US 5273632

I'll search for this specific patent number across USPTO and litigation sources.

Let me look for the full claim set (the provided text truncated at claim 13) and confirm litigation status.

Summary — US Patent 5,273,632

Important note on identifiers (interpreted literally): The number you gave, 5273632, resolves to U.S. Patent 5,273,632. Several other unrelated identifiers share the digit string "5273632" and must not be conflated — notably JP 5273632 B2 (a Japanese patent, registered 2013-05-24, for a sewing machine with an automatic thread-cutting device) and a different, much more recent U.S. "'632" patent appearing in 2026 Netlist/Micron litigation papers (a different patent entirely). Nothing in my searches tied U.S. 5,273,632 to the 2026 Netlist/Micron disputes.

Bibliographic data (from the patent front page / Google Patents and corroborating sources)

Sources: https://patents.google.com/patent/US5273632/en ; https://www.osti.gov/biblio/[869094](/patent/869094) ; https://patents.justia.com/patent/5273632 ; https://uspto.report/patent/grant/5273632

Abstract (verbatim)

"A method and apparatus for sharpening signal peaks in a signal representing the distribution of biological or chemical components of a mixture separated by a chromatographic technique such as, but not limited to, electrophoresis. A key step in the method is the use of a blind deconvolution technique, presently embodied as homomorphic filtering, to reduce the contribution of a blurring function to the signal encoding the peaks of the distribution. The invention further includes steps and apparatus directed to determination of a nucleotide sequence from a set of four such signals representing DNA sequence data derived by electrophoretic means."

Independent claims — plain language

The claim set has 33 claims. Claims 1, 8 and 18 are confirmed independent (claims 2–7 depend from 1; 9–17 from 8; 19–28 from 18 or chains thereof). The text of claims 29–33 was not retrievable from my sources, so I cannot confirm whether they include a separate apparatus claim despite the "and apparatus" in the title — flagging that as an open uncertainty rather than guessing.

Claim 1 — core cepstral peak-sharpening method. Take a signal representing a chromatographic distribution of components from a biochemical mixture; transform the signal from its original space domain into a cepstrum; multiply/manipulate the cepstrum with a lifter function chosen to substantially attenuate the portion of the cepstrum attributable to a blurring function, producing a liftered cepstrum; then transform back to the original space domain to yield a deconvolved "lane" signal. This is the broadest claim: it recites the cepstrum-transform/lifter/de-transform idea generically, without the full homomorphic recipe.

Claim 8 — full homomorphic-filtering recipe. A narrower, step-by-step method: (a) Fourier-transform the signal to an FS (frequency spectrum) signal; (b) take the log of the FS signal to get a CLS (complex log-spectrum) signal; (c) inverse-Fourier-transform the real portion of the CLS to get a cepstrum; (d) multiply the cepstrum by a lifter chosen to reduce the blurring-function contribution, giving a liftered cepstrum; (e) Fourier-transform that to an LLS (liftered log-spectrum) signal; (f) add back the imaginary portion of the log-spectrum signal to make a liftered CLS; (g) take the inverse log (antilog) to get a liftered frequency-spectrum signal; and (h) inverse-Fourier-transform it to produce the deconvolved signal. (Note the literal wording quirk: step (f) recites "the imaginary portion of the LS signal," and (g) parenthetically calls the result a "liftered FS (liftered frequency spectrum) signal.")

Claim 18 — DNA base-calling pipeline. A method of determining the nucleotide sequence of a DNA molecule from an electrophoretic migration pattern of four sequencing lanes (A, T, G, C): (a) provide a set of lane signals with peaks representing relative amounts of fragment sizes; (b) identify putative peaks in all lanes and compute a putative spacing = the average across all putative peaks; (c) align the lane signals to establish a provisional peak ordering; (d) build a three-dimensional matrix whose elements are coordinate pairs of (spacing value; signal location) for each occurrence of a peak pair, a "peak pair" being two non-identical peaks adjacent to each other in the provisional ordering; (e) derive a peak-pair spacing function for each peak-pair category; (f) select a reference lane; (g) insert/average samples within each lane as needed so that its spacing function substantially matches the reference lane's, yielding fully aligned lanes; and (h) read the DNA sequence from the peak order.

Notable dependent claims (grounding the "how")

• Claim 2 adds a noise-filtering step; claims 3–4 put that filtering in conjunction with the de-transform and specify a Gaussian low-pass filter with bandwidth "between about 0.024A and about 0.072A, where A = the number of samples comprising π" (literal).
• Claims 5–7 define the lifter shape: a first portion attenuating the cepstrum in a low-quefrency region, a second portion ≈1 in a high-quefrency region, a 50% cosine taper reaching ordinate 0.50 at a point near where the cepstrum plateaus.
• Claims 9–11 place noise removal after generation of the FS signal and specify a Gaussian low-pass filter "bandwidth equivalent to between about 50 and 150 frequency samples when π is 1024 samples."
• Claims 12–13 require the lifter to be high-pass and to normalize cepstrum amplitude to a chosen range (the patent stresses this is what lets all four lanes of a gel be normalized to each other automatically).
• Claims 19–20 (depending from 18) fit the spacing function to a straight line and establish the putative-peak threshold function; claim 25 adds blind deconvolution before peak identification; claims 26–28 recite the homomorphic deconvolution/noise-filter sub-steps within the sequencing method.

Litigation / CAFC check (as requested)

I searched for U.S. 5,273,632 in litigation and Federal Circuit contexts and found no evidence of any CAFC 2026 docket, appeal, IPR, or district-court assertion involving this patent. My search returned only (i) 2026 Netlist/Micron/Samsung dockets referencing a different "'632" patent, and (ii) the unrelated JP 5273632 B2 sewing-machine patent. I cannot rule out that some litigation exists in records not surfaced by these searches — treat "no litigation found" as a search result, not an affirmative certification.

Uncertainty flags

• Claims 29–33: text not obtained; I cannot confirm whether an apparatus claim exists (the specification does describe apparatus — scanner 400, microprocessor controller 402 with complex-log/antilog lookup tables, FFT/inverse-FFT circuits 410/412 such as TI TMS 320C10/25/15/30 DSP chips, and a workstation embodiment 440).
• Claim 4 vs. claim 11 numeric ranges: the two bandwidth expressions differ under a straight reading (0.024A–0.072A with A = samples per π, versus 50–150 samples at π = 1024). Reported literally; I have not reconciled them.
• Corroborating technical description of the same reader appears in Ives, Gesteland & Stockham, "An automated film reader for DNA sequencing based on homomorphic deconvolution," IEEE Trans. Biomed. Eng., vol. 41, no. 6 (June 1994) — a post-grant publication, not prior art to this patent.

I'll research litigation involving US patent 5273632. Let me run several targeted searches.

Let me run more targeted searches on litigation databases.

Let me try a few more focused searches.

Litigation Search Report — U.S. Patent No. 5,273,632

Prepared for: Counsel / client
Date: April 26, 2026

Bottom line

I found no litigation involving U.S. Patent No. 5,273,632. No district court action, Federal Circuit appeal, ITC investigation, or PTAB proceeding identifying the '632 patent as a patent-in-suit was returned in my searches. I am reporting this as "none found," not as an affirmative certification that none exists — see the caveats below.

The patent at issue (as confirmed in the source records)

Sources: Google Patents, https://patents.google.com/patent/[US5273632](/patent/US5273632)/en; OSTI/DOE patent record 869094, https://www.osti.gov/doepatents/biblio/869094.

What was searched and what came back

1. General web searches for "5273632" litigation / "5,273,632" lawsuit / assignee-based searches. No docket, complaint, judgment, or news item naming US 5,273,632 as an asserted patent.
2. University of Utah Research Foundation enforcement activity. The searches returned substantial UURF litigation — but exclusively on other patents. Specifically:
   • In re BRCA1- and BRCA2-Based Hereditary Cancer Test Patent Litigation, MDL No. 2:14-md-02510-RJS (D. Utah), including Univ. of Utah Research Found. v. Ambry Genetics Corp., 2:13-cv-00640-RJS (D. Utah), and the Federal Circuit appeal at 774 F.3d 755 (Fed. Cir. 2014). Patents asserted there were U.S. 5,709,999; 5,747,282; 5,753,441; 5,837,492; 6,033,857; 5,654,155; 5,750,400; 6,083,698; 5,693,473; 5,710,001; 6,492,109; 6,051,379 — none of which is 5,273,632.
   • University of Utah Research Foundation v. Corbett / Idaho Technology — patents 6,787,338; 7,238,321; 7,081,226; 6,174,670; 6,245,514; 6,569,627 — again, not 5,273,632.
3. Google Patents page for US5273632. The page lists classifications, family, and legal status but displays no litigation field / no "Litigation" entries for this patent.
4. Unified Patents / PTAB materials. Searches surfaced Unified Patents litigation and IPR practices generally, but no Unified Patents case list entry, IPR, or PGR naming US 5,273,632.

Look-alike identifiers encountered (flagged, not conflated)

Because you instructed that identifiers be read literally, I note these hits that share the string "5273632" but are unrelated to US Patent 5,273,632:

• JP 5273632 B2 (特許第5273632号) — a Japanese patent granted Aug. 28, 2013 to Janome Sewing Machine Co., for a sewing machine with an automatic thread-cutting device. Different jurisdiction, different number, different subject matter.
• Nota Nr. 5273632 in a Brazilian brokerage statement (Migalhas PDF).
• Pranešimo Nr. 5273632 in a Lithuanian company-registry notice.
• The "382 patent" discussed in O2 Micro v. MPS/ASUSTeK (Fed. Cir., Sept. 2013) is a different patent (a 6‑digit "382" number), not 5,273,632.

None of these should be treated as US 5,273,632.

Assessment

The absence of litigation is consistent with the record:

• The patent expired November 19, 2012 and is not in force. Even if a cause of action accrued earlier, the window for new assertions closed more than a decade ago.
• The claims are narrow signal-processing method claims (homomorphic/cepstral blind deconvolution of chromatographic lane signals, with claims directed to lifter construction, Gaussian low-pass noise filtering, cepstral manipulations). See claims 1–33, e.g. claim 1 (transform to cepstrum → lifter → de-transform) and claim 8 (the FS → CLS → cepstrum → LLS → LFS → inverse-transform pipeline).
• It is a university-owned, DOE-funded patent; such patents are frequently licensed (and the government retained rights under 37 C.F.R. § 401 / 45 C.F.R. § 8, as recited in the specification), so enforcement, where it occurred, may be documented in licenses rather than dockets.

Caveats on my search

• I performed open-web searches; I was not able to complete a direct, query-by-patent-number run against the Unified Patents litigation portal case list, PACER, or a full CourtListener/RECAP docket search for "5,273,632" within this session. My "none found" conclusion therefore should be verified.
• Historical pre‑2000 district court dockets are unevenly digitized; a 1990s-era infringement suit, if any, could evade web search.

Recommended verification steps for a definitive answer:

1. USPTO Patent Center / Patent Assignment records for application 07/978,915 (assignment history can reveal enforcement-related transfers).
2. A Docket Alarm or CourtListener/RECAP full-text search on "5,273,632" and on "07/978,915."
3. A Unified Patents litigation-portal search (https://portal.unifiedpatents.com/litigation/caselist) keyed to 5,273,632.
4. TPL/PatentVector or Lex Machina patent-number lookup, which indexes older cases more completely.

If you need, I can attempt the portal- and docket-specific queries in a follow-up so the "none found" statement can be upgraded to a verified negative.

Proceedings overview

Total AIA trial proceedings on US 5273632: zero (0). The USPTO Open Data Portal returns no IPR, PGR, or CBM proceedings for this patent, and my independent web searches surfaced no Board proceeding, no Federal Circuit appeal from a Board decision, and no ex parte/ inter partes reexamination of this particular patent. Bottom line for a defendant: there is no PTAB-based invalidity shortcut and no PTAB-based estoppel benefit — but there is also no PTAB-based problem. The patent (filed 1992-11-19, granted 1993-12-28, expired 2012-11-19) reached the end of its statutory term more than a decade before AIA trial practice became a routine defense tool, so the absence of PTAB activity is expected rather than a signal of "hardening." The real defensive point is the one the structured data themselves make: this patent is expired, and any current assertion of it must be traced to pre-expiration conduct.

---

Proceedings — none on file

There are no proceedings to render case-by-case. To be explicit about what was checked and what was not found:

• IPR (35 U.S.C. §§ 311–319) — none. Note independently that IPRs are only available for challenges under § 102 / § 103 on patents and printed publications, and a petition cannot be filed until 9 months after issuance. Given the grant date of 1993-12-28, the entire theoretical window has long since passed, and the patent expired on 2012-11-19.
• PGR (35 U.S.C. §§ 321–329) — none. The PGR window is 9 months from grant, which closed in 1994. A PGR is legally unavailable now, and in any event PGR practice did not exist until 2012.
• CBM — none. Not applicable: this patent is not directed to a "financial product or service," it is a signal-processing/molecular-biology patent (classifications G01N27/44721 and G01N30/8631). The AIA § 18 CBM program also sunset for new petitions in September 2020.
• Reexamination (pre-AIA ex parte / inter partes) — no record found of any. I flag this as a search result, not a certainty; reexaminations are recorded in the patent's file history rather than in the ODP trial API, and I did not retrieve the full file wrapper.

No proceeding number, petitioner, panel, FWD, or appeal exists for me to report. I have not invented any. (Note: web results that mention a "527 patent" in litigation — e.g., a 1997 Display Solutions v. Daktronics summary-judgment decision, or the Acadia/MSN " '527 divisional application" appeal — concern unrelated patents/applications and are not US 5,273,632. Do not be misled by that number collision.)

---

Strategic summary

Claim status. Because no AIA trial or reexamination ever reached this patent, all 33 claims remain as issued — none canceled, none confirmed, all untested at the PTAB. The independent method claims include claim 1 (transforming a chromatographic distribution signal to a cepstrum, liftering it, and de-transforming to obtain a deconvolved lane signal) and claim 8 (the full homomorphic-filtering sequence: FFT → complex log-spectrum → inverse FFT to cepstrum → lifter → FFT to LLS → add back imaginary part → antilog → inverse FFT). Dependent claims cover Gaussian low-pass noise filtering (claims 4, 11), the raised-cosine/high-pass lifter shape (claims 5–7, 12), and noise removal (claims 3, 9, 10). None of these claims has been adjudicated by the Board.

Estoppel landscape. There is no § 315(e)(2) estoppel running against anyone, because no petitioner ever went to trial on this patent. Conversely, a current defendant gets no free-ride benefit from prior PTAB work. All prior-art grounds are equally available — but the more consequential point is temporal: with the patent expired, § 102/§ 103 art is only relevant to conduct during the pre-2012-11-19 damages window, and pre-AIA § 102 governs the validity analysis for a 1992 filing. That materially changes the § 102(b)/§ 102(e) landscape compared to a modern patent (e.g., pre-AIA § 102(g) prior invention, and the absence of AIA § 102(a)(1) "otherwise available to the public" art).

Pattern signals. No repeat petitioner, no patent-owner PTAB appeal practice, no defensive aggregator (Unified Patents or similar) in the chain — there is simply no chain. The patent's only publicly visible ownership event is the 1993 assignment from the inventors (Stockham, Ives) to the University of Utah and then to the University of Utah Research Foundation, which is still listed as current assignee. The University of Utah entity later litigated other, unrelated patents (the BRCA primer/method patents in Univ. of Utah Research Found. v. Ambry Genetics), which is sometimes mistaken for activity on this patent; it is not.

---

Recommended next steps

1. Treat the decisive issue as expiry, not PTAB. Confirm the term: filing date 1992-11-19, publication/grant 1993-12-28, and Google Patents' recorded "anticipated expiration" of 2012-11-19. If anyone is asserting US 5,273,632 today, demand that they identify the accused acts and the date range — § 286 damages reach back only six years, and there is no post-2012-11-19 damages window. Reference: US5273632A on Google Patents.
2. Do not spend money on a PTAB defense that does not exist. No petition under consideration can be paired with an existing Board record.
3. Verify the negative before relying on it. Pull the full file wrapper and any reexamination certificate via USPTO Patent Center and search PTAB E2E / PTAB Decisions by patent number; the ODP representation as of this briefing is that there are no AIA trials, and my searches are consistent with that, but a reexam filed in the 1990s would not appear in the trial API.
4. If a demand letter does cite claim 1 or claim 8, the strongest posture is: the patent is expired; validity of the asserted claims has never been tested at the PTAB; and any infringement theory must be grounded in pre-expiration conduct with a § 286–limited damages period. I found no Federal Circuit or district court decision on this patent that would preclusive either side — I state that as a search result, at less than high confidence, and it should be confirmed against the file history and docket before it is relied on in correspondence.

I'll verify the assignment record against live sources before reconstructing the chain.

Assignment Record Reconstruction — US 5273632

Patent: US 5,273,632 A — "Methods and apparatus for analysis of chromatographic migration patterns"
Application: US 07/978,915 · Filed 1992‑11‑19 · Granted 1993‑12‑28
Status of record: Expired – Lifetime (anticipated expiration 2012‑11‑19)

Sourcing caveat (read first): The USPTO Assignment Center index and its legacy mirror (assignment.uspto.gov) were not reachable/searchable for this patent during this session — searches returned no indexed assignment abstracts. Consequently no reel/frame numbers and no correspondent of record are available from any source I could reach, including the Google Patents "Legal Events" table (which lists assignment events but does not print reel/frame or correspondent data). Everything below is reconstructed from the Google Patents legal‑events record, corroborated by the DOE/OSTI patent record (OSTI ID 869094). I will not invent reel/frame values. Where an item is unknown, I say so.

---

Inventors

Both were university researchers funded by DOE grant DE‑FG02‑88ER60700 — the patent text itself carries the government‑interest statement ("This invention was made with Government support under grant number DEFG0288ER60700 awarded by the Department of Energy").

Unusual-pattern check — not present. There is no sign of inventors departing the assignee ahead of a sale. The opposite pattern is present: the inventors assigned to their employer within ~2 months of filing (1993‑01‑25), the standard university obligation‑to‑assign flow. Two contextual notes: (i) Stockham was reported (Irish Times, 2004) to have been diagnosed with Alzheimer's around 1994, retired from the Utah faculty about that year, and had no apparent further patenting activity; (ii) the 1994 companion publication shows the work was still being published from the university lab after grant — consistent with a research, not commercialization, track.

---

Original assignee

University of Utah Research Foundation, Inc. (Salt Lake City, UT) — the 501(c)(3) technology‑transfer / IP‑holding arm of the University of Utah. Primary line of business: management and licensing of university‑owned intellectual property; it does not manufacture or sell products. Product embodying the claims: none identified. The inventive subject matter (homomorphic blind deconvolution of DNA‑sequencing lane signals) was realized as an academic prototype film reader — the "automated film reader" described in the 1994 IEEE paper — not a commercial instrument. Current status: operating; still the assignee of record. Google Patents shows no post‑2012 change and the patent is expired.

Note the two‑step assignment structure (below) is a routine internal university pattern: inventors → The University of Utah → University of Utah Research Foundation. It is not a fire‑sale or licensing‑shell pattern.

---

Assignment timeline

Google Patents records exactly two assignment events, both pre‑grant. No post‑issuance assignments, licenses, security interests, mergers, or change‑of‑name recordings appear in any source reviewed.

• Executed date: not shown / recorded 1993‑01‑25 — Reel/Frame: not retrievable (Assignment Center not searchable; Google Patents does not publish reel/frame)

  • Conveyance: Assignment
  • Assignor: Stockham, Thomas G.; Ives, Jeffrey T. (individually)
  • Assignee: The University of Utah
  • Correspondent: not available — the Assignment Center record (which is the only place correspondent data lives) could not be retrieved. No recurrence assessment possible.
  • Context: Inventors' standard obligation‑to‑assign of patent rights to their employer; recorded while the application was pending.

• Executed date: not shown / recorded 1993‑09‑21 — Reel/Frame: not retrievable

  • Conveyance: Assignment (Google Patents caption: "ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS)")
  • Assignor: The University of Utah
  • Assignee: University of Utah Research Foundation
  • Correspondent: not available
  • Context: Internal reorganization — the university conveyed the application into its affiliated research foundation, the entity that issued on the patent three months later (1993‑12‑28). This is a change of holding entity within the same institutional family, not an arm's‑length transfer.

If Assignment Center in fact holds no additional records beyond these two (which is what the Google Patents legal‑events table implies), then this patent has no post‑issuance assignment chain at all — the University of Utah Research Foundation has owned it continuously since 1993. That is itself the finding.

---

Timeline diagram

timeline
    title Ownership of US 5273632
    1992 : Application filed by University of Utah
    1993 : Inventors assign to University of Utah
         : University of Utah assigns to UU Research Foundation
         : Patent granted 28 Dec
    2012 : Term expires 19 Nov

---

NPE / troll-pattern signals

Additional dispositive fact: US 5,273,632 expired on 1992‑11‑19 + 20 years = 2012‑11‑19 ("Anticipated expiration" per Google Patents; status "Expired – Lifetime"). The subject matter is in the public domain and cannot be asserted by anyone, which is the strongest possible anti‑NPE signal on this record.

---

Verdict

Insufficient data — the chain consists of nothing but the original two assignments, so the record cannot support an NPE classification.

Justification: The only recorded events are the 1993‑01‑25 assignment from Stockham and Ives to The University of Utah and the 1993‑09‑21 assignment from the University of Utah to University of Utah Research Foundation — a routine two‑step university IP‑consolidation flow, with the assignee unchanged for over 30 years and no post‑issuance recording of any kind. No shell entity, no known asserter, no litigation, no bankruptcy, and no defensive‑aggregator acquisition appears (reel/frame and correspondent data could not be retrieved, so signal 3 is untestable rather than affirmatively negative). The patent additionally expired 2012‑11‑19, making it permanently unassertable — so even a hypothetical hidden chain would have no assertion value today.

Verification link: https://assignmentcenter.uspto.gov/ (search by patent number 5273632) and the Google Patents legal‑events tab at https://patents.google.com/patent/US5273632/en — a manual pull of the Assignment Center abstract is recommended to capture the two reel/frame numbers and the correspondent of record, the only items this analysis could not obtain.

I'll research US patent 5273632 and its prior-art citations.

Let me search for the specific citation list on this patent.

I need the specific "Referenced Cited" section from the patent's front page. Let me search more precisely.

Let me try one more targeted search for the patent's front-page citation list.

I'll research the prior art landscape for this patent before assessing obviousness. Let me run several targeted searches.

Let me search for additional prior art references, including patents and the DNA-sequencing gel-reading literature.

Key finding: later patent US 5,748,491 (Allison et al.) expressly incorporates US 5,273,632 and criticizes it. Let me dig for the pre-1992 prior art more specifically.

Obviousness Analysis — U.S. Patent No. 5,273,632

Under 35 U.S.C. § 103 (pre‑AIA, priority date November 19, 1992)

---

0. Two threshold flags before I begin

(a) The page you gave me has no "Prior Art" section. The Google Patents text supplied contains only a Prior‑art keywords block (signal, lane, cepstrum, function, peaks) and one prior‑art citation in the body: Oppenheim & Schafer, Digital Signal Processing, Prentice‑Hall (1975), Ch. 10. The Google Patents "Patent Citations," "Cited By," and "Similar Documents" tables are not in the fetched text. So I could not "use the results from the Prior Art section of this page" as instructed — there was no such section to use. I say this plainly rather than padding the record. What follows reconstructs the prior art from (i) the patent's own admissions, (ii) the reference list of the inventors' own later journal paper, and (iii) open‑web searching. Every reconstructed reference below should be verified against the original document before it is relied on in a filing.

(b) The claim set is truncated. The page supplies only claims 1–13, and claim 13 is cut off mid‑sentence ("The method of claim 12 wh"). Claims 14–33 — which, given the "33 claims" count and the specification, very likely carry the DNA base‑calling/alignment subject matter — are not before me and are not analyzed. Any definitive § 103 opinion must address them.

(c) Contradiction check against the prior work product. The previously generated section is a litigation report (no litigation found; patent expired Nov 19, 2012). It contains no prior‑art findings, so there is nothing to contradict. One cross‑reference it does supply is useful and consistent: the '632 patent is expired and non‑assertable today, which matters for practical significance but not for the § 103 question, which is assessed as of 1992.

---

1. Legal framework applied

Because the application was filed November 19, 1992, pre‑AIA §§ 102/103 govern. The critical dates:

Consequences:

• References published before Nov 19, 1991 are § 102(b) statutory bars and are fully available for § 103.
• References published before Nov 19, 1992 but not more than a year earlier are available under § 102(a) if "by others," or as § 102(e) art if they are U.S. patents/applications.
• Crucially, an inventor's own publication from more than one year before filing is a § 102(b) bar and may be used both to anticipate and to combine for obviousness. This matters enormously here, because the primary reference is inventor Stockham's own 1975 paper.
• The KSR Int'l Co. v. Teleflex Inc., 550 U.S. 398 (2007) standard applies to motivation: a "finite number of identified, predictable solutions" and "design incentives," plus the "obvious to try" doctrine and the rule that optimizing a parameter of a known method is obvious unless a criticality is shown.

Graham v. John Deere factors — scope/content of the prior art; differences; PHOSITA level; secondary considerations — are walked through in § 5.

---

2. Person having ordinary skill in the art (PHOSITA)

A PHOSITA here is a team‑level artisan: (a) an electrical engineer/computer scientist with 2–4 years' experience in digital signal processing, including Fourier/cepstral methods and the FFT; working alongside (b) a biochemist/geneticist familiar with dideoxy (Sanger) sequencing, slab‑gel/autoradiographic or fluorescent detection, and the known artifacts of band broadening, band compression, and lane‑to‑lane spacing variation. This is the artisan the '632 specification itself addresses, and it is the artisan the Federal Circuit would find on this record.

---

3. The claims on the table (as literally supplied)

Internal inconsistency to flag now: claim 4's range (0.024A–0.072A; at A=1024 → ≈24.6–73.7 samples) and claim 11's range (50–150 samples at π=1024) are only partially coextensive, and the specification (col. disclosing the Gaussian filter) says "between about 50 and about 150 frequency samples when π is 1024." Claim 4 therefore appears to be a drafting/transcription error. Per your standing rule I read the numbers literally and do not auto‑correct, but this inconsistency is itself a validity vulnerability (written‑description/definiteness, § 112), independent of § 103.

---

4. The prior art

4.1 Reference table

Grounding note on R11/R12/R14 dates: these numbers and their substance come from the full‑text snippets returned by the searches (https://patentimages.storage.googleapis.com/b9/5b/10/bc1c835b72ab30/US4807148.pdf and the US 4,941,101 PDF). I did not independently confirm their exact issue dates. Treat the dates as "to be verified."

4.2 A non‑prior‑art reference that is highly probative (later art)

US 5,748,491 (Allison et al., The Perkin‑Elmer Corp.) — "Deconvolution method for the analysis of data resulting from analytical separation processes" (https://www.freepatentsonline.com/[5748491](/patent/5748491).html). This is after the '632 priority date and is therefore not prior art. But it is powerful objective evidence of the state of the art, because it expressly states:

"Present deconvolution methods applied to data resulting from analytical separations use a 'blind' deconvolution method with homomorphic filtering, e.g., U.S. Pat. No. 5,273,632, incorporated herein by reference… Blind deconvolution methods utilize a constant PSF which is chosen without any reference to a particular application… blind deconvolution methods suffer from a number of important limitations…"

Why this matters for § 103: it confirms (i) that by the 1990s homomorphic blind deconvolution of analytical‑separation (chromatography/electrophoresis) data was regarded as conventional, and (ii) that the '632's contribution was understood as a baseline application of the technique, later improved upon — not as a pioneering insight. Post‑filing criticism of a patent's own approach tends to undercut "unexpected results."

---

5. Claim‑chart / obviousness analysis

5.1 The primary combination — R3 (Stockham 1975) + R1/R2 (the homomorphic‑filtering canon) + R4 (Elder 1986) [or R5–R9]

Claim 1. Every element is supplied:

The only element not literally in R3 is the choice of a chromatographic/electrophoretic signal. And R3 already reaches beyond audio: it expressly claims "results for the case of images degraded by some common forms of blur," and its stated motivation for the DSP approach is that the "distortion operator is assumed stationary" and that the technique requires only a general, non‑specific model of the blur — precisely the assumptions the '632 specification repeats about the gel blur.

Motivation to combine.

• The gel‑reading art (R4, R7, R8, R9) had identified band broadening/overlap as the accuracy bottleneck, and R9 explicitly taught that reconstructing the undistorted band shape improves peak resolution — i.e., the field was asking for exactly the tool R3/R1 supply. KSR: "the identified, predictable solution."
• R3 and R1 were the standard, well‑indexed tools for the "blur = convolution" problem, and R3 was written by a named co‑inventor of the '632 patent (Stockham) — so the artisan of skill had every reason to look there.
• R10 shows the art had already transplanted Fourier‑domain filtering onto electropherograms and then called bases — so the last mile (DSP → sequencing trace) had already been walked by others.

Reasonable expectation of success. High. R3 explicitly frames the method as blind — no need to know the blur — and states the distinguishing assumptions (stationarity, small‑extent impulse response). The '632 specification concedes the same about the gel blur ("the blurring function need not be precisely known"). The art's own reports (R8's maximum‑entropy reconstruction; R9's band‑shape reconstruction) had already obtained resolution gains before the '632 filing.

Conclusion on claim 1: Obvious under § 103 over R3 in view of R4 (and/or R1, R2). Additionally, R3 alone is a § 102(b) statutory bar that a POSITA would read as describing the claimed method applied to any linear shift‑invariant blur — making claim 1's novelty turn entirely on the nominal field‑of‑use recitation in the preamble.

5.2 Claim 8 (the explicit FS→CLS→cepstrum→LLS→LFS pipeline) — obvious

Every step is textbook. R2 and R1 disclose, in order: FT of the signal; complex logarithm to convert the convolution product into a sum (the entire point of homomorphic systems); cepstrum = inverse transform of the log‑spectrum; liftering; transform back to the log‑spectrum; exponentiate; inverse transform. R3 discloses the same sequence in the same order and, critically, treats the real part / log‑magnitude as the operative quantity and handles phase separately (the "Hilbert‑transform/minimum‑phase" discussion of the log‑phase and the separate real/imaginary treatment). The '632's steps of "taking an inverse Fourier transform of the real portion" and "adding the imaginary portion back" are not inventive embellishments — they are the unavoidable bookkeeping of the complex cepstrum described in R1/R2/R3. Motivation: implement the known algorithm. Expectation of success: the algorithm was standard.

5.3 Claims 2, 3, 9, 10 (noise filtering; low‑pass on the liftered FS) — obvious

• R3 is largely about noise: it analyses the effect of additive noise on homomorphic estimates, compares the homomorphic estimator with the direct power‑spectrum estimator, and states the latter is "about 22 percent more stable." R1 Ch. 10 discusses noise and the need for filtering in the liftered log‑spectrum.
• R10 claims and describes Fourier‑domain filtering of electropherogram data.
• Low‑pass filtering to suppress high‑frequency noise after boosting is the most routine step in the DSP canon. Motivation: the patent itself says noise filtering "is generally required to produce satisfactory results" — a POSITA would supply it as a matter of course.

5.4 Claims 4 and 11 (Gaussian low‑pass filter; numerical bandwidths) — obvious under KSR

• Selecting a Gaussian (or raised‑cosine, Hamming, Hanning, Butterworth) low‑pass is a design choice among a finite, identified, predictable set of smoothing filters.
• Fixing the cutoff at "50–150 samples out of 1024" (claim 11) or "0.024A–0.072A" (claim 4) is optimization of a known parameter, routinely done by one of ordinary skill by balancing "peaks do not overlap" against "no significant gaps between adjacent peaks" — the very criterion the '632 specification recites as the goal. No criticality (no showing that values outside the range fail) is disclosed. KSR: "[A] result‑effective variable… a court should ask whether the range claimed was obvious to try."
• Also note the claim 4 vs. claim 11 vs. specification inconsistency flagged in § 3 — this further undermines any argument that the numeric ranges reflect a deliberate inventive discovery.

5.5 Claims 5, 6, 7, 12, 13 (lifter = high‑pass; 50 % raised‑cosine taper reaching 0.5 near the cepstral plateau) — obvious

• R1 and R2 teach that the low‑quefrency region carries the system/blur response and that liftering should suppress the low‑quefrency portion while passing the high‑quefrency portion — that is claim 5 and claim 12.
• The 50 % raised‑cosine taper is a textbook window/taper (identical in form to the Hann/raised‑cosine window ubiquitous in DSP); using a smooth taper rather than a brick‑wall cut to avoid ringing is standard practice. Choosing the taper's inflection near where the cepstrum "levels off" is a direct reading of the cepstrum, per R3's own instruction to shape the separating filter based on the observed signal/system content.
• The '632 specification itself concedes alternatives ("Other lifter designs are possible, including lifter functions designed to satisfy a desirable error criterion (for example Wiener filtering). However… such other designs would not achieve significantly better results"). That is an admission that the disclosed lifter is one of a small set of known alternatives with no significant performance difference — the definition of obviousness.

5.6 An independent (alternative) combination that does not rely on R3 at all

Even setting aside Stockham's own work, the art independently renders the general concept obvious:

R2 + R1 (homomorphic filtering / liftering) + any one of R11/R12/R13/R14 (deconvolution of overlapped chromatographic peaks) + R4/R10 (DNA‑sequencing lane signals and FT filtering/base calling).

• R11, R12, R13 and R14 establish that deconvolving overlapped chromatographic peaks via Fourier/transform‑domain methods was a mature, crowded field by the mid‑1980s. R11's model, g(t) = h(t)*x(t) + n(t), is identical to the '632's stated model ("convolution of two or more separate signal functions… with an additional contribution of random background noise").
• R4/R10 establish that chromatographic data of this kind includes electrophoretic DNA‑sequencing lane signals and that such signals were already being FT‑filtered and base‑called.
• Motivation: resolve overlapped peaks; the combination is the union of two recognized solutions to one recognized problem. Expectation of success: R11/R12 report successful resolution of fused peaks; R8 reports reconstruction of sequencing‑gel autoradiographs.

Under this combination, the only thing distinguishing the '632 from the art is the specific selection of a blind, homomorphic deconvolution over the alternative (Wiener, iterative/Janson, factor‑analysis, maximum‑entropy) deconvolution methods. But:

• Stockham's R3 specifically teaches that blind homomorphic deconvolution is preferable where the blur is unknown and must be estimated without a reference measurement — exactly the '632's stated advantage; and
• the '632 specification's own summary explains that "blind" merely means "the blurring function need not be precisely known… nor is there any need for a noise reading… Once the parameters… are defined for a particular chromatographic technique, no further information… is needed."

Choosing the blind variant precisely to avoid per‑sample calibration is a motivated, predictable selection — obvious.

---

6. Secondary considerations (Graham factor 4)

I found no evidence supporting non‑obviousness, and some against it. Counsel should, however, run this down independently:

---

7. Bottom line

Claims 1–13, as supplied, are, in my assessment, vulnerable to a § 103 challenge — and claim 1 additionally to a § 102(b) challenge.

1. Claim 1 is the weakest. Inventor Stockham's own 1975 Proc. IEEE paper (R3) discloses blind deconvolution by homomorphic filtering, including filtering in the transformed (cepstral) domain to separate a desired signal from an unknown blur, explicitly extended to "images degraded by blur." The sole delta is the field‑of‑use preamble, supplied by R4 (1986) — with R9 (1991) and R8 (1990) supplying the motivation ("reconstruction of the undistorted band shape… substantially improves the resolution of peaks"). Because R3 predates filing by 17 years, it is a § 102(b) bar usable both to anticipate and to combine.
2. Claim 8 is the heart of the invention and is entirely disclosed by the homomorphic‑filtering canon (R1, R2) and R3; the real/imaginary handling is inherent bookkeeping of the complex cepstrum.
3. Claims 2–4, 9–11 (noise filtering; Gaussian; numeric bandwidths) are routine DSP steps and mere parameter optimization — KSR obvious‑to‑try. Claim 4 also conflicts with claim 11 and the specification.
4. Claims 5–7, 12–13 (high‑pass lifter; 50 % raised‑cosine taper) are textbook lifter/window design, and the specification admits equivalent alternatives.
5. An independent combination built on the chromatographic deconvolution art (R11–R14) plus the sequencing‑signal art (R4, R10) reaches the same result without relying on R3 at all — which is important, because it means the claims do not survive merely by attacking R3's applicability.

Strongest non‑obviousness counter‑positions (for the patent owner, and to be tested before any challenge is filed):

• That no pre‑1992 reference teaches applying homomorphic/cepstral, blind deconvolution specifically to electrophoretic DNA‑sequencing lane signals — i.e., the specific selection of the transform pair and field is the point of novelty. This is a real but, in my view, weak argument under KSR's "predictable solution" and "obvious to try" doctrines given R3's express extension to blur generally.
• That the numeric ranges in claims 4 and 11 are critical to avoiding overlapped‑but‑gapless peaks. This argument is presently unsupported: no data in the specification shows that values outside the ranges fail. Absent such a showing, the ranges are merely optimized.
• The claims I cannot see (14–33) may contain genuine novelty — e.g., the 4 × 4 lane‑alignment matrix, the 3‑D spacing matrix with interpolation, the overlap‑add segment alignment, or the specific threshold‑function construction. Do not form a final opinion without them. The alignment/refinement subject matter is where the '632's distinct contribution most plausibly lies, and it was not part of the "signal analysis" prior art surveyed above.

---

8. Verification steps I recommend before relying on this

1. Pull the original PDFs of R3 (DOI 10.1109/PROC.1975.9800), R2, and R1 Ch. 10, and confirm the exact page cites for: liftering; the low/high‑quefrency separation rationale; the noise analysis; and the "blurred images" application. Quote these verbatim into a claim chart.
2. Confirm issue/priority dates for R11 (US 4,941,101), R12 (US 4,807,148), and R14 (US 4,353,242) — my dates are inferred, not verified.
3. Confirm publication dates for R7, R8 and R9 (all Electrophoresis) — they are the strongest motivation/expectation‑of‑success references and need to be dated precisely against the Nov 19, 1991 § 102(b) line.
4. Obtain claims 14–33 and any prosecution history (in the file wrapper) and foreign counterparts (the WO 98/11258 family suggests a later continuation strategy that may reveal applicant's own characterization of the '632 over the art).
5. Confirm the § 102(e) status of US 5,098,536 (granted Mar 24, 1992, before the '632 filing) against the '632 — it is close enough to the claimed FT‑filter‑then‑base‑call architecture to be worth a standalone anticipation check.
6. Re‑run the searches I could not complete (Unified Patents/IPR history on the later University of Utah sequencing patents), in case the '632 was cited as prior art against them — which would be useful extrinsic evidence of scope.

Caveats restated: I had no "Prior Art" section on the page to work from; the claim set was truncated at claim 13; and every reconstructed reference and date above should be confirmed against the original document before this analysis is relied upon.