US 4130729

I'll search for information on this specific patent number, including any litigation or docket activity.

Search scope note

I searched for U.S. Patent No. 4,130,729 specifically. I found no indication of any CAFC 2026 docket activity involving this patent. A web search cannot prove a universal negative, so treat this as "no evidence found," not "confirmed none exists." Two important look-alike hits must not be conflated with this patent:

• JP 4130729 B2 — a Japanese patent (published 2008-08-06) on infectious cDNA clones of North American PRRS virus. Unrelated subject matter and a different jurisdiction.
• FR trademark 4130729 ("LA COUPOLE") — a French trademark record, not a patent.

Only US 4,130,729 A is the patent you asked about.

Bibliographic data — US 4,130,729 A

Prior art cited on the face: US 3,715,512 (Bell Telephone Laboratories, "Adaptive predictive speech signal coding system"); US 3,973,081 (TRW Inc., "Feedback residue compression for digital speech systems"); non-patent literature: P. Cummiskey et al., "Adaptive Quantization in Differential PCM…," Bell System Tech. J., Sept. 1973, pp. 1105–1118. Family citation: GB 1,318,985 (Nippon Telegraph & Telephone, "Audio response apparatus").

Abstract (as granted)

A compressed speech system using a residual encoder that transmits or stores a digitized speech signal with one bit of resolution, from which the original waveform can be reconstituted without a vocal-tract analog. The digital signal is the remainder/error between the original input and its predicted value. The predictor uses an analog delay line of sample-and-hold modules feeding correlation multipliers; the summed multiplier outputs estimate the next input value and are fed back to the delay-line input, forming a recursive filter continuously tuned to the input's correlation statistics. Duty cycles for the correlation coefficients are obtained by four-quadrant multiplication of an analog signal with the digital speech output, implemented with analog gate pairs. Finite-DC-gain integrators prevent transmitter/receiver divergence and provide immunity to moderate bit errors. An absolute-value detector scales the system to follow input amplitude variations.

Plain-language overview of the independent claims

Claim 1 — the overall compression system.
A speech compression system that cuts speech down to a minimum number of binary digits by exploiting redundancy ("prediction of redundency," sic). It requires:

• Input means — e.g., a microphone — producing an analog input signal from speech.
• Predictor means — produces a signal predicting the input waveform.
• Comparison means — compares input vs. predicted and outputs a difference signal.
• Conversion means — an A/D stage producing a "synonymous" binary error signal (form changed, information content preserved).
• Reconversion means — a D/A stage reproducing the difference signal.
• Reconstitution means — sums the predicted signal and the reproduced difference to regenerate the input signal.

The claim closes with the feedback loop: the predictor receives both the error signal and the reconstituted input signal, and the prediction is a function of both. In plain terms: this is a closed-loop differential (residual) coder whose decoder-side reconstruction is fed back into the predictor so encoder and decoder stay matched.

Claim 15 — the receiver (decoder) standing alone.
A receiver for a residual encoder that receives a one-bit-resolution output signal. It requires:

• Reconversion means — turns the received bit stream into a "synonymous received difference signal reproduction."
• Predictor means — generates a predictive signal.
• Reconstitution means — sums the reproduced difference and the prediction to output a received input-signal reproduction that is substantially identical to the original input.

Again the predictor takes both the received output signal and the reproduced input signal as inputs, making the prediction a function of both. In plain terms: an independent decoder-side claim to the expander — notable because the specification states the receiver is "substantially identical" to the compressor module, and the patent discloses that removing a jumper and applying the residual bit stream to the receiver input converts the compressor circuit into the receiver.

The remaining 14 claims are dependent, adding: a data link between remote locations plus second reconversion/predictor/reconstitution means (claim 2); local and remote output means (claims 3–4); predictor implementation detail — delay line of sample-and-hold modules with correlation multipliers generating duty cycles (claim 5), summing integrator (claims 6–7), leaky integrators buffering bit errors (claim 8), analog gate pairs for four-quadrant digital-analog multiplication (claim 9), and an analog delay line (claim 10); and absolute-value detectors placed at various points (claims 11–14); claim 16 adds remote output means to the receiver claim 15.

Practical / technical notes

• Terminal disclaimer of the family: the sibling divisional US 4,140,876 shares the same priority date; both are listed as expired for failure to pay maintenance (anticipated expiration 1997-09-19), so both are long expired and in the public domain.
• Datasheet caveats (apparent typographical inconsistencies in the source text, useful if you are relying on the specification):
  • The comparator is described both as a "type 1709 comparator 104" and as a "type 709 uncompensated operational amplifier."
  • Reference numeral 114 is used for both the "clock flip-flop" and the "jumper connector."
  • The FIG. 3A discussion appears to reference "comparator 66" where the figure labels a comparator 67.
  • The summing-integrator feedback is described as "capacitors 154 and 165" in one passage while elsewhere the pair is 154 and 156 (2,000 pF nominal).
• Bit-rate figures in the specification: clock 12.8 kHz (twice the bit rate); bit rate normally 6.0 or 6.4 kHz, workable down to 4.8 kHz with integrator capacitors increased to 2,700 pF, and improved fidelity at 9.6 kHz with 1,300 pF.

Uncertainty: I have the authoritative granted text for US 4,130,729 (Google Patents / FreePatentsOnline), so the bibliographic and claim content above is well grounded. I do not have authoritative data from USPTO PatentCenter or a CAFC docket showing any 2026 litigation or appeal for this patent number; the CAFC calendar hits returned by search were unrelated generic summaries.

I'll search for litigation involving US patent 4130729 across patent litigation sources.

Let me search additional litigation-specific sources.

Litigation Search Report — U.S. Patent No. 4,130,729

Bottom Line

I found no litigation involving U.S. Patent No. 4,130,729. No district court cases, no CAFC appeals, no PTAB proceedings, and no ITC investigations could be identified in the sources I searched. I'd characterize this as "no known litigation" rather than an affirmative proof of none, for the reasons explained below.

What the Patent Is (for identification confidence)

Before reporting a negative, it's worth confirming we're talking about the right document, since the digit string "4130729" collides with several unrelated identifiers (e.g., JP 4130729 B2, a Japanese PRRS virus patent; DE 4130729 A; French trademark 4130729; a Finnish lighting SKU; a Florida FEI number). The U.S. patent at issue is:

• Patent: U.S. 4,130,729, "Compressed speech system"
• Inventor: Richard T. Gagnon
• Assignee: Scitronix Corp.
• App. No.: US 05/834,642
• Filed / Priority: September 19, 1977
• Granted: December 19, 1978
• Status: Expired – Lifetime (anticipated expiration September 19, 1997)
• Related: U.S. 4,140,876 ("Compressed speech system and predictor"), a division filed March 29, 1978, granted February 20, 1979, same priority date

Search Coverage

Case-by-Case Table

There is nothing to report for plaintiff, defendant, jurisdiction, case number, filing date, or outcome, because no case was found.

Important Caveats

1. Search-tool limitations. I was working from open web sources. PACER's Case Locator requires a registered/fee-based account and is not fully indexable by these searches, and Westlaw/Lexis litigation-history ("KeyCite"/CourtLink) reports are behind paywalls. A definitive negative would require running the patent number through those databases directly. I therefore cannot rule out a pre-2000 paper-era suit that was never digitized.
2. Citation ≠ litigation. The patent appears in later patents' "References Cited" and "Cited By" lists (e.g., it is cited in U.S. 4,472,832, U.S. 5,673,364, and several hearing-aid patents). That is prior-art citation activity, not litigation. Similarly, its appearance in the Espacenet search-report citations for US4472832 is patent-office citation, not a lawsuit.
3. Design-around/Term. The patent term expired September 19, 1997. Given the ~20-year useful life and the narrow, analog-circuit-specific disclosure (a residual encoder using analog sample-and-hold delay lines and analog gates), there is no obvious modern assertion window — which is consistent with finding no litigation.
4. Related patent. If you are looking for enforcement history in the same family, you may want to separately check U.S. 4,140,876, the divisional. I did not find litigation on it either, but flag it as the more likely family member to check if the '729 yields nothing.

Recommendation

If you need a litigation-grade clearance statement rather than a research summary, this should be confirmed through a formal litigation-history search on Westlaw (KeyCite) and Lexis (CourtLink) covering the patent's full life (1978–1997 plus post-expiration citations), and a PACER Case Locator run under the "4,130,729" and party names "Scitronix" / "Gagnon." I can draft the specific search strings for those services if that would help.

Proceedings overview

Zero AIA trial proceedings are on file for US 4,130,729 — as of the USPTO ODP ingest reflected in the "PTAB proceedings on file" block (0 IPR, 0 PGR, 0 CBM), with no IPR/PGR/CBM petition found in web searches of PTAB-related sources as of 2026-09-26. The bottom-line defensive posture is therefore not "hardened by prior IPRs" and not "claims canceled" — it is that the patent is untested at the Board and expired, so the real defense is not a PTAB record but the patent's own status: priority/filing 1977-09-19, grant 1978-12-19, and an anticipated expiration of 1997-09-19 (Google Patents). Important caveat on the "absence is a signal" heuristic below: it applies weakly here, because the patent predates the AIA by ~35 years and has been off-term for ~29 years.

Proceedings

None. There is no proceeding to report at claim-level granularity. I am not going to invent an IPR number, a petitioner, or an FWD that does not exist. Specifically:

• Structured source (canonical): ODP returns no AIA trial proceedings for US 4,130,729.
• Web cross-check: searches for PTAB/IPR material keyed to "4130729" and "Scitronix" surfaced only patent-family and citing-art records (e.g., the sibling division US 4,140,876A, "Compressed speech system and predictor," filed 1978-03-29, priority 1977-09-19, also on the same Google Patents family page) — no petition, institution decision, FWD, or appeal. Nothing found for the '876 division either.
• No judge panel, no grounds, no institution decision, no FWD, no settlement, no § 315(e)(2) estoppel, no CAFC appeal exists to describe, because there is no proceeding.

Why the trial types are foreclosed (contextual, not a finding of any filing)

• PGR: statutorily unavailable. 35 U.S.C. § 321(c) requires filing within nine months of grant, and PGR is limited to patents with an effective filing date on or after 2013-03-16. This patent's effective filing date is 1977-09-19. A PGR here was never possible.
• CBM: the transitional CBM program (AIA § 18) sunset on 2020-09-16 and in any event reached only patents with claims for "performing data processing or other operations used in the practice, administration, or management of a financial product or service." A 1970s analog residual speech coder is not CBM-eligible.
• IPR: legally available in theory — the Board does institute on expired patents — but practically pointless and never invoked here. Note the key expiry consequence: for an expired patent the claims cannot be amended in an IPR, and infringement liability (and therefore any § 286 six-year damages lookback) ended long ago.

Strategic summary

Claim status: the whole set (claims 1–16) is UNTESTED at the PTAB, and the patent is expired. Claims 1–7 (the core residual-encoder/system claims), claims 8–10 (leaky integrators, analog-gate four-quadrant multipliers, analog delay line), claims 11–14 (absolute-value-detector placements), and claims 15–16 (receiver for a residual encoder) have never been construed, instituted against, or canceled anywhere on the public PTAB record. There is no "claims canceled" list to hand a defendant. Note that some claims are highly vulnerable on their face if anyone ever bothered — claims 11–14 are essentially functional placement claims for a scaling detector, and the specification's own admissions about prior-art vocoder and adaptive-PCM work (including the examiner-cited art US 3,715,512 (Bell Telephone, 1973) and US 3,973,081 (TRW, 1976), plus the non-patent citation to Cummiskey et al., Adaptive Quantization in Differential PCM, Bell System Tech. J. (Sep. 1973)) frame the predictive-coding field. But that is a district-court § 102/§ 103 analysis, not a PTAB record.

Estoppel landscape: none, because no petitioner exists. § 315(e)(2) estoppel only attaches to a petitioner that obtains an FWD (or terminates after institution). Since there is no IPR petitioner, no party — and no privy — is estopped from raising any prior-art ground against this patent. Practically, the absence of estoppel is worth nothing, because there is nothing left to invalidate around: the patent's term ran to 1997-09-19 (consistent with the URAA transitional rule giving patents in force on 1995-06-08 the greater of 17-years-from-issue (1995-12-19) or 20-years-from-filing (1997-09-19)). One data caution: a Google Patents "Legal Events" block on a page that lists US 4,130,729 as citing art shows a maintenance-fee lapse entry with an effective date of 2012-04-08 — that entry belongs to the citing patent (US 7,356,153), not to the '729 patent, whose own structured record shows anticipated expiration 1997-09-19. Do not brief the 2012 date to the '729 patent.

Pattern signals. There is no pattern to read: no repeat petitioner, no patent-owner appeal activity, no defensive aggregator (Unified Patents or similar) in the chain for this patent. Assignee history is a single entry — SciTronix Corporation, original and current assignee, inventor Richard T. Gagnon; the only related family member is the divisional US 4,140,876A. The citing-art trail shows the patent being used as background art by later AT&T/Bell Labs, Dolby, VMX, Sony, and NEC speech/audio-coding patents — i.e., it was a citation target, never a challenge target.

Recommended next steps

1. If you are a defendant and a demand letter asserts US 4,130,729, lead with expiry, not with an IPR. Confirm the term position from the front page and legal-status record: filing 1977-09-19, grant 1978-12-19, anticipated expiration 1997-09-19 (Google Patents; USPTO PatentCenter). Any infringement theory premised on post-1997-09-19 activity is time-barred, and any damages model is capped by § 286's six-year lookback measured from a date ~29 years past. There is no PTAB FWD to cite, because none exists.
2. Verify the negative independently before relying on it. Pull the patent number through PTAB E2E / PTAB Center (https://ptacts.uspto.gov/ptacts/) and the USPTO Open Data Portal trial-proceedings endpoint, and check CourtListener for any appeal of a nonexistent FWD (there shouldn't be one). My search coverage was web-based plus the ODP block; I did not query E2E directly, so treat "zero proceedings" as high-confidence but not independently double-sourced by me.
3. No trial-stage milestones to track. With no petition there is no § 314(b) institution deadline, no oral hearing, and no § 316(a)(11) one-year FWD deadline running. If you are nonetheless contemplating an IPR/§ 112-style attack for defensive-record purposes, remember (a) no claim amendments are permitted against expired claims, and (b) IPR is limited to § 102/§ 103 on patents and printed publications — § 112 and § 101 challenges would have to be run in district court.
4. Treat the absence of PTAB activity as weak evidence here. For a live, well-asserted patent, "no IPRs" is a meaningful signal that the claims are hard to attack. For a 1978 patent expired in 1997, it means only that nobody had a reason to pay the fees. Do not tell a client the patent "survived two IPRs" — it survived nothing, because nothing was ever filed.

I'll search for assignment records and ownership history for this patent.

Let me check the Justia assignee page and search for any recorded transfers.

Let me search for the assignment records directly and investigate Scitronix Corp.'s corporate history.

Assignment & Ownership Analysis — U.S. Patent No. 4,130,729 ("Compressed speech system")

Methods caveat (read first). The USPTO Assignment Center (https://assignmentcenter.uspto.gov/) and the legacy Assignment Search (https://assignment.uspto.gov/patent/index.html) are interactive, JavaScript-driven applications whose result sets are not exposed to the open-web search tools available to me. I therefore could not retrieve verbatim reel/frame records for this patent, and I will not invent reel or frame numbers. What follows is (a) what the retrievable secondary sources show, (b) explicit labeling of what is unverified, and (c) the NPE-signal table populated on the basis of that record. The corroborating sources I could reach are Google Patents' bibliographic/legal-events data, the Justia assignee page for Scitronix Corporation, uspto.report's grant pages for US 4,130,729 and US 4,140,876, and Espacenet's bibliographic/assignee fields.

Cross-reference note (contradiction flagged). An earlier section of this analysis states in "Practical / technical notes" that the family patents are "listed as expired for failure to pay maintenance (anticipated expiration 1997-09-19)." The bibliographic table in the same document states the status as "Expired – Lifetime." Those are different things: "Expired – Lifetime" denotes expiry of the statutory term, not a fee lapse. The PTAB section separately warns that a 2012 maintenance-fee-lapse legal event on a Google Patents page belongs to a citing patent (US 7,356,153), not to '729. I therefore treat the "failure to pay maintenance" characterization as unsupported and proceed on the record's own status: term expiry at 1997-09-19. This matters for ownership analysis because a term-expired patent (rather than a fee-lapsed one) means Scitronix held the asset for its full life and never monetized or abandoned it early.

Inventors

Basis for the employer inference. The patent issued to Scitronix Corp., and the sibling divisional US 4,140,876 ("Compressed speech system and predictor," filed 1978-03-29, granted 1979-02-20) lists the same inventor and the assignee as "Scitronix Corp. (Ann Arbor, MI)" (uspto.report/patent/grant/4140876). The same company filed US 3,836,717, "Speech synthesizer responsive to a digital command input," application filed 1971-03-01 and granted 1974-09-17 — putting Scitronix in the speech-synthesis business in Ann Arbor at least six years before the '729 filing. A sole inventor filing in the company's core technical area, with the company as named assignee, is the ordinary employee-invention pattern.

Pattern flags: I found nothing unusual. Specifically:

• No evidence of multiple inventors departing the assignee within 12 months of filing (there is only one inventor).
• No evidence Gagnon assigned away personal rights (i.e., no assignment record of record showing Gagnon as assignor post-issuance was surfaced).
• No evidence Gagnon appears as inventor on later, unrelated patents that would suggest he left and went elsewhere. I did not run a full inventor-name search across all USPTO grants, so treat this as "no evidence found," not "confirmed none."
• Do not conflate Gagnon the inventor with "Gagnon" the voice talent. A cultural-history essay on the Votrax speech synthesizer discusses a broadcast voice attributed to a "Gagnon" whose vocal characteristics were used in Votrax devices (kresgeartsindetroit.org PDF, "Hello My Name Is Votrax"). That is a voice-source attribution in a popular-music/TTS context and is not evidence about Richard T. Gagnon's employment or the '729 patent. I flag it only because the surname collision could mislead a casual search.

Original assignee

Scitronix Corp. (rendered on the Google Patents record as "SCITRONIX CORP"; elsewhere as "Scitronix Corporation"), Ann Arbor, Michigan.

• Line of business. Speech synthesis and speech-processing electronics. Its 1971 filing, US 3,836,717, is a phoneme-generating speech synthesizer using digital phoneme addressing with analog-controlled formant filters — squarely a speech-technology shop, not a general electronics firm. The '729/'876 pair sits in the same technical lane (speech compression/bandwidth reduction).
• Did it ship a product embodying the claims? Unclear / not established. US 3,836,717 describes a speech synthesizer architecture that plausibly maps to a commercial device, but I found no documented commercial product carrying the '729 residual-encoder technique (an analog sample-and-hold delay line with analog-gate four-quadrant multipliers). Note that the Votrax line of speech synthesizers was associated with Votrax Inc., a Detroit-area operation — a different corporate chain. I have no evidence Scitronix's compression technology reached a shipping product, and I explicitly decline to bridge Scitronix to Votrax without a document.
• Current status. Unclear. Scitronix Corp. was a small private concern; it does not appear in my searches as an active registrant, an SEC filer, an acquirer, or a party in any assignment/litigation record tied to this patent. I could not confirm whether it is dissolved, dormant, or simply still exists. Its patents expired in 1997 and 1995 (17-years-from-issue for the '876) without any recorded transfer, which is consistent with a small company quietly winding down rather than a fire-sale.
• Do not conflate these namesakes:
  • Scitronix (Beit Shemesh, Israel; founded 1995; defense/electro-optics sales rep) — a completely unrelated 1995 Israeli company surfaced as a search hit.
  • Scitronix Corporation (Princeton, New Jersey) — appeared in an earlier section's summary; I found no evidence connecting it to the Ann Arbor speech-compression entity. Treat Princeton vs. Ann Arbor as an unresolved discrepancy pending a state corporate-registry check.
  • Scitronix (Sarnoff/defense-imaging lineage, later eras) — unrelated.

Assignment timeline

Recorded post-issuance assignments located: none.

I could not obtain reel/frame entries from the Assignment Center interface. From every indexed source reachable, the ownership picture is a single-hop chain: inventor → issuer, with no subsequent recorded conveyance:

• 1977-09-19 — Application 05/834,642 filed; applicant/inventor Richard T. Gagnon; assignee Scitronix Corp.
• 1978-03-29 — Divisional application 05/891,242 filed (same priority, same inventor, same assignee).
• 1978-12-19 — US 4,130,729 A granted to Scitronix Corp.
• 1979-02-20 — US 4,140,876 A (divisional) granted to Scitronix Corp.
• 1997-09-19 — Anticipated expiration (20 years from filing).

Google Patents lists the original assignee and the current assignee as the same entity — SCITRONIX CORP — and displays no assignee-change legal event for this patent. That is the standard signature of a patent that was never assigned after issue.

Two points of precision:

1. The absence of a recorded post-issuance assignment is itself the finding. It means the original assignee — Scitronix Corp. — was, on this record, the owner for the patent's entire life, including at expiry in 1997. There is no chain to trace, no LLC hop, and no asserter in the current assignee field.
2. A pre-issuance inventor assignment may nonetheless exist. Employee-invention assignments are routinely recorded with the USPTO and are often not reflected in Google Patents' simplified "original/current assignee" fields. The '729's front page as reproduced in this analysis does not carry an "Assignee" line at all, yet Google Patents and Espacenet both attribute the grant to Scitronix Corp., which implies either a recorded pre-issuance assignment or a declaration of assignment filed during prosecution. The reel/frame and correspondent for that filing — if any — are the one thing a live Assignment Center query would add, and I could not obtain it. Do not treat the blank reel/frame above as confirmation that no such record exists.

Per the task's instruction: because no records were obtained, I proceed cautiously through the remaining sections rather than stopping, but every NPE finding below is capped at what the record supports.

Timeline diagram

timeline
    title Ownership of US 4130729
    1971 : Scitronix files speech synthesizer patent
    1977 : Application filed by Scitronix Corp
         : Sole inventor Richard T Gagnon
    1978 : US 4130729 granted to Scitronix Corp
         : Divisional US 4140876 filed
    1979 : Divisional US 4140876 granted
    1997 : Anticipated term expiration
         : No assignment transfer recorded

NPE / troll-pattern signals

On the "expired but never asserted" posture. The patent's full useful life ran 1978–1997 and produced no recorded transfer, no litigation, and no PTAB challenge. That is the profile of an internally held, never-monetized asset — but note carefully that this is not the same as the "defensive aggregator" verdict, because no defensive buyer was involved. It is also not evidence of technical merit, for the reason the PTAB section gives: for a 1978 patent that expired in 1997, "no IPRs" means only that nobody had a reason to pay the fees.

Verdict

Insufficient data.

The verdict is driven by the state of the record, not by the merits: the only ownership evidence available is the single hop from sole inventor Richard T. Gagnon to original assignee Scitronix Corp., which Google Patents shows as both the original and the current assignee with no assignee-change legal event; the divisional US 4,140,876 likewise issued to Scitronix Corp. (Ann Arbor, MI). There is no recorded post-issuance assignment, no chain, no successor entity, and therefore no NPE footprint to assess — and the patent expired 1997-09-19, twenty-nine years ago. Every NPE signal above is either "not present" on a record with no transfers, or "unclear" because the underlying reel/frame and correspondent data sit behind an interface I could not query.

Verification link (base entry points — search by patent number "4130729" and by assignor "Gagnon, Richard T" / assignee "Scitronix"):

• https://assignmentcenter.uspto.gov/
• https://assignment.uspto.gov/patent/index.html

What a paid/full query would add, and my confidence in each: (i) whether a Gagnon→Scitronix pre-issuance assignment was recorded, and at what reel/frame with what correspondent — high value, not currently knowable to me; (ii) confirmation that no post-issuance conveyance exists — my confidence is high based on Google Patents' assignee-change silence, but not independently double-sourced through Assignment Center or PEDS/ODP; (iii) the prosecuting firm of record — would resolve the "correspondent" signal properly; (iv) the Ann Arbor vs. Princeton discrepancy in the assignee's address — worth a Michigan and New Jersey corporate-registry pull.

I'll search for the authoritative USPTO records and the cited prior-art references.

Let me verify the USPTO record directly and research the non-patent literature citation.

Prior-Art Analysis — U.S. Patent No. 4,130,729 ("Compressed speech system")

Scope note and how I identified the references

I first confirmed the target record so I would not import art from a look-alike. The only U.S. patent matching 4130729 is US 4,130,729 A — "Compressed speech system," Richard T. Gagnon, Scitronix Corp., App. No. 05/834,642, filed/priority 1977-09-19, granted 1978-12-19 (Google Patents; FreePatentsOnline). This is consistent with the earlier bibliographic section — no contradiction found. (As previously flagged, JP 4130729 B2, DE 4130729 A, and FR trademark 4130729 are unrelated and are excluded. Also note a minor date inconsistency between the earlier sections: the header of this task gives the analysis date as 2026-04-26, while the patent fetch is stamped 2026-09-26; it does not affect the prior-art analysis.)

The complete list of prior-art references cited on the face of US 4,130,729 is short — two U.S. patents (both marked "cited by examiner," i.e., actually applied during prosecution) plus one non-patent publication. Google Patents lists them as follows (US4130729A):

Also listed is GB 1,318,985 A (Nippon Telegraph & Telephone, "Audio response apparatus"), but only under "Family Cites Families" — i.e., it is cited by a family member, not on the face of the '729 patent itself. I therefore treat it as a family-level citation, not a §102 reference against '729.

All three references pre-date the '729 filing (1977-09-19) by more than a year and are thus available as prior art under pre-AIA 35 U.S.C. §102(a) and particularly §102(b) (the '729 application is a 1977 filing, so pre-AIA law governs).

Reference 1 — US 3,973,081 A (the most relevant reference)

Full citation: US 3,973,081 A, "Feedback residue compression for digital speech systems." Inventor: Sandra E. Hutchins. Assignee: TRW Inc. Application filed 1975-09-12; patent issued 1976-08-03. (Google Patents; uspto.report/patent/grant/3973081; scispace summary)

Brief description. A digital speech compression system built around a predictive feedback loop, expressly aimed at simplifying the hardware of prior adaptive-predictive coders. The sequence is: speech source → A/D converter → first adder → quantizer → compression logic → transmission channel. A predictor loop runs from the compression logic through a second adder and a digital predictive filter; the filter output is fed negatively to the first adder and positively to the second adder. The quantizer "may, for example, consist of a two-valued limiter" — yielding a one-bit residual — with compression logic either inside or outside the feedback loop. The system compresses the residue within the quantizer loop. A receiver (FIG. 3) has a converter 26 (maps the compressed code C back to Q), an adder 27, and a digital predictive filter 28 identical to filter 18, with the filter output fed back so the receiver's reconstruction tracks the transmitter's.

Why it is the closest art. It is the same closed-loop differential/residual architecture the '729 claims recite — comparison of input against a prediction, a quantized (one-bit-capable) error output, and a receiver-side predictor that is re-driven by the reconstructed signal — differing from '729 primarily in that '973,081 is a digital (not analog) implementation and does not use an analog sample-and-hold delay line, correlation multipliers, duty cycles, analog gates, or an absolute-value scaler.

Claims potentially anticipated under §102 (single-reference, every element):

Reference 2 — US 3,715,512 A (secondary reference)

Full citation: US 3,715,512 A, "Adaptive predictive speech signal coding system." Inventor: James Michael Kelly. Assignee: Bell Telephone Laboratories, Incorporated. Application No. 05/209,655, filed 1971-12-20; patent issued 1973-02-06. (Google Patents PDF; FreePatentsOnline; uspto.report/patent/grant/3715512)

Brief description. An adaptive predictive coder / linear-prediction vocoder for speech. The transmitted signal consists of an rms value, a pitch signal, a voice/unvoiced indication, and several parameter signals that adjust the coefficients of a linear predictor. The invention improves transmission by generating a low-rate pulsive signal and shaping its spectrum with the parameter signals, so the pulsive signal acts as a carrier for the parameters. The required bandwidth is substantially less than the original speech and somewhat less than for predictively coded signals. The system uses a linear-prediction vocoder analyzer and a linear-prediction vocoder synthesizer at the receiver (claim 1), and it relies on prediction of the present value of the signal from its past. It builds on the Atal predictive-coding work (referring to B. S. Atal US 3,631,520 and US 3,624,302).

Why it does not anticipate. '512 is a parameter/vocoder scheme: it transmits spectral-envelope and pitch/voicing parameters and reconstitutes speech through a dedicated synthesizer. There is no one-bit residual that "contains all of the parametric information necessary to reconstruct the input signal," and no reconstitution "without a vocal tract analog" as the '729 specification requires. Its predictive-coding content (compare input against a prediction, transmit a remainder) is generic background rather than the claimed combination.

Claims potentially anticipated under §102: None. In particular it does not disclose claim 1's "synonymous binary error signal," the reconversion/reconstitution chain, or the predictor-receives-both-signals limitation; nor claim 15's residual-decoder structure; nor any of claims 5–14's analog/duty-cycle/absolute-value elements. Its role is as §103 background art on adaptive predictor means alone.

Reference 3 — Cummiskey, Jayant & Flanagan (1973) (secondary reference)

Full citation: P. Cummiskey, N. S. Jayant, and J. L. Flanagan, "Adaptive Quantization in Differential PCM Coding of Speech," Bell System Technical Journal, Vol. 52, No. 7, pp. 1105–1118, September 1973; DOI 10.1002/j.1538-7305.1973.tb02007.x. (Bell Laboratories, Murray Hill, NJ.) (DOI record; JASA abstract of companion paper)

Brief description. The classic ADPCM reference. Describes a differential-PCM coder in which the quantizer is automatically adapted to the error variance: digital logic examines the transmitted bit stream and adjusts the quantizer step size to reduce slope overload and granular distortion, improving quality over PCM and fixed DPCM at a given bit rate. The paper expressly states that "the receiver incorporates adaptation logic identical to that of the transmitter, and reconstructs a signal estimate similar to that made by the transmitter," with digital hardware implementation and transmission performance demonstrated at 18, 24, and 32 kb/s.

Why it does not anticipate. The disclosed coder is a multi-level (n-bit, e.g., 2–4 bit) DPCM/ADPCM system; it does not produce a one-bit residual, does not use a recursive analog predictor with a sample-and-hold delay line, and does not disclose the duty-cycle/analog-gate estimator or absolute-value scaler of '729.

Claims potentially anticipated under §102: None standing alone. Its utility is §103 background: it evidences the transmitted-difference-signal-plus-matching-receiver reconstruction concept central to claims 1 and 15, and the notion of transmitter/receiver adaptation matching that the '729 specification itself invokes when discussing receiver tracking.

Ranking and bottom line

Most relevant prior art: US 3,973,081 A (Hutchins / TRW). It is the only cited reference that discloses the full closed-loop residual-encoder architecture — input comparison, quantized (one-bit-capable, via a two-valued limiter) error output, and a receiver-side predictor re-driven by the reconstruction — and it is therefore the reference that potentially anticipates independent claims 1 and 15, and dependent claim 2, and that supplies the strongest §103 case against the remaining claims. The '729 patent's likely patentable distinction over it lies precisely in the dependent claims 5–14: the analog sample-and-hold delay line, correlation multipliers generating duty cycles via analog gate pairs for four-quadrant multiplication, the summing integrator, the finite-gain ("leaky") integrators that guard against transmitter/receiver divergence and bit errors, and the absolute-value detector scaling.

US 3,715,512 A (Kelly / Bell) and Cummiskey et al. (1973) are secondary: useful for §103, not anticipatory. Note also that the Atal patents (US 3,631,520, US 3,624,302) surfaced in the '512 text are not cited on the face of '729 and should not be treated as §102 art against it without a separate justification.

Not prior art (do not use): US 4,140,876 (the same-family divisional, filed 1978-03-29) postdates '729's filing and shares its priority/inventor/assignee; and the ~18–21 documents in the "Cited By" lists postdate '729 and are later art, not prior art against it.

Caveats on completeness

• Negative limitation on the record. Google Patents displays only these two patent citations and one NPL item on the face of US 4,130,729. The USPTO PatentCenter/Public-PAIR original file wrapper could contain additional art considered but not printed on the face (it is common for examiners to consider references that are not listed). I could not independently pull the original 1977–78 file-wrapper contents from the open sources searched, so treat the three-item list as what is published, not necessarily everything the examiner saw.
• §102 vs. §103 framing. The mappings above describe potential single-reference anticipation. Because claim 1 and claim 15 are drafted in broad means-plus-function language, whether a given reference "anticipates" versus renders obvious can turn on how broadly the means-plus-function limitations are construed; I have flagged where a reference meets an element only equivalently (e.g., '081's digital/distributed "reconversion means").
• Identifiers preserved literally. Terms such as "redundency," "comparitor," and "temporial quantizer" appear as written in the '729 claims/specification and have not been corrected.

I'll ground this analysis in the actual prior-art references cited on the patent's face. Let me retrieve their content.

Let me ground the non-patent literature reference and the sibling/related patents.

Obviousness Analysis — U.S. 4,130,729 (35 U.S.C. § 103)

Scope note. This is a technical analyst's assessment based on the prior art appearing in the "Prior Art" section of the patent page (the two U.S. patent citations, the Cummiskey non‑patent literature, and the family citation GB 1,318,985). I do not have the prosecution file history (examiner's reasons for allowance, applicant arguments, amendments), so I cannot tell you why the examiner allowed the claims over these very references. Treat the ratings below as risk estimates, not validity opinions.

A. Statutory posture

• Pre‑AIA patent: filed/priority 1977‑09‑19, granted 1978‑12‑19. Obviousness is judged as of the 1977 filing date under § 103 as it then read, applying the Graham v. John Deere factors. The KSR rationales (known technique/known elements, predictable variation, design incentive) are routinely applied to pre‑AIA claims and I use them below.
• Every reference of record predates the 1977 filing by more than one year, so each qualifies as prior art under pre‑AIA § 102(b) and is fully available for § 103 combination:
  • US 3,715,512 — granted 1973‑02‑06
  • US 3,973,081 — granted 1976‑08‑03
  • Cummiskey, Jayant & Flanagan, BSTJ 52(7):1105–1118 — Sept. 1973
  • GB 1,318,985 — published 1973‑05‑31
• Caution on the database lists: the "Cited By," "Families Citing," and "Similar Documents" entries on the page (e.g., US 4,472,832, US 4,547,600‑era items, the 2002 hearing‑aid applications) post‑date the 1977 filing and are not prior art. They are useful only as evidence of field and of how the disclosure was later read.

B. What the prior art of record actually teaches

C. Claim 1 — the primary combination

Proposed combination: US 3,973,081 (primary) in view of US 3,715,512 and Cummiskey et al.

Every element of claim 1 maps to US 3,973,081:

Assessment: high obviousness risk. On this record claim 1 is, at minimum, an obvious implementation of the US 3,973,081 predictive‑feedback loop. The only real distinctions the patent could press are (i) the order of operations — the '729 converts the difference to digital, whereas US 3,973,081 digitizes first and subtracts digitally; and (ii) the analog character of the '729 front end. Point (i) is the classic "mere rearrangement of known parts" situation. Point (ii) is not even required by claim 1 (the analog limitation lives in claim 10), so it cannot save claim 1.

D. Claim 15 — the receiver (independent)

Proposed combination: US 3,973,081 (primary) in view of US 3,715,512.

The Fig. 3 receiver of US 3,973,081 — converter 26 → adder 27 → predictive filter 28, with "a feedback loop at the receiver" — is the same architecture claim 15 recites: reconversion means (26), predictor means (28), reconstitution means (27), with the predictor receiving both the received output signal and the reproduced input signal. The "one bit of resolution" limitation is met by the two‑valued limiter.

Assessment: high obviousness risk, and arguably close to § 102 anticipation of the receiver claim when claim 15 is given its broadest reasonable construction. The only wedge is that US 3,973,081's receiver is described as converting C into Q and adding a digital prediction, while claim 15 is silent as to signal domain — so the wedge is thin.

Note the self‑inflicted exposure: the '729 specification itself states the receiver "would be substantially identical to the circuit disclosed in FIGS. 4A and B" and that one need only "remove jumper connector 114 … and apply the residual or error bit stream to terminal 118." A claim to a decoder that is the encoder's own reconstruction loop is difficult to distinguish from the reconstruction loop already present in US 3,973,081's transmitter and its receiver.

E. Dependent claims

F. Motivation to combine (the § 103 linchpin)

A combination of US 3,973,081 with US 3,715,512 and Cummiskey et al. is supported by several independent rationales:

1. Same field / analogous art. All three are speech‑bandwidth‑compression via predictive (residual/DPCM) coding. No field‑of‑use barrier.
2. Explicit problem statement in the primary reference. US 3,973,081's own background identifies "the complexity of the required hardware" in adaptive predictive coding as the problem and states an object of "a digital speech compression system … characterized by greater simplicity." The '729's stated goal — compression using "relatively low cost analog circuitry" — is the same design objective. That is the classic KSR "design incentive" and "known technique to improve similar devices."
3. Reconstruction‑in‑the‑loop is already taught. Both US 3,715,512 (predict on reconstructed samples; prevent error accumulation) and the receiver loop of US 3,973,081 supply the motivation for claim 1's and claim 15's "predictor receives the error signal and the input‑signal reproduction."
4. Amplitude tracking is already taught. Cummiskey et al. supplies the motivation for claims 11–14's amplitude‑sensing scaling.
5. Predictable variation. KSR permits combining a known technique (analog sample‑and‑hold delay line / analog multiplier) with a known device (the predictive feedback loop) where the result is no more than the predictable use of a known technique for its known purpose (cost/complexity reduction), provided the analog building blocks are shown in some reference — which the face record does not do.

G. Where the obviousness case is weakest (the patentee's best rebuttal)

1. All three face references are digital/algorithmic. None discloses an analog delay line of sample‑and‑hold modules, analog gate pairs performing four‑quadrant analog‑digital multiplication, or duty‑cycle generation by comparing an integrator output against a sawtooth. Claims 5, 9, 10 (and the duty‑cycle/analog‑gate limitations they carry) rest on architecture the cited art does not show. A challenger must find that art elsewhere; with only the face references, the case against these claims is incomplete.
2. The "no side information" distinction. US 3,715,512 transmits pitch, voicing, rms and predictor parameter signals in addition to the residual (see the claim set quoted in the search results: a "combined signal" of spectral‑envelope plus pitch and voicing signals). The '729's stated advance is that "only a single digital error output signal need be transmitted," each bit carrying "all of the parametric information." To that extent US 3,715,512 alone does not render claim 1 obvious; US 3,973,081 (pure residue transmission) must carry the primary burden — which it largely can.
3. The examiner had US 3,715,512, US 3,973,081 and the Cummiskey paper before him and still allowed the claims. That is evidence — of limited weight — that the combination as claimed was not considered facially clear. It is a real consideration but, in my view, a weak one given the narrow disclosure and the analog‑specific claim construction.
4. No evidence of secondary considerations either way. I found no commercial‑success or long‑felt‑need evidence, and none appears in the record; the specification's assertion of high vocoder processing cost is a motivation, not objective evidence.

H. Art not of record that would materially strengthen a challenge

These are not in the "Prior Art" section I was asked to use, but a rigorous § 103 challenge should obtain them; I flag them as leads, not as findings:

• D.L. Cohn & J.L. Melsa, "The residual encoder — an improved ADPCM system for speech digitization," IEEE Trans. Commun., COM‑23:935–941, Sept. 1975 — pre‑dates the filing and is directed at exactly the "residual encoder" concept the '729 specification itself invokes. Highly relevant to claims 1 and 15.
• US 3,550,004 (Candy, Bell Labs) — direct‑feedback one‑bit coders, weighted step control placed outside the loop; relevant background on one‑bit residual coding and loop stability (relevant to claims 8 and 15). Not on the '729 face.
• Contemporaneous analog sample‑and‑hold / bucket‑brigade delay‑line and analog‑multiplier art — required to complete the motivation on claims 5, 9 and 10. Absent such art, those claims are the most defensible.
• The parent/divisional US 4,140,876 shares the identical disclosure and priority; its file history, if obtainable, may reveal the examiner's reasoning applied to the same subject matter and is worth pulling.

I. Bottom line

The single most effective § 103 combination on this record is US 3,973,081 in view of US 3,715,512, with Cummiskey et al. added for claims 11–14 and to reinforce the amplitude‑adaptation rationale. The motivation is supplied by US 3,973,081's own express statement that hardware complexity is the problem to be solved and that simplicity is the object — the same objective the '729 pursues. The broadest claims (1 and 15) are the most exposed; the analog‑implementation claims (5, 9, 10 in particular) are where the patent retains the greatest (though not decisive) nonobviousness footing, because the cited references are uniformly digital and do not show the analog gate‑pair multiplier or analog sample‑and‑hold delay line.

Reliability/uncertainty. Bibliographic data, claim text, and the identity and dates of the prior‑art references are grounded in the authoritative granted text and the cited documents; the claim‑element mapping to US 3,973,081 is grounded in that patent's published description and claims (verified via the granted PDF and FreePatentsOnline/Google Patents pages). My principal uncertainty is the missing prosecution history — without the examiner's reasons for allowance I cannot state why these references were overcome, and the ratings above are probability estimates rather than conclusions.