Chord Hugo M Scaler - Stereophile Review (measurements also)

[RichB]

Is there comprehensive data on the audibility of digital reconstruction filters?
Any valid test would have to include speakers/headphones immune from modulating ultrasonics into the audible range.

There are folks on this site with a clear preference for MQA (lossy high-res audio) and high-res audio.
Folks going at Rusty should have equal passion for all.
For the record, I loath MQA and am agnostic to Hi-Res audio because I believe ultrasonics are inaudible (by definition)

I'd love the try this device with a A/B switch level matched but not enough to spend the money.

- Rich

 

[KSTR]

@RichB,
For grins you might want to start with this, a very exaggerated example ;-)

 

[Music1969]

You seem to be missing the point of my examples. It is entirely possible that one's listening is not sensitive enough and there is entirely the possibility a well conducted blind AB is skewed by a sound preference.

Yes but where is this properly conducted double blind test with M-Scaler?

The one you described that you did is not proper... level differences not accounted for....

 

[RustyGates]

Yes but where is this properly conducted double blind test with M-Scaler?

The one you described that you did is not proper... level differences not accounted for....

There is going from Green (2FS) to 4FS (Blue) up sampling modes which had an (or largest) audible improvement in what I would describe as staging & "tighter" or "clearer" separation with the TT2.

The biggest difference would probably happen if someone ran an upsampler through a NOS DAC.

But, that's just me. Until more people do more tests on the mscaler, HQPlayer, dCS upsampler, there's really no point rambling on about this topic anymore.

 

[MaxBuck]

I still don't really understand exactly what this thing is supposed to be doing.

 

[Purité Audio]

It separates the gullible from their money.
Keith

 

[KSTR]

It separates the gullible from their money.
Keith

Besides that, @MaxBuck , technically it replaces the on-chip digital filter of your DAC with a 'precision' version, an almost perfect sinc filter. Your DAC is fed with an already upsampled and filtered signal so that its own upsampling and filtering is irrelevant.

Is it audible? I personally doubt it (tried a similar upsampling with software -- a 0$ solution -- no audible difference relevant to me).
So, the M-Scaler IMHO definitely is a 'yes we can' sort of statement product with probably little, if any, real life value.

 

[bhobba]

I still don't really understand exactly what this thing is supposed to be doing.

It is simple. What Rob believes, and others as well is that high res recordings sound better. I will not argue if they do - but these people think they do. They believe 96k sounds better than 48k, 192k is better than 96k etc. - up to many MHz - with a law of diminishing returns, of course. Why is this if we can only hear up to 20khz - most people lower? They think it is time smear. The higher the sampling rate, the less time smear. So upscaling as much as possible makes things sound better by reducing time smear. Rob, in his DAC's, sometimes upsamples to many MHz. He uses a sinc filter to upsample because Shannon's sampling theorem says it reconstructs a limited frequency signal perfectly. However, he also believes the quality of the sinc filter is of vital importance especially during the early upsampling stages. His calculations showed that a 1 million tab filter was necessary for at least 16 bit-perfect accuracy in the upsampling process. The question is such perfect accuracy essential?

You can't argue with the measurements - he did what he claimed. You can only say, do people hear a difference? Rob claims it is night and day. Others think so too. I have one and think it is better but not night and day. Others claim no difference, and others I know think it is worse. There is only one way to decide - a blind listening test. Anyone can do a simple one of those by just closing their eyes and having a friend switch it in and out. People can even do a proper one involving several people. Having been involved in a couple of those, they are a nightmare to do correctly, and because of the egos involved, little things that could have affected the outcome, some will latch onto to try and invalidate it, if it does not go the way they want. I remember one that took ages and a lot of work to organise was invalidated because the earthing arrangement of one DAC was different - no kidding - how the earthing was done. All that time and effort for naught. So best to do it yourself and make up your own mind. I have given up on blind tests with several people - too much of a pain.

Thanks
Bill

 

[Frank Dernie]

It is simple. What Rob believes, and others as well is that high res recordings sound better. I will not argue if they do - but these people think they do. They believe 96k sounds better than 48k, 192k is better than 96k etc. - up to many MHz - with a law of diminishing returns, of course. Why is this if we can only hear up to 20khz - most people lower? They think it is time smear. The higher the sampling rate, the less time smear. So upscaling as much as possible makes things sound better by reducing time smear. Rob, in his DAC's, sometimes upsamples to many MHz. He uses a sinc filter to upsample because Shannon's sampling theorem says it reconstructs a limited frequency signal perfectly. However, he also believes the quality of the sinc filter is of vital importance especially during the early upsampling stages. His calculations showed that a 1 million tab filter was necessary for at least 16 bit-perfect accuracy in the upsampling process. The question is such perfect accuracy essential?

You can't argue with the measurements - he did what he claimed. You can only say, do people hear a difference? Rob claims it is night and day. Others think so too. I have one and think it is better but not night and day. Others claim no difference, and others I know think it is worse. There is only one way to decide - a blind listening test. Anyone can do a simple one of those by just closing their eyes and having a friend switch it in and out. People can even do a proper one involving several people. Having been involved in a couple of those, they are a nightmare to do correctly, and because of the egos involved, little things that could have affected the outcome, some will latch onto to try and invalidate it, if it does not go the way they want. I remember one that took ages and a lot of work to organise was invalidated because the earthing arrangement of one DAC was different - no kidding - how the earthing was done. All that time and effort for naught. So best to do it yourself and make up your own mind. I have given up on blind tests with several people - too much of a pain.

Thanks
Bill

Every time I have done a blind test it has confirmed science, and I have done quite a few now on DACs and cables, so I don't feel the need to question science yet again, personally.

To be honest one thing that makes me sceptical is somebody claiming something is "night and day" different.
In my blind tets of DACs I could hear a difference between some of the reconstruction filters but the difference was quite hard to be completely sure of, and cartainly nowhere near enough to influence my musical enjoyment.

 

[bhobba]

Every time I have done a blind test it has confirmed science, and I have done quite a few now on DACs and cables, so I don't feel the need to question science yet again, personally.

I have seen some strange things with very experienced listeners and blind tests. I would count myself as experienced, but not as experienced as these guys. It is very humbling. Admittedly one of those people builds and designs speakers for a living. I have seen him when designing a speaker say - it is 3 dB down there. Measure it - and guess what - it is 3 dB down where he said it was. He believes that measurements are the most important thing, but also there are some things he can hear that measurements do not show. However, they are rare. It is very humbling.

Thanks
Bill

 

[BDWoody]

but also there are some things he can hear that measurements do not show.

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

 

[SIY]

It is simple. What Rob believes, and others as well is that high res recordings sound better. I will not argue if they do - but these people think they do. They believe 96k sounds better than 48k, 192k is better than 96k etc. - up to many MHz - with a law of diminishing returns, of course. Why is this if we can only hear up to 20khz - most people lower? They think it is time smear. The higher the sampling rate, the less time smear.

When ignorant people make claims like that, it's just a bit sad, but fixable with some education. When knowledgeable people make claims like that, it's more than a bit dishonest, and not fixable.

 

[sergeauckland]

I have seen some strange things with very experienced listeners and blind tests. I would count myself as experienced, but not as experienced as these guys. It is very humbling. Admittedly one of those people builds and designs speakers for a living. I have seen him when designing a speaker say - it is 3 dB down there. Measure it - and guess what - it is 3 dB down where he said it was. He believes that measurements are the most important thing, but also there are some things he can hear that measurements do not show. However, they are rare. It is very humbling.

Thanks
Bill

That's what experience teaches. When I first started out in audio engineering, I saw older, wiser and much more experienced engineers doing that sort of thing, wow and flutter in tape machines, dips or peaks in frequency response, EQ errors, they were all pretty close.

However, that's very different from claiming that one can hear things that can't be measured.

S.

 

[Music1969]

His calculations showed that a 1 million tab filter was necessary for at least 16 bit-perfect accuracy in the upsampling process.

Genuine query - has Rob ever shown this calculation?

If it's just sinc interpolation, someone should be able to do a derivation and get a figure at least in same order of magnitude?

 

[RustyGates]

bhobba said:
It is simple. What Rob believes, and others as well is that high res recordings sound better. I will not argue if they do - but these people think they do. They believe 96k sounds better than 48k, 192k is better than 96k etc. - up to many MHz - with a law of diminishing returns, of course. Why is this if we can only hear up to 20khz - most people lower? They think it is time smear. The higher the sampling rate, the less time smear.

When ignorant people make claims like that, it's just a bit sad, but fixable with some education. When knowledgeable people make claims like that, it's more than a bit dishonest, and not fixable.

There is a big difference between capturing sound at >44.1kHz on the ADC, and digital filtering (i.e. by interpolating) 44.1kHz into the MHz. Raw 44.1kHz without filtering is not good enough, i.e. the DAC painting a voltage every 22.67uS and holding, otherwise we would all just be listing to NOS dacs for "good" sound.

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

 

[SIY]

There is a big difference between capturing sound at >44.1kHz on the ADC, and digital filtering (i.e. by interpolating) 44.1kHz into the MHz. Raw 44.1kHz without filtering is not good enough, i.e. the DAC painting a voltage every 22.67uS and holding, otherwise we would all just be listing to NOS dacs for "good" sound.

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

Is it Monty time?

 

[BDWoody]

Raw 44.1kHz without filtering is not good enough, i.e. the DAC painting a voltage every 22.67uS and holding, otherwise we would all just be listing to NOS dacs for "good" sound.

Have you ever looked into how sampling theory works? You seem to think of it as a connect the dots exercise, with more dots always better than less.

If you're interested, here's the oft referenced 'Monty' video (mentioned by @SIY ) that does a nice job of making some sense of it.

 

[mansr]

Genuine query - has Rob ever shown this calculation?

If it's just sinc interpolation, someone should be able to do a derivation and get a figure at least in same order of magnitude?

I believe the idea is to find the length required for interpolation using a truncated sinc function to get within ½ LSB of the ideal value. Now simply truncating the sinc function is a naive approach that results in a narrow transition band while providing poor stop-band rejection. Using more advanced filter design techniques, and allowing a transition band of 1 kHz or so (the sample rate is chosen with some margin), it is easy to achieve much better stop-band rejection using only a few hundred taps.

 

[charleski]

They think it is time smear. The higher the sampling rate, the less time smear.

The problem here is that I have no idea what is meant by 'time smear'. A proper upsampling filter should be linear phase, but there's nothing special about that. The number of taps needed by a linear phase FIR filter is given by:

Which one of these variables is responsible for "time smear"? The conventional linear phase oversampling filters used by competent DACs already do a decent enough job in all these parameters, so how does throwing a ludicrous number of taps at the problem produce any meaningful difference? All the measurements show is that the MScaler produces an extremely sharp brick wall response, which is exactly what you'd expect from the theory (and is more than we really need, since the Redbook spec gives a couple of kHz for the transition band). But I can't see how any of this could be related to "time smear".

 

[bennetng]

Now simply truncating the sinc function is a naive approach that results in a narrow transition band while providing poor stop-band rejection.

Like this?

Review and Measurements of the Chord Hugo 1 DAC & HP Amp

I tested a Hugo 2 a while back but it was with my old analyzer. So a member kindly sent me an original Hugo DAC and headphone amplifier to remeasure. I wa shocked to find out that the Hugo cost US $2,500 when it came out!!! Man that is expensive for a portable DAC and Amp. Here is a shot of...

www.audiosciencereview.com