Which kinds of error does Rob consider audible, and which does he consider benign?

The organizing idea

In December 2015, writing about DAVE, he said his work was centered on "finding out which distortions and aberrations are critical to the brain being able to make sense of what is going on audibly" Head-Fi, Dec 2015. In March 2016 he gave a rule of thumb for his design listening: "if I can measure it, and it is signal dependent error, then its audible" Head-Fi, Mar 2016. In November 2021 he wrote that the brain "can deal with simple distortions" but not digital errors "like noise floor modulation, small signal accuracy, or transient timing errors" Head-Fi, Nov 2021. In April 2022 (DAVE thread) he put the split this way: "The brain has adaptive coping mechanisms to deal with simple harmonic distortion", but "it can't deal with small signal distortions, nor noise floor modulation, anharmonic distortions or transient timing distortions" Head-Fi, Apr 2022.

Treated as benign or low-impact

Fixed noise (hiss). In April 2014, on Hugo, he wrote that "fixed noise from jitter sources is not audible as such (in terms of SQ)" and that noise floor modulation is "much more damaging" Head-Fi, Apr 2014. In October 2015, on pulse array and jitter, he wrote "a fixed noise is subjectively unimportant" Head-Fi, Oct 2015. In April 2016, about DAVE, he wrote "if you can't hear the hiss, then it has no bearing whatsoever" Head-Fi, Apr 2016. In January 2018, in the Qutest thread, he wrote "Fixed resistor noise is just hiss; it does not actually degrade the SQ" Head-Fi, Jan 2018. At the Nintronics open day in November 2018 (talk published December 2018) he said "a fixed amount of noise is unimportant because it's just a background hiss" Pursuit Perfect System 9:36. On Hugo TT 2 he wrote in a post dated about 23 April 2019 (estimated from the post ID) that residual noise in either gain setting "will be inaudible through a power amp" Head-Fi, Apr 2019, and in a post dated about 22 February 2020 (also estimated) "It's only when the noise modulates with the signal that fatigue can set in" Head-Fi, Feb 2020. In March 2021, comparing TT 2 with another maker's DAC, he wrote that "SNR or residual noise of -127 dB is perfectly fine, and will not create SQ problems - so long as the noise is fixed", that he could lower the noise but "to do so would increase distortion", and that "127 dB is inaudible" Head-Fi, Mar 2021. The same day he wrote "residual noise that is static will not" change the sound quality Head-Fi, Mar 2021. At CanJam NYC in February 2022 (video published March 2022), on Mojo 2, he said "A tiny fixed noise is perfectly okay" MFG Audio Reviews 5:54, and in July 2022, on DAVE, "Noise is irrelevant if it's inaudible and fixed" Head-Fi, Jul 2022.

Simple harmonic distortion: coped with, not harmless. At CanJam RMAF in October 2017 (talk published December 2017) he said "the brain is used to dealing with straightforward harmonic distortion" because the ear generates it RMAF 12:07. In April 2018 (DAVE thread) he wrote that the brain is "used to dealing with simple harmonic distortion (the ear is very non-linear as a transducer)", and that with Blu Dave you can play a 1930s mono recording and enjoy it "even though you can hear huge levels of distortion and noise" Head-Fi, Apr 2018. In that post the point is that the brain copes with such distortion while still hearing it Head-Fi, Apr 2018. In July 2022, on DAVE, he wrote that a -198 dB power-supply error "would induce simple harmonic distortion only, you could consider it as inaudible" Head-Fi, Jul 2022. In the April 2022 post he added that the brain coping "does not mean that transducer distortion is not important" and "I always get worried when a headphone has more than 0.1% distortion at low frequencies" Head-Fi, Apr 2022. Deliberately added second harmonic he treats as a coloration: in August 2014 adding it for softness was "fine, whatever suits you - but it's not what I try to do" Head-Fi, Aug 2014; in March 2016 added low-frequency second harmonic "will give a false impression of a soft warm bass" Head-Fi, Mar 2016; and in September 2023 (DAVE thread), of processing that creates second harmonic only, "I don't like what it does as it damages instrument separation and focus" Head-Fi, Sep 2023. In a podcast published June 2020 (recording date not stated) he said "You certainly don't want odd order harmonics" because "they sound worse than even" Part-Time Audiophile 32:35.

Constant (linear) phase or delay, and frequency response. In October 2016 (DAVE thread) he described phase errors and timing differences with frequency response as "linear errors, and are inconsequential" Head-Fi, Oct 2016. In July 2017 he wrote of conventional non-linear phase, "I have no evidence that this type of behavior is important subjectively (unless it's gross in which case FR would be affected too)" Head-Fi, Jul 2017. In June 2019 he wrote "A fixed timing error is not important - so a shift of 1us at 5 kHz say is inaudible", while a constantly changing one is audible Head-Fi, Jun 2019. In July 2020 he wrote that he had no evidence conventional phase non-linearity is important and was "open minded or agnostic on it"; in an edit he added "I have evidence that phase shifts, so long as they don't change with time or amplitude, are not very audible", citing a high-frequency filter swap that "sounded pretty much the same" Head-Fi, Jul 2020. In January 2021 he wrote "As to phase accuracy - I am not convinced this is important", unless there are strong frequency response changes Head-Fi, Jan 2021. In May 2021 (M Scaler thread) he wrote that phase or frequency response errors are "not subjectively damaging" Head-Fi, May 2021, and in July 2021 he had "no evidence that a phase shift that is constant (that is does not vary with amplitude or time) degrades depth perception" Head-Fi, Jul 2021. Later statements are firmer. In June 2024, on Mojo 2, he wrote "linear changes don't matter": with his bandwidth-limiting FIR filter from the ADC project in place, switching in the Mojo 2 high-frequency filter gave "no change whatsoever" Head-Fi, Jun 2024. In February 2025 (DAVE thread) he described a similar test and wrote that "a -1dB or so at 20k with attendant phase shifts are inaudible (at least to my old ears)" Head-Fi, Feb 2025. In an interview published October 2026 (recording date not stated, though the transcript places it before May 2026) he said "A linear phase shift, in my experience, is inaudible" HiFi Buys 1:58:09 and that a transient timing offset that is "rock solid, even if it's wrong in a linear fashion or a consistent fashion, that won't matter" HiFi Buys 1:58:18. On rooms, he wrote in March 2021 that the brain "has evolved to deal with room acoustics" but "can't deal with noise floor modulation, aliasing and small signal errors" Head-Fi, Mar 2021; in March 2016 he wrote that frequency response "plays a major role with transducers, not so with electronics" Head-Fi, Mar 2016.

Pre-ringing of interpolation filters. In March 2025 he wrote that any ringing from an interpolation filter is "at an inaudible 22kHz (for 44.1k sample rate)", so "pre-ringing for an interpolation filter isn't an issue at all", but that pre-ringing from filtering in the audio band, such as EQ, would be audible Head-Fi, Mar 2025.

Source jitter, in particular designs. He said in June 2014 that on Hugo "the jitter levels does not matter, as everything gets re-timed against a low jitter clock" Head-Fi, Jun 2014, and in March 2019, of Qutest, that "source jitter certainly is irrelevant", with RF and correlated noise the remaining concern Head-Fi, Mar 2019. These are statements about his designs, not about jitter in general.

Treated as audible

Noise floor modulation and other signal-dependent noise. In April 2014 he called noise floor modulation much more damaging than fixed noise Head-Fi, Apr 2014. In December 2015 he wrote "With gross noise floor modulation you hear it as grain in the treble" Head-Fi, Dec 2015. In March 2018 he wrote "it does not matter how small it is, careful listening tests exposes it as very audible", and that in small amounts it adds brightness Head-Fi, Mar 2018. In May 2020 he wrote that his listening tests showed it from -180 dB to -200 dB as "easily audible" and called its audibility "the only contentious issue" Head-Fi, May 2020. In March 2021 he described a signal-dependent artefact at -350 dB, seen in simulation, that was "upsetting instrument separation and focus, timbre variation and clarity"; he said how it could be audible "is beyond me", reported a blind test with his son, and attributed the problem to "correlated noise (that is signal dependent noise)" rather than fixed noise Head-Fi, Mar 2021. In talks he said at CanJam NYC in February 2022 (video published March 2022) that it "causes enormous subjective problems" MFG Audio Reviews 5:54 and at CanJam Singapore in March 2023 that residual noise floor modulation below measurable levels is "still audible" Eric's Hi-Fi 32:01. In an interview published May 2026 (recording date not stated) he contrasted it with harmonic distortion, which the brain has "mechanisms" to compensate for: "The problem with noise floor modulation is it's not harmonically related", and instruments come out "brighter than they should be" Audio Force 12:18Audio Force 11:27. In the HiFiBuys interview published October 2026 (recording date not stated) he said it is "when noise varies with signal" that it becomes a major issue HiFi Buys 52:21.

Small-signal errors (depth). In January 2016, on DAVE, he wrote "it does not matter how small the error is it will have an impact" on depth perception Head-Fi, Jan 2016. In September 2016 he wrote that he "could (maybe) be persuaded" that going from a 200 dB to a 300 dB noise shaper was audible, thought 300 to 340 dB "surely inaudible", but "with careful listening I can indeed perceive the change" Head-Fi, Sep 2016. In April 2022 he agreed that "the idea that we need accuracy at -301dB is plainly nuts", but said the evidence of "hundreds of listening tests" is that "any distortion, however small, has audible consequences on depth perception", and that he follows "the evidence, not what my brain tells me as being sensible or not" Head-Fi, Apr 2022. At CanJam NYC in February 2023 (video published March 2023) he said that for DAVE going from 330 dB to 350 dB noise-shaper performance still gave "a tiny improvement in depth perception"; readers should check the figures against the recording The Headphone Show 44:02.

Timing errors that vary with amplitude or with sampling. In October 2016 he wrote that when a transient's timing depends on how it was sampled, "this sampling dependent uncertainty is highly audible" Head-Fi, Oct 2016. In January 2017 he described a change in his own thinking: "I used to think 1uS errors would be inaudible, now I take the view that any timing error" of any size "is important" Head-Fi, Jan 2017. In July 2020 he said phase shifts of a few nS that vary with amplitude were audible in his listening tests Head-Fi, Jul 2020. In February 2023 he called transient reconstruction the most important of three issues, the others being small signal accuracy and noise floor modulation Dusty Audio 18:36, and at CanJam NYC in February 2023 (video published March 2023) he said of transient timing "it doesn't matter how small the errors are, they are perceivable" The Headphone Show 40:09.

Higher-order and anharmonic distortion. At CanJam RMAF in October 2017 (published December 2017) he said higher-order distortion is very audible and "Anharmonic distortion is very audible", because the brain cannot subtract it as it does the ear's own harmonics RMAF 12:07. In January 2018, in the Qutest thread, listing the metrics a design of his must pass, he wrote that anharmonic distortion is "much more important than harmonic distortion" Head-Fi, Jan 2018. At CanJam NYC in February 2023 (video published March 2023) he said "Higher order distortion is a massive problem" The Headphone Show 40:57. In January 2025 he wrote "higher order odd harmonics are extremely nasty" and "higher order harmonics can degrade transient timing too" Head-Fi, Jan 2025. In the HiFiBuys interview published October 2026 (recording date not stated) he said "the brain can't deal with anharmonic distortion because it doesn't happen naturally" HiFi Buys 39:45.

Aliasing and out-of-band artifacts. In April 2016 he wrote "The conventional view is that aliasing products in the 20-22.05 kHz range is inaudible. But it is not", because aliasing degrades the timing of transients Head-Fi, Apr 2016; in August 2017 he made the same point about ADC aliasing Head-Fi, Aug 2017. In June 2019 he wrote that an out-of-band artifact "can adjust the timing of transients", so that components inaudible on their own become audible Head-Fi, Jun 2019. In January 2021, from ADC work, he wrote that aliasing "if it's observable on a digital domain measurement, is easily audible" Head-Fi, Jan 2021, and in May 2021 that "aliasing at -350dB is audible" Head-Fi, May 2021.

Measurement and audibility

In August 2014 he wrote "There are plenty of distortions that have sonic consequences but are below the current ability to measure" Head-Fi, Aug 2014. His March 2016 rule of thumb was limited to error that is measurable "and it is signal dependent" Head-Fi, Mar 2016. In November 2021 he wrote "I am certainly not saying that measurements are all important, but they do cover gross deficiencies" Head-Fi, Nov 2021. At CanJam NYC in February 2023 (video published March 2023), answering "what levels of distortion is audible?", he said "Generally, if I see something that's measurable, then it will have an audible consequence" The Headphone Show 40:57. Fixed noise is the class he exempts: it is measurable, yet he calls 127 dB of it inaudible Head-Fi, Mar 2021.

What he has not said

He gives no level below which noise floor modulation, small-signal error, aliasing or varying timing error becomes inaudible; for those classes he says the opposite Head-Fi, Mar 2018Head-Fi, Jan 2016Head-Fi, Jan 2017Head-Fi, May 2021. The figures he ties to inaudibility are 127 dB of fixed noise Head-Fi, Mar 2021 and a -198 dB error producing only simple harmonic distortion Head-Fi, Jul 2022. This archive does not give a single ranked list of every error class from most to least audible.

Quotes

“it does not matter how small it is, careful listening tests exposes it as very audible”2018-03-11 Head-Fi, Mar 2018

“SNR or residual noise of -127 dB is perfectly fine, and will not create SQ problems - so long as the noise is fixed.”2021-03-29 Head-Fi, Mar 2021

“A fixed timing error is not important - so a shift of 1us at 5 kHz say is inaudible; but a 1us shift that is constantly changing is audible”2019-06-14 Head-Fi, Jun 2019

“Anharmonic distortion is very audible.”2017-10 RMAF 12:07

“A linear phase shift, in my experience, is inaudible.”2026-10-02 HiFi Buys 1:58:09

Dated positions

Caveats and open points

• These are Watts's own listening-test claims; this archive contains no independent measurement or blind-test data, and his blind test with his son is a statement, not data.
• The dates of Head-Fi, Apr 2019 (about 23 April 2019) and Head-Fi, Feb 2020 (about 22 February 2020) are estimated from post IDs.
• Several posts are in the general Watts Up thread and do not name a product; the product is left blank in those positions rather than assumed.
• The RMAF talk was given in October 2017 and published as video in December 2017; the Nintronics talk was given in November 2018 and published in December 2018 (sources.json).
• CanJam NYC 2022 and 2023 were held in late February of those years (show dates unverified in sources.json); the videos were published in March. The quote dates here follow when he spoke, except for the Audio Force and HiFiBuys interviews (2026) and the Occasional Podcast (2020), whose recording dates are not stated, so their dates (including the 2026-10-02 date on the in_his_words entry) are publication dates.

Topics: Error classes and which are audible, Auditory processing: how the brain hears