceiling"]
    SCEV --> BYP["Shunt bypass action"]
  end
  subgraph DC["Data centre HVDC or 48 V plant"]
    RB1["Rectifier or battery string"] --> BW1["Busway span section 1"]
    BW1 --> TAP["Tap box"]
    TAP --> SCDC["Section monitor - insulation rating"]
    SCDC --> ORF["ORing-FET ramp and contactor sequence"]
  end
  subgraph RAIL["Rail 750 VDC"]
    SS1["Substation 1"] --> CS1["Catenary section"]
    CS1 --> NS["Neutral section"]
    CS1 --> SCRA["Section monitor - EN 50122 touch voltage"]
    SCRA --> REG["Substation voltage regulation and sectionalising"]
  end

---

D16.6 — Sensor fusion, digital twin, and DLT-verified section telemetry A4

Enabling description.

(a) AI sensor fusion. Section voltages are fused with module temperatures, irradiance, and an I-V curve sweep into a per-module health estimate; a 1-D CNN on the swept I-V curve classifies the module state (healthy, PID, cracked cell, soiled, bypass-diode shorted) on the LMU MCU in < 20 ms. The section monitor then predicts which module will next push a section over the ceiling and pre-positions its set-point — the section-monitoring method upgraded from reactive to predictive by the fusion layer.

(b) Digital twin. A per-plant twin (single-diode model per module, thermal model per section, cable model per string) runs in the cloud and is continuously reconciled against section telemetry; the residual between twin and measurement triggers a model-update (parameter re-identification) rather than an alarm. Disclosed: twin-based validation of section-voltage monitoring.

(c) DLT. As in D8.6, the section-voltage vector is hashed per 15-minute epoch and Merkle-anchored, giving an auditable record that specific sections never exceeded the ceiling — which is exactly the record a string-level safety certification or an insurance claim needs. Disclosed for both safety certification evidence and warranty adjudication.

What this forecloses: CNN I-V classification feeding section monitoring; twin-residual-based validation; anchored section-voltage vectors as safety-certification evidence.

sequenceDiagram
  participant SEC as Section monitors
  participant LMU as LMU I-V sweep and CNN
  participant TW as Cloud digital twin
  participant ANCH as Merkle anchor
  SEC->>TW: section voltage vector every 1 s
  LMU->>TW: module health class from I-V curve CNN
  TW->>TW: single-diode plus thermal plus cable model residual
  TW->>SEC: re-identified parameters and predicted next ceiling breach
  TW->>ANCH: SHA-256 of 15 min section-voltage vector
  ANCH-->>TW: anchor tx id for safety file

---

D16.7 — Inverse: monitored-failure modes for the two-section method A5

Enabling description. A deliberately instrumented failure architecture.

1. Two-of-three voting across the three independent section measurements (per-LMU isolated modulator, terminal divider, reflectometry estimate); disagreement beyond 2% drops the outlier and derates the ceiling by 5% until the next calibration window.
2. Sum-check interlock. If Σ sections ≠ terminal voltage beyond tolerance, the monitor declares a measurement fault and moves the whole string to the conservative ceiling rather than trusting a possibly-low reading. (A low reading is the dangerous failure — hence the conservative response is asymmetric, which is disclosed explicitly.)
3. Blind mode. If all section monitors fail, the controller falls back to nameplate Voc × n × temperature coefficient at the coldest forecast temperature — a worst-case calculation that requires no sensing at all.
4. Blocking-vs-limiting ambiguity resolution. Disclosed: when a section exceeds the ceiling, the method prefers limiting the last module upstream of that section (the module whose contribution actually crosses the section) over limiting every module, because limiting all modules discards yield. The tie-break rule when several modules could be chosen: pick the one with the lowest marginal power loss per volt reduced, computed from its current I-V slope.
5. Zero-yield safe state. If the ceiling cannot be respected even with all modules at 25% output, the monitor disconnects the string entirely (a deliberate total-loss state chosen over an over-voltage condition).

What this forecloses: voting/sum-check/derate logic; asymmetric conservative response to low readings; the marginal-power tie-break rule for choosing which module to limit — a genuinely non-obvious optimisation the applicant community has been heading toward.

stateDiagram-v2
  [*] --> Fused
  Fused --> Derated : outlier beyond 2 percent
  Fused --> Conservative : sum-check fails
  Fused --> Blind : all section monitors fail
  Derated --> Fused : calibration window passes
  Conservative --> LimitLast : choose module with lowest marginal loss per volt
  Blind --> WorstCase : nameplate Voc with cold coefficient
  LimitLast --> TotalDisconnect : ceiling unreachable at 25 percent output
  WorstCase --> TotalDisconnect : ceiling unreachable
  TotalDisconnect --> [*]

---

D16.8 — Sampling, granularity, and hysteresis parameter expansions A2

Enabling description. Parameter-space disclosures around the two-section method, each with concrete numbers:

• Temporal granularity: section monitoring at 10 Hz minimum, 1 kHz preferred, 1 MSps for the arc band; disclosed as a continuous family rather than a point value, with the trade-off quantified (1 kHz costs ~2 mW per monitor; 1 MSps costs ~300 mW and needs a dedicated AFE).
• Spatial granularity: sections defined at every module (n sections), every k modules (n/k sections, k = 2–20), or every substring. Disclosed: granularity as a design variable, with the k = 4–8 range preferred for cost/benefit in 1000 V systems (the reasoning: a 4-module section still bounds the excess to ~200 V, while cutting monitor count by 4×).
• Hysteresis band: enter limiting at Vlim, exit at Vlim − 2% (a 5 V band at 250 V, 12 V at 600 V); disclosed to prevent the inverter's MPPT from chasing the clamp.
• Rate limit: cap the section voltage's rate of change by limiting module set-point slew to 2 V/s per module, disclosed to bound the dV/dt seen by the inverter's input capacitors.
• Quantisation: set-points quantised to 0.1 V; section-voltage measurements reported to 0.05 V; disclosed to define the granularity at which "limiting" is meaningfully distinguishable from "not limiting."

What this forecloses: any later claim reciting a specific sampling rate, section granularity, hysteresis band, or set-point quantisation in a two-section bus-monitoring method — the disclosed family spans a 10 Hz-to-1 MSps range and a 1-to-n section granularity.

flowchart LR
  subgraph TEMP["Temporal"]
    T1["10 Hz baseline"] --> T2["1 kHz preferred"]
    T2 --> T3["1 MSps arc band"]
  end
  subgraph SPAT["Spatial granularity k modules per section"]
    K1["k = 1"] --> K4["k = 4 to 8 preferred at 1000 V"]
    K4 --> K20["k = 20"]
  end
  subgraph HYST["Threshold behavior"]
    H1["Enter limit at Vlim"] --> H2["Exit at Vlim minus 2 percent"]
    H2 --> H3["Slew limit 2 V per s per module"]
  end
  TEMP --> DEC["Unified decision engine"]
  SPAT --> DEC
  HYST --> DEC
  DEC --> ACT["Module set-point vector"]

---

6. COMBINATION PRIOR ART SCENARIOS (patent × open standard)

Each scenario below is a combination prior-art disclosure: the target patent's mechanism is joined with an existing, publicly documented standard in a way that is enabling and reproducible, so that a later applicant cannot claim the combination.

---

C1 — US 8,102,074 × SunSpec Modbus (SunSpec Alliance register-map ecosystem)

Combination disclosed. Expose the per-module voltage limiter and the section-voltage monitor as SunSpec Modbus data points on a shared RS-485 or TCP channel. Map: string/segment voltage → the SunSpec DER measurement model's voltage points; per-LMU set-point and duty → a vendor-specific model block addressed by module serial; the site-level ceiling → the inverter/DER nameplate voltage; the section-voltage vector → a repeating group of points. Control flow: a plant SCADA polls the SunSpec map at 1 Hz, and the single controller remains the sole authority that converts the site ceiling into per-module set-points; SunSpec is the interface, not the authority — and the disclosed architecture explicitly constrains SunSpec writes to a "budget request" that the controller may accept or refuse. Combine with the SunSpec rapid-shutdown signalling convention for transmitting the "limit to 0 V" command along the power conductors, so a single signalling layer carries both the operational set-point and the emergency disconnect.

Why this is enabling. SunSpec Modbus is a published register-map standard with a defined transport (Modbus RTU/TCP), a defined model/point addressing scheme, and published conformance tests; a POSITA can implement the mapping without further invention. The novel combination disclosed here — a SunSpec-facing voltage-budget request arbitrated by a per-module controller with a hard ceiling — is the part that must be disqualified from later patenting.

flowchart LR
  SCADA["Plant SCADA - SunSpec client"] -->|"Modbus RTU 19200 8N1"| GW["SunSpec gateway"]
  GW -->|"budget request"| CTRL["Single controller - sole authority"]
  CTRL -->|"per-module setpoint"| L1["LMU 1"]
  CTRL -->|"per-module setpoint"| L2["LMU 2"]
  L1 --> BUS["String bus"]
  L2 --> BUS
  L1 -.->|"Vout, duty, serial as SunSpec points"| GW
  GW -.->|"Model map to SCADA"| SCADA
  CTRL -->|"rapid-shutdown signalling on power conductors"| L1

---

C2 — US 8,102,074 × IEEE 2030.5 / CSIP × IEEE 1547-2018 × OpenADR 2.0b

Combination disclosed. Bind the per-module limiting authority to a grid-services control chain. IEEE 1547-2018 defines the voltage ride-through and must-trip curves and the Volt-VAr / Volt-Watt functions; IEEE 2030.5 / CSIP (Common Smart Inverter Profile) defines the DER control resources (e.g., fixed VAr, Volt-VAr, limit-max-power, connect/disconnect) over a RESTful interface with a DER program abstraction; OpenADR 2.0b provides the VTN/VEN event channel for price- and reliability-driven curtailment. The disclosed combination: an OpenADR/2030.5 event or DER control is translated by the single controller into a string-bus voltage budget, and the per-module limiters implement the budget so that the point of common coupling satisfies the 1547 ride-through envelope while the DC-side ceiling is simultaneously respected. A further disclosed refinement: the controller uses the 1547 trip curve as a feed-forward term — it knows the point on the curve where the inverter would trip and pre-emptively caps the string voltage to stay inside the "continuous operation" region, thereby preventing an unnecessary trip.

Why this is enabling. All three standards are public, versioned, and implementable; the mapping from a DER program event to a DC-side string budget is disclosed with the translation table (event → power limit → DC voltage budget → per-module set-point vector).

sequenceDiagram
  participant VTN as OpenADR VTN
  participant VEN as Site VEN or 2030.5 client
  participant CTRL as Single controller
  participant LMU as LMU string
  participant INV as Inverter at PCC
  VTN->>VEN: event: limit to 60 percent for 2 h
  VEN->>CTRL: DER control DERControl limitMaxPower
  CTRL->>CTRL: translate to DC voltage budget using 1547 ride-through envelope
  CTRL->>LMU: setpoint vector
  LMU-->>CTRL: Vout feedback
  INV-->>CTRL: PCC voltage and frequency
  CTRL->>LMU: adjust budget to remain in continuous operation region

---

C3 — US 8,102,074 × MQTT Sparkplug B (Eclipse Foundation) × OPC UA over TSN (IEC/IEEE 60802)

Combination disclosed. Put the single controller and every LMU on an industrial publish/subscribe fabric. Each LMU is a Sparkplug Edge Node (or a Device of the controller's Edge Node) with a birth certificate declaring its capabilities (max set-point, sensor channels, firmware CRC) and a death certificate published within 1.5× the configured rebirth interval; the controller publishes a STATE topic carrying the string budget, and each LMU's NDATA/DDATA carries Vout, Iout, temperature, and duty. For the deterministic part — the 1 ms set-point commit — use OPC UA PubSub over TSN with a configured 802.1Qbv gate schedule, while the non-deterministic telemetry rides MQTT. The combined architecture is thus dual-fabric: deterministic control plane (TSN) + lossy telemetry plane (MQTT), with an explicit rule that the control plane never awaits the telemetry plane, and that a missing birth certificate forces the LMU to its hardware default duty (see D1.9).

Why this is enabling. Sparkplug B's own documentation defines the topic namespace, payload protobuf, and state machine; IEC/IEEE 60802 profiles TSN for industrial automation. The disclosed combination maps a voltage budget onto a Sparkplug STATE topic and a set-point commit onto a TSN-scheduled OPC UA PubSub frame.

flowchart TD
  subgraph DET["Deterministic control plane"]
    CTRL["Single controller OPC UA PubSub publisher"] -->|"802.1Qbv scheduled 1 ms"| L1["LMU 1 subscriber"]
    CTRL --> L2["LMU 2 subscriber"]
  end
  subgraph TELE["Telemetry plane"]
    L1 -->|"NDATA"| BRK["MQTT broker"]
    L2 -->|"NDATA"| BRK
    BRK --> HIST["Historian and SCADA"]
  end
  BRK -->|"birth or death certificate"| CTRL
  CTRL -->|"STATE: string budget"| BRK
  L2 -.->|"missing birth certificate - hardware default duty"| SAFE["Safe 40 percent duty"]

---

C4 — US 8,102,074 × IEC 61850-7-420 (DER logical nodes) with GOOSE

Combination disclosed. Model the plant in IEC 61850-7-420 logical nodes: the per-module limiter as a component of the DPVM / DPVC-family DER measurement and control nodes; the string controller as a DER unit controller; the section monitors as measurement nodes with quality flags propagated end-to-end. Then use GOOSE (IEC 61850-8-1) for the time-critical exchange: a section over-voltage publishes a GOOSE message with a retransmission scheme (2 ms, 4 ms, 8 ms, … to 1 s keep-alive) so that every LMU receives the "cap now" state within < 4 ms on a 100 Mbit/s switched Ethernet segment. Disclosed refinement: encode the voltage budget in the GOOSE dataset as an integer in 0.1 V units alongside a sequence number and a time-allowed-to-live, so that a stale GOOSE is detectable and the LMU can enforce its hardware default. This gives a hard, sub-4 ms, standards-based path for the limiting action that does not depend on a polling SCADA.

Why this is enabling. IEC 61850-7-420 is a published object model and GOOSE is a published, widely implemented protocol with defined retransmission and TTL semantics; the combination with per-module limiting is the disclosed subject matter.

sequenceDiagram
  participant SEC as Section monitor
  participant STR as String controller
  participant SW as Substation Ethernet
  participant LMU as LMU 1 to LMU n
  SEC->>STR: MV voltage with quality, section exceeded
  STR->>SW: GOOSE publish budget cap, seq 1042, TTL 300 ms
  SW->>LMU: multicast GOOSE retransmit 2 4 8 ms
  LMU->>LMU: enforce cap within 4 ms
  LMU-->>STR: MMS report Vout and status
  STR-->>SEC: acknowledgement and revised section limit

---

C5 — US 8,102,074 × open-source software stack (Zephyr RTOS / RISC-V control core, OSQP-style QP solver, Hyperledger DLT)

Combination disclosed. Build the controller entirely from open-source components with published interfaces: a RISC-V control core running Zephyr RTOS with a deterministic scheduling policy (Zephyr's cooperative threads + a hardware timer at 1 kHz for the control task, preemptible threads for telemetry); an open-source convex QP solver (an OSQP-style operator-splitting or an active-set solver generated with an open code-generation tool) solving the voltage-budget allocation at 200 µs; an open-source MPPT and control library; and Hyperledger (or an open DLT) for the epoch-anchoring described in D8.6/D16.6. The disclosure specifies the task set and priorities (control task at priority 0, cooperative, 500 µs period; estimator at priority 1, 1 ms; telemetry at priority 2, 10 ms; anchoring at priority 5, 15 min) and the solver warm-start strategy (previous solution as the initial point, reducing iterations from ~40 to ~6 and therefore solve time below 100 µs), which is the enabling detail that makes an on-MCU QP feasible.

Why this is enabling. Every component is public and versioned; what must be disqualified from later patenting is the specific composition — an open-source RISC-V/Zephyr control core running a warm-started QP voltage-budget solver on an MPC horizon, with DLT anchoring of the resulting telemetry — disclosed here with the timing budget.

flowchart TD
  RISCV["RISC-V core"] --> ZEPHYR["Zephyr RTOS scheduler"]
  ZEPHYR --> T0["Priority 0 control task 500 us - cooperative"]
  ZEPHYR --> T1["Priority 1 estimator 1 ms"]
  ZEPHYR --> T2["Priority 2 telemetry 10 ms"]
  ZEPHYR --> T5["Priority 5 DLT anchor 15 min"]
  T0 --> QP["Warm-started QP solver, 6 iterations, under 100 us"]
  QP --> ACT["Module setpoint vector"]
  T1 --> QP
  T2 --> MQTT["MQTT and OPC UA uplink"]
  T5 --> DLT["Hyperledger anchor of epoch hash"]
  ACT --> LMU["LMU string"]

---

7. Foreclosure matrix

---

8. Verification gaps and cautions (do not smooth over)

1. Claim 16–19 text. Claim 16 is truncated mid-sentence in the authoritative Google Patents rendering supplied in this prompt, and I could not recover claims 17–19 verbatim from the searches run in this session. The claim-16 anchor used above is the wording recorded in the earlier "Patent summary" section of this file (sourced there from uspto.report). Before filing or publishing this document, re-read claims 16–19 from the granted PDF (patentimages.storage.googleapis.com/08/f0/2a/d8aea7cc33c70c/US8102074.pdf) and confirm that D16.x occupies the granted limitations, particularly any "maximum regulatory safety voltage" wording in claim 16 and the scope of dependent claims 17–19.
2. No derivation above is a characterization of the patent owner's actual products. D1.10(a) reads on a published Tigo Smart-Curve application note (v1.3.4), but that document is cited only as evidence that a commercial embodiment of temperature-independent output-voltage capping exists. It is not prior art to US 8,102,074 (it post-dates it) and it is not a prior-art reference for anything in this document. Do not cite it as § 102 art.
3. Standard identifiers are used as illustrative hosts. SunSpec Modbus, IEEE 2030.5/CSIP, IEEE 1547-2018, OpenADR 2.0b, MQTT Sparkplug B, OPC UA, IEC/IEEE 60802, IEC 61850-7-420/8-1, IEC 60270, UL 1699B, UN ECE R100, FMVSS 305, EN 50122, NEC Article 690, IEC 62548, AS/NZS 5033, and USB PD 3.1 are named as publicly documented interfaces. I have not, in this session, verified the current edition, clause number, or numeric limit in each. Verify every numeric limit against the current published edition before this document is filed as evidence; the disclosure's enablement does not depend on the values being correct, but its credibility as § 102 art does.
4. Numerical parameters are design examples, not measured data. Converter efficiencies, gate timings, solver times, and sensor accuracies above are engineering-plausible values stated to make the disclosure enabling. Where a claim of performance is material, mark it as a design target.
5. Cross-reference flag (repeated from earlier sections). The front page of the grant prints Int. Cl. H04B 3/54; the Google Patents full-text rendering in this prompt lists CPC H10F 77/955 and Y02E 10/50. Both are recorded as-is. Nothing in this document depends on the classification.
6. Ownership does not affect this document. The assignment section of this file records a chain ending in a 2026-01-20 recording to Tigo Energy Innovations LLC following a December 2025 patent purchase agreement. Whether the current owner is Tigo Energy, Inc., Newlight Capital LLC, or Tigo Energy Innovations LLC has no effect on the prior-art status of this disclosure. It does affect who would be motivated to assert the patent against the subject matter above.
7. Date note. Task-prompt date April 26, 2026 versus environment date 2026-09-29. The disclosure is drafted to be published at a date certain (with the hash anchoring in §1) so the effective prior-art date does not depend on which date governs.
8. No search result contradicted this document's factual premises. The patent number "8102074" resolved to the Tigo Energy grant on the first attempt; the only non-patent collision found (FAERS case 8102074) is excluded above.

---

Deposit checklist for publication:

• Finalise claims 16–19 verification (gap 1 above)
• Re-verify every standard's numeric limit (gap 3 above)
• Render all 31 Mermaid diagrams and confirm they compile in the target renderer
• SHA-256 the PDF; anchor via OpenTimestamps
• Publish to IP.com and/or Zenodo with DOI; mirror to an indexed defensive-publications channel
• Record the anchor transaction ID and DOI on the face of the document
• Serve a courtesy copy on the current owner of record so that intent to publish is documented