What does Rob mean by depth, and what in a DAC improves or collapses it?

What he means by depth

Watts defines it plainly: “When I talk about depth it's just simply how far away it sounds” (July 2020), and he tests it with an organ recorded in a cathedral, listening only for whether one setting sounds farther away than another Head-Fi, Jul 2020. His recurring benchmark is live sound: an organ 100 to 200 feet away sounds that far away live, while reproduced it sounds at best a few feet back Head-Fi, Nov 2015MFG Audio Reviews 9:52MFG Audio Reviews 21:17. In May 2024 he wrote that lateral imagery is “pretty much perfect compared to live sound - but depth is clearly nothing like as good” Head-Fi, May 2024.

He names several cues the brain uses: the ratio of reverberant to direct sound, the high-frequency absorption of air (so brighter sounds seem closer), and, in later statements, the timing difference between direct and reflected sound Head-Fi, Dec 2015Head-Fi, Jul 2020Head-Fi, Jun 2024ecoustics 31:05ecoustics 32:34. In 2016 he also wrote that small echo cues tell the brain a sound is further away Head-Fi, Sep 2016. He presents the timing cue as his own reasoning: “this is not something in textbooks” (interview published October 2026; its transcript places the recording before May 2026, recording date not stated) HiFi Buys 1:25:41.

He insists better depth must not be added depth. Close-miked recordings with no reverb should stay flat, and he uses that as proof that Dave is not adding reverb Head-Fi, Apr 2016HIFICLUBE.NET JVH 9:00Part-Time Audiophile 23:27. He also warns that a wider soundstage usually signals worse depth: “that normally tells me that the perception of depth is truncated” (September 2018, on Hugo 2 sources), and in December 2016 he wrote that when depth is degraded “you invariably perceive a wider sound-stage” Head-Fi, Sep 2018Head-Fi, Dec 2016. In April 2019 he qualified this: better amplitude linearity makes depth better and perceived width smaller, while better filtering can improve subjective width on recordings with phase encoded Head-Fi, Apr 2019Head-Fi, Jul 2018.

Cause 1: small-signal amplitude accuracy (stated from 2014 onward)

In October 2014, on Hugo, he wrote that signal-related noise in the ground plane making a -120 dB signal read as -119 dB has “the audible effect of confusing the brains ability to perceive depth of sound-stage” Head-Fi, Oct 2014. In December 2015 he separated two noise-shaper errors: noise floor modulation, heard as smoothness and warmth (and instrument separation and focus), and fundamental (small-signal) linearity, of which he said “Now this only affects depth perception, it does not do anything else” and that “small signal non-linearity is the only factor that I have correlated with depth perception” Head-Fi, Dec 2015. In April 2016 he said depth compression in DACs has “absolutely nothing to do with ringing or time domain behaviour” Head-Fi, Apr 2016.

The Dave noise-shaper story is the core evidence he cites. Hugo's noise shapers were about 200 dB; with Dave's larger FPGA he went to 220 dB, 240 dB and onward over about 90 days (“several months” in one post), and each step gave deeper depth, ending at a 17th-order noise shaper of 350 dB Head-Fi, Nov 2015Head-Fi, Dec 2015Hifistatement 3:05. In March 2016 he qualified this: each improvement “largely” gave deeper depth, “sometimes it didn't” Head-Fi, Mar 2016. One September 2015 post gives 360 dB as the end point Head-Fi, Sep 2015; later statements give 350 dB. His description of the top step changed with time. In September 2015 he found it easy to hear: “how easy it was to hear a 330 dB noise shaper against a 360 dB one - but only in terms of depth perception” Head-Fi, Sep 2015. In November 2015 and June 2017 he wrote that he could still hear an improvement from 330 dB to 350 dB Head-Fi, Nov 2015Head-Fi, Jun 2017. In February 2023 he called that step “the first time when it started to become a struggle to hear the difference” Dusty Audio 24:38. He reported Dave's depth at the end as about four times Hugo's: “its about four times deeper with recordings that have good depth than Hugo for example” (October 2015) Head-Fi, Oct 2015Head-Fi, Nov 2015.

He hedges on why. He first suspected the noise-shaper figure was a proxy for something in the analog stage, then reported the same result with purely digital truncation noise shapers (56 bits to 24 bits) Head-Fi, Sep 2015Head-Fi, Jun 2017Pursuit Perfect System 26:51. In June 2017 he conceded his tests were not double blind and that confirmation bias is something “I can't rule that out as a possibility”, while saying repeated tests convinced him Head-Fi, Jun 2017. Later talks say he has also run blind checks: he describes handing a listener (usually his son) the toggle switch without saying which setting is which, and getting the same description of better depth Darko.Audio 29:21HIFICLUB 21:57ecoustics 39:09MFG Audio Reviews 45:45. These are spoken accounts without trial counts or scores. From this he made a design rule: every digital module must reproduce a -301 dB signal accurately Head-Fi, Mar 2016Head-Fi, Sep 2017Head-Fi, Feb 2018.

By product, he says Hugo 1 and Mojo have similar depth because they share the same 4e pulse array, noise shapers and truncators (February 2018) Head-Fi, Feb 2018. He says Hugo 2 is deeper than Hugo because of better small-signal linearity (September 2017) Head-Fi, Sep 2017. In May 2018 he said TT2 has a lot better depth than Hugo 2, which he had expected from comparing their noise shapers, but that TT2 does not match Dave's depth because “there is more to depth and transparency than just the noise shapers” Head-Fi, May 2018Head-Fi, May 2018. In an interview published October 2026 (its transcript places the recording before May 2026; recording date not stated) he called Dave “still by far the best depth resolving DAC that you can get at any price point” HiFi Buys 1:23:25.

Cause 2: timing and phase of small signals (from 2017, firmed up 2020 to 2023)

The M scaler filter in the Blu Mk2 (2017, later used in the Hugo M Scaler) broke his experience that filter tap length never changes depth: he called the soundstage opening out “massive”, heard big changes in depth between 0.5 M and 1 M taps, and said the soundstage collapses in the 2/3 M tap video mode with the same architecture Head-Fi, May 2017Hifistatement 2:57. He offered a hypothesis that the filter's roughly 1.4 seconds of data (1.438 seconds in a 2018 post) matches concert-hall reverb time (RT60), but repeatedly doubted it: “I am sceptical that this is indeed the reason” (October 2017) and “but I am not convinced this is the explanation” (December 2018) Head-Fi, Oct 2017Head-Fi, Dec 2018.

By July 2020 he wrote that phase shift that changes with amplitude, rather than with frequency, matters for depth Head-Fi, Jul 2020. In April 2022 he described the test behind this: adjusting time delays in the original Mojo noise shaper gave one setting with clearly best depth, and simulation at -301 dB showed phase-with-amplitude errors of 2 uS (worst), 100 nS, and 10 nS (best). He concluded “both timing and amplitude accuracy are both important for depth perception” Head-Fi, Apr 2022. In March 2023 he dated the discovery to about five years earlier, soon after the M Scaler work, said it improved Mojo 2's noise shapers but not those of Hugo 2 to Dave (which have no such phase error at -301 dB), and said “the ultra small signal phase shift only influences the perception of soundstage depth” Head-Fi, Mar 2023. In a podcast recorded in October 2025 (published March 2026) he instead put the discovery at “three or 4 years ago” and linked it to the timing of reflections; that dating comes ecoustics 41:49. His stated tolerance is 0.001 dB in amplitude and 0.001 degree in phase at -301 dB Head-Fi, Jul 2021Head-Fi, Sep 2023Head-Fi, Apr 2022. A January 2022 podcast transcript gives 0.002 of a degree as still audible; that figure is and should be checked against the recording Darko.Audio 25:21.

This revises his 2015 to 2016 position that depth was about amplitude only, a change he states himself: before the M Scaler, depth “was only about amplitude” (CanJam Singapore, March 2023) Amplify Audio 7:04Head-Fi, Mar 2023Head-Fi, Apr 2016. In the Quartet era he goes further: in May 2024 he wrote that hearing Quartet convinced him “transient timing was absolutely essential for giving the impression of real instruments playing in space”, and in a podcast recorded in October 2025 (published March 2026) he said locating instruments in space, particularly depth, depends on transient timing Head-Fi, May 2024ecoustics 21:04.

What he says degrades or collapses depth

• Signal-correlated noise: noise linked to the audio signal that enters the DAC's ground plane (from a computer, from FPGA switching, or from GHz-range logic noise) degrades small-signal accuracy and so depth. He says random RF noise mainly affects warmth and brightness, and “Normally RF noise does not change sound stage depth too much” (June 2014, Hugo) Head-Fi, Jun 2014Head-Fi, Oct 2014Head-Fi, Nov 2017Head-Fi, Feb 2019Head-Fi, Nov 2019. He recommends galvanic or optical isolation and choosing the source that sounds warmer with deeper depth Head-Fi, Sep 2014Head-Fi, Sep 2018Head-Fi, Aug 2020.
• Analog parts: metal-to-metal contacts, solder joints and passive components add small-signal nonlinearity he says he could hear but, with one power-amp exception, not measure, and “no analogue component can ever be perfect in small signal terms” (March 2016) Head-Fi, Oct 2015Head-Fi, Jul 2015Head-Fi, Mar 2016Head-Fi, Aug 2022. Analog or added passive volume controls degrade depth, which is why his DACs use digital volume Head-Fi, Mar 2016Head-Fi, Mar 2020Head-Fi, Feb 2022.
• Truncation, dither and DSP: on Blu 2's truncators, “TPDF had the worst sound quality with a reduction in depth” (February 2018), with Gaussian dither about halfway to the 768 kHz noise shapers; in June 2024 he gave the same ranking, rectangular worst, then triangular, then Gaussian, with noise shaping best for depth Head-Fi, Feb 2018Head-Fi, Jun 2024. He warns that software volume and DSP that truncate degrade depth, though he expects the loss to be smaller than with an analog volume control Head-Fi, Mar 2020Head-Fi, Feb 2020Head-Fi, Feb 2018. He said in 2022 that Mojo 2's UHD DSP has no loss in depth MFG Audio Reviews 42:27.
• DAC architecture: he says DSD64's noise shaper (about 120 dB) cannot resolve the small signals depth needs and sounds flat; R2R DACs have large low-level errors from resistor mismatch; and chip-DAC noise shapers are about 140 dB Head-Fi, Sep 2015Head-Fi, Aug 2017Head-Fi, Apr 2016Head-Fi, Nov 2015Passion for Sound 11:25.
• Noise floor modulation: in 2015 he tied it to smoothness, separation and focus rather than depth Head-Fi, Dec 2015, but in a 2022 talk transcript he said that with it present “The sound stage collapses, it sounds flat and unengaging”, adding that small signals lose amplitude accuracy MFG Audio Reviews 30:09. He separately says dynamic range is not the cause: “People often make the comment that a black background is due to lower noise floor and actually that's not the case.” (November 2016) Head-Fi, Nov 2016.

Beyond the DAC

On headphones he says early reflections from the cups make sounds seem close, and that crossfeed on Hugo 2, TT2 and Dave greatly improves depth Head-Fi, Sep 2017Head-Fi, Mar 2017Head-Fi, Jun 2019Head-Fi, Jun 2024. He says the ADC is now the weak link Head-Fi, Dec 2015Head-Fi, Mar 2022The Source AV 20:30. From his own Blumlein recordings (March 2022), a cello about 5 meters from the mic sounded about 2.5 meters away on Hugo 2 with Stealth headphones, about 50%, with an unnamed scaler and Dave giving around 66% of the real depth (“perhaps better but no more than 75%”) Head-Fi, Mar 2022. Talk transcripts give other figures: in September 2022, a cello 4 meters away heard at about 2.5 meters on Dave with the prototype of a new scaler and about 2 meters on Hugo 2; in the podcast recorded in October 2025 (published March 2026), about 2 meters on a standard Dave and 2.5 to 3 meters with Quartet and Dave The Source AV 18:49The Source AV 19:00ecoustics 45:04. These transcript figures should be checked against the recordings.

What he has not said: this archive has no published result of the blind ADC depth test he planned in 2017 with two Davina units, no measurement that captures depth directly, and no explanation he accepts for why errors at these levels are audible Head-Fi, Jun 2017Head-Fi, Mar 2016Head-Fi, Jul 2021. In an interview published October 2026 (recording date not stated) he said two test recordings had been made through Quartet's pulse array ADC and that he had expected it to give better soundstage depth; in that passage he reports a surprise improvement in vocal pitch but states no depth result Distorsiuni Interioare 11:57Distorsiuni Interioare 12:51.

Quotes

“When I talk about depth it's just simply how far away it sounds”2020-07-09 Head-Fi, Jul 2020

“small signal non-linearity is the only factor that I have correlated with depth perception”2015-12-01 Head-Fi, Dec 2015

“both timing and amplitude accuracy are both important for depth perception”2022-04-07 Head-Fi, Apr 2022

“transient timing was absolutely essential for giving the impression of real instruments playing in space”2024-05-04 Head-Fi, May 2024

Dated positions

Caveats and open points

• His listening tests are mostly sighted; in 2017 he said he could not rule out confirmation bias (Head-Fi, Jun 2017). Later he describes blind checks with his son or another listener, but gives no trial counts or scores.
• The final Dave noise-shaper figure is 350 dB in most statements but 360 dB in one 2015 post (Head-Fi, Sep 2015).
• His position changed: amplitude-only through 2016, amplitude plus phase or timing after the M Scaler, as he states himself. His own dating of the phase discovery varies (about 2018 in a 2023 post, three or four years before an October 2025 podcast recording, per its transcript).
• Distance figures from his recordings (meters, percentages), the 0.002 degree phase figure and the podcast's dating of the phase discovery come partly; check the recordings.
• His explanations for why the M Scaler improves depth (RT60, correlation in the brain) are offered as hypotheses he doubts.
• No depth result from his pulse array ADC recordings appears in this archive as of October 2026.

Topics: Depth and soundstage, Small-signal accuracy and resolution, Noise floor modulation