Chord Hugo M Scaler - Stereophile Review (measurements also)

[DSJR]

This was the track of the moment when I heard the M-Scaler in and bypassed.

 

[RustyGates]

Being called "best DAC" at an audio show is sort of like being called best dressed at the local hoedown. Nothing says I am full of crap like claiming you can extract the sound of only a DAC from unfamiliar speakers, with unfamiliar amplification in an unfamiliar and awful room. Audio reviewers claim that all the time. It's meaningless.


You will excuse me if I doubt the engineering chops of anyone who would post those two paragraphs or who spends more on a DAC then their car.

I spend money on full custom designs, with fantastic build quality and western manufactured, that perform extremely well (prob still the best measuring DAC on the APx555 out there), because you know why? I like it. I value that a lot more than my A to B machine from 2005.

I would love a Merc AMG or Porsche with a hand built engine, just far too out of my price range to care at this stage.

And if you don't like the fact that I like it then... sue me or something, but it has nothing to do with my engineering "chops".

 

[RustyGates]

You shouldn't talk here in their speak. They should not even be quoted or mentioned here.
You will be punished immediately

edit: as usual mansr beat me to it.

Fine fine. Noted.

 

[audio2design]

I spend money on full custom designs, with fantastic build quality and western manufactured, that perform extremely well (prob still the best measuring DAC on the APx555 out there), because you know why? I like it.

No one is questioning the quality of Chord DACs. However I do question an engineer bringing up "best in show" on an engineering oriented forum given the meaningless of the accolade. Most of what you have written has been pretty much marketing speak. You have yet to justify from an engineering standpoint how this is "audible". Parroting the company/designer does not make for a justification.

 

[solderdude]

This was the track of the moment when I heard the M-Scaler in and bypassed.

One simple question... did you know when it was bypassed or not ?

 

[RustyGates]

No one is questioning the quality of Chord DACs. However I do question an engineer bringing up "best in show" on an engineering oriented forum given the meaningless of the accolade. Most of what you have written has been pretty much marketing speak. You have yet to justify from an engineering standpoint how this is "audible". Parroting the company/designer does not make for a justification.

There is nothing marketing about the on-paper performance of increasingly complex digital filters.

As for the psychoacoustic component of it, I am not in a position to comment further. I've already talked about my own experience with both Mscaler and HQPlayer, maybe other proper ABx testing results can be posted here by others, but there are none so far. But by continually arguing about this is like arguing with someone why I bought a DAC which measures better than 115dB SINAD, which is already apparently pushing it to the absolute limit from a psychoacoustic perspective and ABx testing is not going to show a difference, or trying to get a justification from a DAC designer as to why they bothered with such performance.

 

[Music1969]

But in the end of the day, the M-Scaler is the closest implementation of Whittaker–Shannon interpolation there is, a perfect audio-band brick wall filter.

This is not correct

HQPlayer's sinc-L closed form sinc interpolator is 2M taps at PCM 705kHz (131k taps x conversion ratio) with similar stop band attenuation to M-Scaler

Non-apodizing like M-Scaler

Way cheaper than M-Scaler and double taps (if that's really important)

 

[audio2design]

There is nothing marketing about the on-paper performance of increasingly complex digital filters.

As for the psychoacoustic component of it, I am not in a position to comment further. I've already talked about my own experience with both Mscaler and HQPlayer, maybe other proper ABx testing results can be posted here by others, but there are none so far. But by continually arguing about this is like arguing with someone why I bought a DAC which measures better than 115dB SINAD, which is already apparently pushing it to the absolute limit from a psychoacoustic perspective and ABx testing is not going to show a difference, or trying to get a justification from a DAC designer as to why they bothered with such performance.

On paper performance is meaningless. It's akin to the ridiculous claims about cables, in fact in this case, even more ludicrous in many ways as at least inductance and capacitance can create differences in the realm of audibility, if someone wants to make a cable that is a tone control.

However, in this case, it is just that ... paper. Where are the published data with DACs, that clearly show a marked improvement in THD, or IMD, cause if they can't improve those with other competently designed (and note I did not say expensive) DACs, then its just a solution in search of a problem. Oh, by the way, I have a Chord DAC, and an RME, a Topping, several Benchmarks. I have HQplayer, and whip up my own upsampling too. The internal processing in modern DACs is good enough to render HQPlayer (and the Hugo Scaler) pretty much useless. DACs usually have slightly different performance at different input sample rates (function of internal logic noise, impacts of noise on analog sections, DAC performance at a given speed, etc), that will far outweigh any impact of the resolution of the digital upsampling. In most cases, the THD+N is maximized not at the maximum input sample rate at all but at lower sample rates.

Then throw in this BS about time smear. Give me a break. Pure marketing speak. Speaker and room will grossly outweigh any so called improvements in this area. And not to mention the "paper" exercise does not even include the very necessary final processing step, i.e. the analog filter in the calculations.

You call it arguing, I call it holes in an argument large enough to drive a bus through.

 

[ahofer]

Anyone who drops the “y’all are just hurt (‘jelly beans’) cuz you can’t afford it” should simply be ignored as an immature poser.

 

[bhobba]

That would be a first...in the world.

Not really - it just means they have not investigated it enough to devise something to measure it - yet.

Thanks
Bill

 

[bhobba]

The problem here is that I have no idea what is meant by 'time smear'.

It's the response to a Dirac impulse. Dirac in - how wide is what comes out. MQA tries to keep it under 20us. I have even heard them claim to be able with modern recording techniques (beats me how even though I know how MQA in principle works) keep it under 3us. Trouble is they, of course, never occur in practice. Shannon's sampling theorem guarantees perfect reconstruction with a sinc filter with a lot of ripples when fed a Dirac impulse. This is Rob Watts view. But then, time smear is a non-issue. One could call the whole thing BS. Except I think Rob means less than perfect sinc filters colour the signal in ways measurements show is very low. But he believes it is still audible. Rob holds funny views like that. Are they true? Well, I do like the M-Scaler, but that may not be the reason why - it may simply be lower jitter or something else.

Thanks
Bill

 

[bhobba]

Then Robert W will say something like, you've now compromised the transients to achieve better stop-band rejection.

Then (as per one of his slides) a piano may sound like a trumpet...

That's where the good old blind test needs to be done.

Thanks
Bill

 

[Music1969]

That image LOL

LOL Rob is laughing the loudest though - all the way to the bank with sales...

I read an interview that said Mojo alone has sold over 100k units globally.

Just imagine if he gets $20 per unit sold - just Mojo !

 

[bhobba]

Of course, >44.1kHz native format in the context of NOS (no filtering) would be much better, and the higher the better.

What both Rob and the MQA people claim is recordings are all noise above 50 kHz so quality downsampling to 96k then upsampling should be audibly the same. If it is true is another matter. Where both camps differ is in how that downsampling and upsampling is done.

Rob uses a sinc filter that measures perfect or close to it. He believes that accuracy is audible.

The MQA guys are tricky:

MQA Time-domain Accuracy & Digital Audio Quality

Even at high sample rates, standard PCM audio ‘smears’ important timing information. A new digital format, MQA, promises vastly improved time-domain accuracy — without the huge file sizes.

www.soundonsound.com

The much higher sampling rate of the recording is resampled to 96k using a triangle function. That introduces a slow roll-off above 20 kHz which should be inaudible. It also folds the higher frequencies into the 0-48khz band (aliasing). But since it is below the noise floor it should be inaudible. The last 8 bits are chucked away so it does not even appear in the final MQA recording. When transmitted they do not use a sinc filter to upsample - just good old linear interpolation. It is highly inaccurate of course - but since it is noise who cares - at least that is their reasoning. That is the simple explanation. In practice, they use spline instead of linear functions.

What sounds better?. The only real way to tell is a blind listening test. I think I mentioned in another post I have been involved in setting such up and they are such a pain I personally will not do them again. I will participate in them but not be involved in setting them up. The easiest way is simply to close your eyes and have a friend change stuff. I know this will probably go down like a lead balloon here but I also have learned to trust my ears more and don't subject what I hear to blind tests much anymore. But that is just me - it is the only scientifically valid method.

Thanks
Bill

 

[charleski]

It's the response to a Dirac impulse. Dirac in - how wide is what comes out. MQA tries to keep it under 20us. I have even heard them claim to be able with modern recording techniques (beats me how even though I know how MQA in principle works) keep it under 3us. Trouble is they, of course, never occur in practice. Shannon's sampling theorem guarantees perfect reconstruction with a sinc filter with a lot of ripples when fed a Dirac impulse. This is Rob Watts view. But then, time smear is a non-issue. One could call the whole thing BS. Except I think Rob means less than perfect sinc filters colour the signal in ways measurements show is very low. But he believes it is still audible. Rob holds funny views like that. Are they true? Well, I do like the M-Scaler, but that may not be the reason why - it may simply be lower jitter or something else.

Thanks
Bill

So then any filter with a massive number of taps (like MScaler or HQPlayer) would end up with a correspondingly huge amount of “time smear”. MQA (and “NOS” DACs) go in the opposite direction and use extremely short filters with very few taps (or none) and end up leaking loads of ultrasonic junk. It’s certainly better to err on the side of having more taps, but when designing the filter you just need to decide what levels of passband ripple and stopband rejection you want and then provide the corresponding number of taps.

None of this has anything to do with transient response, which is locked in by the sampling frequency of the original signal. I hope it’s obvious that oversampling per se can’t improve the time resolution of the signal, since that would involve magicking up information from nowhere. The function of oversampling is simply to make things easier for the analog electronics of the output stage.

 

[Frank Dernie]

The thing that makes the whole DAC and software development refinement pointless in any way other than an intellectual excercise is record players.
If the level of performance being claimed to be "night and day" was even of minor consequence a record player would be completely un-listenable whereas it can actually be quite pleasant and there are plenty of HiFi fanatics who will go on about its superiority.

Even they wouldn't be able to convince themselves if any actual SQ improvement occured due to this device (or HQplayer).

 

[Music1969]

In most cases, the THD+N is maximized not at the maximum input sample rate at all but at lower sample rates.

This is true for many DAC chips (per their datasheets) however if you ask @Miska (developer of HQPlayer) THD+N is not everything and going to highest sample rates improves other measurements, even if one (THD+N) drops a little bit.

Presence of digital images on the DACs analogue output is one.

There was some analysis done on this forum that did a test on how digital images affects distortion inside the audible band but can't find it. Yes possibly inaudible but would be nice to see more objective data. Even if its just an intellectual discussion (I'm not an intellectual to lead any discussion - I'm just learning bits and bobs from others about this stuff as I go along and asking questions).

Fortunately I never had issues with identifying a piano from a trumpet, even with low taps !

 

[charleski]

This is true for many DAC chips (per their datasheets) however if you ask @Miska (developer of HQPlayer) THD+N is not everything and going to highest sample rates improves other measurements, even if one (THD+N) drops a little bit.

Presence of digital images on the DACs analogue output is one.

There was some analysis done on this forum that did a test on how digital images affects distortion inside the audible band

It sounds like you're talking about IMD. A filter with poor stopband rejection will leak aliased images that can then intermodulate with the signal and produce products in the audible range. But I'd have thought the 19/20k CCIF IMD test would show this up very well.

 

[Music1969]

It sounds like you're talking about IMD. A filter with poor stopband rejection will leak aliased images that can then intermodulate with the signal and produce products in the audible range. But I'd have thought the 19/20k CCIF IMD test would show this up very well.

Yep, would be great if @Miska could show 19/20k CCIF IMD measurement with RME ADI-2.

PCM44.1kHz input vs DSD256 input (HQPlaying PCM44.1k -> DSD256 conversion, DSD Direct mode, fully by-passing AKM DSP).

I found these IMD plots for iFi micro iDSD and some other ADI-2 plots:

Does DSD sound better than PCM?

Thanks. I consider myself very objective and skeptical of pretty much anything without researching. It was more a curiosity, as well as that's how these "audiophile" recordings are actually released. Like I said, I like vinyl for reasons other than the sound quality. Although, there can be crap...

www.audiosciencereview.com

 

[ahofer]

Rob holds funny views like that.

In a carefully-designed box made of commercial self-interest, I suspect.