US 4933675

The first search confirms the patent's bibliographic data from multiple independent sources (FreePatentsOnline, Google Patents, Espacenet, Justia). The CAFC docket search returned no results. Let me run a few more targeted searches to confirm whether any CAFC 2026 proceeding exists for this patent number.

US Patent 4,933,675 — Summary

Search results note

• USPTO/patent databases: Bibliographic data confirmed across Google Patents, FreePatentsOnline, Justia, and Espacenet (all consistently identify this patent as US4933675A / US 4,933,675 — "Audio digital/analog encoding and decoding").
• CAFC 2026 dockets: I searched for CAFC/Federal Circuit 2026 docket entries for "4933675" and found no results. No pending or recent Federal Circuit appeal, PTAB, or litigation docket tied to this patent number surfaced in any search. Given the patent's expired status (see below), this is unsurprising, but I cannot rule out an unindexed proceeding — treat the absence of CAFC activity as not found, not confirmed absent.

Bibliographic data (verified across multiple sources)

Sources: https://patents.google.com/patent/US4933675/en; https://patents.justia.com/patent/4933675; https://FreePatentsOnline.com/4933675.html

Abstract (as published)

Apparatus and an associated method are described for encoding an analog signal to a digital representation thereof and then decoding the same to reconstruct the original analog signal with reduced quantization noise and error. The analog signal is first adaptively pre-emphasized. A series of samples of the pre-emphasized signal are then obtained and encoded to create a series of digital representations which have a lower order resolution than the samples. The difference between each sample and its corresponding lower resolution digital representation is obtained and combined with the next sample. Decoding of the combined signals takes place in a complementary manner to create an approximate analog output signal, which is then de-emphasized in a manner complementary to the pre-emphasis to produce an analog output signal closely approximating the original analog signal.

Independent claims (plain-language overview)

The patent has 10 claims. Independent claims: 1, 3, 5, 6, 8, and 9. Claims 2, 4, 7, and 10 are dependent.

Claim 1 — Method (encode + decode): Pre-emphasize the analog signal to boost high frequencies; non-linearly A/D-encode it into digital representations at sample points; at each sample, compute the difference between the quantized (digital) representation and the actual analog value; add that difference into the next sample before encoding it (error carryback); decode the digital stream with a non-linear D/A converter complementary to the encoder; then de-emphasize with a complementary circuit to reconstruct an output closely approximating the original signal. (Dependent claim 2 adds adaptive pre-/de-emphasis.)

Claim 3 — Encoding apparatus: A pre-emphasis circuit; a non-linear A/D converter; a differencing means that measures the error between the analog equivalent of each digital representation and the actual sample value; and means for adding each sample's error to the next sample value before encoding. (Dependent claim 4 adds adaptive pre-emphasis.)

Claim 5 — Decoding apparatus: A non-linear D/A converter that decodes the digital representations in a manner complementary to the original encoder; and a de-emphasis circuit, complementary to the pre-emphasis used at encoding, that produces an output closely approximating the original analog signal.

Claim 6 — Hybrid (analog/digital) encoding apparatus: The full feedback loop in analog form: pre-emphasis circuit → analog adder (adds the previous sample's residual quantization error) → sample-and-hold → first non-linear circuit → linear A/D converter (output = the encoded digital signal) → linear D/A converter → second non-linear circuit (exactly complementary to the first) → analog subtractor that computes the error between the reconstructed quantized signal and the actual sample; a connection feeds that error back to the adder's second input for the next sample.

Claim 8 — Hybrid encoding apparatus (simpler variant, no error carryback): Pre-emphasis circuit → sample-and-hold → non-linear circuit → linear A/D converter producing the non-linear digital output. (This is the special case described in the specification where the residue carryback is set to zero, keeping decoder compatibility.)

Claim 9 — Hybrid decoding apparatus: A linear D/A converter receives the non-linear digital representations; a non-linear circuit removes the non-linearity to produce a linear quantized analog signal (complementary to the encoding non-linearity); a de-emphasis circuit, complementary to any pre-emphasis applied at encoding, produces the reproduction of the original analog signal.

(Dependent claims 7 and 10 add predictor circuits to the hybrid encoder/decoder structures.)

Technical essence

The invention combines three elements to reduce quantization noise: (1) non-linear (quasi-logarithmic) encoding so step size — and thus quantization error — scales with signal amplitude (leveraging psychoacoustic masking); (2) error carryback — the quantization residue from each sample is fed forward and added to the next sample, strongly suppressing low-frequency noise and giving accurate error cancellation at zero crossings; and (3) pre-emphasis/de-emphasis of high frequencies, which exploits the fact that the carryback shifts noise to higher frequencies. The specification describes fully digital implementations (e.g., 16-bit A/D → 8-bit non-linear output via lookup tables/EPROMs, with a separate error lookup table) and hybrid analog/digital implementations, plus optional predictor-based variants (0th/1st/2nd order predictors).

Uncertainty caveats

• Assignee: Google Patents lists the original assignee as "Individual" and the 2005 reassignment to Terry D. Beard Trust; I did not independently verify the USPTO Assignment Database record (search step limit reached), so the assignment details are as reported by Google Patents/Espacenet.
• CAFC: No 2026 CAFC docket entries found for this patent number; given the 2007 expiration, active litigation is highly unlikely, but my searches do not constitute a complete docket-sheet audit.
• Claim text quoted above is from the Google Patents/Justia records provided and matches the full patent text supplied.

Based on my searches, I found no litigation involving US patent 4,933,675 ("Audio digital/analog encoding and decoding," inventor Terry D. Beard).

Search results summary

I ran the following searches:

• "4933675" patent litigation — no litigation results; only a non-litigation hit showing US 4,933,675 cited as prior art in another patent (US 5,351,048).
• "US4933675" lawsuit — no results.
• "patent 4933675" case — no litigation results.
• "4933675" docket OR complaint OR "civil action" OR defendant — no results.
• "493,675" OR "4933675" patent infringement lawsuit unifiedpatents — returned only the unrelated case Applera Corp. v. MJ Research Inc. (D. Conn., 363 F. Supp. 2d 261), which involves different patents (PCR/thermocycler patents designated '188, '675, and '493 — not the audio patent 4,933,675). That result is not relevant to this patent number.
• Terry Beard patent litigation "audio digital/analog" — no results.

Conclusion

I am not aware of, and could not find, any litigation naming US patent 4,933,675 as an asserted patent. This is consistent with the patent's history: it expired in 2007 (anticipated expiration noted June 12, 2007) after a full term with maintenance fees paid through year 12, and the assignee (Terry D. Beard Trust) does not appear in any reported district-court or Federal Circuit action I could locate.

Caveats on coverage: My searches were general web searches and did not fully exhaust paywalled litigation databases (PACER, DocketNavigator, RPX, Unified Patents' full caselist portal, Westlaw/Lexis). It is possible that an older, unreported, or dismissed case exists that is not indexed in the sources I could reach. If you need absolute certainty, a docket-text search of PACER and the Unified Patents litigation portal for "4,933,675" would be the definitive check.

Proceedings overview

The USPTO Open Data Portal returns zero AIA trial proceedings (IPR / PGR / CBM) for US 4933675 as of the most recent ingest, and independent web searches surfaced no older or recently-filed petitions against this patent. Breakdown by status: 0 active, 0 claims invalidated, 0 claims sustained, 0 settled, 0 institution denied — there is simply nothing on file. The bottom-line defensive posture: this is a clean-slate patent from a PTAB standpoint — no petitioner has ever challenged it, but equally no claim has ever been battle-tested before the Board, so the "hardened" label does not apply either.

Per-proceeding details

None on file

• Type: N/A
• Filed: N/A
• Status: No proceeding exists in USPTO ODP data or public search results for US 4933675.
• Judge panel: N/A
• Petition grounds: N/A
• Institution decision: N/A
• Final Written Decision: N/A
• Settlement / termination: N/A
• Appeal: N/A
• Defensive value: There is no IPR decision, no FWD, and no CAFC affirmance to cite. A defendant cannot invoke estoppel or a PTAB claim cancellation against this patent because none has ever occurred.

Note on diligence: Web searches for US4933675 inter partes review, "4933675" IPR, and "Beard" "audio digital/analog encoding" IPR returned only unrelated patents with "493" in the number (e.g., ClearOne v. Shure IPRs on US 9,565,493; a Sony petition on US 7,944,493). Nothing referencing the Beard '375 patent was found.

Strategic summary

Claim status: all 10 claims UNTESTED (not canceled, not sustained). US 4933675 has ten claims — independent claims 1, 3, 5, 6, 8, 9 and dependent claims 2, 4, 7, 10. None has ever been subjected to an AIA trial. The patent is a 1989 divisional of application 07/027,747 (parent issued as US 4,862,168), and per its file wrapper it expired by non-payment / reached its statutory term and is listed "Expired - Lifetime" (Google Patents shows anticipated expiration 2007-06-12, with maintenance fees paid through the 12th year). That timeline matters: the patent died in June 2007, more than five years before the AIA trial regime launched in September 2012. No IPR/PGR/CBM could have been filed while the patent was alive, and no one has bothered to challenge it post-expiry — which is common, because post-expiry challenges only make economic sense when damages for past infringement are actually at stake in live litigation.

Estoppel landscape. Section 315(e)(2) estoppel is a non-issue here because there is no prior petitioner, no instituted ground, and no FWD. For a defendant facing assertion today, every § 102/§ 103/§ 112 ground remains available — nothing has been consumed by estoppel. If anything, the defendant's position is better than if an IPR had been litigated: the patent owner has never obtained a favorable validity ruling from the Board, so there is no presumption or track record of sustainability to overcome. The realistic defenses are the ordinary ones: (1) the patent is expired and, if it expired for non-payment of maintenance fees, damages may be capped or unavailable for post-expiry conduct; (2) the claims are old (priority 1987-03-19) and highly functional, so decades of intervening prior art — including later digital-audio, noise-shaping, delta-sigma, and companding systems — are available as § 102/§ 103 grounds in district court.

Pattern signals. There is no pattern to detect: no repeat petitioner, no Unified Patents or defensive aggregator involvement, no PO-side PTAB appeals, no CAFC history. The only notable chain detail is that this patent shares its family with US 4,862,168 (parent) and US 4,882,585 (a related Beard high-resolution A/D-A/D transformation patent) — the family, not this individual patent, is the entity that has drawn citations over the years (the '375 patent is cited by Pacific Microsonics' HDCD family and by the B&W Loudspeakers patents). If a non-practicing entity is asserting '375 today, the interesting question is what other family members it is asserting alongside it, since any prior art that reads on '375 likely reads on the parent '168 as well.

Recommended next steps

• There is no FWD to link and no disposition to quote — say so plainly to anyone expecting a PTAB victory lap. If your demand letter cites this patent, the correct response is not "claim 1 was invalidated in IPR" (it wasn't) but rather: the patent expired on 2007-06-12, and any damages theory must be confined to pre-expiry conduct (35 U.S.C. § 154's term and the maintenance-fee record control here — the file shows maintenance fees paid through year 12, so check the USPTO maintenance-fee history for whether the expiration was for non-payment, which could further limit remedies).
• No pending proceeding, so there are no trial-stage milestones (no institution-decision deadline, no oral hearing date, no FWD due date) to track.
• If you are actually being sued or threatened, do not assume the absence of an IPR means the patent is strong — it means no one has tried. Since the patent is expired, an IPR petition is procedurally available only as a shield against ongoing damages exposure and is rarely cost-justified; a district-court validity attack on summary judgment (using the voluminous post-1987 prior art) or a § 101/§ 112 challenge is the more efficient route. Verify the current owner chain (the assignment history shows a 2005-12-16 transfer from Terry D. Beard to the Terry D. Beard Trust) and check whether the demand comes from the Trust or a later assignee — that will tell you whether you are dealing with the original inventor's entity or a monetization vehicle.
• Confirm the ODP absence independently before trial: search PTAB's public E2E/PTACTS docket by patent number (ptacts.uspto.gov) at the time you need to rely on it, since ODP ingestion can lag. As of 2026-08-27, both ODP and web search agree: no AIA trial proceedings exist for US 4933675.

I could not directly query the USPTO Assignment Center (the tool environment returned no live assignment records), so the assignment data below is reconstructed from the USPTO-sourced legal-events record embedded in the Google Patents file for US 4,933,675. I flag every place where the correspondent field could not be verified rather than inventing it.

Inventors

• Terry D. Beard — sole named inventor. Employer at the time of filing: not determinable from the patent record. The patent issued to "Individual" (i.e., Beard himself, with no corporate assignee), and the parent application (US 07/027,747 → US 4,862,168, filed 1987-03-19) and the related companion patent US 4,882,585 (Beard, "Method and apparatus for high resolution analog-digital-analog transformations," filed 1988-03-17) are likewise single-inventor, individually owned filings. Pattern note: there is no corporate employer to lose or divest, so the "inventors departed the assignee" fire-sale tell is structurally inapplicable here.

Original assignee

• "Individual" (Terry D. Beard) per Google Patents' Original Assignee / Current Assignee fields. The patent never carried a corporate original assignee, so there was no operating company shipping a product embodying the claims. The invention is an audio codec method/apparatus (non-linear A-D/D-A encoding with error carryback and pre-emphasis).
• Current status: the patent reached anticipated expiration on 2007-06-12 after maintenance fees were paid through year 12 (FPAY events on 1993-10-25, 1998-03-20, and 2001-11-08), and the recorded rights were transferred in 2005 to the Terry D. Beard Trust (see below). No product, no litigation, no known licensing program.

Assignment timeline

The USPTO-sourced legal-events record shows exactly one recorded post-issuance assignment for US 4,933,675:

• 2005-12-12 (executed) / recorded 2005-12-16 — Reel 017125/0399
  • Conveyance: Assignment of Assignor's Interest (recorded free text: "ASSIGNMENT OF ASSIGNORS INTEREST; ASSIGNOR: BEARD, TERRY D.")
  • Assignor: Terry D. Beard
  • Assignee: Terry D. Beard Trust (California)
  • Correspondent: not available — the Google Patents legal-event free text does not include the correspondent, and I could not reach the Assignment Center directly to capture the attorney/agent of record. Flagged as unverified rather than guessed.
  • Context: personal estate-planning transfer from the individual inventor to his own eponymous trust, recorded ~15 years after issuance and ~18 months before expiration; not an acquisition, fire-sale, or transfer to an assertion vehicle.

Prior to that, the patent issued to the inventor with no corporate assignment of record (the "Original Assignee: Individual" entry). No security agreements, mergers, changes of name, releases, or corrections appear in the legal-events record.

Verification link: https://assignmentcenter.uspto.gov/ (search Patent Number 4933675; reel/frame 017125/0399 should resolve to the Beard → Terry D. Beard Trust conveyance). Because I could not load the Assignment Center directly, the correspondent and the completeness of the record are the two items I could not independently confirm.

Timeline diagram

timeline
    title Ownership of US 4933675
    1987 : Filed by Terry D Beard
    1989 : Divisional application filed
    1990 : Patent issued to inventor
    2005 : Assigned to Terry D Beard Trust
    2007 : Patent expired

NPE / troll-pattern signals

1. Shell-entity transfer — Not present. The only post-issuance transfer (reel 017125/0399, recorded 2005-12-16) runs to "Terry D. Beard Trust," an eponymous personal trust bearing the inventor's own name — consistent with estate planning, not with a licensing-only LLC. There is no evidence of a registered-agent-service address, a single-member Delaware/Texas LLC, or a licensing-only purpose.
2. Known asserter in the chain — Not present. Neither Terry D. Beard nor the Terry D. Beard Trust appears on any public NPE/asserter directory I can reach, and no litigation naming this patent was found in the prior analysis.
3. Repeat correspondent across the chain — Unclear / not established. There is only one recorded conveyance, so no recurrence within the chain can exist; and I could not retrieve the correspondent for reel 017125/0399 from available sources. No finding either way.
4. Cascading transfers — Not present. A single conveyance 15 years post-issuance is the opposite of chained LLC transfers within 24 months.
5. Pre-litigation transfer — Not present. No infringement suit ever naming US 4,933,675 was located, so the 2005 trust transfer cannot be tied to suit preparation (and it post-dates the expiration-relevant period by only ~18 months before lapse).
6. Bankruptcy fire-sale — Not present. No corporate debtor; the chain is inventor → personal trust.
7. Privateering — Not present. No operating company is in the chain to delegate assertion to an NPE.
8. Defensive aggregator (anti-NPE) — Not present. The chain terminates at the Terry D. Beard Trust, not at RPX, AST, LOT, Unified Patents, or OIN.

Verdict

Insufficient data

Every NPE signal that could be evaluated is not present: the chain is a single transfer from the individual inventor to his own personal trust (reel 017125/0399, recorded 2005-12-16), the patent expired in 2007, and no litigation or public assertion activity exists. The classification is limited to "insufficient data" only in the narrow sense that I could not load the USPTO Assignment Center directly to (a) capture the correspondent of record for reel 017125/0399 and (b) rule out additional unindexed conveyances — so I cannot certify the record as complete with full confidence. On the available evidence, there is no NPE pattern, no operating-company assertion, and no defensive-aggregator terminus; the most accurate description is an expired patent held by an individual inventor's estate-planning trust.

Verification: https://assignmentcenter.uspto.gov/ (Patent No. 4933675; confirm reel/frame 017125/0399 and pull the correspondent field).

Prior Art Analysis — US Patent 4,933,675 ("Audio digital/analog encoding and decoding")

1. Verification note on the USPTO record

I confirmed the USPTO record for US 4,933,675 via Google Patents, FreePatentsOnline, and Espacenet (all keyed to the exact number 4,933,675 — no similar-number results are included in this analysis). The patent's own front page (authoritative source supplied with this task) lists 10 U.S. patent citations and 2 non-patent citations. Live searches verified the content of several of these in detail (US3973199, US4130729, US4133976, US4270177); the remaining references are reported from the patent's citation list, with content characterized from general knowledge and flagged where unverified. I was unable to complete a USPTO PatFT full-text search of each reference before the search step limit — the citation list itself is from the granted patent and is authoritative as to which references were cited and when.

2. Applicable §102 framework (pre-AIA)

US 4,933,675 was filed June 9, 1989 (divisional of US 4,862,168, filed March 19, 1987), so pre-AIA 35 U.S.C. §102 governs. Because the divisional is entitled to the parent's filing date under §120:

• §102(b) statutory bar – references patented or published before March 19, 1986 (one year before the effective filing date).
• §102(e) – U.S. patents by others filed before the invention date (presumed no later than March 19, 1987).
• §102(a) – patents/publications before the invention date.

For anticipation, a reference must disclose every claim element, arranged as claimed.

3. Cited references — full citations, descriptions, and §102 claim analysis

(A) Most relevant references

1. US 4,700,362 — "A-D encoder and D-A decoder system" (Dolby Laboratories Licensing Corporation)

• Filed: Oct. 7, 1983; Published: Oct. 13, 1987.
• Description: A Dolby analog-to-digital encoder / digital-to-analog decoder system — a system-level A-D/D-A architecture with complementary encoding/decoding, in the same technical space as 4,933,675. The 4,933,675 specification itself invokes "Dolby" noise reduction (Type B) as the model for its adaptive pre-emphasis.
• §102 analysis: Filed Oct. 7, 1983, before the effective filing date → qualifies under §102(e) (and §102(a) if the invention post-dates 1983). As a complementary A-D/D-A system it is the strongest system-level candidate against claims 1, 3, 5, 6, 8, and 9. Caveat: I could not retrieve the full text in the available search steps; whether it discloses the specific per-sample error carryback (claim 1(d), claim 6 elements (h)-(i)) and pre-emphasis/de-emphasis (claims 1(a)/(f), 3(a), 5(b)) could not be confirmed. Treat its anticipation potential as high but unverified.

2. US 4,133,976 — "Predictive speech signal coding with reduced noise effects" (Bell Telephone Laboratories; Atal & Schroeder)

• Filed: Apr. 7, 1978; Published: Jan. 9, 1979. (Verified via live search.)
• Description: Predictive speech coder in which the input is lowpass-filtered, high-frequency pre-emphasized (6 dB/octave from 700 Hz), and sampled; a quantizer is surrounded by an adaptive feedback filter that combines the quantizing-error signal with the difference signal to concentrate quantization noise in formant regions so it is masked by the speech signal. A matching decoder reconstructs and (complementarily) filters the signal.
• §102 analysis: Published 1979 → §102(b) art (pre-1986). This is the closest verified art to claim 1's combination: pre-emphasis before encoding (1(a)); quantizing-error feedback to modify the coded signal (1(c)-(d), albeit via a filtered formant-weighted feedback rather than a literal "add previous sample's residue to the next sample"); predictive coding; complementary decode with de-emphasis (1(e)-(f)). Likely anticipates claims 1, 3, 5, and 9 if the error-feedback loop is read as the claimed carryback; the pre-emphasis is fixed, not adaptive, so claim 2/4 (adaptive) is probably not anticipated. It also bears on predictor-dependent claims 7 and 10.

3. US 3,973,199 — "Prediction differential pulse code modulation system with adaptive compounding" (U.S. Philips Corporation)

• Filed: Sep. 3, 1973; Published: Aug. 3, 1976. (Verified via live search.)
• Description: DPCM speech transmission system with a non-uniform (companding) quantizer, a predictor, a difference producer, a local receiver inside the transmitter (expander + predictor), and dynamic (adaptive) control of the quantizer's dynamic range based on averaged predictor-input magnitude.
• §102 analysis: §102(b) art (1976). Discloses non-linear/companded quantization and adaptive range control but no high-frequency pre-emphasis (it uses a 0.3–4 kHz band-pass filter) and no per-sample quantization-residue carryback to the next sample (it adapts the quantizer range instead). Best fit is against the non-linear-quantization-plus-prediction elements of claims 1, 6, 7 — but it likely fails full anticipation of any independent claim on the missing pre-emphasis/carryback elements. Relevant as an obviousness anchor for the companding + prediction combination.

4. US 4,385,393 — "Adaptive prediction differential PCM-type transmission apparatus and process with shaping of the quantization noise" (L'Etat Francais, représenté par le Secrétaire d'Etat)

• Filed: Apr. 21, 1980; Published: May 24, 1983.
• Description: Adaptive-prediction DPCM apparatus with shaping of the quantization noise spectrum (feedback of quantization error through filters to push noise where it is least perceptible), with adaptive prediction and complementary transmit/receive processing.
• §102 analysis: §102(b) art (1983). Closely parallels 4,133,976 and 4,933,675's noise-shaping/error-feedback concept. Relevant to claims 1, 3, 6, 7, 10; like 4,133,976 it likely lacks the literal next-sample residue adder and the pre-emphasis/de-emphasis combination required by claims 1/6, so full anticipation is doubtful but the noise-shaping element is squarely disclosed.

5. US 4,130,729 — "Compressed speech system" (Scitronix Corporation; Gagnon)

• Filed: Sep. 19, 1977; Published: Dec. 19, 1978. (Verified via live search.)
• Description: Residual-encoder speech compressor: an adaptive predictor (analog delay line of sample-and-hold stages with correlation multipliers) generates a predicted signal; the difference (residual) between input and prediction is A/D-converted to a one-bit digital stream; a local D/A reconstructs the residual, which is summed with the prediction to reconstitute the signal; a remote receiver mirrors the local reconstitution.
• §102 analysis: §102(b) art (1978). Discloses predictive residual encoding with local reconstruction and feedback — the architecture class of claims 6/8 (minus pre-emphasis and non-linear companding). Relevant to predictor-based claims 7 and 10 and to the hybrid encoder/decoder structures of claims 6, 8, 9; full anticipation of the independent claims is unlikely (no pre-emphasis, no companding non-linear A/D-D/A, no residue carryback to the next sample as claimed).

(B) Secondary / supporting references

6. US 4,270,177 — "Digital amplitude control for digital audio signal" (Tokyo Shibaura Denki K.K.; Endoh et al.)

• Filed: June 20, 1979; Published: May 26, 1981. (Verified via live search.)
• Description: Digital volume control: a potentiometer's output is A/D-converted and used as an address into a memory storing a non-linear (e.g., logarithmic) amplitude-control transfer function, then multiplied digitally with the audio signal.
• §102 analysis: §102(b) art (1981). Relevant only to the memory/lookup-table non-linear mapping implementation of 4,933,675 (FIGS. 12–13; claims 1, 3, 5 elements). It does not disclose pre-emphasis, error carryback, or complementary A/D-D/A encoding; no realistic standalone anticipation of any claim.

7. US 4,353,060 — "Analog to digital converter system with an output stabilizing circuit" (Tokyo Shibaura Denki K.K.)

• Filed: July 13, 1979; Published: Oct. 5, 1982.
• Description: An A/D converter system with an output-stabilizing circuit (suppressing fluctuation/instability of the digital output).
• §102 analysis: §102(b) art (1982). Tangential — bears only on the accuracy/stability of the A/D converter used inside the 4,933,675 system (the spec stresses matched, high-accuracy converters). No anticipation of the claimed combination. Content not independently verified in available searches.

8. US 4,430,670 — "Reconstruction of quantized DPCM or PCM signals" (Bell Telephone Laboratories)

• Filed: Mar. 12, 1982; Published: Feb. 7, 1984.
• Description: Decoder-side techniques for improving the reconstruction of quantized DPCM/PCM signals from their quantized representations.
• §102 analysis: §102(b) art (1984). Relevant to the decoding-side claims 5, 9, 10 (non-linear complementary decoding, reconstruction of an analog output). Likely lacks the claimed complementary non-linear D/A + de-emphasis arrangement, so full anticipation is doubtful. Content not independently verified in available searches.

(C) Non-prior-art citations (same inventor / same disclosure — not §102 art)

9. US 4,862,168 — "Audio digital/analog encoding and decoding" (Terry D. Beard) — Filed Mar. 19, 1987; Published Aug. 29, 1989.

• This is the parent application of which 4,933,675 is a divisional. Same disclosure, same inventive entity, same effective filing date. It cannot be §102 prior art against its own divisional and cannot anticipate any claim.

10. US 4,882,585 — "Method and apparatus for high resolution analog-digital-analog transformations" (Terry D. Beard) — Filed Mar. 17, 1988; Published Nov. 21, 1989.

• Same inventor; filed after the parent's March 19, 1987 filing date. Because the divisional's invention date is no later than March 19, 1987, this later-filed, same-inventor patent is not §102(a)/(e) prior art. Cited as related art only.

(D) Non-patent citations

11. Precision Monolithics, Inc., "Linear and Conversion Products 1986/1987 Data Book" (COMDAC devices)

• Printed publication; 1986/1987. If publicly distributed before March 19, 1986 → §102(b) art.
• Description: Documents Precision Monolithics' COMDAC non-linear A/D and D/A converters — the very prior-art devices the 4,933,675 specification discusses ("One solution to this problem has been the use of non-linear digital encoders/decoders such as those commercially sold by Precision Monolithics, Inc. under the trademark COMDAC").
• §102 analysis: Strongest art on the non-linear (companding) A/D-D/A conversion element of claims 1, 3, 5, 6, 8, 9 — but COMDAC lacks pre-emphasis and error carryback, so it does not anticipate the full claimed combination.

12. The Engineering Staff of Analog Devices, Inc., "Analog-Digital Conversion Notes," 1980, p. 64

• Printed publication; §102(b) art (1980). General A/D-D/A conversion principles (likely quantization-error and conversion techniques). Weak standalone anticipation value; useful as background/obviousness art for the converter elements.

4. Ranking — most relevant prior art

5. Bottom line

The single most relevant prior-art family is the combination of pre-emphasis + predictive/DPCM coding + quantization-noise feedback already present in the 1970s–80s speech-coding art (US 4,133,976 being the closest verified disclosure, since it explicitly pre-emphasizes high frequencies and feeds the quantizing error back to shape noise), together with US 4,700,362 as the closest system-level A-D/D-A reference (unverified content) and US 3,973,199/US 4,385,393 for adaptive companding DPCM with noise shaping. No single cited reference appears to disclose the full claim 1/6 combination — in particular the literal per-sample operation of adding the previous sample's quantization residue to the next sample before encoding and the complementary non-linear decoding + de-emphasis structure — which is consistent with the patent issuing with all claims allowed over this art.

6. Caveats

• Per the operating rules, the citation list and dates above come from the granted patent's own front page (authoritative). Live searches independently confirmed the substance of US 3,973,199, 4,130,729, 4,133,976, and 4,270,177. The full text of US 4,700,362, 4,353,060, 4,385,393, 4,430,670, and 4,882,585 was not retrievable within the available search steps — their descriptions and anticipation assessments are based on the citation list and general knowledge and should be verified against the USPTO full-text images (PatFT) or Google Patents before being relied on in an invalidity position.
• All §102 analysis assumes pre-AIA law (application filed 1989). A definitive anticipation conclusion for any claim would require claim charts against the full text of each reference.

Obviousness Analysis of US Patent 4,933,675 Under 35 U.S.C. § 103

Patent under analysis: US 4,933,675 ("Audio digital/analog encoding and decoding," Beard; divisional of US 4,862,168; priority March 19, 1987; filed June 9, 1989; issued June 12, 1990; expired 2007).

Analytical frame: Obviousness is assessed as of the priority date (March 19, 1987). All ten claims are at issue — independent claims 1, 3, 5, 6, 8, 9 and dependent claims 2, 4, 7, 10. The analysis is grounded in the Prior Art section of the patent's Google Patents page (10 U.S. patents + 2 non-patent literature items), supplemented by the verified content of those references.

I. What the claims actually require (element extraction)

The claims decompose into three technical pillars, and the obviousness analysis is really an analysis of whether each pillar — and their combination — was within the grasp of a PHOSITA in 1987:

Critically, the patent itself concedes the state of the art for two of the three pillars: the specification states that non-linear encoders/decoders "such as those commercially sold by Precision Monolithics, Inc. under the trademark COMDAC" were known, and that adaptive pre-emphasis filters "are well known in the art; the 'Dolby' noise reduction type B is a typical example." That leaves P3 (error carryback) as the only element a challenger must source to the prior art.

II. The prior art landscape (as cited on the face of the patent)

A. Non-linear/companding conversion art (P1)

1. Precision Monolithics, Inc., Linear and Conversion Products 1986/1987 Data Book (NPL) — The COMDAC family (e.g., DAC-76, first shipped ~1975–1977) is a monolithic companding D/A converter with a quasi-logarithmic transfer: "the step size increases as the output changes from zero scale to full scale," unlike linear converters. The PMI data book explicitly documents compressing A/D conversion using the DAC-76 plus a comparator, exclusive-OR gate, and successive-approximation register — i.e., a complete non-linear A/D encoder and complementary D/A decoder pair, exactly the P1 structure of claims 5, 8, and 9. The data book also documents μ-255 (Bell-law) companding and audio applications. This is the reference the '375 patent itself identifies as the starting point. (Sources: bitsavers.org scan of the 1977 PMI Linear and Conversion IC Products data book, DAC-76 "Basic Encode Operation (Compressing A/D Conversion)"; chipsetc.com PMI DAC-76 history; seekic.com reprint of PMI's companding A/D application note.)

2. US 3,973,199 (Philips, "Prediction differential pulse code modulation system with adaptive compounding," filed 1973, issued 1976) — A DPCM transmission system whose quantizer is non-uniform/companding ("piecewise linear compression characteristic according to the CEPT standard"), with a local receiver in the transmitter (expander circuit 10 + adder 9 + predictor 8) that reconstructs the quantized signal, a difference producer 11 that forms the error between the input sample and the prediction, and dynamic range control of the quantizer. It expressly teaches that a non-linear quantizer output "must undergo an expansion, which is reciprocal to this compression" before recombination — i.e., complementary non-linear encode/decode. (Source: FreePatentsOnline.com/3973199; Google Patents full-text.)

3. US 4,700,362 (Dolby, "A-D encoder and D-A decoder system," Todd & Gundry, priority Oct. 7, 1983, issued 1987) — An A/D-D/A system in which a control signal derived from the input sets the optimum step size of the converter (syllabically adaptive companding), with time-delay compensation, and a decoder that expands using transmitted step-size information. This is a Dolby-family companding codec reference and is highly relevant to adaptive aspects (claims 2, 4). (Sources: FreePatentsOnline.com/4700362; Espacenet biblio.)

4. US 4,130,729 (Scitronix, "Compressed speech system," 1977) and US 4,270,177 / US 4,353,060 (Toshiba, digital audio amplitude control and A/D stabilization, 1979) — general companding/compressed-speech and digital-audio converter art corroborating that non-linear amplitude mapping and high-performance audio A/D were routine by the late 1970s.

B. Quantization-noise feedback/shaping art (P3)

1. US 4,385,393 (French State/CNET, "Adaptive prediction differential PCM-type transmission apparatus and process with shaping of the quantization noise," filed 1980, issued 1983) — This is the single most damaging reference for P3. It explicitly defines the quantization noise as "the difference between the restored signal ŷ(t) and the incident signal y(t)" and teaches shaping the quantization-noise spectrum to parallel the signal spectrum: "in the ranges where the signal y(t) is weak, the quantization noise is also weak and in the ranges where y(t) is strong, the noise assumes relatively high values without this being prejudicial to the transmission quality." It achieves this by feeding the restored error signal back into the prediction loop (AMAR modeling using both restored signal samples and restored error samples, with coefficients readjusted at each sample time). That is precisely the "noise proportional to signal, small near zero crossings" property that the '375 patent touts as the benefit of combining non-linear encoding with error carryback. (Sources: Google Patents full-text US4385393; uspto.report.)

2. US 4,131,976 (Bell Labs, "Predictive speech signal coding with reduced noise effects," 1978) and US 4,430,670 (Bell Labs, "Reconstruction of quantized DPCM or PCM signals," 1982) — Bell Labs work on predictive coding with explicitly reduced quantization-noise effects and on improved reconstruction of quantized DPCM/PCM signals at the decoder. These corroborate both the noise-reduction motivation and the decoder-side reconstruction techniques.

3. Analog Devices, Inc., Analog-Digital Conversion Notes (1980), p. 64 (NPL) — The standard converter-design text of the era; the cited page concerns quantization-error behavior, the textbook basis for residue feedback / noise shaping in converter loops (and the general knowledge that feeding quantization error back into the loop spectrally shapes it).

C. Pre-emphasis/de-emphasis art (P2)

1. The '375 patent's own admission — "Such adaptive pre-emphasis filters are well known in the art; the 'Dolby' noise reduction type B is a typical example." Pre-emphasis/de-emphasis was ubiquitous in audio by 1987 (Dolby B/C, FM transmission, tape noise reduction). US 4,700,362 (Dolby) also supplies a Dolby-family adaptive signal-processing pedigree. This pillar needs no sourcing beyond the patent's own text.

D. Family references — not usable as § 103 prior art, but noteworthy

1. US 4,862,168 (parent, Beard, filed Mar. 19, 1987) and US 4,882,585 (Beard, filed Mar. 17, 1988) — The parent shares the '375 priority date; the sibling was filed before '375's actual filing date but after the '375 priority date and is commonly owned with overlapping disclosure. Neither is clean § 103 prior art for claims entitled to the 1987 priority date. They are relevant to a double-patenting analysis, not obviousness. A challenger should not waste a ground on them.

III. Level of ordinary skill in the art

A PHOSITA circa March 1987: an electrical engineer (B.S./M.S.) with 2–5 years of experience in audio signal processing and A/D–D/A converter design, familiar with (i) companding codecs (COMDAC, μ-law/A-law), (ii) DPCM/ADPCM and predictive coders, (iii) delta-sigma / error-feedback noise-shaping concepts, and (iv) audio noise-reduction and pre-emphasis/de-emphasis practice (Dolby, FM). All four bodies of knowledge are represented in the cited prior art, so the "ordinary skill" bar is comfortably met by the references themselves.

IV. Element-by-element mapping of the primary combination

Primary combination for claims 1, 3, 6: PMI COMDAC data book (non-linear codec) + US 4,385,393 (quantization-noise feedback/shaping) + known pre-emphasis/de-emphasis (Dolby practice; US 4,700,362) + US 3,973,199 (DPCM with companding quantizer and local reconstruction).

Claim 3 (apparatus) is the same combination in hardware terms: pre-emphasis circuit (known), non-linear A/D (COMDAC), differencing means (US 4,385,393's restored-vs-incident comparison; US 3,973,199's difference producer), and residue-add-to-next-sample (US 4,385,393's error feedback).

Claim 6 (hybrid analog encoder with full feedback loop) maps onto US 3,973,199's FIG. 1 transmitter almost element-for-element once the error-feedback path of US 4,385,393 is added: pre-emphasis → analog adder (adds previous residue) → dual sample-and-hold → first non-linear circuit → linear A/D → linear D/A → complementary non-linear circuit → analog subtractor → feedback to adder. The dual sample-and-hold (one-word delay) is the textbook way to avoid a combinational feedback loop — a routine design choice, not an inventive step.

Claims 5, 8, 9 (decoder and no-carryback encoder): These need only P1 + P2. Claim 8 is literally the PMI compressing-A/D block diagram (non-linear circuit + linear A/D) preceded by a pre-emphasis filter; claim 9 is the PMI COMDAC decode path (linear D/A + non-linear expander) followed by a de-emphasis filter; claim 5 is the same decode pair. The '375 specification itself describes claim 8's configuration as the "special case ... in which the error residue carried back is set at zero" — i.e., the patent concedes this claim is the prior-art COMDAC structure with a filter added.

Dependent claims 2 and 4 (adaptive pre/de-emphasis): US 4,700,362 (Dolby) teaches control signals derived from the input signal to adapt conversion characteristics, and adaptive pre-emphasis was admitted known (Dolby B). Obvious.

Dependent claims 7 and 10 (predictors): US 3,973,199 and US 4,385,393 both teach predictors, including the 0th-order "previous sample" predictor (US 3,973,199: "predictors 8 and 8' are formed as storage elements ... holding a signal applied thereto during one sampling period") and higher-order linear predictive combinations. The '375 specification's predictor formulas (T₀ = T₋₁; T₀ = 2T₋₁ − T₋₂; T₀ = 3T₋₁ − 3T₋₂ + T₋₃) are the standard Newton/linear-prediction equations of the DPCM art. Obvious over the base combination.

V. Motivation to combine — Graham/KSR analysis

1. Scope and content of the prior art

Every structural element of the claims existed in the cited art: companding codecs (PMI COMDAC, CEPT/μ-law), predictive DPCM with local reconstruction (US 3,973,199), quantization-noise shaping by error feedback (US 4,385,393), adaptive step-size/companding codecs (US 4,700,362), and audio pre-emphasis/de-emphasis (Dolby practice, admitted in the specification).

2. Differences between the prior art and the claims

The only arguable difference is the specific error-carryback topology of P3 (residue of sample n added to sample n+1 before quantization, in a non-linear coder, with pre-emphasis). US 4,385,393 feeds the restored error into an adaptive predictor rather than adding the raw residue to the next input sample, and US 3,973,199 feeds back a prediction rather than a quantization residue. A challenger must show this gap is a routine design choice — and the art supports that.

3. KSR motivation — this is a textbook "known elements, predictable results" case

• Same field, same problem, same solution space. All references address the identical problem: reducing audible quantization noise in coded speech/audio without increasing bit rate. US 4,385,393 states the goal in words the '375 patent later repeats almost verbatim: noise should be small where the signal is small and large (masked) where the signal is large.
• Express suggestion to combine. US 3,973,199 already merges companding with DPCM prediction — it is the combination of P1 and P4. A PHOSITA seeking still better low-level noise performance would naturally add the noise-shaping error feedback of US 4,385,393 to that companded DPCM loop; the '375 patent's contribution is essentially swapping US 4,385,393's adaptive predictor for a one-sample delay line.
• Design incentive / marketplace pressure. In the mid-1980s, consumer digital audio (CD, DAT, digital recording) drove demand for wider dynamic range at low bit rates; companding codecs (COMDAC, μ-law) were the established way to get more dynamic range from fewer bits, and noise shaping was the established way to push residual quantization noise into masked frequency bands. Combining them, plus the audio-ubiquitous pre-emphasis, was the obvious engineering path. The '375 specification's own rationale ("low frequency noise is substantially suppressed and shifted to higher frequencies," so "it is appropriate to pre-emphasize the higher frequencies") is exactly the rationale a PHOSITA would derive from US 4,385,393's noise-shaping teaching plus standard Dolby/FM pre-emphasis practice.
• Predictable results. Each element performs its known function: the compander scales step size with amplitude; the feedback loop shapes the noise spectrum; the pre/de-emphasis pair trades noise between frequency bands. There is no synergism beyond the sum of these known effects; under KSR, the combination is the "combination of familiar elements according to known methods" that § 103 forbids patenting.
• Obvious-to-try. The design space (non-linear quantizer + feedback of quantization error + spectral weighting) was small and well mapped; the references point to the same optimum.

4. Secondary considerations

None of the usual objective indicia are present in the record: no long-felt need (companding codecs and noise shapers already existed), no commercial success evidence, no copying evidence, no unexpected results (the claimed benefit — masking of noise by signal — is the stated goal of US 4,385,393). The patent expired in 2007 after routine maintenance-fee payment, which is neutral but provides no countervailing evidence of value.

VI. Best counterarguments (the honest caveats)

1. The error-carryback topology itself. No single cited reference shows the exact "add previous residue to next input sample" topology in a companding codec. US 4,385,393 shapes noise through an adaptive predictor; a strict examiner might say the motivation to replace the predictor with a fixed one-sample residue adder is not expressly disclosed. The rebuttal: error-feedback coders were textbook (Analog Devices notes; delta-sigma art), and US 4,385,393's "parallel spectra" teaching gives the express reason to do it.
2. The '375 insight about zero crossings. The patent claims a specific benefit: because the companding step size is smallest near zero, carryback gives "highly accurate cancellation of accumulated errors at zero crossings." This is a why it works explanation. Under KSR, however, an explanation of why a known combination works does not make it non-obvious; and the benefit is squarely the US 4,385,393 teaching (noise weak where signal weak).
3. Claim 6's detailed analog loop. If challenged element-by-element without the benefit of US 3,973,199's local-receiver figure, claim 6's full feedback loop looks intricate. But US 3,973,199's FIG. 1 transmitter contains nearly all of it; adding the residue-feedback path and pre-emphasis yields the rest. The dual sample-and-hold is a routine one-word-delay implementation.
4. Family double-patenting. Because the parent '168 and sibling '585 share the disclosure, a challenger should also consider whether '375's claims are patentably distinct from the family under obviousness-type double patenting — a validity avenue independent of § 103 prior art.

VII. Conclusion

US 4,933,675 is vulnerable under § 103 on the cited prior art, with the strongest case built as follows:

• Claims 5, 8, 9 (and 2, 4): obvious over the PMI COMDAC data book (non-linear A/D-D/A codec, compressing A/D and expanding D/A explicitly documented) in view of known audio pre-emphasis/de-emphasis (Dolby practice; US 4,700,362 for the adaptive variants). The specification itself describes claim 8 as the carryback-disabled version of the known COMDAC-type structure.
• Claims 1, 3, 6 (the carryback claims): obvious over US 3,973,199 (companded DPCM with local quantized-signal reconstruction — P1 + P4) in view of US 4,385,393 (quantization-noise shaping by feeding restored error into the coding loop — P3) and known pre-emphasis/de-emphasis (P2). US 4,385,393 supplies the express motivation: make quantization noise track signal amplitude so it is masked. The '375 patent's error carryback is a one-sample-delay simplification of that teaching, and the "small steps near zero → accurate zero-crossing cancellation" benefit is the stated objective of the French reference.
• Claims 7, 10: obvious over the same combination, since both US 3,973,199 and US 4,385,393 teach predictors (0th- and higher-order) and the claimed predictor equations are standard DPCM linear prediction.

A PHOSITA in 1987, armed with the COMDAC data book, the Philips companded-DPCM patent, and the CNET noise-shaping patent — all cited on the face of '375 — would have had every element and every motivation to arrive at the claimed invention, with a reasonable expectation of success. The only genuinely contestable step is the precise error-carryback feedback topology, and the prior art's explicit noise-shaping teachings substantially undercut even that argument.

Key sources: US3973199 (FreePatentsOnline.com/3973199.html), US4385393 (patents.google.com/patent/US4385393), US4700362 (FreePatentsOnline.com/4700362.html; Espacenet), PMI DAC-76/COMDAC (bitsavers.org 1977 PMI data book; chipsetc.com; seekic.com PMI companding A/D application note), and the full text of US4933675 as provided.