US 7436903

I'll search for this specific patent on USPTO and court dockets, then summarize.

Let me do more targeted searches on litigation and the Federal Circuit docket.

Summary — US Patent No. 7,436,903 (interpreted literally as "7436903")

Sources consulted: The authoritative full text fetched from Google Patents (https://patents.google.com/patent/[US7436903](/patent/US7436903)/en, fetched 2026-09-26) and a secondary drug/patent database record at Unified Patents (https://portal.unifiedpatents.com/patents/patent/US-7436903-B2).

Bibliographic data

Abstract (verbatim)

"A multicarrier transmitter may transmit cyclically delayed linear combinations of two or more data streams with three or more antennas. In some embodiments, the multicarrier transmitter may transmit cyclically delayed linear combinations of three data streams with four antennas."

Plain-language overview of each independent claim

Claim 1 (apparatus). A multicarrier transmitter has IFFT circuitry that generates first and second time-domain samples from first and second data streams. Cyclic shifting and combining circuitry then (a) cyclically shifts the first samples and combines them with the second samples for transmission on a first antenna, and (b) cyclically shifts the second samples and combines them with the first samples for transmission on a second antenna.

Claim 9 (apparatus). A multicarrier transmitter with IFFT circuitry generating first and second time-domain samples, plus cyclic shifting and combining circuitry made up of first and second cyclic shifters, and a controller. The controller instructs each shifter whether or not to cyclically shift its respective time-domain samples (which represent the first and second data streams). The emphasis is on controllable/per-stream shifting.

Claim 13 (apparatus). A multicarrier transmitter with IFFT circuitry and cyclic shifting/combining circuitry that includes a first cyclic shifter (shifting the first samples) and combiner circuitry that linearly combines the cyclically-shifted first samples with the second samples for a first antenna. A second cyclic shifter shifts the second samples before combining. The transmitter also includes first RF circuitry to transmit the combined output on the first antenna, second RF circuitry to transmit the first time-domain signal without a cyclic shift on a second antenna, and third RF circuitry to transmit the second time-domain signal without a cyclic shift on a third antenna.

Claim 14 (method). A method of transmitting multicarrier signals: generating first and second time-domain samples from first and second data streams; cyclically shifting the first samples; combining the shifted first samples with the second samples for transmission on a first antenna; cyclically shifting the second samples; and combining the first samples with the shifted second samples for transmission on a second antenna.

Claim 23 (method). A method that cyclically shifts first samples and combines them with second samples for transmission; cyclically shifts the second samples before linearly combining with the shifted first samples to produce output samples for a first antenna; and concurrently transmits the output samples on a first antenna, the first signal with no cyclic shift on a second antenna, and the second signal with no cyclic shift on a third antenna.

Claim 24 (method). A method of transmitting N data streams with N+1 antennas, where N is either 2 or 3. First and second time-domain samples (corresponding to first and second data streams) are cyclically shifted and combined in a first manner for a first antenna, in a second, different manner for a second antenna, and in a third, different manner for a third antenna.

Specification highlights (context for the claims)

• The transmitter is described for OFDM/MIMO systems; cyclic time-domain shifting is equivalent to a linear phase shift in the frequency domain.
• The description covers worked examples: 2 streams on 3 antennas (becoming a 2×2 composite channel at the receiver), 1 stream on up to 4 antennas (2×1 composite channel), 2 streams on 4 antennas (2×2 composite channel), and 3 streams on 4 antennas (2×3 composite channel).
• Cited prior art includes US 6,842,487 B1 (Ericsson, "Cyclic delay diversity for mitigating intersymbol interference in OFDM systems") and US 2005/0254592 A1 (Naguib, "Time varying cyclic delay diversity of OFDM"); non-patent literature includes Bauch, Dammann, and Huebner on cyclic delay diversity.

Litigation / CAFC docket status — uncertainty noted

I searched for a Court of Appeals for the Federal Circuit 2026 docket and for USPTO post-grant (IPR/PGR) or district-court activity tied specifically to patent number 7436903. I could not find any authoritative record of a CAFC 2026 appeal, IPR/PGR, or district-court litigation involving this specific patent number. The general Federal Circuit PTAB statistics results returned by my searches (Finnegan summaries, etc.) are not specific to 7436903 and should not be attributed to it. I did not locate the patent in the Delaware district-court exhibit list that appeared in results, nor in any 2026 docket.

Caveats: (1) Absence of a search hit is not proof that no litigation or appeal exists — docket coverage in public web search is incomplete. To confirm, the definitive checks are the USPTO Patent Center/Patent Trial and Appeal Board (PTAB) APIs for 7436903 and the Federal Circuit's own docket (and PACER/ CourtListener) searched by patent number, which I could not fully query here. (2) The patent's stated "Active" status and 2026-11-23 adjusted expiration come from the database record and are assumptions, not legal conclusions. (3) The bibliographic details above are drawn from the supplied authoritative patent text; examiner names come only from the secondary Unified Patents record.

I'll research litigation involving US Patent 7,436,903. Let me search multiple sources.

The first search surfaced a case involving patent 7,463,703 — but note that is a different number from 7,436,903. Let me dig deeper with targeted searches.

The searches consistently point to patents 7,463,703 and 7,593,492 in the Collision/Samsung case — not 7,436,903. Let me run more targeted searches to confirm whether 7,436,903 appears in any litigation.

Litigation Search: US Patent No. 7,436,903

Bottom Line

Based on the searches performed, I found no litigation asserting or involving US Patent No. 7,436,903 specifically. I was unable to identify any district court case, ITC investigation, PTAB proceeding, or CAFC appeal in which this patent number appears as an asserted patent, a patent-in-suit, or a subject of validity challenge.

Because I could not run exhaustive docket checks (PACER, Docket Navigator, RPX, etc. sit behind paywalls and Unified Patents' litigation database was not fully searchable via the tool), this conclusion is qualified: no known litigation was located, rather than an affirmative certification that none exists.

Important Number-Confusion Warning (do not auto-correct)

My searches repeatedly surfaced a large, high-profile case — Collision Communications, Inc. v. Samsung Electronics Co., Ltd. et al. — but the patent asserted there is US 7,463,703, not 7,436,903. These are different patents:

Per the operating rules, I am not treating these as the same patent. The Collision v. Samsung case is not litigation involving 7,436,903, and I flag it here only because it is the likely source of any confusion (the digits "703" recur, and one Chinese-language article even renders the Samsung patent as "US7463703").

For Reference (different patent — flagged, not attributed to 7,436,903)

The only significant wireless-patent litigation that surfaced in these searches concerned 7,463,703:

• Case: Collision Communications, Inc. v. Samsung Electronics Co., Ltd. and Samsung Electronics America, Inc.
• Court: U.S. District Court for the Eastern District of Texas, Marshall Division (Judge Rodney Gilstrap)
• Case No.: 2:23-cv-00587 (some syndicated reports erroneously state 2:23-cv-00500 — the docket-filed number is 2:23-cv-00587)
• Filed: December 2023
• Patents-in-suit (four): 7,463,703; 7,920,651; 7,593,492; 6,947,505
• Outcome: Oct. 10, 2025 jury verdict of $445.5M for willful infringement; post-trial injunction motion regarding the '492 patent; appeal to the Federal Circuit anticipated.
• Related PTAB proceedings: IPR2024-01247 ('492), IPR2024-01248 ('071), IPR2024-01249 ('505), IPR2025-00011 ('651).

Again — this case does not involve 7,436,903.

Patent 7,436,903 — Prosecution / Ownership History (no litigation found)

For context on the patent itself (from the authoritative patent record):

• Application: US 10/954,734, filed 2004-09-29; granted 2008-10-14
• Original assignee: Intel Corp
• Current assignee: Tahoe Research, Ltd. (assignment recorded 2022-08-15)
• Adjusted expiration: 2026-11-23
• Forward citations: ~107 (many by Sharp, Qualcomm, Marvell) — indicating technical relevance, but citations are not litigation.

No litigation, ITC, or PTAB activity tied to the 7,436,903 number was located.

Caveats and Recommended Next Steps

1. Scope limitation: I reached the tool step limit; further searches (e.g., Docket Navigator, RPX Insight, Unified Patents' full litigation portal, PACER's "Patent/Trademark" docket reports) could reveal litigation I could not reach.
2. The patent has only ~7 months of term remaining (expiration 2026-11-23) and is a Tahoe Research (Intel-derived) asset, so any assertion activity would likely be as part of a broader Intel/Tahoe or acquirer licensing campaign rather than a standalone suit.
3. Verification suggestion: Search PACER/the Unified Patents litigation portal for the literal string "7,436,903" and "10954734" (the application number), and check Intel/Tahoe Research enforcement actions, to confirm the negative result.

Conclusion: No known litigation involving US Patent 7,436,903 was found. If a source appears to link 7,436,903 to the Samsung/Collision matter, that source has almost certainly transposed the digits of 7,463,703 — a different patent that should not be conflated with the subject patent.

I'll verify the structured block's negative result against public sources before finalizing.

Let me run additional targeted checks before finalizing.

Final confirmation checks on PTAB/CourtListener naming conventions.

Proceedings overview

Zero AIA trial proceedings are on file against US 7,436,903. The structured "PTAB proceedings on file" block — the canonical USPTO Open Data Portal (ODP) list supplied with this task — returns an empty set: no IPRs, no PGRs, no CBM reviews, and therefore no institution decisions, Final Written Decisions, terminations, or PTAB-to-CAFC appeals. My independent web searches (PTAB/PTACTS document corpus, Docket Alarm PTAB records, Google Patents and FreePatentsOnline litigation/trial tabs, Unified Patents' patent page at https://portal.unifiedpatents.com/patents/patent/US-[7436903](/patent/7436903)-B2) likewise surfaced no proceeding number naming this patent.

Bottom line for a defendant: this is an untested patent, not a hardened one. Nothing has been canceled, nothing has been sustained, and no ground has been estoppel-barred. Every one of the 24 claims — including independent claims 1, 9, 13, 14, 23 and 24 — is UNTESTED at the Board. The absence of PTAB activity is not a sign of strength for the patent owner; it reflects that the asset has not been asserted (consistent with the litigation section above finding no suits) and that it is a Tahoe Research (Intel-derived) portfolio patent nearing end of term.

There are therefore no per-proceeding sections to write, and I will not invent any. Instead, below are the number-confusion traps I actually hit while searching, which are the material risk to your analysis.

Do-not-conflate register (different patents — NOT 7,436,903)

Every search hit containing "…903" that looked like PTAB activity was a different patent number. Flagging these explicitly because transposition is the single likeliest source of a false positive:

Conclusion on the register: no IPR/PGR/CBM proceeding against 7,436,903 exists, and every apparent hit is a number-transposition or same-surnames artifact.

Strategic summary

Claim status. All 24 claims of 7,436,903 are UNTESTED before the PTAB. None are CANCELED. None are SUSTAINED by any tribunal. There is no FWD to cite, no certificate of cancellation, and no claim-level disposition to argue from. If you are drafting an invalidity position, you start from a blank agency record — which cuts both ways: you face no § 325(d) discretionary-denial risk from a prior Board record, but you also cannot point the examiner or a jury to any prior administrative finding of unpatentability.

Estoppel landscape. § 315(e)(2) estoppel is empty. Because no IPR/PGR was ever instituted against this patent, no petitioner and no privy is barred from raising any § 102 or § 103 ground in a civil action, and no petitioner is estopped before the Office. Practically, this means the full universe of printed publications and patents — including the very references the patent itself cites (US 6,842,487 B1 to Larsson/Ericsson on CDD; US 2005/0254592 A1 to Naguib; US 2003/0099304 A1 to Gore; the Bauch, Dammann and Huebner non-patent literature) — remains available to you without an estoppel shadow. That is unusual and favorable to a defendant.

Pattern signals. There is no pattern, because there is no activity: no repeat petitioner, no defensive aggregator (Unified Patents does not appear in any proceeding; its portal merely hosts bibliographic data on the patent, and that page shows no litigation or PTAB campaign attached), and no patent-owner appeal history. The forward-citation cluster (~107 citations, heavily Sharp, Qualcomm and Marvell) evidences technical relevance in the CDD/MIMO literature, but citations are not adversarial events. Separately, the ownership signal is notable: assignment to Tahoe Research, Ltd. recorded 2022-08-15 out of Intel Corp, with Unified Patents tagging the patent "Likely Sold." Monetization of a near-expiry, never-asserted asset is the scenario in which a demand letter could appear with no PTAB history behind it.

Recommended next steps

1. If you receive a demand letter citing 7,436,903, you have a clean slate. There is no FWD to link and no canceled claim to quote — because none exists. Do not represent to a court or an adversary that any claim has been invalidated by the Board. Search PTAB E2E (https://ptacts.uspto.gov/) and the USPTO Open Data Portal by application number 10/954,734 and patent number 7,436,903 to re-confirm the null result on your own docket date.

2. Time is the dominant constraint — and it is short. The adjusted expiration is 2026-11-23 (Google Patents record, https://patents.google.com/patent/US7436903/en). From today (2026-09-28) that is roughly eight weeks of remaining term. A new IPR petition filed now would realistically reach an FWD well after expiry; the patent owner would have every incentive to disclaim or let the term lapse, and the Board has historically declined to institute or has found proceedings moot where the challenged claims expire before a decision can issue. An IPR is very likely not a cost-justified defense for this patent — the practical defense is term expiry plus non-infringement, not Board cancellation. Confirm the expiration and any maintenance-fee lapse in Patent Center before relying on the date.

3. If you nonetheless need an agency record, note that the one-year bar of § 315(b) attaches only if you (or a privy) were served with a complaint alleging infringement more than one year ago. Because the litigation section above found no suit on this patent, treat § 315(b) as un-triggered as of today — but verify service dates before filing.

4. Do not let a source conflate the numbers. If any search result or demand letter links 7,436,903 to the Collision/Samsung matter or to any PTAB petition, it has transposed 7,463,703 or picked up 7,402,903 / 9,747,903. This is the single error most likely to corrupt your analysis.

Confidence and limits. I state the null result with high confidence on the strength of the ODP-derived structured block, corroborated by targeted searches of the PTACTS document corpus, Docket Alarm, Google Patents/FreePatentsOnline trial data and Unified Patents. The residual uncertainty is ordinary negative-search uncertainty: I could not directly query PTAB E2E or PACER APIs in this session, and a very recently filed, not-yet-indexed petition could in principle elude all of these sources. The definitive verification is a PACER/PTAB E2E lookup by patent number and application number 10/954,734.

I'll research the assignment chain. Let me start with targeted searches on the assignment records and the Intel→Tahoe transfer.

Good data. Let me get the reel/frame details for the original Intel assignment and check whether 7436903 was part of the downstream Tahoe cascade.

Now let me pin down the reel/frame and correspondent details, plus the Powerwatch/Equitable IP structure.

I hit the tool-call limit before I could pull the correspondent-of-record line from every reel, so I'll flag precisely what is documented vs. not retrievable. Below is the assignment-chain reconstruction, built on the authoritative patent text and the assignment/INPADOC records surfaced.

Ownership-chain analysis — US Patent No. 7,436,903

Verification entry point: https://assignmentcenter.uspto.gov/ (search by patent number 7436903 or application 10/954,734; the bulk Intel→Tahoe reel/frame below is the key record to open).

Inventors

Both named inventors were Intel Corporation employees at the time of the 2004-09-29 filing — the patent was filed as an Intel-originated application, and an index of Minnie Ho's patents lists Intel Corporation (Santa Clara, CA) as assignee across the family (radaris inventor index).

Unusual-pattern note: The only observable oddity is timing, not personnel. The inventor→Intel assignment did not appear on Google Patents' legal-events timeline until 2007-03-23 (roughly 2.5 years after filing) — late relative to the 2004 filing. I found no evidence of either inventor leaving Intel within 12 months of filing, and no evidence of a distressed or fire-sale transfer by the inventors. The 2004–2007 gap is flagged as a data point, not a finding; a confirmatory late-filed assignment is a common Intel prosecution pattern and I cannot distinguish it from a re-recorded/confirmatory assignment on the records retrieved.

Original assignee

Intel Corporation (Santa Clara, CA) — the entity named on the issued patent (granted 2008-10-14).

• Business: operating semiconductor company (microprocessors, logic, connectivity/Wi-Fi, memory/storage). Not a holding or licensing shell — at all relevant times a products company.
• Products embodying the claims: The claims cover a multicarrier (OFDM/MIMO) transmitter that cyclically shifts/combines multiple data streams with cyclic delay diversity. Intel's 802.11n-era Wi-Fi silicon (e.g., the PRO/Wireless 4965AGN and successors) used MIMO-OFDM with cyclic delay diversity and plausibly practiced these claims. I did not locate a product-mapping or standard-essentiality record tying the claims to a specific Intel part, so treat "practiced" as plausible, not confirmed.
• Current status: Intel is still an operating company (under significant financial restructuring during 2022–2025). It did not enter bankruptcy, and this patent was not sold in any Chapter 7/11 proceeding.

Assignment timeline

Two recorded post-filing events tie to this patent. (The record is short; that brevity is itself informative — see Verdict.)

1. 2007-03-23 (recorded) — Reel/Frame not retrieved

• Conveyance: Assignment (assignment of assignors' interest) — recorded as a "reassignment" event on the Google Patents legal-events timeline.
• Assignor: Minnie Ho; Sumeet Sandhu (the named inventors).
• Assignee: Intel Corporation.
• Correspondent: Not retrievable from the sources I could reach. Recommended check: open this entry in Assignment Center by application 10/954,734 to capture the reel/frame and correspondent of record.
• Context: Inventor-to-employer assignment (initial vesting in Intel). Timing is late (≈2.5 years post-filing).

2. Executed 2022-07-18 / recorded 2022-08-15 — Reel 061175/0176

• Conveyance: Assignment ("ASSIGNMENT OF ASSIGNOR'S INTEREST").
• Assignor: Intel Corporation.
• Assignee: Tahoe Research, Ltd. (Ireland) — current assignee.
• Correspondent: Not captured in the records I retrieved. This is the highest-value item to confirm — the correspondent on a bulk NPE acquisition often recurs across the whole portfolio family.
• Context: Transfer-to-monetization/asserter entity (privateering). This reel/frame is the master bulk assignment for the ~5,000-patent Intel→IPValue transaction: the identical reel/frame 061175/0176 and identical effective date 20220718 also appear on unrelated Intel patents (e.g., the gate-electrode alloy publication US 2007/0096163), confirming a single portfolio-wide assignment rather than a patent-specific transfer.
• Deal background: IPValue Management, Inc. announced on 2022-08-08 that ~5,000 Intel patents were transferred to "Tahoe Research Limited, a newly formed company within the IPValue Management Group," which will "license the Tahoe portfolio to its established and growing network of licensees." Press coverage (EE News, The Register, Fierce Sensors, Techdirt) characterized this as Intel handing patents to a monetization vehicle for licensing/enforcement, with Intel retaining freedom-to-operate — the classic privateering structure.

No later assignment to any downstream LLC appears in the record I retrieved for 7,436,903. Per Google Patents, the current assignee remains Tahoe Research, Ltd. (Note: Unified Patents' page shows a heuristic flag reading "Likely Sold" under Parent Company "Tahoe Research Ltd" — that is a database inference, not a recorded assignment, and should not be treated as one.)

Timeline diagram

timeline
    title Ownership of US 7436903
    2004 : Application filed by Sandhu and Ho
         : Intel is assignee of record
    2007 : Inventor assignment to Intel recorded
    2008 : Patent issued to Intel
    2022 : Assigned to Tahoe Research Ltd
         : IPValue takes over Intel portfolio
    2026 : Adjusted expiration on 23 Nov

NPE / troll-pattern signals

1. Shell-entity transfer — PRESENT.
Reel 061175/0176 (executed 2022-07-18; recorded 2022-08-15) moves this patent from operating company Intel to Tahoe Research, Ltd., which IPValue's own press release describes as a newly formed company created to hold and license the portfolio. It has no products in commerce; the transaction is a pure licensing vehicle. This satisfies the signal on recorded evidence, not on naming alone.

2. Known asserter in the chain — PRESENT (documented via IPValue).
Current assignee Tahoe Research, Ltd. is the licensing arm of IPValue Management Group, a recognized patent-monetization firm (CEO John Lindgren, formerly Conversant), repeatedly characterized in press as an NPE/PAE vehicle. More concretely, the same Tahoe portfolio was flowed into a litigation vehicle in 2026: Intel-origin patents US 7,281,079 and US 7,432,731 went Tahoe Research → Southfork IP Holdings → Powerwatch Systems LLC, and Powerwatch (parent: Equitable IP Corporation) sued Nanya Technology in E.D. Tex. on 2026-09-24 (PowerWatch Systems LLC v. Nanya Technology Corp., No. 2:26-cv-00871; PACER clerk's-docket reflects "Corporate Parent Equitable IP Corporation" and counsel Timothy Devlin). Note: that's an E.D. Tex. plaintiff-side NPE; Devlin is a well-known NPE litigation/recording name. This is portfolio-level, not a recorded event for '903.

3. Repeat correspondent across the chain — UNCLEAR / not retrievable.
I could not capture the correspondent of record on either recorded entry (2007 or 2022). Flag for follow-up: pull the correspondent on reel 061175/0176 and compare it against the correspondent on the Southfork/Powerwatch recordings (2026) for '079/'731. If the same attorney/firm filed the 2022 bulk reel and the 2026 cascade, that is a discrete repeat-correspondent finding.

4. Cascading transfers — NOT PRESENT for this patent; PRESENT for sibling assets.
'903 has only two links (inventors→Intel→Tahoe) with no chained LLCs. However, the same Tahoe portfolio shows the canonical cascade in 2026 (Tahoe → Southfork IP Holdings → Powerwatch Systems, Aug–Sep 2026), so the family exhibits the pattern even though '903 has not (yet) been moved.

5. Pre-litigation transfer — NOT PRESENT.
The 2022-07-18/2022-08-15 transfer is not within 6 months of any suit naming '903; consistent with the earlier finding that no litigation involving 7,436,903 was located. Caveat from the prior section stands: absence of a hit is not proof of absence.

6. Bankruptcy fire-sale — NOT PRESENT.
Assignor Intel was not in bankruptcy. The transfer was a strategic monetization deal (Intel "freedom to operate," revenue-sharing), not an insolvency sale.

7. Privateering — PRESENT.
Intel (operating company) transferred ~5,000 patents to IPValue/Tahoe Research, a licensing NPE, while retaining freedom-to-operate/licensing rights, with IPValue openly stating it will license the portfolio to third parties. This is the textbook privateering structure and is corroborated by IPValue's press release and multiple independent press accounts.

8. Defensive aggregator (anti-NPE) — NOT PRESENT.
The chain terminates at Tahoe Research / IPValue, which is a monetization entity, not RPX, AST, LOT Network, Unified Patents, or OIN. The patent has not been neutralized; the reverse — the same portfolio demonstrably feeds an asserting plaintiff (Powerwatch, 2026).

Verdict

NPE — high confidence.

Two strong, documented signals drive this: (i) shell-entity transfer — reel 061175/0176 (effective 2022-07-18, recorded 2022-08-15) moved '903 from operating-company Intel to Tahoe Research, Ltd., IPValue's newly formed licensing vehicle with no products; and (ii) privateering — Intel's own transaction materials state the portfolio is to be licensed/enforced by IPValue while Intel keeps freedom-to-operate. The asserter-network signal is reinforced at the portfolio level by the 2026 Tahoe → Southfork IP Holdings → Powerwatch Systems (Equitable IP) cascade and the Nanya suit. Material caveats: (a) I could not retrieve the correspondent of record or the reel/frame of the 2007 inventor assignment, so the repeat-correspondent signal is unresolved; (b) no suit naming this specific patent has been located, and the patent is very near expiry (adjusted expiration 2026-11-23), which limits practical assertion value. Verify the two open items at the USPTO Assignment Center (patent 7436903 / reel 061175/0176): https://assignmentcenter.uspto.gov/

Scope note: SEC filing citations (Intel 10-K/10-Q describing the IPValue transaction) were not located within the search budget; the deal terms above rest on the IPValue press release and press coverage, not on a primary SEC document.

I'll analyze the prior art cited on the face of US 7,436,903. Let me verify details of each cited reference.

Let me verify the remaining two cited references.

Prior-Art Analysis — US Patent No. 7,436,903

Scope note (literal-ID rule). I am analyzing the prior art cited on the face of US 7,436,903 (application 10/954,734, filed 2004‑09‑29, granted 2008‑10‑14, "Multicarrier transmitter and method for transmitting multiple data streams with cyclic delay diversity," inventors Sandhu & Ho, Intel/Tahoe Research). I did not substitute any other number. This is the six-reference "Patent Citations" list printed on the patent, plus the three non‑patent citations the examiner also listed.

Important limitations. (1) I do not have the file wrapper/office actions for the '903 prosecution, so I cannot state which references the examiner actually applied and how; the following is a technical analyst's §102/§103 mapping. (2) Anticipation under 35 U.S.C. §102 requires every element of a claim in a single reference; reference dates below assume the applicant's invention date is the 2004‑09‑29 filing date (pre‑AIA patent). (3) I could not fully retrieve the disclosure of US 2003/0099304 (Gore) — that entry is flagged.

The six cited patent references (roster)

Per-reference analysis

1. US 6,842,487 B1 — Larsson (Ericsson) — the core CDD reference

Description (verified). Splits an OFDM signal across M antenna paths; in each path cyclically shifts the time‑domain samples a predetermined number of positions, wrapping samples shifted off the end into the beginning; copies the last n samples into a cyclic prefix; transmits from each antenna with a different cyclic delay per path. Claims 1–21 cover, e.g., shifting samples with wrap‑around (cl. 1, 10, 11), per‑antenna different shift (cl. 2, 9), and CP addition (cl. 13–15). Source: https://uspto.report/patent/grant/[6842487](/patent/6842487) ; https://patents.google.com/patent/[US6842487B1](/patent/US6842487B1)/en.

Potential §102 mapping. Larsson discloses, in a single reference, "cyclically shifting… time‑domain samples," "delaying some of the time‑domain samples within the group," "removing one or more of the time‑domain samples from an end of the group and adding the removed… to a beginning of the group," and "adding a cyclic prefix… comprising repeating a portion of the output time‑domain samples" — language matching claims 4, 5, 7, 18, 19, 21 and the shift mechanics of claims 3, 17.

But it does not anticipate any claim as a whole. Every one of those claims depends from an independent claim (1, 14, 24) that requires two different data streams to be cyclically shifted and linearly combined (e.g., "combine the cyclically‑shifted first time‑domain samples with the second time‑domain samples"). Larsson splits one signal across antennas — it never discloses combining a first data stream with a second data stream. Larsson therefore anticipates no independent claim (1, 9, 13, 14, 23, 24) and, because anticipation requires all elements in one reference, no dependent claim as an integrated whole. It is nonetheless the strongest §103 reference for the cyclic‑shift/CP subject matter and the closest art of record.

2. US 2004/0141548 A1 — Shattil

Description (verified). Carrier‑Interferometry (CI) multicarrier transmitter focused on PAPR reduction; describes generating time‑domain waveforms, cyclic prefixes/guard intervals, and "cyclically shifting copies of the generated time‑domain signal" and combining (summing) waveforms to synthesize signals. Source: https://worldwide.espacenet.com/publicationDetails/biblio?CC=US&NR=2004141548A1 ; https://uspto.report/patent/app/20040141548.

Potential §102 mapping. Its mention of cyclically shifting time‑domain signals and combining them touches the general concepts behind claims 1/14 and the "shift‑then‑combine" idea. However, Shattil's stated purpose is PAPR/CI coding, not dual‑data‑stream transmit diversity, and it does not disclose the specific per‑antenna mapping of shifting stream 1 and combining with stream 2 on antenna 1, and shifting stream 2 and combining with stream 1 on antenna 2. No clear §102 anticipation of any claim; at most §103 art.

3. US 2003/0026348 A1 — Liang & Chin (National University of Singapore)

Description (verified). Combined beamforming + transmit diversity for frequency‑selective fading: space‑time encode one signal into two signals, pass each through (zero‑forcing) pre‑equalizers, apply transmit beamforming weights, sum, and transmit from multiple antennas; receive at a single antenna and space‑time decode. Source: https://scholarbank.nus.edu.sg/handle/[10635/34925](https://assignmentcenter.uspto.gov/search/patent/reelFrameDetail?reelFrame=10635-34925) ; granted as US 7,167,526.

Potential §102 mapping. Discloses multi‑antenna transmission of space‑time‑encoded streams and combining/pre‑equalizing — background relevant to claims 1/14 (multi‑antenna transmit diversity). It does not disclose cyclic shifting or cyclic‑delay diversity, and does not map to the claimed shift‑and‑combine arrangement. No §102 anticipation; relevant as §103 background showing transmit diversity of multiple streams was known.

4. US 2003/0099304 A1 — Gore

Description. Title indicates transmit diversity based on transmission‑channel delay spread (delay‑diversity‑type scheme). Publication 2003‑05‑29 places it in §102(b). I was unable to retrieve this document's specification/claims before reaching my research limit, so I explicitly do not characterize its disclosure beyond its title and dates. If it discloses per‑antenna delays (as opposed to cyclic shifts) adapted to the channel delay spread, it would be §103 art for the diversity concept but still would not, on its face, anticipate a claim requiring cyclic shifting plus linear combination of two different data streams. Unable to conclude anticipation; flagged for verification.

5. US 2005/0254592 A1 — Naguib & Agrawal (Qualcomm), "Time varying cyclic delay diversity of OFDM"

Description (verified). Applies a time‑varying cyclic delay to OFDM symbols transmitted from one or more antennas; a same symbol may be sent from antennas A1…A4 with no delay / 1 / 2 / 3 samples respectively (its FIG. 4); delay may vary linearly, randomly, or by a function; CP added before/after the delay; an optional phase‑shift equivalent is disclosed (its cl. 48–63); IFFT + CP + delay elements (cl. 15, 25). Source: https://uspto.report/patent/app/20050254592 ; https://patents.google.com/patent/US20050254592.

Potential §102 mapping. Directly relevant to the "cyclically shifting… in a first/second/third manner" concept of claim 24 and to the per‑antenna different‑cyclic‑shift features. However, Naguib replicates a single symbol/stream across antennas with different cyclic delays; it does not disclose linearly combining a first data stream with a second, different data stream (the heart of claims 1, 14, 24) nor the specific claim‑13/claim‑23 arrangement transmitting different streams unshifted on separate antennas. Its publication (2005‑11‑17) and filing (2005‑03‑10) both post‑date the '903 filing; it can only be §102(e) art if the 2004‑05‑17 provisional supports the relied‑on subject matter. No §102 anticipation of the independent claims; strong §103 candidate and §102(e)/§103 reference for the shift‑attribute limitations.

6. US 2006/0067421 A1 — Walton et al. (Qualcomm), spatial spreading + STTD/SFTD/OTD

Description (verified via abstract/spec). Techniques for transmitting data using a combination of transmit‑diversity schemes including spatial spreading, continuous beamforming, cyclic delay diversity, STTD, SFTD, and OTD. Processes N_D data symbol streams into N_C coded symbol streams, then spatially spreads to N_T transmit symbol streams for N_T antennas; explicitly covers configurations such as 3 data streams sent as 4 coded streams from 4 antennas. Sources: https://patents.google.com/patent/US20060067421 ; https://uspto.report/patent/app/20060067421 ; EP family EP2802087 (priority 2004‑09‑03 / 2004‑09‑08).

Potential §102 mapping. This is the most relevant of the six to the multi‑stream/multi‑antenna claims, because it discloses (a) multiple data streams, (b) cyclic delay diversity as a transmit‑diversity scheme, and (c) mapping data streams to more antennas than streams (e.g., 3 streams → 4 antennas), which is the architecture of claims 13, 23 and 24 (N data streams, N+1 antennas; and the no‑shift/different‑stream antenna assignments). It is the only cited reference with the potential to reach the independent method claims (14, 23, 24) and possibly apparatus claim 1. Caveat: its US publication (2006‑03‑30) and filing (2005‑09‑02) both post‑date the '903 filing; a §102(e) date exists only if the 2004‑09‑03 provisional (60/607,371) supports the specific relied‑on cyclic‑shift‑and‑combine mapping. Spatial spreading (matrix weighting) is not the same as cyclically shifting one stream and linearly combining it with a different unshifted stream on the same antenna; I could not confirm the provisional/spec discloses that exact arrangement. Potential §102(e) anticipation of broad claims 14/23/24 (and material §102/§103 art for claims 1 and 13) — but unconfirmed and recommended for verification against the provisional.

Non‑patent citations (also on the '903 face; §102 printed‑publication art)

• Bauch, "Orthogonal frequency division multiple access with cyclic delay diversity," 2004 ITG Workshop on Smart Antennas, 2004‑03‑18, pp. 17–24 — §102(a) (published <1 yr before filing). OFDMA + CDD; likely the closest printed‑publication art for the cyclic‑delay‑diversity concept, but again single‑stream CDD rather than the claimed two‑stream linear combining.
• Dammann, "Beamforming in combination with space‑time diversity for broadband OFDM systems," ICC 2002, 2002‑04‑28–05‑02, pp. 165–171 — §102(b).
• Huebner, "A simple space‑frequency coding scheme with cyclic delay diversity for OFDM," 5th EPMCC, 2003‑04‑22–25, pp. 106–110 — §102(b). Space‑frequency coding with CDD; §103 background.

Synthesis — exposure by claim

Bottom line

• Most relevant cited prior art overall: US 6,842,487 B1 (Larsson/Ericsson) for the cyclic‑shift/cyclic‑prefix mechanics, and US 2006/0067421 A1 (Walton/Qualcomm) for multi‑stream transmit diversity that expressly includes cyclic delay diversity.
• No cited reference, as disclosed on its face, anticipates any independent claim (1, 9, 13, 14, 23, 24) in its entirety under §102 — chiefly because the cited CDD references (Larsson; Naguib) cyclically shift/replicate a single signal across antennas and do not linearly combine two different data streams, which the '903 independent claims require.
• §102(e) dates to watch: Naguib (provisional 2004‑05‑17) and Walton (provisional 2004‑09‑03) only qualify if their provisionals support the relied‑on subject matter. This should be verified against the provisionals — it materially changes whether Walton can be §102 art against claims 13/23/24.
• Unverified: US 2003/0099304 A1 (Gore) — I could not retrieve its disclosure and make no anticipation assertion about it.

Recommended next steps: (1) pull the '903 file wrapper to see the examiner's actual §102/§103 rejections; (2) obtain provisionals 60/572,137 and 60/607,371 to fix Naguib's and Walton's §102(e) dates and scope; (3) retrieve US 2003/0099304 A1 in full.

I'll ground my analysis by confirming what the nine prior-art references of record actually teach, then build the § 103 combinations.

Let me confirm the teachings of the two closest-in-time references (Naguib and the Qualcomm spatial-spreading application) and the Shattil/Gore references.

Obviousness Analysis — U.S. Patent No. 7,436,903 under 35 U.S.C. § 103

Builds on the earlier claim summary and litigation sections. The prior-art universe analyzed here is exactly the "Prior Art" section of the patent record: the six U.S. patent documents in "Patent Citations (6)" and the three non-patent references in "Non-Patent Citations (3)." Claim numbering follows the 24 claims already mapped.

1. Legal framework and level of ordinary skill

The application (US 10/954,734) was filed 2004-09-29, so pre-AIA § 103(a) governs. The analysis applies Graham v. John Deere Co., 383 U.S. 1 (1966), and KSR Int'l Co. v. Teleflex Inc., 550 U.S. 398 (2007), using the MPEP 2143 rationales — chiefly (A) combining prior-art elements according to known methods to yield predictable results, and (C) use of a known technique to improve a similar device in the same way.

PHOSITA: a communications engineer with a B.S. in EE plus 2–3 years (or an M.S.) working on OFDM/MIMO physical-layer design as of September 2004 — familiar with cyclic delay diversity ("CDD"), cyclic-prefix insertion, per-stream IFFT bank architectures, spatial spreading/precoding, and space-time/space-frequency coding, all of which were mature by then.

The central question is narrow: the '903 claims require generating two different data streams (each its own IFFT), cyclically shifting in the time domain, and then linearly combining the shifted samples of one stream with the samples of the other for transmission on different antennas. The novelty is therefore CDD applied per-stream plus linear combining (spreading) of distinct streams across antennas — not CDD itself.

2. References of record, dates, and key teachings

3. Analysis of the independent claims

Claim 1 (apparatus)

Element map:

• Two data streams, each with IFFT → Naguib (FIG. 2: data streams → per-stream IFFT blocks 226a–226t → per-stream CP generators); Qualcomm spatial-spreading; Huebner (multiple coded streams).
• Cyclically shift first samples; combine with second samples → antenna 1; cyclically shift second samples; combine first with shifted second → antenna 2 → the CDD mechanics are in Larsson '487, Naguib, Bauch, and Huebner; the linear combination of distinct streams across antennas is taught by Dammann (beamforming = linear antenna combining combined with CDD/ST diversity), Qualcomm (spatial spreading), and Huebner (SF-coded streams sent as shifted versions per antenna).

Primary combination: Larsson '487 + Naguib + Dammann (optionally + Qualcomm). Larsson '487 supplies the full cyclic-shift/CP/wrap mechanics and the express time↔frequency phase equivalence; Naguib supplies the multi-stream MIMO-OFDM front end (one IFFT per stream) with per-antenna differing cyclic delays; Dammann supplies the express motivation and mechanism to linearly combine streams/antenna signals together with CDD.

Claim 9 (apparatus with controller)

Adds first/second cyclic shifters plus a controller instructing each shifter whether (and how much) to shift. Larsson '487 discloses per-path shifting including a zero-shift path (first path undelayed), i.e., selective shifting; Naguib discloses per-antenna delay elements with feedback-based delay selection ("delay periods… determined based upon feedback information"); Shattil supplies the software-adaptable, reconfigurable multicarrier transmitter. Motivation: reconfigurability to adapt diversity vs. throughput — a known design goal. Obvious.

Claim 13 (apparatus: hybrid shifted/unshifted antennas)

Requires a second cyclic shifter, combiner circuitry, and RF circuitry transmitting un-shifted first signal on antenna 2 and un-shifted second signal on antenna 3 (the "2 streams / 3 antennas" pattern depicted in the '903's own Eq. 402). Larsson '487 (first path unshifted, others shifted) + Naguib (FIG. 4: same symbol undelayed on A₁, delayed on A₂/A₃/A₄) supply the unshifted/shifted mix; the N+1-antenna redundancy rationale (below) supplies the motivation. Obvious.

Claims 14, 23 (methods)

Claim 14 is the process counterpart of claim 1 — same combination. Claim 23 recites concurrent transmission on three antennas (one combined/shifted output plus two unshifted streams). Larsson '487 and Naguib both disclose simultaneous multi-antenna transmission of shifted and unshifted versions. Obvious.

Claim 24 (N streams on N+1 antennas, three different shift manners)

The distinct novelty is the arithmetic "N data streams with N+1 antennas, N = 2 or 3," with three different cyclic-shift-and-combine treatments. Larsson '487 teaches different cyclic shifts per antenna path and expressly contemplates up to M antennas for the stream; Naguib teaches multiple antennas for the streams; Huebner and Qualcomm teach spreading multiple streams across more antennas than streams. Using one extra antenna for redundancy/diversity is a predictable, well-motivated design choice (see § 4). Obvious.

4. Motivation to combine (the crux)

A PHOSITA would have combined these references for concrete, articulated reasons:

1. Same field, same problem. All nine references address multipath/flat Rayleigh fading and transmit diversity in OFDM/MIMO — squarely analogous art.
2. Express combination teaching. Dammann's title and content are the motivation: combine linear antenna combining (beamforming/precoding) with space-time diversity and CDD for broadband OFDM. Huebner likewise builds a space-frequency coding scheme with CDD. That is precisely "cyclically shift distinct streams and linearly combine them across antennas."
3. Known problem, known fix, predictable result. CDD converts spatial diversity into frequency diversity (Bauch, Larsson '487), and spatial spreading decorrelates streams (Qualcomm). Applying per-stream cyclic shifts and then spreading the streams yields an expected composite channel with improved diversity — the same result the '903 specification itself describes (2 streams/3 antennas → 2×2 composite channel, etc.) without asserting anything unexpected.
4. KSR rationale (A): combining known elements (per-stream IFFT, cyclic shifter, adder/combiner, per-antenna RF chain), each performing its known function, to yield a predictable improvement in fading robustness.
5. Design incentive / "obvious to try." Transmitting N streams over N+1 antennas provides redundancy and richer diversity without new bandwidth — an ordinary capacity/coverage trade-off. Larsson '487's zero-shift-plus-shifted structure and Naguib's mixed-delay antennas make the N+1 arrangement a routine extension.
6. Domain equivalence removes any implementation gap. Larsson '487 and Naguib both expressly note that a time-domain cyclic shift equals a frequency-domain linear phase, so implementing the combination in the time domain (as claimed) versus the frequency domain (as Qualcomm's spreading matrix does) is a mere design choice.

5. Anticipation overlaps and date caveats (must be flagged)

• Larsson '487 alone anticipates several dependent claims (e.g., claims 5, 19: "removes samples from an end of the group and adds them to a beginning"; claims 7, 21: cyclic-prefix addition; claims 4, 18: delaying samples within a group), and would support § 102 challenges as well as § 103.
• Naguib US 2005/0254592 is prior art only under § 102(e), and its date is contested. Its non-provisional was filed 2005-03-10, after the '903 filing (2004-09-29). It can only antedate the '903 as § 102(e) art if the relied-upon disclosure is supported by provisional 60/572,137 (2004-05-17). Any reliance on Naguib subject matter not carried in the provisional fails as prior art. This materially weakens Naguib-based combinations.
• Qualcomm US 2006/0067421 is likewise § 102(e)-only as of its 2004-09-03 priority — a mere 26 days before the '903 filing. If the applicant can show an earlier invention date (conception + diligence), this reference drops out.
• By contrast, Larsson '487 (§ 102(e) 2000 / § 102(b)), Gore '304 and Singapore '348 (§ 102(b)), and the Dammann, Huebner, and Bauch NPL (§ 102(b)/§ 102(a)) are solidly prior art on their faces.

Net effect: the durable obviousness case rests on Larsson '487 + Dammann + Huebner (+ Bauch), all of which pre-date the '903 by more than one year except Bauch (2004-03-18, ~6 months, § 102(a)). Naguib and Qualcomm are the strongest multi-stream references but are the most date-fragile.

6. Potential rebuttals and secondary considerations

• Teaching away (weak): Bauch notes space-time block codes outperform CDD at higher complexity — but that is a performance/complexity trade-off, not a teaching away from combining CDD with multi-stream/spreading. Dammann and Huebner affirmatively combine them, negating any teaching-away argument.
• Secondary considerations: I found no evidence of nexus-bearing objective indicia — no litigation, licensing campaign, industry praise, or unexpected-results data tied to claim 1's combination. (Consistent with the earlier section's finding of no known litigation involving 7,436,903; the Samsung/Collision matters concern the different patent 7,463,703 and are not probative here.) The patent has ~7 months of term remaining (adjusted expiration 2026-11-23), further undercutting any commercial-success narrative.
• Prosecution posture: the examiner allowed over these references, but the record does not show the examiner ever applied the multi-stream + CDD + linear-combining combination (Dammann/Huebner + Larsson '487 + Naguib). That combination, not any single reference, is the strongest § 103 case.

7. Conclusion

Bottom line: Every independent claim would have been obvious over the cited art. The single most persuasive combination is Larsson '487 (CDD mechanics) + Huebner (multi-stream space-frequency coding with CDD) + Dammann (express teaching to combine linear antenna combining with space-time/CDD), reinforced by Naguib for the per-stream-IFFT MIMO-OFDM transmitter. The claims' only arguable contribution — shifting distinct streams and then linearly combining them across N+1 antennas — is the predictable union of CDD (Larsson '487, Bauch) with spatial spreading/precoding (Qualcomm) and multi-stream CDD coding (Huebner), performed for the known purpose of multipath-diversity gain.

Caveats: (1) Where the analysis depends on Naguib (US 2005/0254592) or Qualcomm (US 2006/0067421), those references are prior art only as § 102(e) art and only to the extent their relied-upon disclosure is supported by their earlier provisional/priority filings; both are vulnerable to an earlier-invention-date showing. (2) Reference content for the Qualcomm '421, Shattil '548, Singapore '348, and Gore '304 documents is taken from the patent record's identifiers/titles and general field knowledge; I did not independently retrieve their full texts within this session, so those specific mappings are stated at lower confidence than the Larsson '487, Naguib, Dammann, Huebner, and Bauch mappings, which I verified against source documents.