Let's develop an ASR inter-sample test procedure for DACs!

[JSmith]

I am not sure how the volume on the SMSL DO100 works but it does not appear to work in the digital domain

The DO100 uses digital "volume" attenuation as far as I'm aware.


JSmith

 

[levimax]

The DO100 uses digital "volume" attenuation as far as I'm aware.


JSmith

If so then why would I hear obvious digital clipping even when the volume is set to 20 out of 99? I hear the exact same clipping at the exact same spots in the sweep whether the volume is set to 10 or 99. If I set the attenuation to -8 dB either in the convolver or in Foobar2000's volume control then the clipping stops.

 

[amirm]

After reading this thread I thought rather than use the attenuation in the convolver I would just set the convolver attenuation to 0 dB and turn down the volume on the DAC until the 0 dB sweep did not audibly clip. This did NOT work.

Of course it doesn't work. Unless the two subsystems are integrated, you need headroom allowance in each part of the pipeline. Once you clip in the convolver, there is nothing the DAC can do to correct that.

 

[levimax]

Of course it doesn't work. Unless the two subsystems are integrated, you need headroom allowance in each part of the pipeline. Once you clip in the convolver, there is nothing the DAC can do to correct that.

OK that makes sense.... never really thought about "convolver clipping" as something separate from other digital clipping before. So that means there are multiple sources of potential digital clipping. Do you think a 0 dB sweep is a good test of the entire system or is it "too strict"?

 

[amirm]

Do you think a 0 dB sweep is a good test of the entire system or is it "too strict"?

It is as energy in music drops off quickly above bass. So I test with music that has strong bass and if that doesn't clip, nothing else will either. You usually see that in my headphone EQ settings.

 

[KSTR]

It is as energy in music drops off quickly above bass. So I test with music that has strong bass and if that doesn't clip, nothing else will either. You usually see that in my headphone EQ settings.

Averaged energy yes, but for short transients there is no real frequency dependence.

 

[popej]

Us forcing it will mean all DACs will eventually lose 3 dB of dynamic range or whatever this needs to be.

It took me some time to think about it
It is not true. You get 3dB dynamics at high level and lose 3dB at noise level. It affects standard dynamic measurements, but not DAC capabilities.

I looked at one song by Cranberries, since I know they are heavy clipped. Oversampling in Audacity showed many many ISP there, up to 1dB. I expect this is quite common. Without headroom, you lose this 1dB. Maybe it is not audible, but +3dB of noise at -100dB is not audible either.

3dB headroom is 1/2 bit of resolution. Seems an analog to 1/2 bit used for dither. Both make the output technically correct.

 

[Lambda]

Us forcing it will mean all DACs will eventually lose 3 dB of dynamic range or whatever this needs to be. So best think about what you are asking.

You can also say the opposite "Us" caring about 1khz THD+N so mush we will eventually lose all headroom for ISO because manufacturers fighting to get the higest SINAD number
While it is audible completely irrelevant if a DAC has 123.1dB or 123.8dB SINAD

There are may DACs Starting from 100$ That have more then enough dynamic.
In the >300$ range DACs starting to reach the limit of what the AP analyzer can resolve.

There are many many DACs that can almost perfectly (audible perfectly) play a 1k sin wave with low distortion and noise.
So why not add a test to differentiate between them?

 

[mdsimon2]

Non-computer source. And if your DAC has an ASRC upstream of the volume control (cough, cough miniDSP) then what?

Michael

@amirm Just in case you didn't respond to this question because you didn't have the data, here it is.

11.025 kHz +3 dBFS tone sent to a miniDSP 2X4HD at a device volume control position of 0 dB.

Changing the device volume control position to -4 dB decreases the signal level but doesn't solve the nastiness because the ASRC is upstream of the volume control.

This behavior is common to all miniDSP products, including the SHD and Flex which were well reviewed here. miniDSP thinks this is a serious enough issue that they are attempting to modify their firmware to put volume control upstream of the ASRC, unfortunately their attempts to do this so far have also come with other issues. It remains to be seen if they can correct this issue without some sort of hardware change.

In almost every DAC review members are clamoring for manufacturers to add DSP to DACs, if / when they do this if they also add an ASRC this may become a much more common issue.

Michael

 

[Ken Tajalli]

I must admit I haven't fully followed this thread.
The intersample overload should only be an issue on an oversampling DAC, so I am sure the engineer who design them would allow for that.
But music produced by engineers should not suffer any IS overs. Music is captured at high sampling rates at 32bit res. to allow for a lot of headroom.
The engineer would allow for clipping before decimation to 24 or 16 bit for final output.
BTW, digital volume control can not always guarantee protection against clipping, I think.
If the signal is processed for oversampling before volume control, then clipping may have occured already!
If the player does any upsampling, then Amir's remedy of -3db takes care of it. If the DAC does the upsampling, the designer should allow the 3dB headroom.
I know, all Chord DACs, which do upsample, do allow for a 2.76dB headroom, partly for IS over runs. The volume control comes after DSP.

 

[melowman]

Why? Roon player will output correct values over SPDIF just as well. Or do you mean from a non-computer source? If so, once again, buy a DAC with volume control. That is the correct solution regardless of this issue anyway.

Lots of contradictions overall, if I may:

. You don’t want to sacrifice SNR, but you do apply a constant negative gain. And you are overall all aware that applying DSP requires to make some headroom.

. You don’t want people to buy new DACs because of inter-sample overs (you want a software solution), but you do say ‘buy a DAC with digital control’. So why not buy a DAC which gracefully handles ISOs with some built-in headroom? What if a DAC with digital volume control is not an option to buy?

. You - and many here - also fall in the trap of connecting loud music and ISOs; I’m pretty perplex as to why it is so. You get plenty of ISOs with all kinds of loudness. And the ‘funny’ part is that when they clip, they have a higher probability to be heard with softer music! Louder material is less problematic regarding ISO mishandling.
It is basically completely off to bring the loudness discussion on the table.

-
The software solution is not viable because issues in DACs are tailored to each DAC design. You don’t want to apply a software attenuation for a DAC which don’t have ISO misbehavior, do you? So how do we know if our DAC misbehaves? By testing it!

Also, a software solution, which would attenuate the volume on a song by song basis, isn’t viable because it would create volume discrepancies that would go all over the place.

Also, what if other DSP is applied (by the operating system for example) after you’ve lowered the volume in your software?

-
I think that it is very wise to say that the design of a DAC has to take into account any DSP they decide to apply. Oversampling is one. As you said, when applying EQ you leave enough room; so when a DAC design applies oversampling, why doesn’t it leave enough headroom? PCM signals coming in are perfectly valid, but some DAC designs mess them up with their own DSP.

The responsibility is clearly defined. (and clearly, it has nothing to do with loudness, as said above)

 

[MRC01]

... You - and many here - also fall in the trap of connecting loud music and ISOs; I’m pretty perplex as to why it is so. You get plenty of ISOs with all kinds of loudness. And the ‘funny’ part is that when they clip, they have a higher probability to be heard with softer music! Louder material is less problematic regarding ISO mishandling.
It is basically completely off to bring the loudness discussion on the table.

I never saw a realistic ISO example (music, not a test signal) that could not be fixed by lowering the level a few dB. Do you have examples of music recordings having digital peak levels -3 dB or lower that have ISOs?

 

[melowman]

You’re talking peak, I’m talking loudness (RMS/LUFS).

 

[MRC01]

You’re talking peak, I’m talking loudness (RMS/LUFS).

That's fine, my request remains.

My point is, whatever the loudness levels, if you limit the peak levels (use overall recording levels that give you a few dB of headroom) you won't get ISOs, or at least they will be so rare that the whole issue is a nothing sandwich. I'm willing to change my belief based on evidence, but every ISO example I've seen has high peak levels, which is why I ask for examples.

 

[melowman]

(You keep talking peak, I’m talking loudness.)

You, as well as I, as well as Amir, the OP, etc. do say there’s a headroom issue. Now the question: who’s making the move?

So far, 3 solutions are discussed:
- leave headroom in DAC designs (OP’s position, mine too, John from Benchmark’s also, etc., other manufacturers too like the one you mentioned)
- pre-attenuate the volume in digital audio software (Amir’s position)
- pre-attenuate the volume with a DAC having a digital volume control (Amir’s too, etc.)

Another position is: leave as is, there’s no issue. (“I can’t hear it” kind of non-issue, or “I don’t believe there’s an issue” kind of non-issue).

 

[MRC01]

(You keep talking peak, I’m talking loudness.)

I talk peak because it demonstrates that ISOs are intrinsically related to loudness wars. Any music recording that has ISOs will be "fixed" simply by reducing its overall level by a few dB. This is true no matter its loudness/RMS/LUFS. And because ISOs are so easy to avoid, when they happen it must be intentional (assuming most recording & mastering folks are competent professionals).

You mention 3 possible solutions - I'll add one more, the "do nothing in the DAC" solution:

The "headroom issue" is not a limitation of the recording format or the DAC. The root cause is recording & mastering folks intentionally abusing the recording format to make recordings as loud as possible. Addressing this in the DAC doesn't address the root cause, but only puts a band-aid over the effects. Doing so would likely be counterproductive, since if we "fix" it in the DAC then those folks will only up the ante and make recordings even louder.

 

[splitting_ears]

The root cause is recording & mastering folks intentionally abusing the recording format to make recordings as loud as possible.

Do you know when the first international standard (recommendation, really) for true-peak metering was created?
2006. That is more than two decades after the commercial introduction of CDs.

Even today, most DAWs on the market only display sampled values and require the user to manually change them to true-peak. Oversampling is also likely to be disabled by default in any DSP processor, to accommodate lower-end computers.
In practice, this means that you can work with too little headroom... and never see a warning, even with conservative production choices.

Take this well known track for example. Does it look/sound smashed to you?

But don't make me say things I didn't say. This is just to illustrate the idea that ISPs are not merely a by-product of compression and limiting. Although, in this particular example, I wouldn't care about only 5 clipped samples out of almost 12 million. See, I'm not that unreasonable after all

Yes, smashing things creates more 'opportunities' for ISP (simply because you spend most of the time close to the maximum), but that is *not* the only reason. And yes, we can rant all we want about engineers not using enough headroom. But it won't change the way music has been created and distributed for decades.
Fortunately, most things have changed for the better: true-peak is now a more familiar concept, oversampling is easier to implement, content normalisation is widespread, and computing resources allow better-sounding implementations.

Sorry to pick on you, but please don't perpetuate and repeat the tired cliché that audio engineers deliberately push things because they just enjoy destroying audio. And remember, they are first and foremost service providers, whose clients are artists and labels. They get the last word because they pay for the service.

Again, ISPs are not a 'problem' per se, but a logical consequence of the sampling theorem, which requires headroom at every stage because the sampled values are always less than the original/reconstructed values. There's no debate about this fact. The question is what to do with this knowledge, especially for music released at a time when this was not an intuitive idea.

I've seen Amir's replies (thanks, by the way), and while I disagree with the SNR argument (@popej expressed my opinion perfectly in post #167), I understand how difficult it is to find the right stance on this topic for ASR reviews as a whole. But what I do know for sure is that it was extremely informative and entertaining to discover @edechamps' reviews, in which he made great use of ISP test signals to better understand the behaviour and overall design choices of the equipment he was testing

 

[Adis]

I talk peak because it demonstrates that ISOs are intrinsically related to loudness wars. Any music recording that has ISOs will be "fixed" simply by reducing its overall level by a few dB. This is true no matter its loudness/RMS/LUFS. And because ISOs are so easy to avoid, when they happen it must be intentional (assuming most recording & mastering folks are competent professionals).

You mention 3 possible solutions - I'll add one more, the "do nothing in the DAC" solution:

The "headroom issue" is not a limitation of the recording format or the DAC. The root cause is recording & mastering folks intentionally abusing the recording format to make recordings as loud as possible. Addressing this in the DAC doesn't address the root cause, but only puts a band-aid over the effects. Doing so would likely be counterproductive, since if we "fix" it in the DAC then those folks will only up the ante and make recordings even louder.

Of course we will.

 

[Adis]

Do you know when the first international standard (recommendation, really) for true-peak metering was created?
2006. That is more than two decades after the commercial introduction of CDs.

Even today, most DAWs on the market only display sampled values and require the user to manually change them to true-peak. Oversampling is also likely to be disabled by default in any DSP processor, to accommodate lower-end computers.
In practice, this means that you can work with too little headroom... and never see a warning, even with conservative production choices.

Take this well known track for example. Does it look/sound smashed to you?

View attachment 323038

But don't make me say things I didn't say. This is just to illustrate the idea that ISPs are not merely a by-product of compression and limiting. Although, in this particular example, I wouldn't care about only 5 clipped samples out of almost 12 million. See, I'm not that unreasonable after all

Yes, smashing things creates more 'opportunities' for ISP (simply because you spend most of the time close to the maximum), but that is *not* the only reason. And yes, we can rant all we want about engineers not using enough headroom. But it won't change the way music has been created and distributed for decades.
Fortunately, most things have changed for the better: true-peak is now a more familiar concept, oversampling is easier to implement, content normalisation is widespread, and computing resources allow better-sounding implementations.

Sorry to pick on you, but please don't perpetuate and repeat the tired cliché that audio engineers deliberately push things because they just enjoy destroying audio. And remember, they are first and foremost service providers, whose clients are artists and labels. They get the last word because they pay for the service.

Again, ISPs are not a 'problem' per se, but a logical consequence of the sampling theorem, which requires headroom at every stage because the sampled values are always less than the original/reconstructed values. There's no debate about this fact. The question is what to do with this knowledge, especially for music released at a time when this was not an intuitive idea.

I've seen Amir's replies (thanks, by the way), and while I disagree with the SNR argument (@popej expressed my opinion perfectly in post #167), I understand how difficult it is to find the right stance on this topic for ASR reviews as a whole. But what I do know for sure is that it was extremely informative and entertaining to discover @edechamps' reviews, in which he made great use of ISP test signals to better understand the behaviour and overall design choices of the equipment he was testing

Recommendations aren't obligatory, make it the Law! Make it punishable, and then we'll listen. Maybe.

 

[MRC01]

... But don't make me say things I didn't say. This is just to illustrate the idea that ISPs are not merely a by-product of compression and limiting. Although, in this particular example, I wouldn't care about only 5 clipped samples out of almost 12 million. See, I'm not that unreasonable after all

I agree - 5 ISOs among 12 million samples (4.5 minute song) is not a practical problem. And the track you showed looks way better that many others that I have seen. If they were all like that, we would not be having this discussion. But even those 5 ISOs will disappear if you lower the overall level by 1-2 dB.

Again, show me a music track (not a test signal) with peak levels -3 dB or lower having ISOs. I've never seen one, what I know of reconstruction theory tells me they would be extremely rare and I wonder if any real-world examples exist at all.

... Yes, smashing things creates more 'opportunities' for ISP (simply because you spend most of the time close to the maximum), but that is *not* the only reason. And yes, we can rant all we want about engineers not using enough headroom. But it won't change the way music has been created and distributed for decades.

Which is another way of saying they are causing the problem by recording too loud. Put differently, there are 2 ways ISOs happen:

1. The recording or mastering engineer makes the recording too loud.
2. DA reconstruction raises a waveform more than 3 dB above a sampled amplitude.

(2) has only been demonstrated with carefully constructed test signals, and while theoretically possible with music, is so rare we don't have any actual examples. Even if they did exist, they would be so rare as to be a non-issue.

... Sorry to pick on you, but please don't perpetuate and repeat the tired cliché that audio engineers deliberately push things because they just enjoy destroying audio. And remember, they are first and foremost service providers, whose clients are artists and labels. They get the last word because they pay for the service.

It's not a tired cliche, but a well known fact that audio engineers in certain genres of popular music are deliberately pushing things to sound as loud as possible. Just look at a random sample rock/pop songs produced or remastered in the last 5-10 years and you'll hear it and see it in wave analysis. Plenty of examples posted here at ASR by myself and many others. Just because they do this intentionally, does not imply they enjoy it. Even if they are serving the desires of their studios & artists, who want their song to sound louder than the others people listened to on their streaming service, they are contributing to the vicious cycle of the loudness wars.

If you are one of the exceptions, then congratulations, you have my respect. I wish there were more like you.

... I've seen Amir's replies (thanks, by the way), and while I disagree with the SNR argument (@popej expressed my opinion perfectly in post #167), I understand how difficult it is to find the right stance on this topic for ASR reviews as a whole. But what I do know for sure is that it was extremely informative and entertaining to discover @edechamps' reviews, in which he made great use of ISP test signals to better understand the behaviour and overall design choices of the equipment he was testing

All this discussion aside, I share your curiosity about how different DACs handle this.