What did Rob design before Hugo, and how did the WTA filter and pulse array come about?

Analog beginnings

Watts says he started assembling electronics as a boy, beginning with a matchbox radio article, Apr 2018. In 2022 he said he could not afford a hi-fi on his paper-round money, so he built from DIY designs, and that at university he designed a power amp "which featured my own compound Class A output stage" article, Mar 2022article, Mar 2022.

In 2018 he said he "started as an analogue design engineer" and had designed "thick film hybrid op-amps for audio" with a 0.5 GHz gain-bandwidth product article, Apr 2018. In 2016 he described himself as "a designer of analogue electronics (started designing my own gear in 1980)" Head-Fi, Mar 2016. In 2017 he called a DPA 200S "one of my designs from the early 1990's", saying it used one of his thick film hybrid amps Head-Fi, Aug 2017. In an interview published in March 2026 (recording date not stated) he summarized the original work as "analog design, preamplifiers and power amplifiers" Audio Force 7:36.

Asked in 2022 how Deltec Audio, the first company he co-founded, came about, he answered "With some friends that I met at university" article, Mar 2022. In the same interview he tied the PDM1024 DAC to Deltec article, Mar 2022. Asked in 2023 whether, looking back on his early days at Deltec, he would have done anything differently, he said "I would have had a business partner that knew what he was doing and had some capital", because he ended up running the whole company himself Eric's Hi-Fi 43:30Eric's Hi-Fi 43:38.

1989 to the early 1990s: first DACs

Watts said in 2014 that in early 1989 he was "dead against all things digital" until he heard the first PDM CD player, which used a new Philips chip, and that the DAC he designed with that chip "went into production in 1989 - the PDM1" article, Aug 2014. In another 2014 post he named the chip as the Philips Bitstream SAA7320 Head-Fi, Apr 2014. In 2017 he put his first DAC design in 1989, "PDM or a DSD 256 DAC" article, Dec 2017. In a 2023 Q&A he said DACs up to 1989 were R2R and "sounded dreadful", and that the Philips Bitstream chip, which "turned out to be the same as DSD256", left him "just blown away" Soundnews 20:12Soundnews 20:24Soundnews 20:51.

The chip still bothered him. In 2023 he described the problem as 256 one-bit pulses and asked "How is that good enough to describe a 20 kHz waveform?" Soundnews 21:31. In the Passion for Sound interview (published in parts December 2019 to January 2020; recording date not stated) he gave the same worry as the reason for the PDM1024, in which each DAC chip was fed "with a random source" and the outputs summed in the analog domain Passion for Sound 1:36. In 2017 he said that in 1992 he used "multiple PDM noise shapers with each chip having it's own randomised digital data", and that the better sound "proved" his intuition that PDM or DSD had fundamental resolution problems, which led him to design his own noise shapers Head-Fi, Dec 2017. In 2017 he also said that building the PDM1024 was what drove him to create his own DAC technology: it "sounded much better, but indicated that fundamentally better performance was possible" article, Dec 2017.

His descriptions of the PDM1024 vary. In 2014 he said it had a number of Philips noise shaper chips, each fed a different dither signal, and in another 2014 post that it also used "digital filtering (delay and add)" against jitter sensitivity and out-of-band noise article, Aug 2014Head-Fi, Apr 2014. In 2016 he dated it to 1992 and said it used multiple DSD256 noise shapers, each fed its own random noise Head-Fi, May 2016. In that interview he said "four of these devices, or maybe you used eight, I can't remember" Passion for Sound 1:36. Without naming a product, he said in 2015 that he used 8 PDM noise shapers with different dither Head-Fi, Apr 2015, and in February 2019 he described "32 element DSD conversion to analogue" in the early 1990s Head-Fi, Feb 2019. These are different descriptions, not a stated change of view.

In 2016 he gave a second motive for building his own DAC technology: his early-1990s DSD DACs were inconsistent, each measuring and sounding a bit different, and "the only way of solving these issues was to take control over every element within a DAC" Head-Fi, Jun 2016.

1994 or 1995: pulse array

Watts most often places pulse array in 1995, in statements from 2014 to 2024 article, Aug 2014Head-Fi, Apr 2014Head-Fi, Dec 2017Head-Fi, Mar 2021HIFICLUB 44:59. He has also given 1994. In 2016 he wrote that "in 1994 I acquired digital design skills and started designing my own DAC's creating pulse array DAC technology" Head-Fi, Mar 2016, and in 2017 that he had used his own pulse array technology "from 1994" Head-Fi, Jun 2017. In a podcast published in October 2026 (recording date not stated) he called it "something that I developed in 1994 using FPGAs" Part-Time Audiophile 38:10, and a HiFi Buys interview published the same month (recording date not stated; its transcript places it before May 2026) has him saying "it was developed in 1995 or 1994, 1995" HiFi Buys 1:41:42.

In 2017 he wrote that because of DSD's limitations, "in 1995 I developed my own DAC using FPGA’s, and invented the pulse array DAC" article, Dec 2017. In 2014 he said that in 1994 larger FPGAs were becoming available and he realized the only way to solve the resolution and jitter problems was to design his own noise shaper: "So in 1995, pulse array DAC was born, which had a 16 element discrete DAC and an FPGA noise shaper" article, Aug 2014. In 2022 he recalled designing around 1994 or 1995 with Actel anti-fuse FPGAs, which could be programmed only once, so each design change meant a new device Head-Fi, Aug 2022.

On what it was meant to do, he said in 2015 that he wanted much higher resolution, so "the noise shaper OP is 5 bits not 1 bit", run at 2048 times the sample rate rather than 256 Head-Fi, Apr 2015. In 2023 he said the intent "was to make it sound as good as DSD 256, but solve all the other problems that DSD has got" The Headphone Show 44:50. In 2017 he said the part of DSD he wanted to keep was that "one resistor doesn't create distortion problems" RMAF 40:50.

In 2018 he listed five problems with DSD that pulse array was meant to avoid: extreme sensitivity to master clock jitter, strong correlation between switching activity and signal, poor signal resolution, transient errors that depend on input amplitude, and modulation instability Head-Fi, Nov 2018. In 2022 he described the aim as solving small-signal non-linearity and noise floor modulation MFG Audio Reviews 12:31.

He said in 2017 that he did not patent WTA or pulse array Head-Fi, Dec 2017. In 2024 he said "it's just strange that nobody else has copied it" HIFICLUB 44:59. That line is, so check the recording.

1999: the DAC64 and the WTA filter

Where the idea came from. Watts traces the WTA idea to his university days. In 2021 he wrote "I have been working on the theory of interpolation since 1981" Head-Fi, Jun 2021, and in an interview published in August 2022 (recording date not stated) he also dated it to 1981 AVstore 1:09. In 2018 he said he reasoned from sampling theory that a filter accurate to 16 bits would need about a million taps, which was impossible then MajorHiFi 14:33article, Apr 2018. His view at the time, he said in 2018, was "forget digital, it's never going to sound good, you need a million taps, stick to vinyl" c w 17:48. The figure is stated differently across tellings: about 700,000 taps ("something like a million") in 2017, and three quarters of a million in the Passion for Sound interview (published December 2019; recording date not stated), when he dated the calculation to 1980 miceblue425 27:06Passion for Sound 10:57.

How it got built. In 2014 he said that when he first used FPGAs (mid 1990s) "you could do a DAC but no filter", and that in the late 1990s you could do a DAC and a filter Head-Fi, Apr 2014. In 2023 he said his 1995 design used an external silicon interpolation filter, and that in 1999 the Xilinx FPGAs "had enough processing power to do the interpolation filters as well" Soundnews 18:09Soundnews 18:49. He has dated the start of his WTA filter design work to 1998: in 2021 he wrote that "in 1998 I started actually designing and listening to these filters built on those principles" Head-Fi, Jun 2021, and in a 2017 talk he said "about 1998", when he needed three FPGAs for an FIR filter miceblue425 20:34.

In 2017 he described the sequence: a 1990s FPGA filter of 2048 taps with a closed-form design that preserved the original samples; a delay before he tried a non-aliasing design, because "it did not feel right, modifying the data"; and then a "guaranteed non-aliasing filter" that became the first WTA, where "a 256 tap WTA sounded much better than a 2048 tap conventional filter" Head-Fi, Feb 2017. In 2015 he gave the same comparison against a 2048-tap half-band Kaiser filter, placing the work in the late 1990s Head-Fi, Jun 2015.

First product. He says the first WTA filter, and the first product with one, was the Chord DAC64 in 1999 article, Dec 2017c w 14:48. In 2018 he gave its tap count as 1,024 and the name as Watts Transient Alignment article, Apr 2018, and in other 2018 talks as about a thousand Pursuit Perfect System 6:44. In 2017 he said that on the DAC64 the WTA and pulse array were split across different FPGAs Head-Fi, Dec 2017. In 2020 he defined the WTA algorithm as "a unique windowing function designed to maximise the accuracy of recovering transient timing from the original analogue signal" Head-Fi, Nov 2020.

How his own thinking about WTA moved. In 2018 he said the first WTA version had a limited tap length and he was most concerned about image aliasing, and that this focus gave its large benefit. With more taps, he found two filters with identical image rejection could sound very different, and that the closer a filter was to a sinc function the better it sounded. He summed this up as his thinking over ten years having "gone full circle back to the theoretical ideal that I thought in the 1980's" Head-Fi, Jul 2018.

QBD76. In April 2014 he said the QBD76 was planned with 4096 taps in the DAC64's filter structure, and that listening tests showed the tap length of the secondary stages mattered, so the filter was overhauled into "a flat 8 times over sampled (OS) structure" and the algorithm changed completely, reaching 18k taps. He said all production QBD76 had the 18k filter and that the work was done at the prototype stage Head-Fi, Apr 2014. In the same post he said the QBD76 pulse array had 16 elements against 4 on Hugo, and described the QBD76 as seven years old Head-Fi, Apr 2014.

Tap history. In 2018 he gave the sequence as roughly a thousand taps for the DAC64, 18 thousand for the QBD76, then 26 thousand for Hugo and 164,000 for DAVE Pursuit Perfect System 9:06c w 15:36. In April 2018 he also cited 1,015,808 taps for the Blu MKII (2017) article, Apr 2018. In a podcast published in October 2026 (recording date not stated) he said "we're at 4 million taps" Part-Time Audiophile 46:08. That last figure is.

1999 to 2012: Chord, silicon consulting and other designs

On Chord, he said in 2014 that Chord is John Franks' company and he is "an independent design consultant, with Chord as one of my clients" Head-Fi, Jun 2014. In an interview published in March 2026 (recording date not stated) he said he met Franks at a show in Las Vegas in 1995 and that they brought out the DAC64 in 1999 Audio Force 3:19.

He says he stepped back from audio after the DAC64. In 2015 he wrote that "after the DAC 64 in 2001 I gave up on the audio business and put my main focus on silicon design and consultancy", with Chord work part time and for fun Head-Fi, Oct 2015. In 2016 he said that from 2001 he worked with silicon companies, mostly "creating IP and patenting it, then selling the patents", and that he had stopped that work the year before Head-Fi, Mar 2016; in 2023 he put the stop in 2015 The Headphone Show 54:58. In 2022 he put the consulting at 14 years Head-Fi, Feb 2022. He said in 2014 that when Hugo was being designed only 10% of his income came from Chord Head-Fi, Oct 2014. In a podcast published in June 2020 (recording date not stated) he said that with Hugo, "everything completely changed. Chord's business took off" Part-Time Audiophile 4:40.

On digital amplifiers, he said in 2018 "I invented the technology behind DDFA", and that the design under discussion was very different from DDFA, the key difference being a non-switching output Head-Fi, Jan 2018. In 2019 he said DDFA did not use pulse array, because it is a PWM switching output Head-Fi, May 2019. In an interview given at the Munich High End show (the source says the most recent one before publication, likely May 2024; published February 2025) he said his semiconductor work included "digital amplifiers with feedback which I patented" audio-2G 1:20.

Other pre-Hugo projects he mentions: the Davina ADC project, whose "first prototype was in 2001" Head-Fi, Feb 2016; and the digital phase-locked loop, which he began in 2000 and "perfected" after about six or seven years Head-Fi, Jun 2016. On upgrades after release, he said in 2015 that "The DAC 64 had a lot of changes, QBD less, Qute even less, Hugo none" Head-Fi, Oct 2015. In a 2017 post he dated his first listening to Hugo to October 2012 Head-Fi, Jun 2017; a post of 20 October 2014 says "Hugo is one year old today" and dates the first listening test of the finished code to "Sunday 20th Oct" a year earlier Head-Fi, Oct 2014. No item in the archive relates the two dates.

Where his accounts differ

First DAC work. He usually says 1989 article, Dec 2017Audiophilesdoteu 3, but in 2017 he wrote "I have been designing DAC's from 1988" Head-Fi, Jun 2017.

First Hugo listening. October 2012 in a 2017 post Head-Fi, Jun 2017; the first listening test of the finished code a year before October 2014 in a post marking Hugo's first birthday Head-Fi, Oct 2014. They may describe one event or two.

Pulse array year. Mostly 1995, sometimes 1994, as set out above Head-Fi, Dec 2017Head-Fi, Mar 2016Part-Time Audiophile 38:10.

First FPGA use. In 2014 he said "mid 90's" Head-Fi, Apr 2014, and in 2015 that FPGAs were "just becoming available (1994 now)" Head-Fi, Apr 2015. In 2016 he wrote that he used them with the PDM1024 "way back in 1992" Head-Fi, May 2016. In 2022 he said around 1994 or 1995 Head-Fi, Aug 2022. In 2023 he said he started using simple Actel anti-fuse FPGAs in 1992 for parallel dithered noise shapers, and that they became big enough for his own noise shapers, and pulse array, in 1995 Soundnews 17:11Soundnews 18:09. In 2024 he said he "first started getting involved in FPGAs in 1995" Audiophilesdoteu 3. In an interview published in March 2026 (recording date not stated) he said "by 1992 I was using FPGAs" Audio Force 7:36.

When his Chord work began. He said "since 1996" in 2018, "since 1997" in a podcast published in June 2020 (recording date not stated) and again in an interview published in October 2026 (recording date not stated), and in 2023 that he started consulting with Chord in 1999 article, Mar 2018Part-Time Audiophile 3:56HiFi Buys 2:01Soundnews 18:09. He does not reconcile these years.

FPGAs in the DAC64. He said four in 2018 MajorHiFi 10:27, three at CanJam London in 2018 c w 14:48, "partitioned into four FPGAs" in 2020 Head-Fi, Nov 2020, and "five FPGAs" in an interview published in October 2026 HiFi Buys 16:16.

What he has not said

He has not given a complete list of the products he designed before Hugo. He gave no launch year for the QBD76 beyond calling it seven years old in 2014, and no dates for when Deltec started or closed. He did not say whether the DPA 200S was a Deltec product, though in 2022 he said a mid-1990s design "would have been dpa digital then" Head-Fi, Apr 2014Head-Fi, Aug 2017Head-Fi, Aug 2022. In a podcast published in June 2020 (recording date not stated) he mentioned having "worked with one other small high end company" without naming it Part-Time Audiophile 3:56. He has not explained why his stated years for first FPGA use, pulse array and the start of his Chord work vary.

Quotes

“my first DAC design was in 1989, which was PDM or a DSD 256 DAC”2017-12-09 article, Dec 2017

“in 1995 I invented pulse array as this eliminates correlated jitter”2016-09-04 Head-Fi, Sep 2016

“I have been working on the theory of interpolation since 1981.”2021-06-04 Head-Fi, Jun 2021

“a 256 tap WTA sounded much better than a 2048 tap conventional filter”2017-02-10 Head-Fi, Feb 2017

“No I did not patent WTA or pulse array.”2017-12-15 Head-Fi, Dec 2017

Dated positions

Caveats and open points

• Several statements (DAC64 and early-FPGA remarks, the 2022 1981 date, the million-tap figures, the 2024 'nobody else has copied it' remark, the October 2026 '4 million taps' and '1994' remarks, and several Chord start dates) come of talks and interviews; check the recordings for exact wording and figures.
• Almost everything here is Watts recalling events from 10 to 40 years earlier; dates and counts (first DAC year, pulse array year, first FPGA use, DAC64 FPGA count, PDM1024 chip count, Chord start year) differ between tellings, and the differences are reported rather than reconciled.
• Statements are scoped to the product named: the 18k-tap filter and 16-element pulse array are QBD76 statements, the 1,024-tap filter is the DAC64, and later tap counts for Hugo, DAVE, the Blu MKII and later products are context only. The 2015 '8 noise shapers' and 2019 '32 element' descriptions do not name the PDM1024.
• No complete list of his pre-Hugo products appears; the answer is limited to what he named. The Weekly ASCII article's product list is the author's, not his, and was not used. Interviewer framing (for example that Deltec folded) is not used as his statement.

Topics: Watts's career and design history