Review and Measurements of Chord Mojo DAC and Amp

[Simon P]

As a maybe parting and hopefully good natured greeting, for those apparently convinced they are supported by rigorous and scientific thinking, time domain response of a filter occurs by convolution with the digital signal. A short filter as used by many, probably most, dacs, produces inter-sample time domain errors compared with an accurate full sinc function required by the Nyquist criteria for reconstruction. If I'm wrong, explain please, since apparently I don't think much..
A short, narrow windowed FIR filter can give an error in contribution on inter-sample reconstruction say 100 samples from the function peak, and the error, in terms of contribution from an individual digital sample, is frequently off by tens of dB compared to what it should be. These errors are very largely hidden in the frequency domain when measured by FFT, i.e. spectrum analyser.
This may be why, for some of us, on a bad day, digital sounds like a very small man with a nail gun is riding on the reconstructed waveform. Not an ideal analogy.

 

[mansr]

Uh-huh

 

[Veri]

since Chord Dacs have a proprietary DAC section design and extensive oversampling filtering, unlike almost any other available DAC, then as you put it, it would be a nigh impossible task to find a Comparable DAC or Comparable filters.

Fairly certain I read a blog post maybe even by mansr himself demonstrating the total overkill of the millions of taps, and how easy it is to simulate the Chord 'WTA filter' in approximation...

 

[mansr]

Fairly certain I read a blog post maybe even by mansr himself demonstrating the total overkill of the millions of taps, and how easy it is to simulate the Chord 'WTA filter' in approximation...

This post, perhaps: https://www.audiosciencereview.com/...-rme-adi-2-dac-fs-tap-count.22124/post-734918

 

[Ken Tajalli]

Fairly certain I read a blog post maybe even by mansr himself demonstrating the total overkill of the millions of taps, and how easy it is to simulate the Chord 'WTA filter' in approximation...

Yes, as I did say, software are available to do comparable function.
Though I am not certain how long these softwares have been around? do you?

 

[Veri]

Yes, as I did say, software are available to do comparable function.
Though I am not certain how long these softwares have been around? do you?

Mansr's post I was referring to and which he kindly found the link to, is mostly maths. Maths have been around for quite some time if I recall correctly I mean a lot of these filters and their respective responses can just be simulated in Matlab or similar software. I know you refer to upsampling software but I meant to call into question the absurdity of the length of these Chord sinc filters. Which the above post delightfully illustrates.

 

[Ken Tajalli]

Mansr's post I was referring to and which he kindly found the link to, is mostly maths. Maths have been around for quite some time if I recall correctly

Fantastic. but putting maths function into a marketable product (hardware or software) takes some expertise .
BTW, maths has been available since the big bang! probably before . . .

 

[BDWoody]

This may be why, for some of us, on a bad day, digital sounds like a very small man with a nail gun is riding on the reconstructed waveform.

Sure, all someone needs to do is demonstrate they can hear it. No luck so far, just more repeated claims with descriptions of why they might be hearing what they claim to hear. No evidence though... It shows how a good story can go a long way.

 

[AdamG]

digital sounds like a very small man with a nail gun is riding on the reconstructed waveform

Pneumatic, Electric or Powder charge? Each have their own unique pitch/timbre

 

[Ken Tajalli]

Pneumatic, Electric or Powder charge? Each have their own unique pitch/timbre

The type Indian street vendor dentists might use. See here:

 

[Simon P]

Fairly certain I read a blog post maybe even by mansr himself demonstrating the total overkill of the millions of taps, and how easy it is to simulate the Chord 'WTA filter' in approximation...

Fantastic. but putting maths function into a marketable product (hardware or software) takes some expertise .
BTW, maths has been available since the big bang! probably before . . .

Reading Roger Penrose on the relationship between math(s), physical reality and consciousness is interesting, see 'The Road to Reality', near the beginning.

 

[Simon P]

Mansr's post I was referring to and which he kindly found the link to, is mostly maths. Maths have been around for quite some time if I recall correctly I mean a lot of these filters and their respective responses can just be simulated in Matlab or similar software. I know you refer to upsampling software but I meant to call into question the absurdity of the length of these Chord sinc filters. Which the above post delightfully illustrates.

These are all frequency domain results, which is how filters are usually designed, and this is the point Watts takes issue with. Audio is normally considered in the frequency domain, and for very understandable reasons, but not necessarily definitive reasons.

 

[Simon P]

Nyquist as a proof does not work for anything other than a hard brickwall anti-alias filter with zero passband ripple, zero passband phase shift vs. frequency, zero transition band frequency range and infinite stopband attenuation. The input signal must of necessity be time bounded but the encoded file must be unbounded in time. The reconstruction filter must be the same, i.e. a sinc filter with infinite impulse response in time. The DAC (commonly termed reconstruction-) filter is a time domain interpolation filter in terms of what you want to achieve: most of us listen to music in the time domain, I believe.... If this is not done perfectly we get out of band images in the frequency domain. Attempt are made to assess reconstruction filter performance by examining them in the frequency domain, and this has some validity: it measures the ability to reject Fourier images. However, the point is, our primary aim is surely to reconstruct a music waveform in the time domain. Frequency domain measurements are just an attempt to predict adequate time domain performance.

Any departure from perfect implementation will affect both frequency and time domain performance. The implications for audible time domain distortion of known departures from perfection of frequency domain characteristics is not well defined. So several filter designers have made time domain impulse response a fundamental in their thinking.

Now I'd agree the (short tap) filters are often so good in the FD you might not think they'd have an audible effect, but actually, we don't know, and an exercise in thinking out the signal convolution in time is very instructive.

To get a definitive analytical mathematical answer here to the time domain effects of our departures from Nyquist-Shannon-Whitakker is going to be incredibly difficult if not impossible.

 

[Simon P]

Pneumatic, Electric or Powder charge? Each have their own unique pitch/timbre

Actually depends on what mood the very small man is in more than the type of gun. The man himself manifests as a fixed DC offset.

 

[Frank Dernie]

Nyquist as a proof does not work for anything other than a hard brickwall anti-alias filter with zero passband ripple, zero passband phase shift vs. frequency, zero transition band frequency range and infinite stopband attenuation. The input signal must of necessity be time bounded but the encoded file must be unbounded in time. The reconstruction filter must be the same, i.e. a sinc filter with infinite impulse response in time. The DAC (commonly termed reconstruction-) filter is a time domain interpolation filter in terms of what you want to achieve: most of us listen to music in the time domain, I believe.... If this is not done perfectly we get out of band images in the frequency domain. Attempt are made to assess reconstruction filter performance by examining them in the frequency domain, and this has some validity: it measures the ability to reject Fourier images. However, the point is, our primary aim is surely to reconstruct a music waveform in the time domain. Frequency domain measurements are just an attempt to predict adequate time domain performance.

A gearbox only works at a fixed distance between the shaft axes which must be parallel to each other and the geometry only remains correct whilst transmitting zero load.
If engineers insisted on making them more and more perfectly spaced and parallel they would not work any better than they do by allowing manufacturing tolerances that mean the parts can be made at reasonable cost by designing them to work over the actual range of requirements they face.

Software filters that achieve accuracy way beyond any human's hearing capacity may be smart bits of intellectual self indulgence but they are just that, not a necessary requirement.

If they were working to perfect record heads for tape recorders which didn't saturate at high levels of high frequencies or to have a surface shape avoiding the bass uneveness they naturally have they may make an audible improvemeny on analogue music reproduction because both of these leave artefacts which are clearly audible to humans, that would be worthwhile.

If they were working to design a linear magnetic circuit for pickup cartridges, such that its output was actually closer to the groove signal, this also would be readily audible to humans that would be worthwhile.

Once you get to a level of accuracy where further improvement is inaudible any more is a pointless indulgence, stupid even for anything but marketing to the ignorant.

IMO even if you get to a point where an improvement is only audible in carfully controlled conditions on particularly challenging music you are getting closer than any tape recording ever was.

If you were railing about how technically compromised and audibly imperfect every tape recorder ever made is.
If you were disappointed by the poor performance of all pickup cartridges used on record players in every way, (including in the time domain).
If you were railing against the audible speed fluctuations of many tape recorders and record players.
If you were railing against the audibility of acoustic and structural vibration pickup when playing an LP.

That would make some sense, not much, but some. But you aren't so you have completely missed the point.

Luckily people recording music haven't missed the point and we have lots to enjoy.

I'll leave you to continue tilting at windmills Mr Quixote.

 

[Simon P]

A gearbox only works at a fixed distance between the shaft axes which must be parallel to each other and the geometry only remains correct whilst transmitting zero load.
If engineers insisted on making them more and more perfectly spaced and parallel they would not work any better than they do by allowing manufacturing tolerances that mean the parts can be made at reasonable cost by designing them to work over the actual range of requirements they face.

Software filters that achieve accuracy way beyond any human's hearing capacity may be smart bits of intellectual self indulgence but they are just that, not a necessary requirement.

If they were working to perfect record heads for tape recorders which didn't saturate at high levels of high frequencies or to have a surface shape avoiding the bass uneveness they naturally have they may make an audible improvemeny on analogue music reproduction because both of these leave artefacts which are clearly audible to humans, that would be worthwhile.

If they were working to design a linear magnetic circuit for pickup cartridges, such that its output was actually closer to the groove signal, this also would be readily audible to humans that would be worthwhile.

Once you get to a level of accuracy where further improvement is inaudible any more is a pointless indulgence, stupid even for anything but marketing to the ignorant.

IMO even if you get to a point where an improvement is only audible in carfully controlled conditions on particularly challenging music you are getting closer than any tape recording ever was.

If you were railing about how technically compromised and audibly imperfect every tape recorder ever made is.
If you were disappointed by the poor performance of all pickup cartridges used on record players in every way, (including in the time domain).
If you were railing against the audible speed fluctuations of many tape recorders and record players.
If you were railing against the audibility of acoustic and structural vibration pickup when playing an LP.

That would make some sense, not much, but some. But you aren't so you have completely missed the point.

Luckily people recording music haven't missed the point and we have lots to enjoy.

I'll leave you to continue tilting at windmills Mr Quixote.

You'd think so based on measurement, I agree.

 

[Ken Tajalli]

A gearbox only works at a fixed distance between the shaft axes which must be parallel to each other and the geometry only remains correct whilst transmitting zero load.
If engineers insisted on making them more and more perfectly spaced and parallel they would not work any better than they do by allowing manufacturing tolerances that mean the parts can be made at reasonable cost by designing them to work over the actual range of requirements they face.

Software filters that achieve accuracy way beyond any human's hearing capacity may be smart bits of intellectual self indulgence but they are just that, not a necessary requirement.

If they were working to perfect record heads for tape recorders which didn't saturate at high levels of high frequencies or to have a surface shape avoiding the bass uneveness they naturally have they may make an audible improvemeny on analogue music reproduction because both of these leave artefacts which are clearly audible to humans, that would be worthwhile.

If they were working to design a linear magnetic circuit for pickup cartridges, such that its output was actually closer to the groove signal, this also would be readily audible to humans that would be worthwhile.

Once you get to a level of accuracy where further improvement is inaudible any more is a pointless indulgence, stupid even for anything but marketing to the ignorant.

IMO even if you get to a point where an improvement is only audible in carfully controlled conditions on particularly challenging music you are getting closer than any tape recording ever was.

If you were railing about how technically compromised and audibly imperfect every tape recorder ever made is.
If you were disappointed by the poor performance of all pickup cartridges used on record players in every way, (including in the time domain).
If you were railing against the audible speed fluctuations of many tape recorders and record players.
If you were railing against the audibility of acoustic and structural vibration pickup when playing an LP.

That would make some sense, not much, but some. But you aren't so you have completely missed the point.

Luckily people recording music haven't missed the point and we have lots to enjoy.

I'll leave you to continue tilting at windmills Mr Quixote.

You do know that most recording studios have gone digital all the way? no more archaic tape recorders or turntables or grooves ....
Recording is mostly 32bit 384kB I believe, and mastering is all the way digital.
This is why our resident Don Quixote (if I understand your jest) is working on advanced ADC implementation and design, to further the cause of sound quality, as he believes there is nothing else to be done on DAC front. He is also eyeing microphone and transducer design!
Man needs to keep busy, man needs to eat too.

 

[Simon P]

You do know that most recording studios have gone digital all the way? no more archaic tape recorders or turntables or grooves ....
Recording is mostly 32bit 384kB I believe, and mastering is all the way digital.
This is why our resident Don Quixote (if I understand your jest) is working on advanced ADC implementation and design, to further the cause of sound quality, as he believes there is nothing else to be done on DAC front. He is also eyeing microphone and transducer design!
Man needs to keep busy, man needs to eat too.

Yes. Barring old analogue recordings and the odd remastered-from-analogue LP, it's a technicality if analogue preserves something lost in digital, except to perfect the digital, if you think it needs perfecting. Which quite a few people still do. I'm aware Rob Watts was looking at studio ADCs.

 

[Simon P]

....and to get significantly further in terms of the Watts approach to DAC windowing, to reduce inter-sample error leak, as it were, you'd need unrealistic amounts of processing power.

 

[BDWoody]

You'd think so based on measurement, I agree.

Also based on the utter and profound lack of any evidence to the contrary.