DAC audible differences are finally proven with ABX?

[mcdn]

Still remains polarity. Devices have shown to be variable in this regard. Although this device seems to retain it. But DACs can vary.

Good point. Polarity switching is easy for balanced inputs (just add more relays), but complicated for single ended, as it requires an inverting buffer.

 

[RandomEar]

I'm catching up and saw this. And, oh no...

And there's that one too:

View attachment 547637

So, with a 2Vrms input and an INL of -0.1LSB to -0.3LSB, even if the full scale output is level matched at 0.1dB, the above means a potential variance of up to 0.2dB at -20dBFS, that would be a quieter moments in music. That with high distortion and crosstalk too, also virtually increasing the noise floor of one of the two devices under test if they required level matching.

No need to investigate further, that alone means the real device under test was the box, but we suspected it from the beginning.

Select a fade to noise moment, push the volume and get 100% result.

All the DACs in use acted as "perfect generators" to feed this low resolution device.

Bottom line: It was a not so interesting testing of the AD5262 digital potentiometer. Reading the datasheet would have been enough to guess the final results in the context. Like Erin wrote, the box can be useful to play with 2 amplifiers of very different output impedances to hear something. Beyond that...

Could you elaborate a little bit what the graphs shows? I'm not familiar with this form of presenting linearity errors. From what I could get from the datasheet and other posts in this thread, this was measured for a fixed input signal with the attenuation (which is defined by the CODE) being varied, right?

So let's say that we operate at CODE 48, where the INL is around 0.6 bits: Does this mean our linearity is skewed by the same factor over the full input voltage range? Or are lower voltages affected proportionally more than higher ones, because the error is an offset and not a factor?

 

[welwynnick]

I'm catching up and saw this. And, oh no...
And there's that one too:
View attachment 547637
So, with a 2Vrms input and an INL of -0.1LSB to -0.3LSB, even if the full scale output is level matched at 0.1dB, the above means a potential variance of up to 0.2dB at -20dBFS, that would be a quieter moments in music. That with high distortion and crosstalk too, also virtually increasing the noise floor of one of the two devices under test if they required level matching.

Yes, it's an ugly looking chart, but as KSTR has already explained, it doesn't matter. The relatively low precision of the resistance setting is inside the amplitude control loop.
The potentiometer is set to whatever it has to be set to, in order for the output levels to be within 0.1dB as measured by an external device.
In Josh's application and use, it's a fixed, closed loop, control of the output level, not an open loop control.

 

[solderdude]

1LSB = the smallest step = 0.033dB attenuation per step so there is about 15% variation between the intended and measured step. Small enough not to matter at least not for channel matching ?
Once set, during the level adjustment which is monitored by external metering so even if the variance was large it can still be set correctly and after that the resistance won't change.
I don't see the potmeter as a problem nor the relays for these low level signals.

 

[mcdn]

Good point. Polarity switching is easy for balanced inputs (just add more relays), but complicated for single ended, as it requires an inverting buffer.

Although, if we are talking about DACs then the signal can be inverted at the digital source.

 

[solderdude]

Although, if we are talking about DACs then the signal can be inverted at the digital source.

Yes, but the point is it should be detected before a test was done and if the test were about listening if inverse polarity is audible the signal would have to be inverse polarity (easy enough for electronics to do)

 

[pjug]

I don't think/suppose that @Josh83 is likely to help invalidate his test by either posting his test files or recorded output. Because if that were to show some tells his show is over and needs to revise his article and redo it in a way some nerds at ASR would like it.
I hope I am wrong about this and at least posts the actual up/down sampled test files he used.
That would be a good place to start to investigate the anomalous findings.

From there it would probably be a good idea to record the output of each DAC (no ABX box yet) and see if there is clipping or ISO not handled correctly.
Step, by step.

That could potentially show why (if tells were present) how he could tell them apart which would at least validate his findings.

@Josh83 did say he will follow up and that should answer some questions. But it seems to me it would be better to just make a few recordings and share them. Then we could eliminate some possibilities and focus on any remaining ones that show up.

 

[welwynnick]

OK, I'm going to think about this (a SOTA ABX hardware frontend for balanced and unblalanced line level signals, to be used with a computer -- via isolated USB-MIDI, for example, for control interface -- where the control software runs). But it would be a slow process as I still have a day job as electronic design engineer. I would provide stuffed and tested PCBs, to be built into some case.

That's a critical point indeed. Precision A/B level matching requires better precision L/R channel matching to begin with, but that sort of alters the DAC when viewed as a black box. Same for polarity inversion.

The notion of routing the output of a state of the art DAC through a much lower quality analog device just doesn't sit well with me. When testing products like this, I listen to very low level signals that I then highly amplify. A comparator which adds noise and linearity errors totally messes with this.

Maybe one could use a pair of good audio pre-amps to do the level-matching? These days there are more and more pre-amps that use relay-switched-resistor network volume controls, though most of those have 0.5 or 1.0 dB precision. They have the advantage of good L/R balance control.

There are a few with 0.1dB precision, like Esoteric Grandioso C1X, Mark Levinson No 326S or TAD C700, but they're a bit pricy. The Topping Pre900 isn't exactly cheap, but it does have 0.1dB volume steps, it's available off the shelf, and it's performance is a known quantity. Just a thought.

You can't use just one of them for level matching, because each signal path must be the same.

Thinking out loud a bit more, you might not need a fancy pre-amp, just one with a conventional volume control. The requirement is to set the levels to within 0.1dB, and any good preamp might be able to do that, as long as you measure the output level right. A pair of Topping L30 II anyone? Then it simplifies to a switching task.

Alternatively, use a pair of preamps ahead of the inputs to an AVA ABX Comparator, and by-pass the attenuators altogether.

 

[Plcamp]

Would a top performance opamp with a field of about 6 precision resistors selected by dip switches not be the simplest very low noise and distortion solution for level matching? Could also use an opamp to invert signal.

 

[Roland68]

I'm catching up and saw this. And, oh no...

And there's that one too:

View attachment 547637

So, with a 2Vrms input and an INL of -0.1LSB to -0.3LSB, even if the full scale output is level matched at 0.1dB, the above means a potential variance of up to 0.2dB at -20dBFS, that would be a quieter moments in music. That with high distortion and crosstalk too, also virtually increasing the noise floor of one of the two devices under test if they required level matching.

No need to investigate further, that alone means the real device under test was the box, but we suspected it from the beginning.

Select a fade to noise moment, push the volume and get 100% result.

All the DACs in use acted as "perfect generators" to feed this low resolution device.

Bottom line: It was a not so interesting testing of the AD5262 digital potentiometer. Reading the datasheet would have been enough to guess the final results in the context. Like Erin wrote, the box can be useful to play with 2 amplifiers of very different output impedances to hear something. Beyond that...

Maybe one could use a pair of good audio pre-amps to do the level-matching? These days there are more and more pre-amps that use relay-switched-resistor network volume controls, though most of those have 0.5 or 1.0 dB precision. They have the advantage of good L/R balance control.

There are a few with 0.1dB precision, like Esoteric Grandioso C1X, Mark Levinson No 326S or TAD C700, but they're a bit pricy. The Topping Pre900 isn't exactly cheap, but it does have 0.1dB volume steps, it's available off the shelf, and it's performance is a known quantity. Just a thought.

You can't use just one of them for level matching, because each signal path must be the same.

Thinking out loud a bit more, you might not need a fancy pre-amp, just one with a conventional volume control. The requirement is to set the levels to within 0.1dB, and any good preamp might be able to do that, as long as you measure the output level right. A pair of Topping L30 II anyone? Then it simplifies to a switching task.

Alternatively, use a pair of preamps ahead of the inputs to an AVA ABX Comparator, and by-pass the attenuators altogether.

Would a top performance opamp with a field of about 6 precision resistors selected by dip switches not be the simplest very low noise and distortion solution for level matching? Could also use an opamp to invert signal.

Any type of active electronics will influence and distort the result, no matter how good the circuitry is.
This will make every test vulnerable to attack, in all directions.

For this reason, our tests place only two resistors (forming a voltage divider) for precise voltage adjustment and a switching element (a high-quality toggle switch, bistable relay, or high-quality DC relay) in the signal path. Nothing more has ever been necessary.

 

[danadam]

however, in our comparative tests, we aim for the highest possible accuracy at the third decimal place

For this reason, our tests place only two resistors

Who exactly is "we", if I may ask?

 

[Plcamp]

Any type of active electronics will influence and distort the result, no matter how good the circuitry is.
This will make every test vulnerable to attack, in all directions.

I don’t agree (unless the resulting comparator design fails to pass a transparency test on its own).

A tested and proven transparent comparator could not be validly deemed ‘vulnerable’ any more than an lpad.

 

[Roland68]

Who exactly is "we", if I may ask?

It is a project involving—among other things—highly precise, level-matched blind tests of equipment, cables, accessories, and so on. The goal is to categorize for ourselves what has an audible impact—and is relevant in terms of measurements—and what doesn't. To put it very simply.
We have been doing this for 26 years now, and we certainly paid our dues in the early years—especially regarding what actually influences test results and what doesn't.

I don’t agree (unless the resulting comparator design fails to pass a transparency test on its own).

A tested and proven transparent comparator could not be validly deemed ‘vulnerable’ any more than an lpad.

Everyone is entitled to their own opinion on the matter. But if you choose to take the easy way out, you have to live with the consequences.
My own view is based on well over a thousand such tests, involving more than a hundred guests over the years who initially questioned everything. On top of that, there have been numerous tests conducted with developers, manufacturers, and companies from various sectors.

 

[Jazigo]

This is a bit of a discussion for a different topic. In the context of this thread, nothing much has been measured at all about the comparator, which is one of the points being criticized.

However, in general the frequency domain and the time domain are interchangeable. That's why you can do a Fast Fourier Transform (FFT) on your data, follow up with an inverse FFT and - apart from tiny errors due to numerical limitations - end up with an identical signal in the time domain.

But there's a but: FFTs do use window functions, so their time resolution is limited. And all the graphs we typically see are from relatively long, time averaged FFTs over seconds and sometimes maybe minutes. That increases the SNR, which is desired. However, if you assume that a device has a hypothetical small glitch right at the beginning of the measurement, that glitch would be "averaged out" the longer you capture data for your individual time-averaged measurement. That's just how it is and is technically not even related to the fact that the frequency domain is being analyzed.

In addition, you do need to see the phase response on top of the frequency response to get the full picture in the frequency domain. That is rarely shown here on ASR. I would assume that's because it is very uneventful most of the time. If you analyze in the time domain and just measure the impulse response, you'll get the full transfer function of the device in a single plot which is very convenient. I'm sure that some types of defects are easier to spot in one domain than in the other, so there is value in both paths of analyis.

But now back to proposed audible differences in DACs!

I am giving Josh the benefit of the doubt and was sort of looking for things that could be missed, in which, could explain why a DAC is not transparent. While separate tests of frequency response, sinad and jitter, could suffice to rule out any doubt, they are still separate tests, which leaves we wondering if something can be missed between those tests? Removing any doubt what so ever would be to do actual measurements in the time domain and null test it. Basically, play a song through a DAC, record it with an ADC an null test the source vs the capture.

I tried a one point to build such a system (with a Topping E50 and a PocketADC) and while I could easily get -98.7 db thd+n (on 16 bit data) in the frequency domain, I could only get a time-domain error (THD+N + Jitter + Alignment): -67 dB null (~0.043% error) on average.

I could not get lower due to clock drift between the DAC and the ADC, but say that I had a DAC and ADC with shared clock, maybe it would only be -80 db or something? That could be audible(?).

 

[welwynnick]

Any type of active electronics will influence and distort the result, no matter how good the circuitry is.
This will make every test vulnerable to attack, in all directions.

I agree about making test equipment as simple, robust and high quality as possible, but there are several reasons for suggesting a pre-amp.

• The quality of the AVA ABX comparator is being called into question; the level-matching variable resistors in particular.
• Pre-amps are know quantities - ASR has tested many of them, and we can have much more confidence that they perform well, compared with the AVA.
• When you dig deep, it's very clear that good pre-amps are actually better than good DACs.

Their SINAD values are only similar because they're limited by the quality of the analogue test signal. DACs don't have that limitation. Good pre-amps are better than any source and any power amp. They're not the weak link in the chain. There's lots of evidence. Here's my favourite test, the Topping A70 Pro. It measures as good as any switch or cable, and better than any DAC. Plus it has >100kHz bandwidth, at least 127dB DNR, high input impedance, low output impedance, and near perfect volume control.

 

[NTTY]

Yes, it's an ugly looking chart, but as KSTR has already explained, it doesn't matter. The relatively low precision of the resistance setting is inside the amplitude control loop.
The potentiometer is set to whatever it has to be set to, in order for the output levels to be within 0.1dB as measured by an external device.
In Josh's application and use, it's a fixed, closed loop, control of the output level, not an open loop control.

Oh yes, very true, me stupid It's a fixed setting in the VR latch and does not vary after initial adjustment. I shall strike through what I wrote before.

Question: since there are two AD5262 for the two channels, assuming the VR has been set in a non-identical INL zone (assuming one is -0.1LSB and the other -0.3LSB), are we not in a similar case of increasing difference between the two outputs as the input voltage lowers from the two DACs (my initial case at -20dBFS)?

Heeelp, I need a break maybe

EDIT: all of that questioning while it would be so quick and easy to measure directly...

 

[RandomEar]

I am giving Josh the benefit of the doubt and was sort of looking for things that could be missed, in which, could explain why a DAC is not transparent. While separate tests of frequency response, sinad and jitter, could suffice to rule out any doubt, they are still separate tests, which leaves we wondering if something can be missed between those tests? Removing any doubt what so ever would be to do actual measurements in the time domain and null test it. Basically, play a song through a DAC, record it with an ADC an null test the source vs the capture.

In essence, you can always test more. A null test is a very direct and easy to grasp proof of identity, but in practice it's also surprisingly difficult to do well.

I tried a one point to build such a system (with a Topping E50 and a PocketADC) and while I could easily get -98.7 db sinad (on 16 bit data) in the frequency domain, I could only get a time-domain error (THD+N + Jitter + Alignment): -67 dB (~0.043% error) on average.

I could not get lower due to clock drift between the DAC and the ADC, but say that I had a DAC and ADC with shared clock, maybe it would only be -80 db or something? That could be audible(?).

Yeah, as far as I have seen you need a master clock and there are various pitfalls. I wouldn't recommend that to any person who hasn't done it before or has access to expert advice while setting it up. I wouldn't do it myself for serious comparisons, because I lack the experience. Here's a good example on ASR with some useful discussion.

 

[Roland68]

I agree about making test equipment as simple, robust and high quality as possible, but there are several reasons for suggesting a pre-amp.

• The quality of the AVA ABX comparator is being called into question; the level-matching variable resistors in particular.
• Pre-amps are know quantities - ASR has tested many of them, and we can have much more confidence that they perform well, compared with the AVA.
• When you dig deep, it's very clear that good pre-amps are actually better than good DACs.

Their SINAD values are only similar because they're limited by the quality of the analogue test signal. DACs don't have that limitation. Good pre-amps are better than any source and any power amp. They're not the weak link in the chain. There's lots of evidence. Here's my favourite test, the Topping A70 Pro. It measures as good as any switch or cable, and better than any DAC. Plus it has >100kHz bandwidth, at least 127dB DNR, high input impedance, low output impedance, and near perfect volume control.

View attachment 547690

Perhaps the function and architecture of an active preamplifier aren't entirely clear.
Input buffer -> Gain stage (single- or multi-stage) -> Output buffer
In between, there is volume control and, potentially, a DC servo or coupling capacitors.
That amounts to a lot of components in the signal path, especially in a high-quality circuit. With balanced cabling, you also add input desymmetrizing stages and output symmetrizing stages.
Every single one of these stages will minimally affect the signal and add noise; there is nothing you can do about that, no matter how good the preamplifier is. That’s physics.

With a good DAC, only the output buffer is needed at the output stage—a buffer capable of holding its own against even the best preamplifiers; nothing else is required.
What makes you think an additional preamplifier might be better in this context? It cannot improve the signal coming from the DAC in any way. How could it possibly do so?

Our setup for DAC testing looks like this:
DAC -> first voltage divider -> second voltage divider -> selector switch -> amplifier
No potentiometer, no other volume control, no op-amp, no capacitors or other components in the signal path. It doesn't get more purist or free from interference than that.

 

[KSTR]

Perhaps the function and architecture of an active preamplifier aren't entirely clear.
Input buffer -> Gain stage (single- or multi-stage) -> Output buffer
In between, there is volume control and, potentially, a DC servo or coupling capacitors.
That amounts to a lot of components in the signal path, especially in a high-quality circuit. With balanced cabling, you also add input desymmetrizing stages and output symmetrizing stages.
Every single one of these stages will minimally affect the signal and add noise; there is nothing you can do about that, no matter how good the preamplifier is. That’s physics.

You can get away with one single stage for all of that, when you can live with a limited gain range (not a true preamp function). With well-designed opamp composites there is absolutely no measurable distortion when all things of the complete design fall into place. Noise is excellent (assuming OPA828 masters) even with moderate/high-Z sources but could be reduced below the self-noise of a 50R resistor for very low-Z sources in case anyone had doubts.
Introducing the "High-Z Rebalancer" with 6...(N+6)dB gain using balanced output, or 0...NdB gain with single-ended output (taking the output from either leg). Gain is set by a single trim resistor (which could be switched type or a precision pot, for a trim range of a few dB). (Note: conceptual and very simplified schematic, but quite close to a working circuit I have).

I'd use input AC-coupling caps that can stay precharged, before the source switching, that't the only way that works. AC-coupling or DC-servo after the source switching doesn't.
The amp would have low offset by design but would offer manual trims for that.
If gross attenuation is required one would need to add buffers after a balanced divider, or inverters with lower than unity gain. Since we have a low impedance environment, a composite inverter with a low noise BJT master (say, a Groner-Polak composite) would neither add noise nor distortion. I'd probably make it strong enough to double up as the monitor headphone amplifier.

Other schemes are possible as well, for example a standard high-Z instrumentation amp front-end (or just simple high-Z buffers) feeding a low-Z FDA-based rebalancer, forming a two-stage design with great flexibility.

Non-technical readers, please excuse all this nerd stuff ;-)

 

[voodooless]

For an ABX box for DACs I would probably opt for a more digital approach. Take the iDSP that @Weeb Labs made as a basis: spdif input, dual spdif output. Those are the A and Bs. Now you can use the software DSP to trim each individual channel, both in volume and delay. This should be more than transparent enough. Add a simple ADC, and a bit of extra software, and you can have it match the DACs perfectly automatically. Next you’ll need no additional active parts. Just the bits that cleanly switch the DAC analog outputs without tells. Obviously this may not be good enough if the output voltages are too far apart, but in most cases this should work out perfectly fine.