Does hi-res audio sound better than CD, and should I buy the 96k or the 44.1k release?

2014 to 2015: the original Hugo narrows the gap

In May and June 2014, writing about the original Hugo, Watts said higher sample rates have innately better timing but let in more high-frequency noise from ADC noise shapers, which he said causes noise floor modulation and a harder or brighter sound Head-Fi, May 2014Head-Fi, Jun 2014. Because Hugo's 26,000-tap WTA filter shrinks the timing problems of 44.1k, he said the timing benefit of higher rates gets smaller while that noise downside becomes more substantial Head-Fi, Jun 2014. He limited this to DACs with WTA filters: "if you are using regular DAC's without a WTA filter, then higher SR recordings should always be better than the same recording but decimated to a lower SR" Head-Fi, Jun 2014.

The same day he summarized: "if the timing is perfectly reconstructed in a 44.1 recording, there is no benefit in using higher sample rates", adding that "In practical terms, some recordings will be better, some worse" and that "some high res are merely up sampled, and they sound awful" Head-Fi, Jun 2014. Later that month he wrote: "With a decent enough interpolation filter then timing is restored, then you no longer need high sample rate recordings - good old CD is good enough." Head-Fi, Jun 2014. On whether to sample-rate convert 24/192 files in the computer before Hugo, he advised "Give Hugo access to native, you can't go wrong", while saying he would be intrigued if someone tried binary decimation such as 192 kHz to 48 kHz Head-Fi, Jun 2014.

In April 2015 he wrote: "No Hugo does not sound better with normal 44.1/16 against 24/192 - but the difference is very, very much smaller than other DAC's", so "you can choose a recording for the musical performance not the SQ" Head-Fi, Apr 2015. Days later he described buying a 96k 24-bit Chandos release that was the original master and comparing it with the 44.1/16 version: "Frankly I was surprised how small the difference was", the master being "a bit smoother, more transparent", and he concluded "the recording is way more important than the format" Head-Fi, Apr 2015. In March 2015 he had flagged the case where the answer differs: "if the master recording was 96kHz, and they used a poor sample rate converter (SRC) to get 44.1 kHz then no amount of increased taps will solve the SRC distortion and noise" Head-Fi, Mar 2015.

2016 to 2017: Dave, the M Scaler, and the recording first

In March 2016, in the Dave thread, he said "generally, all things being equal, higher SR is preferable - but not by much" Head-Fi, Mar 2016. In May 2016: "Generally speaking, a higher rez recording will sound better than redbook - but its by far not the most important factor", and he said a firm answer on "how much redbook loses" would wait for test recordings from Davina, his ADC project Head-Fi, May 2016. In February 2016 he described his experience of hi-res as mixed: "Sometimes it can sound much worse, usually a bit better than redbook, sometimes it can sound extraordinary." Head-Fi, Feb 2016.

In October 2016 he separated his two roles: technically, "For sure 768 kHz or 705.6 kHz 24 bit is the best", but as a listener his full 4TB drive is "almost all redbook", so he needs 16-bit 44.1 to work acceptably and small file sizes to work as well as possible Head-Fi, Oct 2016.

After the M Scaler arrived (first in Blu Mk2, January 2017) he described it as a leveler between formats. He wrote that with it his high res downloads were also better, "but not like 16 bit", and that "sound quality is now almost completely dependent upon the recording process and not the actual sample rate used" Head-Fi, Jan 2017. Two days later: "No I don't think 384/24 is indistinguishable with 48/24 or certainly not 44.1/16 - but my feeling is that the standard is much less important than the recording method", citing 1960s Decca Red Book recordings that do things he said modern 192/24 recordings cannot Head-Fi, Jan 2017. Later that month: "Yes HD recordings benefit too", while "the format is not important as the original recording is way more important" Head-Fi, Jan 2017.

His explanation for why hi-res gains are modest: in January 2017, with a conventional interpolation filter, "doubling the sample rate is only halving the peak error", so with conventional filters even 192k recordings stay far from the better-than-16-bit accuracy he claims for the M Scaler, and "this is why HD recordings do not guarantee great sound" Head-Fi, Jan 2017. By December 2017 he put it as "doubling the sample rate only improves the area of the timing error by a factor of four (2 bits)", against roughly a 16-bit improvement from an M Scaler over a conventional filter Head-Fi, Dec 2017. He used the factor-of-four framing again in 2019 and 2021 posts Head-Fi, Jun 2019Head-Fi, Apr 2019Head-Fi, Sep 2021, in a podcast published June 2020 (recording date not stated) and in a CanJam NYC 2022 seminar (held in late February 2022, show dates unverified; published March 2022) Part-Time Audiophile 39:38MFG Audio Reviews 52:02.

Why he says hi-res helps: timing, not ultrasonics (2017 to 2021)

In February 2017 he wrote that "most people claim that higher sample rates are better because of ultrasonics - and I completely disagree with that", while calling his answer conjecture and adding "I could be wrong about that" Head-Fi, Feb 2017. In January 2019: "Now I do not think any harmonics above 20k are audible at all; the benefit of HD recordings from the SR POV is the reduction in timing uncertainty" Head-Fi, Jan 2019. In June 2019: "There is no real evidence that anything above 20 kHz is audible", with a parenthetical exception for intermodulation in air that a microphone would capture anyway Head-Fi, Jun 2019. In a Passion for Sound interview (Part 4, published January 2020; the series began publishing in December 2019 and the recording date is not stated) he said he can no longer hear above about 15 to 16 kHz, and "it's not about perceiving the sine waves. It's about the timing." Passion for Sound 1:03. In June 2021, in the Hugo TT 2 thread, he wrote that "at this stage there is zero evidence that we need signals above 20 kHz - but what we do need are transients that have the same timing as the original, with zero uncertainty" and that "HD recordings help that process, but are not innately essential to do this", adding that "my conclusions are tentative" until his ADC project can test whether content above the audio band matters Head-Fi, Jun 2021.

2018 to 2023: buy the original-rate release

In February 2018, in the Dave thread, he said non-integer conversion such as 96k to 44.1 cannot be done without measurable distortion and noise and creates timing issues, "so whenever I see the original recording is 96k, I will buy that over the 44.1." He also called today's half-band decimation filters (96k to 48k, say) "pretty poor" and ended: "I still buy the original master recording over a integer decimated or SRC version." Head-Fi, Feb 2018. In June 2018, in the Qutest thread, he separated the two factors: "I am not bothered too much by the lower sample rate, so long as it is binary multiples, but I am bothered by sample rate conversion", giving the rule "if it is a 96k original, buy that rather than the 44.1 version" Head-Fi, Jun 2018. The next day: "I will always buy the 96 in preference to the 44.1, mainly because of the sample rate conversion issue", adding that most of his purchases are classical from Chandos and Presto Classical Head-Fi, Jun 2018.

In September 2018: "Yes go for the original bit perfect master, particularly if it's 96k", with the qualifier that "an M scaled at 705k from 44.1 will beat a 96k source that has not been M scaled" Head-Fi, Sep 2018. In April 2019: "I too always buy the original file, particularly when its an original 96k that gets SRC to 44.1k." Head-Fi, Apr 2019. In September 2020: "So if a download of 96/24 is available I always buy that against the 44.1/16", adding that "when I have bought both copies the difference in SQ was worth the extra cost", and to avoid "the distorted MQA version, nor an upsampled or re-processed version" Head-Fi, Sep 2020. In May 2020 he said Presto's HD files are authentic, judging by FFT, and that an FFT shows whether a file is upsampled Head-Fi, May 2020.

On bit depth alone, in the same September 2020 post: "As to a 44.1/16 against 44.1/24 - the 24 bits will sound better (better depth), but whether it's worth the extra cost for a relatively small improvement in SQ is more of a dilemma." Head-Fi, Sep 2020.

On integer ratios, in June 2019 he said a 96k file preserves timing better than a 48k one, and "a 96k recording will sound better than the 48k version, simply because of the poor quality of decimation filters", meaning the half-band filters in modern ADCs Head-Fi, Jun 2019.

In March 2023, in the Mojo 2 thread: "I haven't said it obviates the need for hi-res, just that the WTA filter reduces the gap from integer higher sample rates". He went on: "My advice is always to use the recording that is closest to the original", and "No amount of WTA filtering will recover the damage done from non integer sample rate conversion." Head-Fi, Mar 2023. In August 2023 he compared his 16-bit 44.1k copy of Johann Johannsson's Arrival with a 48k 24-bit version on Qobuz: "The difference was extraordinary - it sounded like a completely different album", which he attributed to non-integer conversion from 48k to 44.1k, a common practice with film soundtracks Head-Fi, Aug 2023.

How much better: 2019 to 2022

In June 2019: "Sure the HD versions are generally better, but not enormously so", and "for sure 192k alone against an M scaled RBCD, the RBCD would easily win from a SQ POV" Head-Fi, Jun 2019. In a podcast published in June 2020 (recording date not stated) he said Red Book "is capable of extraordinarily good sound quality", called the difference between 192 kHz and 48 kHz "actually quite subtle", and said of 192 against 768 that "768 is a massive change"; he also said a 44.1 kHz recording M scaled would sound better than a 192 kHz recording without an M Scaler, though the transcript is garbled at that point Part-Time Audiophile 39:21Part-Time Audiophile 40:28Part-Time Audiophile 41:12.

In September 2021: "This is why higher sample rates sound better, as the area of the transient errors are getting much smaller." Head-Fi, Sep 2021. In July 2022, in the Hugo M Scaler thread, comparing 16/44 and 24/192 under M scaling, he said "you will hear a bigger relative change in going from 16/44", but "in absolute terms the 24/192 M scaled will be better", assuming the 24/192 is the original master; his stated goal is "to get 24/48 to sound as good as 24/192" Head-Fi, Jul 2022.

The noise downside of hi-res files (2017 to 2025)

He has said that hi-res files carry ADC noise-shaper noise above the audio band, which his HF filters remove. Of Hugo 2's green (HF filter) setting in March 2017: "The green makes a very big difference with 96 and 192 - its easy to hear - but it is very much smaller with 44.1 as there is almost no noise to filter out" Head-Fi, Mar 2017. In June 2019: "there is a serious disadvantage to high res in that the ADC noise present, which isn't there with RBCD", which the HF filter on his DACs reduces Head-Fi, Jun 2019. Of the Hugo 2 filters in March 2022: "So the biggest benefit is with HD recordings, but I still hear a small change when using the HF filter with 16/44.1." Head-Fi, Mar 2022. In March 2025, describing Mojo 2's UHD DSP filter: "The sound quality benefits of bandwidth limiting is substantial, due to reduced noise floor modulation from the random HF content from HD recordings" Head-Fi, Mar 2025. Two days later he wrote that for the best sound quality "you need to very aggressively remove HF content from the source (particularly with HD recordings)" Head-Fi, Mar 2025.

Native 768 kHz recordings (2017 to 2026)

In June 2017 he said a private native 768k test recording "sounded just like an M scaled recording", unlike the DXD recordings he had heard Head-Fi, Jun 2017. In March 2019 he wrote "All the 352.8/384k recordings I have heard sound little better than 44.1", while allowing "That may be down to the recording company or the ADC used rather than the benefits of 768k recording"; the day before he had written that "an M scaled rbcd file sounds just like a native 768k recording" Head-Fi, Mar 2019Head-Fi, Mar 2019; in April 2019 he said he had only one 768k recording Head-Fi, Apr 2019. In January 2022: "768 kHz does seem to be the sweet spot for recordings" Head-Fi, Jan 2022. In a podcast recorded in October 2025 and published in March 2026 he said he had "listened to a few 768 kHz recordings" and that they sound as if played through an M Scaler ecoustics 26:03.

Most recent statements (2024 to 2026)

In July 2024 he wrote: "Today, my new music comes via Qobuz as PCM - a lot being 96/24 or 192/24." Head-Fi, Jul 2024.

In a podcast published in October 2026 (recording date not stated), asked whether CD is good enough, he said: "Is it good enough to enjoy music on? Absolutely.", noting he has "lots of 44.1 16-bit recordings" that sound really good; but "If you want something to close the gap Between live unamplified sound and a reproduced sound, C D is not good enough". He pointed to the 768 kHz recordings he says they have been making, called 768 kHz at 24 or 32 bits the future, and stated "our mastering standard is 768 kHz, 32 bits". Check the recording for exact wording Part-Time Audiophile 48:08Part-Time Audiophile 48:52Part-Time Audiophile 49:27Part-Time Audiophile 49:31. In an interview published in May 2026 (recording date not stated), asked about the future of high resolution audio, he described Chord's pro ADC with 768 kHz and 192 kHz recording and said "What we want to do is for studios to adopt 768 kHz as the standard" Audio Force 16:46.

What he has not said: this archive contains no published result of the test he has long said would settle how much Red Book loses. In 2016 he said Davina test recordings would tell him "how much redbook loses" Head-Fi, May 2016; in January 2017 he described the plan to make 768 kHz recordings, decimate them to 48 kHz and M scale them back for comparison Head-Fi, Jan 2017; in April 2019 he said he had not yet done it but now could Head-Fi, Apr 2019; in May 2021 it was still his goal Head-Fi, May 2021; and in a Soundnews interview recorded in autumn 2026 and published in October 2026 he said they now have the ability to run that comparison, without reporting a result article, Oct 2026.

Quotes

“So my rule of thumb is if it is a 96k original, buy that rather than the 44.1 version.”2018-06-09 Head-Fi, Jun 2018

“the benefit of HD recordings from the SR POV is the reduction in timing uncertainty, and this is just a function of the sample rate”2019-01-26 Head-Fi, Jan 2019

“Sure the HD versions are generally better, but not enormously so.”2019-06-13 Head-Fi, Jun 2019

“My advice is always to use the recording that is closest to the original”2023-03-17 Head-Fi, Mar 2023

Dated positions

Caveats and open points

• Most statements come with his own DACs in mind; his 2014 comparison was made on the original Hugo, and he said that with DACs lacking a WTA filter higher sample rates should always be better.
• His advice to buy the 96k concerns cases where 96k is the original master and the 44.1k is made by non-integer sample rate conversion; he said he is not much bothered by a lower rate in binary multiples.
• The June 2020 passage comparing M scaled 44.1k with 192k is garbled in the transcript.
• The planned 768 kHz to 48 kHz decimation and M scaling test that he said would show how much Red Book loses has no published result; as of an autumn 2026 interview (published October 2026) he said they could now run it.
• His early 2017 framing (doubling the sample rate halves peak error) and later framing (it quarters the area of the timing error) are both his wording, reported as given.
• His February 2018 post says he buys the original master over an integer-decimated version too, while his June 2018 post says he is not much bothered by a lower rate in binary multiples; both are reported as given.

Topics: Sample rates, hi-res and bit depth, Hearing limits and hearing health