Chord Electronics Hugo M Scaler upsampling digital processor

The idea of using digital signal processing (DSP) to convert digital audio data sampled at 44.1kHz or 48kHz to a higher sample rate is not new. I first heard the beneficial effects of upsampling at Stereophile's 1998 hi-fi show in Los Angeles, where a pro-audio dCS 972 digital-to-digital processor was being used to convert 16-bit/44.1kHz CD data to a 24/192 datastream (footnote 1). So persuaded was I of the sonic improvement offered by upsampling that I bought a dCS 972 to upsample CDs to 24/88.2 to feed my then-reference Mark Levinson No.30.6 Reference DAC. As D/A processors improved, my need for upsampling faded away. (While my dCS 972 still sees some use, this is for the opposite purpose: downsampling my hi-rez recordings to 16/44.1 to produce CD masters.)

I subsequently wrote that I was convinced that the sonic improvement I heard with the dCS 972 was due to its using a different oversampling digital reconstruction filter with a different number of taps and, as a consequence, different passband ripple and stopband rejection (footnote 2). Twenty years after those words appeared in print, the Chord Electronics $4795 Hugo M Scaler arrived in my listening room, and I found myself returning to the subject of digital filters and upsampling—or upscaling, as the British company calls it.


The Hugo M Scaler is a small, almost square component powered from a supplied 15V wall wart and controlled with either a small remote or the front-panel buttons. It features five digital inputs: a galvanically isolated Type-B USB, two coaxial S/PDIF on BNC, and two TosLink optical. (DSD data is converted to PCM with a 6dB reduction in level.) The M Scaler doesn't have analog outputs, but it has three digital outputs: one coaxial BNC S/PDIF, one optical, and a pair of galvanically isolated BNC jacks that enable upsampling to 705.6kHz or 768kHz—but only when used with compatible Chord Electronics DACs. The M Scaler will work with D/A processors from other manufacturers, but only, of course, up to the maximum frequency the DAC can accept. (My PS Audio DirectStream DAC indicated it was receiving data sampled at 384kHz when I set the M Scaler to output that rate via an S/PDIF connection, but the DirectStream is limited to 192kHs and there was no sound.)


The front panel offers six of Chord's traditional glass-sphere buttons, which illuminate in different colors according to what the M Scaler has been asked to do. Only four of these are currently functional; the rightmost pair, marked "DX," are intended for a future product design. From left to right, these indicate: whether video mode or automatic input detection is selected; which input is in use; the output sample rate; and the input sample rate.

The M Scaler offers a video mode, as the upsampling introduces a latency up to 600ms, too long for synchronizing audio with video. In video mode, the M Scaler uses an interpolation filter with lower latency. For audio use, a pass-through option, with the output sample rate the same as the input rate, is provided to allow comparison to upsampled output. However, as the upsampling can introduce digital "overs"—interpolated sample values exceeding 0dBFS—both the passthrough and upsampled signals are reduced in gain by about 2.8dB, which complicates such direct comparisons.

The core of the M Scaler is a Xilinx XC7A200T FPGA (field-programmable gate array) on which runs the code for the Watts Transient Alignment reconstruction filter—named for design consultant Rob Watts—first seen in Chord's Blu Mk.II upscaling CD transport. The FPGA has 740 DSP cores; Watts's filter uses 528 of them running in parallel at 16Fs and a bit depth of 56 to achieve a filter length of an extraordinary 1,015,808 taps. For comparison, the WTA filter in Chord's DAVE D/A processor, which I reviewed in June 2017, used 164,000 taps implemented in 166 DSP cores.

In preparing the DAVE review, I asked Watts what is the advantage of using ever-longer digital filters. "If you have a conventional filter with 100 taps, you will recover some of the transient information," he explained. "A 100-tap filter gives you sufficiently good frequency-domain performance, but not in the time domain. . . . Every time you increase the number of taps, you improve the perception of pitch, timbre gets better—bright instruments sound brighter, dark instruments sound darker—the starting and stopping of notes becomes easier to hear, the localization of sounds get better. There is less listening fatigue—the brain has to do less processing of the information presented to it to understand what's going on."

Digital filters and upsampling
In the promotional literature for the M Scaler, Chord writes, "The Hugo M Scaler . . . takes the digital file and repairs it, adding back the information lost between the samples, then it sends the repaired file to the DAC. . . . With 705,600 samples per second, a huge amount of important information that was lost when creating the 44.1 digital file is now recovered. The more samples, the closer you get to the original analog signal. . . . The Hugo M Scaler in essence places 15 additional new musical samples in between each original musical sample, resulting in an astounding improvement in the recreation of the original music signal."

My eyebrows raised, I kept reading. Referring to the figure reprinted here, the text states that "The Hugo M Scaler takes a rough stairstep CD quality waveform and transforms it into a smooth analog-like waveform. That quantum leap in sampling brings a breathtaking leap in detail, accuracy and realism to your music."


Hmm. The measurements I performed to accompany our reviews of the dCS 972 and Purcell definitively showed that upsampling doesn't add information above the Nyquist frequency—22.05kHz with CD data—of the lower sample rate. So what is the M Scaler doing?

In one of the first articles I wrote for Stereophile, "Zen & The Art of D/A Conversion," which was published in September 1986, I discussed how the recovered analog signal is not directly described by the levels of the digital samples. Instead, the interaction between those samples and the impulse response of a digital low-pass reconstruction filter recreates the analog waveform—not just at the sampling intervals but between them (footnote 3). By processing the incoming data with a low-pass filter featuring an extremely long impulse response, the M Scaler makes it possible for the accompanying DAC to more accurately reconstruct the analog signal. In effect, it replaces the DAC's digital filter with its own, as the DAC's filter is now operating at the higher sample rate, and its cutoff is one or more octaves above the original data's Nyquist frequency.

Listening with the Chord Electronics DAVE
I performed most of my auditioning of the M Scaler with a sample of Chord's DAVE D/A processor, sourcing audio data from my Roon Nucleus+ server via USB and sending the upsampled data to the DAVE with a dual-BNC connection. I also used the M Scaler with my PS Audio DirectStream and Mark Levinson No.30.6 DACs, using a single S/PDIF connection. I tried using the M Scaler with my NAD M10 integrated amplifier, sending the latter data upsampled to 96kHz or 192kHz via an optical S/PDIF link. However, while the NAD would play music for a few seconds, it stuttered and then stopped. I was using the M10's Dirac room correction and wondered if that was the problem, so switched off the Dirac filter, but there was no improvement.

Footnote 1: Jonathan Scull reviewed the dCS 972 in February 1999. dCS subsequently introduced a functionally identical but cosmetically improved consumer version, the Purcell.

Footnote 2: In a letter responding to this essay, Bob Katz conjectured that because the inevitable quantization distortion is spread over a wider frequency range with upsampled data, the audibility of this distortion is significantly reduced.

Footnote 3: For a detailed explanation of how a digital filter operates, see an article I wrote in September 2018.

Chord Electronics Ltd.
US distributor: Bluebird Music Ltd.
275 Woodward Avenue
Buffalo, NY 14217
(416) 638-8207

Bogolu Haranath's picture

It would be interesting to see a comparison review of the sound quality of Chord M Scaler with the dCS Vivaldi Upsampler, when the digital data is upsampled :-) .......

PAR's picture

Although a comparison of the M Scaler with a dCS Vivaldi Upsampler may be of interest it would not be a fair one if the comparative price of each was part of the issue as your request indicates. The Vivaldi Upsampler may be more expensive but it is not only a sample rate converter ( which offers DSD rates as well as up to DXD conversion in the PCM domain), it is also a network streamer providing local file streaming, access to Qobuz, Tidal, Deezer and Spotify plus a huge range of internet radio options. It also has an excellent control app which, as well as offering navigation of the above, also provides remote control of the full Vivaldi stack.

Those features all add to the cost of the product and make the dCS and Chord products not truly comparable on a like for like basis.

Bogolu Haranath's picture

I know ..... I was wondering about the sound quality comparison between the M Scaler and the Vivaldi Upsampler, when the digital data is upsampled ........ I know Vivaldi components are the flagship digital products from dCS :-) ........

Bogolu Haranath's picture

I deleted the words 'more expensive' from my original comment :-) .......

Anton's picture

If that were 2K, I'd impulse buy it.

Looks like it would be a great toy!

MhtLion's picture

The measurement is absolutely remarkable!

Ortofan's picture

... the $4795 Chord Hugo M Scaler compare with the 'Advanced AL32 Processing Plus' - which includes upsampling and interpolation - incorporated into certain CD players from Denon, such as the $499 DCD-800NE and the $1199 DCD-1600NE?

Archimago's picture

The beauty with modern digital audio and computer audio is that much can be done as a hobbyist without much expense.

Remember that digital filtering can be done in software including through your own Raspberry Pi 3/4 streamer; there's enough processing power to get the job done upsampling to 384kHz or even 768kHz depending on the capabilities of your DAC.

If you have PiCorePlayer, here's a post from back in 2017 where you can play with this stuff for way less than the asking price - settings for Chord-like, Meridian-like, MQA-like, even NOS-like:

If you want to integrate a few concepts like reducing pre-ringing but still achieve excellent filtering, have a listen to my intermediate phase "Goldilocks" settings:

Have fun.

Bogolu Haranath's picture

It is interesting to see the impulse response of the NOS filter of HoloAudio R2R DAC and, the impulse response of the recently reviewed Metronome DAC's slow and super-slow roll-off filters, with CD resolution ....... Compare the impulse response measurements of HoloAudio, Fig.1 and Metronome DAC's impulse response measurements, Fig.2 and Fig.3 ......... Both those DAC's impulse response measurements show no ringing with CD resolution :-) ........

Bogolu Haranath's picture

Some other examples ....... If it is a short linear-phase filter like, CH precision integrated amp (Fig.18) or, short minimum-phase filter like the music/listen filter of Ayre QX-5 Twenty (Fig.1), they don't show much ringing in the impulse response :-) .........

jeffhenning's picture

I'm riffing here, but wasn't this type of spline-based upsampling available 2 decades ago for about the same price? What was that company's name?

Also, it seems that upsampling is now somewhat ubiquitous so I'm not busting on this unit for not doing a good job, but, rather, wondering why it has to cost so much today. I think Chord makes great stuff, but this seems destined to go the way of the Digital Lense de-jitter processor.

Also, why isn't there a piece of software you can run on your entire music library to upsample them to a new library of 24/96, losslessly compressed files? That would be handy. I'd buy that.

Or an audio plug-in rather than a big hunk of metal that costs about $5K?

Questions, questions, questions permeating the minds of old audiophiles today!

CG's picture

1. Wadia

2. As mentioned, SoX will do this. For free. There's other software that's more user friendly and a little less versatile that costs a bit, but not much.

A great mystery to me for the longest time was why people didn't perform all these digital machinations off line, producing a a new library of modified files as you suggest. So what if it runs overnight out for a couple days? Once complete, you could just play back your modified files without all the possible noise production caused by the processing while listening. You could use a much lower power computer for playback, too. But, very few people seem to like this approach. Best ask them why - not me.

barrows's picture

Perhaps Mr. Atkinson should consider a review of the remarkable HQPlayer software, which offers very sophisticated on the fly oversampling in a much more affordable software package. It is especially good at oversampling to high rate DSD, and there are a wealth of filter and modulator choices to choose from. Best to have a very powerful computer to run it on though!

supamark's picture

that thing rings like a bell. Couple that with the wall wart power supply for a $5k+ (with tax) device... it's a hard nope. Seriously, I've never seen a DAC output ring that bad. Yuck.

CG's picture

That could be a feature, not a bug.


Although many people believe that "perfect" reproduction leads to the best sound quality, that may not be true. Maybe, that's not how it works. I'm sure it depends on a lot of factors. I'm not even qualified to offer a good opinion on that. But, certainly, very, very few recordings are released where it's just the untouched feed from two microphones with no additional tweaking.

Bogolu Haranath's picture

See, Footnote 2 in the review ........ The 'hypothesis' is, since the ringing is spread over a wide area due to upsampling, it may not be audible :-) .......

Bogolu Haranath's picture

That ringing may also become lower in amplitude, which could make it less audible :-) .........

Currawong's picture

I think many of us have, or had, a great deal of misunderstanding about the effect of things such as ringing in digital, often as a consequence of manufacturers telling us things that were flat-out wrong. An impulse response is an illegal signal -- it is something that never occurs in music. It is equivalent to having nothing, then a single-sample pulse. Since more than 2 pulses are required (in essence) to capture a sine wave, a single pulse represents something that is does not exist in digitally encoded music. From a different perspective, it is equivalent to having frequencies far outside the bandwidth being played simultaneously with those within. The ringing response to the filter only serves to show the type of filter being used. This ringing does not occur in music playback, as impulse responses (and square waves) do not occur in natural music.

Secondly, the use of a wall-wart is the consequence of the quality of the power supply in the device. It is actually possible to make a low-noise switching power supply and/or filter out any effects of it on the device. Funnily enough, some Chord customers complained that Rob was wrong, and that there were sound differences between using the included wall-wart with the Hugo 2 and not. It turned out that the wall-wart was injecting noise back into the power board which was being picked up by other components! I think we are so used to the idea of switching power supplies being the domain of cheap products from China that we are often unaware of the underlying aspects of the actual technology and how they affect, and don't affect, electronics.

Bogolu Haranath's picture

Regarding impulse response and ringing ........ Several Stereophile reviewers reported that, when some of those filters are used in the equipment being reviewed, music sounded 'better' ......... Measurements showed lessening or elimination of that ringing, when those filters are used for testing the impulse response :-) .......

Currawong's picture

What's "better"? But regardless, it's a "correlation does not imply causation" issue. The critical point is not what you see with an impulse response, but the passband. In the Chord DACs, you'll see no ringing at the cut-off. I understand what you're saying though. I had the iFi Pro iDSD here, and the short, GTO filter brought the stereo image very forward, which was fun to listen to music with. However that stereo image isn't very accurate.

Bogolu Haranath's picture

That 'causation' problem could be due to the ubiquitous and formidable 'brick-wall filter' ......... See my mention about the AES May 2019 article, 'Modern sampling: A tutorial' by J. Angus, below ....... That article discusses about the various reconstruction filters ...... That article is open access, freely available online :-) ......

Bogolu Haranath's picture

JA1 reviewed the M Scaler with Chord Dave DAC ........ See, Dave measurements ....... Dave comes with HF filter which can be turned on and off ........ Dave does show ringing with CD resolution ........ With HF filter off, Dave does not roll-off smoothly till 384 kHz :-) ..........

Anton's picture

You made me smarter!

Thanks for the information!

SNI's picture

Spot on mr. Currawong.
The ringing is simply not there.
It is the nature of a bandlimited signal.
The impulse response should only be used to show the phase linearity and filter depth.
Finaly someone who knows a bit about digital audio.

tonykaz's picture

Chord's strongest feature is that the device isn't made in China! ( it isn't, is it? )


Schiit folks are now offering a USB advancement to their DAC Range, pricing their Top DAC around $2,500. ( which is way less than a proper Phono Cartridge, for gods sake )

It's exciting to realize that our lovely digital playback will just keep getting better in every way, while I seem to continue to be satisfied with good ole Redbook. Hmmmmmmm

Tony in S.Carolina

davip's picture

This is a question for John principally, as the author of this piece, but perhaps Archie can also pipe-in if he's willing given his wealth of experience. As an academic researcher I've become used to (indeed must) skimming manuscripts to get the main point and see if it's germane to my inquiries and merits detailed consumption. Lightly reading this out of interest (I'm a vinyl guy), one thing jumped out at me. Are you saying 1) that the brick-like process of 16/44.1 digitisation, long derided for being a poor approximation of the original musical waveform, is not really the limiting step that we all thought but it's how you put those bricks back together that's of most importance (quote: "...the recovered analog signal is not directly described by the levels of the digital samples. Instead, the interaction between those samples and the impulse response of a digital low-pass reconstruction filter recreates the analog waveform—not just at the sampling intervals but between them")? Or, are you saying 2) that, in effect, the subjective noises made by this thing are so pleasing that one can look past its (apparent, if '1' is incorrect) something-from-nothing operation (quote: "...[improve] the recreation of the original music signal," as Chord claims, the M Scaler definitely did, with all three D/A processors I tried").

I have big trouble with both of these premises (although they are mine, based on skimming your text) as both '1' and '2' imply rejection of the GIGO truism, that fillet-Mignon CAN be conjured from McDonald's (to reverse-quote another maxim from another Stereophile review), and that something CAN be had from nothing (if not For nothing, given the price). Have I missed the point, John and Archie, or is this what digital naturally devolves to -- ever-increasing levels of deception / sleight-of-hand? Better 'suspension of disbelief', if you prefer.

The first time I heard a CD after having vinyl through my formative years (that I foolishly gave away before going to Uni) I realised that one sounded like real musicians playing in my room and one was a poor, cardboardy pastiche of the real thing. My vinyl system of 40 yrs ago (STD/Hadcock/CA252/AR18s) was one that made the best of the medium and would show the door to many analogue systems today, but in the implied ever-closer representation of the 'original musical signal' I can't help but see this digital 'mathmatistry' (sensu Box, 1976) as anything but a wool-pulling con. I've been fooled once, but I won't be fooled again (as both the saying and song go).

John Atkinson's picture
davip wrote:
Are you saying 1) that the brick-like process of 16/44.1 digitisation, long derided for being a poor approximation of the original musical waveform, is not really the limiting step that we all thought but it's how you put those bricks back together that's of most importance . . . ?

Yes, though higher sample rates and greater bit depths are still better than 16/44.1k in absolute terms.

John Atkinson
Technical Editor, Stereophile

Currawong's picture

....since then, to the point I can actually enjoy listening with music played through the latest Sigma Delta converters to some degree. The issue with brick-wall filters seems to have been the limited processing available for them in the past, resulting in an unnatural sound. I am of the impression that slow-roll-off filters came about more to alleviate this than for any other reason.

I have long been a fan of old-school R2R converters which were simply pleasant to listen to music with, seeming to bring the feeling of it out, much as vinyl seems to. Knowing what I know now, they can be somewhat artificial in how they achieve this. The Chord DAVE was something else altogether, seeming to inject the music into my blood. It was as if other converters that people talk about as having a "black" background were really actually losing something in that black, whereas the DAVE could reproduce that underlying something that makes you feel the music, without having to resort to injecting even-order harmonic distortion to fake it.

My thoughts, anyway.

BillBrown's picture

I was happy to read JA's history of the subject, subjective impressions of the Chord, and measurements.

I loved the quote from David Rich. I think we are there.

I am enthusiastic about reading JA's impressions on room correction at some point. It is something I have been hoping for (esp. when I read his in-room speaker measurements).

This all aligns with my current pursuits: 1- electronics that I think are neutral (I used to be much more subjective in this regard), 2- Digital filtering- I upsample using iZotope (preferring it over SOX) in Audirvana, using a minimum-phase filter replicating as close-as-possible Ayre's "Listen" filter and love it, and 3- in room frequency response.

The AES reference from Jim Austin yesterday was timely!

I second Barrow's mention of HQPlayer. I haven't tried it as I am not entirely convinced re. DSD and don't want to worry about processor speeds, configuration, etc. Certainly some impressive filter-response graphs, though.

Thanks again,


Bogolu Haranath's picture

It would be interesting to see Benchmark make a DAC with different filters :-) ........

BillBrown's picture

I suspect that is EXCEEDINGLY unlikely :)

Bogolu Haranath's picture

Like the saying goes, 'never say never' :-) .........

hollowman's picture

JA said: It's been a long time since I last listened to the No.30.6. Feeding it the Frisell "Grapevine" with the M Scaler in pass-through mode, I was struck by the low-frequency authority and control it exerted on the double bass and kickdrum, as well as the sense of musical momentum."

JA, are you implying the No. 30.6 (stock, w/o scaler), is SUPERIOR in low-frequency authority and control / musical momentum to the PS Audio AND Chord DAC (sans M-scaler)?

John Atkinson's picture
hollowman wrote:
JA, are you implying the No.30.6 (stock, w/o scaler), is SUPERIOR in low-frequency authority and control / musical momentum to the PS Audio AND Chord DAC (sans M-scaler)?

Yes regarding the PS Audio; no regarding the Chord DAVE.

John Atkinson
Technical Editor, Stereophile

Bogolu Haranath's picture

Just a suggestion ....... As Ortofan suggested above, may be Stereophile could review the Denon DCD-1600NE CD/SACD player ($1,200), which has built-in upsampler? :-) ........

hollowman's picture

The M-scaler (or any scaler for that matter) should also be experimented with one of the NOS dacs. Particularly the "high-end" models that have been getting praise and attention lately: HoloAudio, TotalDAC; Denafrisps, AudioGD, etc.

Another issue is jargon and terminology ... whether or not this is market-driven.

For example -- regardless of no. of taps -- "scaling" "upsampling" or SRC (sample rate conversion) needs to be put into CLEAR perspective with topologically similar oversampling (aka digital filtering).
I think JA somewhat addressed this in the opening paragraphs, but (perhaps) a dedicated, Stereophile-approved treatise is needed. As was the case in the series of heavily-commented MQA blogs.

Glotz's picture

In regards to the Chord M Scaler? This unit would appear to be creating information, as well as increasing bit-depth?

John Atkinson's picture
Glotz wrote:
In regards to the Chord M Scaler? This unit would appear to be creating information, as well as increasing bit-depth?

Er . . . I thought my review showed that the M Scaler doesn't add new information. Instead, it allows a much longer digital reconstruction filter to be applied to the original data.

John Atkinson
Technical Editor, Stereophile

Glotz's picture

I was completely mistaken. I'll re-read the review! My apologies to all.

I need to stop smoking pot before posting! (Sheesh)

Bogolu Haranath's picture

Are you (JA1) aware of the AES, May 2019 article by J. A. S. Angus, which was mentioned by one of the readers on another Stereophile forum website? ......... The article title is 'Modern sampling: A tutorial' ....... The article is open access, freely available online :-) ........

Bogolu Haranath's picture

That article discusses about using different types of reconstruction filters :-) ......

shawnwes's picture

I had the opportunity to listen to one for about an hour at a local store a couple of months ago when the Chord rep was passing through giving a public demonstration. It definitely was very obvious when it was in the signal path or being bypassed however I was unfamiliar with most of the music being played & it was all multitrack rock and or overly processed electronic music so I didn't really get a frame of reference. It did seem to improve the overall soundstage & enjoyability of the music being played but I also mentioned that I'd love to hear it with some minimally miked recording like KOB, Waltz For Debbie or some orchestral recording to give it a proper frame of reference but that didn't happen. Without that reference it was hard to tell if it was just having an interesting sounding effect on the signal like "concert hall" on an AVR or actually doing something beneficial. Well worth the listen but I'd suggest taking your own music along before plunking down $5k.

hollowman's picture

From "MBL's Chief Engineer Juergen Reis on Designing DACs | Stereophile" [Jan 30, 2017]. Note what Jurgen says about no. of taps at end of video...

So, not Watts' as-many-as-possible ... more of a Goldilocks approach.

skris88's picture

No wonder there are so many turning to analogue with this kind of digital hogwash being promoted - YOU CANNOT CREATE SOMETHING OUT OF NOTHING!!! You cannot 'put back' what was never captured in the first place. Disappointed in that Stereophile has stooped so low to even discsussing such scams - just to keep their sales up.

RustyGates's picture

Its not creating something out of "nothing". 44.1KHz audio is the NYQUIST RATE used to sample the 22.05KHz limited audio band. Because of that, all the information required to reconstruct the original analogue signal from 1 Hz to 22.05KHz is there. Upsamplers "fill in the dots" by using the right mathematics to do so. The M-Scaler upsampling to 768KHz with that level of bitwise accuracy (FIR, 24-bit), then up to 2048FS (2048 times 44.1KHz or 48KHz) within the Chord DACs themselves (via a secondary 24-bit FIR to 256FS, then three stages of linear interpolation and IIR filtering to get to 6-bit 2048FS) allow the DAC noise shaper / output stage to apply a new and extraordinarily accurate reference voltage up to every ~10nS. Compare this instead to the crappy Zero-and-hold filters that NOS DACs employ, which make the DAC OP apply a reference voltage and holding for 22,675.7nS, resulting in some really bland and distorted sound. As for analogue sources, that is a load of hogwash; inconvenient and noisy.

skris88's picture

Sorry. You're entitled to your opinion but I cannot agree. It simply doesn't make sense. Is the original 24/96 data somehow still in the 16/44 release? It's just snake oil rubbish IMHO.

Of course technically you CAN up-sample, but in no way that 're-created' 24/96 output is in any way identical to the original 24/96 used to create the 16/44. It's just common sense.

In the early days many used to compare 16/44 WAV against their 320kps MP3 files and subtracted the two to say, "Look how much you can still hear the music in the difference file"!

Well, we could do the same with these two 24/96 files and result will be a very audible as well. I am sure it won't be a zero sum game.

Isn't the goal of Hi-Fi to be true to the original?

RustyGates's picture

Just like all computers, the output is dependent on the capabilities of the system, and even more so, on the quality of the input. The greater the bit depth, the greater the bit-wise accuracy of the filter & upsampling computation. Feed an upsampler 16bit data, it should be able to reconstruct sample points to at least 16 bit accuracy in amplitude and phase, or -90.31dBFS small signal resolution, and given 24 bit data should do the same to -138.47dBFS accuracy, though 24-bit is far too complex to be accurate to in the time domain. Even Rob Watts himself states M-Scaler is accurate to 16.7-bit equivalent phase accuracy, but under all input circumstances. The higher the input sample rate, the more efficient the computation. So feeding an upscaler higher amounts of true samples (by the ADC) of course improve upsampling output quality. The output of a 24/96 will be better than the output of a 16/44.1. Of course analogue is, analogue, continuous. But the quality of recreating digital audio is so good these days, you look at the huge inconvenience and mechanical noise analogue sources bring about and you just realize its not worth it. Its not the 1980's anymore. And just to be clear, every single DAC in the world upscales, even NOS DACs by using a primitive zero-order-hold filter (ZOH); they have to, to create a continuous wave form. Its the quality of upsampling that matters.

Here are similar questions to yours, of which Rob answers, and better to get it from him than me.

skris88's picture

I don't see how we should think all DACs up-sample. That's even more snake oil, IMHO.

You can buy a 4K TV. But the 2K ("Full HD" 1,920 x 1,080 pixels) video you are watching is still 2K. The TV being 4K capable does not make what you watch automatically 4K unless you turn On or enable an up-sampling feature - if one exists.

Likewise when you play 16/44 audio through a 24/96 DAC. It is not up-sampling irrespective of the what the (marketing department-inspired) indicator lights may show.

What 24 bits do is give you 144dB of dynamic range. While 16 bits gives you only 96dB of dynamic range. That could mean there is a 'reserve' of 48dB headroom when 16/44 audio is played through a 24/96 DAC if the input level is not altered.

With a lot of material recorded at or near 0dBFS the additional (now un-clipped) headroom is what everyone is hearing, thus claiming 24/96 up-sampling "works".

But I'm happy to be proven wrong!

I think this is a task for the experts at Stereophile Labs to take on and show proof of, one way or the other.

If we talk about accurate Hi-Fidelity reproduction we need to be objective about technology, not fall for subjective opinions that A is better than B.

hb72's picture

"With a lot of material recorded at or near 0dBFS the additional (now un-clipped) headroom is what everyone is hearing, thus claiming 24/96 up-sampling "works"."

Restoring lost dynamic information is not really what upsampling claims to be able to do. Upsampling refers to increased digital data points in a given time interval for better approximation of the original signal, not necessarily gains in dynamic scale, before conversion into analog, essentially relaxing the need for low pass filters possibly interfering with the audible range. Going way up in frequency, and using smooth but musical forms of interpolation apparently makes the difference here.

SNI's picture

I think it will be appropriate to point out the real purpose of up-sampling.
There are two kinds of up-sampling, the asyncronous and the synchronous.
Often the syncronous is called oversampling, and is just an interpolating filter, which calculates new samples between the original samples, thus making the frequency range larger pushing images upwards in frequency, resulting in more simple analog filtering.
The asyncronous up-sampler does not calculate samples between the original ones, it completely trashes all incoming samples, none of those are used in the processed signal.
The way it works is that incoming samples are buffered in a small buffer containing typically arround 100 words.
Next a conceptual interpolation of typically 2^20-1 sample are calculated, yes you are right 1.048.576 - 1 between every incoming sample regardless of Fs, which could be 192Khz or more.
Next a first order hold function is applied to connect the samples in time to a continous signal, we are getten pretty analog here :-).
And then sample a new signal with a new Fs and bitdepth from this continous signal.
There are two purposes for this proces.
The first is to elliminate jitter, which it does to an extent only limited by the local clock, which can be very precise these days.
And the second purpose is to provede a digital signal to the DAC chip, with bitdepth and Fs right in the sweet spot for the DAC chips performance, which mostly is arround 32 bit 100Khz Fs. Surprice :-)
Most chips has the lowest distortion in the vicinity of 100KHz Fs.
So up-scaling/up-sampling has not really anything to do with recreation of lost data, because as per Nyquist nothing is lost in the first place.
Instead up-sampling is an analog proces in the digital domain which reduces jitter significantly, and can be adjusted to optimise the performance of the following DAC chip.

Keine hexerei nur behändigkeit

Bogolu Haranath's picture

So, can your second example of 'asynchronous' up-sampler, create a perfect impulse response and a perfect sine wave, even with a 16/44.1 PCM data? ....... Just curious :-) .......

SNI's picture

Impulse response is only allowed if it does not contain out af band frequencies.
The only ourpose for impulse response measurement on digital processors is to show if it is phase linear.
NOS DACs has a very differnet impulse response as it is not bandlimited, thus sending a cascade of high frequencies downstream, if not filtered out with steep analog filters.
Anyways pulse response filtered digitally, is just the way a bandwithlimited signal looks. If dona in the analog domain it will look the same, if it is kept phase linear, which is really difficult.
The sine wave is recreated perfectly @ Fsx2 of the max bandwith. There is no reason to discus that.
Anyways you have bitdepth which is determining noise, and you have Fs determining frequency response.
Higher bitdepth means lower noise, higher Fs means wider frequency area.
Filtering will create ripple in the stop band, today passband ripple is just a fraction of a fraction of a dB ie PCM1792 from BB has only 0,00001 dB passband ripple.

Bogolu Haranath's picture

JA1 reviewed the Chord M Scaler with the Chord Dave DAC ....... You can see the reviews and the measurements of both those components on the Stereophile website ....... What is Dave DAC not doing? ....... What is M Scaler doing when added to Dave DAC? :-) .......

SNI's picture

The measurements shown in this test proves the scaler to have a very deep digital filter when up-sampling is enabled.
You see that on the impulse response, where the stop band ringing is much larger than the pass through signal.
This is further proved in the comparison with the ML 30.6.
But you have to be aware that all measurements were made in the digital domain. No DA conversion was done at any time during measurement, due to limitations in Fs on the AP system.
What is not shown is the jitter behavior of the Scaler.
Up-sampling is what could be called a jitterhammer, in theory it reduces jitter down to local clock jitter. But the up-sampler needs to be located just upstream from the DAC chip to do so.
If the up-sampled signal has to be serial interfaced, the results will go down the drain.
In case of the Scaler, it also contains a USB interface, which I do not know anything about.

Bogolu Haranath's picture

The blue trace shown in the M Scaler measurements is the effect of short linear phase filter, which shows very minimal ringing ........ DACs, which have this type of filter, when used, sounded more 'musical' as mentioned by the Stereophile reviewers ........

You may already know this ........ Benchmark published an article about 'Asynchronous upsampling', which is available online ....... You can Google that article .........

As I mentioned above in this forum, AES published an article in May 2019, 'Modern sampling: A tutorial' by J. Angus ........ That article is open access, freely available online ........

Wikipedia also has an article about 'Sample rate conversion' :-) ........

SNI's picture

Well I didn´t know of these articles, but I´ve studied datasheets of some of the most widespread DAC chips, and they all perfor at their best @ arround 100KHz.
Because of that our own DAC was made to up-sample to ~100KHz, and the chip sounded better this way.
We also tried out different filters, but I cannot agree that the short filters sound better. It always ended with the sharp roll of standard filter as the prefered one.
And again and again "It has no ringing".
The stopband is excactly what is stopped, the passband is excactly what is passed.
The impulse response is IMO the most misunderstood measurement ever in HI-FI, and it has even lead some of us to NOS DACs, God forbid it :-)

Bogolu Haranath's picture

HoloAudio Kitsune R2R DAC (HR's reference) Stereophile measurements show, no ringing when the NOS filter is used (see Fig.1 in measurements) :-) .........

SNI's picture

I didn´t see any measurements of the Holo, but no ringing is no filter.
A filtered signal will always contain stopband ringing, the steeper the more.
Benchmarks choise of 110KHz Fs looks pretty wise to me.
First of all it is just arround the peak performance of most chips, and second it will avoid any resonance with chips running @ the integer of the distributed 44,1 and 48 formats.

Bogolu Haranath's picture

The one above AD's follow-up is the measurements section of the HoloAudio Kitsune review :-) .......

Bogolu Haranath's picture

BTW ...... Benchmark prefers 'asynchronous upsampling' to 110KHz :-) ........

Bogolu Haranath's picture

If you want to experiment with different reconstruction filters, try the Mark Levinson No.5805, reviewed by Stereophile .......... It has 7 different reconstruction filters in the DAC section ........ One of them is similar to the blue trace, short linear phase filter like the one shown in M Scaler measurements :-) ........

SNI's picture

Well I will be trying out some different filters in the near future, as we are making a new DAC which will probably be based on the AKM 4497 chip.
Then i will try out the onboard filters, but I´m not very excited about it.
There is so much more to gain in the powersupplies and in the topology of the analog stage.
The latter has for some strange reason completely been left out of scope nowadays.

Ali's picture

Thanks for review, It would be nice to also consider Chord owns Qutest to see if M scaler could be a good match.Maybe HR give us a follow up since he using this DAC; Cross my finger...Regards.