What does Rob think of Class D amplifiers, and how do his pulse array power amps differ from them?

Short version

In statements from 2016 to 2026 Watts names two problems with Class D, transient timing errors caused by switching and noise floor modulation; in a podcast published June 2020 he said the second can be solved in feedback Class D designs but the first cannot, and in 2023 and in an interview published October 2026 he says Class D sounds poor or "awful" Head-Fi, Apr 2016Part-Time Audiophile 21:57Head-Fi, Jul 2023HiFi Buys 1:45:17. His pulse array power amps, which he first called the "digital power amp" and later the DX amps or "Power Pulse Array", are in his description linear, non-switching and built like his DACs' single-amplifier output stage Head-Fi, Jun 2016Head-Fi, May 2017Head-Fi, Jan 2017. They have not been released in anything this archive covers; his latest statement on them is from October 2024 Head-Fi, Oct 2024.

What he says about Class D

2016: switching degrades the timing of transients

On 2 April 2016 he wrote that the biggest problem with Class D is switching, because "the very act of switching the OP degrades the timing of transients", and that even at 1.6 MHz "you can't escape from the small timing errors that get created" Head-Fi, Apr 2016. The next day he said Class D has a timing-of-transients problem that "sounds just like" the problem he has with digital sampling, that the output LC filter "can not do an effective enough job" of reconstructing the timing, and that noise floor modulation is "another massive problem" Head-Fi, Apr 2016. In August 2016 he wrote that his biggest worry with switching amplifiers is the timing of transients issue Head-Fi, Aug 2016. In an August 2016 post on DAVE's power supply he put the noise floor modulation of "typical class D" at around minus 120 dB and said that level gives "considerable hardness and glare" Head-Fi, Aug 2016. On 22 April 2016 he said he wanted to get away from the name digital power amp because it is synonymous with Class D, "which is often awful" Head-Fi, Apr 2016.

2017 to 2018: noise floor modulation, in conference talks

In a DAC design masterclass from CanJam RMAF 2017 (given in October 2017, video published December 2017) he said "Class D amplifiers have huge levels of noise floor modulation" RMAF 10:38. At a Nintronics open day talk in November 2018 (video published December 2018) he said large noise floor modulation, which he says tends to come with Class D amplifiers, makes the treble sound grainy and coarse, and gave Class D figures of about minus 115 dB at full output dropping to about minus 130 dB Pursuit Perfect System 10:18Pursuit Perfect System 11:05.

2019 to 2020: poor-quality examples, but a hedge

In March 2019 he described a simple Texas Instruments Class D amp that converts digital data to PWM with no output filter, adding "But of course it's poor quality" Head-Fi, Mar 2019. In May 2020 he said GaN transistors do offer benefits for Class D, but "they won't solve the innate transient timing problems that Class D has" Head-Fi, May 2020.

In a podcast published June 2020 (recording date not stated) his position is more qualified than in the forum posts. He said Class D has "innate problems", that noise floor modulation is a major issue especially for lower-quality Class D, and that he has "worked on Class D amplifiers and designed Class D topologies" with feedback, with extremely good measured performance and no noise floor modulation, so those issues can be solved Part-Time Audiophile 21:57. What he said cannot be solved is transient reconstruction: his Class D designs ran at 800 kHz and some at 1.6 MHz, and the output filters have to recover the timing, so it "will never be as good as a linear amplifier" if the DAC reconstructs timing to nanoseconds Part-Time Audiophile 25:19. He added that feedback Class D can give "audiophile quality performance, but you can't get state of the art" Part-Time Audiophile 26:10.

In July 2020, in the Hugo TT2 thread, he said "Class D would destroy the transient timing accuracy, making the WTA filters almost worthless" and ended "So no it's certainly not Class D..." Head-Fi, Jul 2020.

2022 to 2026: "sounds poor"

In June 2022 he wrote that PWM "by it's very nature will have transient timing that is amplitude dependent", and that an ideal Class D output reconstruction filter would remove the problem but "you would need an impossibly capable filter to do this" Head-Fi, Jun 2022. On 29 July 2023 he wrote "I would never use Class D" and that "absolutely Class D suffers from transient timing distortions, that's why it sounds poor" Head-Fi, Jul 2023. On 6 August 2023 he said Class G and especially Class H are "fundamentally better than Class D, which creates transient timing errors and subsequent poor sound quality" Head-Fi, Aug 2023.

In a September 2024 live stream, talking about the Quartet M Scaler, he said one audiophile could hardly hear the difference (not using DAVE) and that the problem turned out to be his power amp, "which was a Class D" The Source AV 15:29The Source AV 15:42. Of using DAVE with a Class D power amplifier he said "you're going to get an improvement", but "Nothing like the same level" The Source AV 16:28The Source AV 16:37The Source AV 16:42.

In an interview published 2 October 2026 (recording date not stated) he explained that with PWM a high-amplitude transient pulse comes early and a negative one late, and said "That's why Class D amplifiers sound awful" HiFi Buys 1:44:40HiFi Buys 1:45:17. He said the transient timing is destroyed at 250 kHz, 750 kHz or 1.5 MHz switching speeds, and that Class D "will never be a musically good performing amplification system" HiFi Buys 1:45:24. In the same interview, recalling taking the M Scaler to reviewers' homes, he said "If they're using Class D power amplifiers, forget it. You're not going to hear the difference." HiFi Buys 2:24:19.

The pulse array power amps

Origins, 2014 to 2016: the "digital power amp"

In 2014 he said he had been working on digital power amps "for the last ten years" and that driving loudspeakers directly was the goal article, Aug 2014. In March 2016 he called the analog power amplifier by far the biggest loss in his chain and said the digital power amp "will solve it", based on an early prototype that "had amazing depth reproduction" Head-Fi, Mar 2016. On 31 March 2016 he said the digital power amp was a project he would talk about on his blog, "even though that will end up as a product too" Head-Fi, Mar 2016.

In November 2015 he said DAVE's DX digital outputs were for his digital power amps, wrote "These are non-switching, but stillretain the analogue simplicity of two resistors and 2 capacitors plus a single globalfeedback path", and said power would run from 20 W to at least 200 W, with the first DX power amp due the next year at 20 W RMS with a 70 W mono-block option Head-Fi, Nov 2015. In June 2016 he said DAVE's output stage and second order noise shaper topology "came from the digital power amp", which is "the same except for the power supply", and that the actual output is linear (non switching) Head-Fi, Jun 2016.

2017: the name and the design

In January 2017 he described the amps as having one single feedback path from digital data to analog, only four passive components in the direct signal path, and a third order analog noise shaper structure. He said "The output stage is non-switching so timing recovery is set by the DAC", with a first model of 20 W Head-Fi, Jan 2017. In a February 2017 DAVE-thread post that does not name the product it describes, he said "No way is it Class D!", that it is linear Class A with no output reconstruction filters, and that the fastest Class D he had done was at 1.6 MHz and "would still give significant timing errors" Head-Fi, Feb 2017. In May 2017 he said the technology would be called Power Pulse Array Head-Fi, May 2017. In July 2017 he said prototypes were being built, that he had not put all the blocks together in one unit before, and that it would not be released until he was 100% happy with it subjectively and by measurement Head-Fi, Jul 2017.

2018: how they differ from Class D and from DDFA

In January 2018 he described the intent as hundreds of watts "with no loss at all in transparency and refinement", while adding "at least that is the intent" Head-Fi, Jan 2018. He said the DX amps are "just DAC's", pulse array DACs with very high power output stages and "a very clever power supply arrangement", from 20 W to many hundreds of watts, either as a dedicated power amp fed by DAVE's DX output or as an integrated amp Head-Fi, Jan 2018. In a Qutest-thread post the same week he said "it" (the post does not name the product) is very different from NAD's, "actually Qualcomm's", DDFA technology, that he invented the technology behind DDFA, and that "The key difference is that it is a non switching output" Head-Fi, Jan 2018. In 2019 he repeated that DDFA did not use pulse array because "it is a PWM switching output", whereas "My pulse array DACs are linear outputs" Head-Fi, May 2019.

In February 2018 he gave the efficiency argument: power pulse array "shares the efficiency of Class D" while the output stage can be full Class A and the signal path is completely linear, so sliding bias is not needed Head-Fi, Feb 2018. In August 2018 he said the naming was a problem because digital amps mean Class D, hence "power pulse array DAC or DX amp" Head-Fi, Aug 2018.

In May 2018 he said the Hugo TT2 output stage came from the power pulse array design Head-Fi, May 2018, and that the DX amp's earlier, more complex output stage "turned out to be not suitable" and was redesigned Head-Fi, May 2018. In a May 2018 video talk he said the power amplifier range uses "one single amplifier with a single global feedback path", the same topology as in Mojo, TT and DAVE, and that this architecture was under development Darko Audio 2:10.

On product role, in July 2018 he said the DX amps are intended as power amps although headphone drive would be available Head-Fi, Jul 2018, and in November 2018 that "headphones won't need a power pulse array DX amp - but some models will happen to have HP sockets" Head-Fi, Nov 2018. In August 2018 he said DAVE would be more transparent than the first DX amp through headphones Head-Fi, Aug 2018. In July 2018 he said that "judging from my listening tests so far" the DX amps would be a big step forward in transparency Head-Fi, Jul 2018.

2020 onward: efficiency, delays, status

In a podcast published June 2020 (recording date not stated) he said the project let him have the benefits of Class D in efficiency without a switching output stage, with "a linear output stage", capable of "hundreds of watts", "still being class A", using "a class H approach in terms of the power supply"; he hoped for products "in the next year or so" Part-Time Audiophile 26:10Part-Time Audiophile 26:53. His later posts show no product had arrived by 2021 or 2022 Head-Fi, May 2021Head-Fi, Mar 2022. In July 2020 he said he had worried that signal-dependent switching activity coupled through the ground plane could limit performance, that he had "solved the potential issues of signal dependent activity, by eliminating it", and that a DAVE DX amp would take "greater than 1 year, and for sure less than 5" Head-Fi, Jul 2020. In May 2021 he said he was not happy with the first attempt, a new prototype was being designed, and he guessed production would take "a few years" Head-Fi, May 2021; the same day he described the plan as DX amps without headphone outputs for high-power loudspeaker use, fed from M Scalers or DX-output DACs Head-Fi, May 2021. In July 2021 he said there was "No timeline for DX amps" and guessed they would come after the ADC Head-Fi, Jul 2021. In September 2021 he said he had finished a prototype PCB testing one of the new concepts in power pulse array and that progress was "somewhat glacial" Head-Fi, Sep 2021. In March 2022 he said "the DX amp project did get stalled": of two technical issues from the first prototype, getting the second order analog noise shaper topology stable at high voltages had been solved and the other was being tested with another prototype, after which he could take a development proposal to Chord Head-Fi, Mar 2022. A day earlier he said the pressure to do DX amps was off him somewhat because the TT2 was adequate for driving his loudspeakers Head-Fi, Mar 2022.

In July and August 2023 he said higher powers from a single amplification DAC are possible, and that Class H could give high efficiency; he said modulating rails even linearly with Class H may carry sound quality risks and needs care Head-Fi, Jul 2023Head-Fi, Aug 2023. In February 2023 he said the DX program is for future DACs of 100, 200 or 400 watts, "So it's a future program" Dusty Audio 1:01:31. In October 2024 he said "The large power DX DAC concept is certainly not dead in my mind" and that he thinks of M scalers as digital pre-amps feeding headphone DACs or loudspeaker-only DX amps Head-Fi, Oct 2024.

What he has not said

This archive has no statement of a release date, price or final power rating for a pulse array power amp, and none after October 2024 on the project's status Head-Fi, Oct 2024. The measurements he shows for this output stage are for the Hugo TT2, not a DX amp Head-Fi, May 2018Head-Fi, Jul 2020. His 2020 podcast hedge, that feedback Class D can reach audiophile quality, does not recur in his 2023 to 2026 statements above, and this archive shows no retraction of it either Part-Time Audiophile 26:10.

Quotes

“Class D creates subjectively very large timing of transients problems, due to switching activity.”2016-08-29 Head-Fi, Aug 2016

“I need to come up with a better name than digital power amp as it sounds like Class D, and the actual output is linear (non switching).”2016-06-28 Head-Fi, Jun 2016

“it shares the efficiency of Class D - but the output stage can be full Class A, and the signal path is completely linear.”2018-02-23 Head-Fi, Feb 2018

“I would never use Class D - it's not fit for purpose to recreate the sense of real musical instruments playing with a sense of space.”2023-07-29 Head-Fi, Jul 2023

Dated positions

Caveats and open points

• The pulse array power amps were unreleased in everything this archive covers; statements about them describe intent and prototypes, not shipping products.
• Figures such as the 2018 noise floor numbers and the 2020 and 2026 switching frequencies should be checked against the recordings.
• Head-Fi, Feb 2017 (February 2017) and Head-Fi, Jan 2018 (January 2018) do not name the product they describe and carry no quoted question in this archive; context (DAVE and Qutest threads, nearby DX-amp posts) suggests the DX amps but the posts do not say so.
• In the September 2024 live stream the words "switching amp" and "not going to be the same level" are the host's; only Watts's own lines are relied on.
• His 2020 podcast statement that feedback Class D can reach audiophile quality is more qualified than his forum and 2023 to 2026 statements; this archive shows him neither repeating nor retracting it.

Topics: Power amplifier and speaker-drive design, Transient timing error from reconstruction, Noise floor modulation