Does a truly transparent ADC actually exist?

[JamesQuorn0]

Hello,

I'm not the most technical of people, so please be kind with me! I was hoping somebody could help.

In the studio, I often have to transfer from analogue sources (tapes and vinyl, as well as the occasional acetate) and I'm not entirely pleased with the results I've been getting. I've used devices from Prism, Lynx, Apogee, Forsell, and a few others. In level matched tests, it's clear the digital copy suffers from either degradation and/or colouration. Everyone here agrees.

A few questions:

1) Is it normal for ADCs to colour the sound a certain way?
2) I'm getting similar results running digital files through the ADC, with the deficiencies becoming more obvious with each pass through - although the initial pass through seems to show less difference compared to when transferring an analogue source. Is there a technical reason as to why the ADC would play nicer with these files as opposed to a tape/vinyl record being run through the same equipment?
3) Given the above, is it fair to assume that it's the analogue section of these devices that is causing the issue? Or is my hypothesis way off the mark?
4) I've been told that DSD is better for the purpose - I have used a Tascam that had DSD capability and it was pretty awful. Is this genuinely true and the Tascam is just a poor device?
4) Is there such a thing as an actual transparent ADC, or am I setting my standards too high?

If anybody has used an ADC that is truly transparent, then please recommend.

Many thanks

 

[staticV3]

1) Is it normal for ADCs to colour the sound a certain way?

No.

2) I'm getting similar results running digital files through the ADC

You cannot input digital files into an ADC.

4) I've been told that DSD is better for the purpose - I have used a Tascam that had DSD capability and it was pretty awful. Is this genuinely true and the Tascam is just a poor device?

DSD has no inherent advantages over PCM.

4) Is there such a thing as an actual transparent ADC

Amir's APx555 is pretty dang transparent. It's certainly orders of magnitude better than the human ear.
Ditto for stuff like the Cosmos ADC.

 

[MaxwellsEq]

This is a key post where @Blumlein 88 ran audio through an ADC DAC combination 8 times.

Post in thread 'Serious Question: How can DAC's have a SOUND SIGNATURE if they measure as transparent? Are that many confused?' https://audiosciencereview.com/foru...rent-are-that-many-confused.9245/post-1062220

 

[JamesQuorn0]

This is a key post where @Blumlein 88 ran audio through an ADC DAC combination 8 times.

Post in thread 'Serious Question: How can DAC's have a SOUND SIGNATURE if they measure as transparent? Are that many confused?' https://audiosciencereview.com/foru...rent-are-that-many-confused.9245/post-1062220

Very interesting. Thank you for sharing.

 

[dominikz]

In level matched tests, it's clear the digital copy suffers from either degradation and/or colouration. Everyone here agrees.

IMO that sounds like there may be issues elsewhere in the chain - could be a number of things but e.g. perhaps something in the recording chain is being overdriven, or introducing an undesired frequency response deviation (e.g. perhaps an input/output impedance mismatch somewhere?), or the analogue gear is having excessive tolerance between playback attempts, or level matching in the listening comparison isn't perfect...?

Normally the full chain would be measured and optimized to avoid any coloration during transfer to digital. The transfer itself should then be feasible to do even with relatively cheap ADCs, as long as they measure objectively better than the analogue gear.

Note that even 16-bit digital audio has significantly better SNR than either magnetic tape or vinyl, and converter frequency response is usually pretty flat - so any reasonably well measuring ADC shouldn't be a bottleneck for this task.

Amir's APx555 is pretty dang transparent. It's certainly orders of magnitude better than the human ear.
Ditto for stuff like the Cosmos ADC.

IMHO transparency from practical audibility perspective starts much much sooner than APx555 or Cosmos ADC level

 

[JayGilb]

Welcome to ASR !

When you are comparing the analog vs adc are you level matching the output ?
Is your test blind or do you know which version you are listening to ?

 

[JamesQuorn0]

Welcome to ASR !

When you are comparing the analog vs adc are you level matching the output ?
Is your test blind or do you know which version you are listening to ?

Thank you! Yes, the output is level matched, and for most I've been unaware of which was being played (so yes, blind) - not with the Tascam as that wasn't in the current studio. We have a Lavry device being worked on at the moment and I'll repeat the test with that and can document the process.

 

[JamesQuorn0]

or the analogue gear is having excessive tolerance between playback attempts

Thank you for the response. By this, do you mean that the gear (Reel, record player, etc) used is producing different results across different playbacks? I mean it's possible.

 

[dominikz]

Thank you for the response. By this, do you mean that the gear (Reel, record player, etc) used is producing different results across different playbacks? I mean it's possible.

Yes - e.g. to give a simple example: if a record player for any reason spins at a slightly different speed when recording via ADC than it does later when doing comparative listening, the pitch and frequency balance would be slightly shifted between the two takes. There are of course many other imaginable reasons for perceived differences (some physical and some perceptual).

So IMHO there is a small chance that the previous digital recording of a record could be perfectly faithful while still sounding slightly different to a live playback of the same record.

The best test I can think of would be to connect the reel / tape machine at the same time both directly and via a real-time ADC/DAC chain to the playback system.
You could do this via a signal splitter and a mixer or switch, as long as it allows quick switching and very precise (better than ~0,1dB) level matching of the two chains.
You'd need to be careful to minimize AD/DA passthrough latency for the test to make sense, and you'd need to carefully measure and match electrical levels with an objective method (e.g. a true RMS multimeter, scope or another ADC). Level matching with a sound meter / SPL meter is usually not sufficiently precise.

Then you could listen to the same playback attempt with both methods (pure analog and digitized) in real time. If under blind conditions you can still perceive a difference (and especially if you can tell which is which) then next I'd look for gross measurable issues in the setup. Once any/all such measurable issues are resolved you should find the AD/DA conversion becomes completely audibly transparent.

 

[manisandher]

In the studio, I often have to transfer from analogue sources (tapes and vinyl, as well as the occasional acetate) and I'm not entirely pleased with the results I've been getting. I've used devices from Prism, Lynx, Apogee, Forsell, and a few others. In level matched tests, it's clear the digital copy suffers from either degradation and/or colouration. Everyone here agrees.

I certainly agree.

I've spent many years trying to get my vinyl rips to sound identical to the original. No ADC has been up to the job. I've owned all the 'top' machines from Pacific Microsonics, Prism, RME, Motu, Tascam, Korg, and have recorded at all rates and PCM/DSD. There's always been a degradation in SQ.

However...

The E1DA Cosmos is utterly transparent as far as I can tell. I recommend using it with the Scaler for easier impedance-matching.

Of course, others here will disagree. So be it.

Mani.

 

[JamesQuorn0]

I certainly agree.

I've spent many years trying to get my vinyl rips to sound identical to the original. No ADC has been up to the job. I've owned all the 'top' machines from Pacific Microsonics, Prism, RME, Motu, Tascam, Korg, and have recorded at all rates and PCM/DSD. There's always been a degradation in SQ.

However...

The E1DA Cosmos is utterly transparent as far as I can tell. I recommend using it with the Scaler for easier impedance-matching.

Of course, others here will disagree. So be it.

Mani.

Thanks, Mani. I'll check that converter out.

What's interesting is that I find greater transparency if the audio going in has already gone through a DA conversion. When going direct from an analogue source however is where the change happens, so I'm wondering if dominikz is onto something with playback variability.

When you take that converted file from a reel/record and run it through multiple times you still don't match the change from source to initial transfer. It's very strange.

 

[DVDdoug]

Of course, others here will disagree. So be it.

Take a look at this. some people do prefer vinyl, but technically it's inferior.

It's just a fact that digital is better than analog vinyl in every way. A digital copy can sound like vinyl but vinyl can't sound like digital. There is ALWAYS audible surface noise on a record, and sometimes some nasty clicks & pops. There are often frequency response variations and sometimes distortions. (Vinyl can go higher than CDs... into the inaudible ultrasonic range but digital is flatter over the audible range, and most of the ultrasonics are noise. Or, you can sample at 192kHz to capture the ultrasonics.

es - e.g. to give a simple example: if a record player for any reason spins at a slightly different speed

I never HEARD a speed variation from a turntable that wasn't broken.

 

[manisandher]

Take a SOTA DAC, something like an SMSL SU-10. Take a SOTA ADC, something like an RME ADI-2 Pro. Sync the clocks (trivial if using spdif into SU-10).

Capture the analogue output of the SU-10 five times. Compare captures to original in DeltaWave. (DeltaWave will ensure pretty much perfect level-matching.)

Original-to-captures null to around -70dB.
Capture-to-captures null to around -110dB.

Draw your own conclusions.

Mani.

 

[DonH56]

The problem I have seen most often with capturing signals using an ADC is overdriving the ADC and/or insufficient (anti-alias) filtering. Level matching in a gain chain for an ADC means optimizing the signal into the ADC, not necessarily setting other components like DACs to full-scale or whatever. ADCs can respond differently to overloads, both in the distortion generated and how long they take to recover, so taking a look at the analog input (or inferring from the digital output) to ensure the ADC is not overdriven is key. Inter-sample overs and all that jazz can also be an issue.

That does not rule out another issue, of course.

FWIWFM - Don

 

[Blumlein 88]

I don't think it would be happening, but maybe if you're recording at 48 or 44 khz there is some ultrasonic noise causing some difference. So try again with 96 khz and see what happens. Maybe some interaction with filters.

 

[JamesQuorn0]

I don't think it would be happening, but maybe if you're recording at 48 or 44 khz there is some ultrasonic noise causing some difference. So try again with 96 khz and see what happens. Maybe some interaction with filters.

I've tried various sample rates and ADCs and the issues persist.

On a related note, I was speaking to a friend of mine fairly recently who has his own mastering studio. This topic came up and we both agreed that we'd had better results at lower sample rates. Why that is, I do not know.

 

[MaxwellsEq]

I've tried various sample rates and ADCs and the issues persist.

On a related note, I was speaking to a friend of mine fairly recently who has his own mastering studio. This topic came up and we both agreed that we'd had better results at lower sample rates. Why that is, I do not know.

That is strange and should not be correct. Higher sample rates either allow for wider bandwidth or less steep ADC filtering. It makes me wonder if there's some high frequency interaction going on. It might be worth experimenting with having lots of headroom (i.e. more than 6dB). Are you able to look at spectra before and after?

 

[kemmler3D]

I think run-to-run variation is a pretty plausible explanation here.

And to be pedantic, because someone always says it in these threads, you should be sure that the level matching is extremely precise. As I'm sure you know, very small changes in level still create noticeable loudness effects which affect perceived quality. AFAIK the levels need to be matched +/- 0.1dB or so to be really sure it's not just a level mismatch.

 

[Blumlein 88]

How did you level match?

 

[splitting_ears]

On a related note, I was speaking to a friend of mine fairly recently who has his own mastering studio. This topic came up and we both agreed that we'd had better results at lower sample rates. Why that is, I do not know.

This is not as counter-intuitive as it may seem.
Wider bandwidths (i.e. higher sample rates) could result in higher levels of intermodulation distortion when the signal hits any subsequent non-linearity, be it processing or playback systems.