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

[Jazigo]

Hi everyone!

I haven't been able to read all the post here, so please excuse me if the question has already been answered:

If the audio content have been produced (and verified) to have a -1 dBTP (in content), will the DACs still produce intersamples overs?

 

[MRC01]

... If the audio content have been produced (and verified) to have a -1 dBTP (in content), will the DACs still produce intersamples overs?

Yes; with peaks lowered to -1 dB intersample overs are less likely though still possible. IIRC, the question how low the peak level must be to make intersample overs impossible, even in worst case, has been answered here at ASR, but I don't remember what the value was.

Based the Shannon Whittaker reconstruction formula works, I would guess that -6 dB would be quite safe since the contribution from summing all the adjacent sampling point values attenuates quickly as they get further away, making them extremely unlikely to double the value at point being reconstructed. But that's just a guess...

 

[Jazigo]

Yes; with peaks lowered to -1 dB intersample overs are less likely though still possible. IIRC, the question how low the peak level must be to make intersample overs impossible, even in worst case, has been answered here at ASR, but I don't remember what the value was.

Based the Shannon Whittaker reconstruction formula works, I would guess that -6 dB would be quite safe since the contribution from summing all the adjacent sampling point values attenuates quickly as they get further away, making them extremely unlikely to double the value at point being reconstructed. But that's just a guess...

Thank you! I will look closer into the formula.

 

[High Power Dude]

Hi Amir,
I think might be aware of this issue with DACs and I think it would make a fascinating issue to cover: https://www.****************/2025/12/dac-industry-broken-filters-useless-chip-specs/

 

[NTK]

Hi Amir,
I think might be aware of this issue with DACs and I think it would make a fascinating issue to cover: https://www.****************/2025/12/dac-industry-broken-filters-useless-chip-specs/

The issue is a tempest in a teapot. The real problem is not with the DACs. It is with the recordings. Take a look at Archimago's recent post on the subject. Here is his conclusion.

A look at the music library: How much DAC intersample overhead is enough? True Peak analysis of genres in an eclectic digital collection.

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

Recordings like the ones shown in the top waveform plots will give you plenty of intersample overs. And they are the direct results of clipping the signals. How much do you think having a 3 dB margin for intersample overs will help to recover the original unclipped signals? Almost zilch is the answer. So why should I care about DAC intersample overs headroom, when there is an infinitely bigger problem intentionally created by the music producers that I cannot solve?

 

[High Power Dude]

Excellent point.

 

[popej]

So why should I care about DAC intersample overs headroom, when there is an infinitely bigger problem intentionally created by the music producers that I cannot solve?

I guess, with that convincing logic, you have made most of ASR topics irrelevant

 

[Mnyb]

I guess, with that convincing logic, you have made most of ASR topics irrelevant

Yes the hierarchy is usually the recording your room and then then the hifi ... but arguments can be made that with modern speakers with controlled directivity your brain can overcome some room issues ( except for bass ) and you can listen trough somewhat , but if your room is truly terrible what gives '?

 

[Sokel]

The issue is a tempest in a teapot. The real problem is not with the DACs. It is with the recordings. Take a look at Archimago's recent post on the subject. Here is his conclusion.

A look at the music library: How much DAC intersample overhead is enough? True Peak analysis of genres in an eclectic digital collection.

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

Recordings like the ones shown in the top waveform plots will give you plenty of intersample overs. And they are the direct results of clipping the signals. How much do you think having a 3 dB margin for intersample overs will help to recover the original unclipped signals? Almost zilch is the answer. So why should I care about DAC intersample overs headroom, when there is an infinitely bigger problem intentionally created by the music producers that I cannot solve?

I would expect more clipped samples of the original copy to tell the truth:

 

[MRC01]

The issue is a tempest in a teapot. The real problem is not with the DACs. It is with the recordings. ...

Exactly. When record producers intentionally abuse the format we shouldn't blame our equipment for it. It's ironic that article appears in a site with "honesty" in the name, because the honest appraisal is that there's nothing wrong with the equipment.

One could say that SOTA DACs have plenty of SINAD (120 dB or more) to handle this without issue. But designing equipment to handle this intentional abuse could be counterproductive in the long term since it only encourages them to squash it even more.

The DIY solution is to do what the engineers should have done in the first place and shift the recording levels downward. This can be done in real-time with digital volume controls in front of the DAC, or permanently by editing the recording. This is a 100% fix if the recording has no clipping but only intersample overs. But if it has clipping there is no recovery; information is lost/missing.

 

[tccalvin]

Hi everyone!

I haven't been able to read all the post here, so please excuse me if the question has already been answered:

If the audio content have been produced (and verified) to have a -1 dBTP (in content), will the DACs still produce intersamples overs?

Most likely not.

 

[Blew]

Exactly. When record producers intentionally abuse the format we shouldn't blame our equipment for it. It's ironic that article appears in a site with "honesty" in the name, because the honest appraisal is that there's nothing wrong with the equipment.

One could say that SOTA DACs have plenty of SINAD (120 dB or more) to handle this without issue. But designing equipment to handle this intentional abuse could be counterproductive in the long term since it only encourages them to squash it even more.

The DIY solution is to do what the engineers should have done in the first place and shift the recording levels downward. This can be done in real-time with digital volume controls in front of the DAC, or permanently by editing the recording. This is a 100% fix if the recording has no clipping but only intersample overs. But if it has clipping there is no recovery; information is lost/missing.

The record industry doesn't care what audiophiles want, unless they can sell the product at a premium, and even then they usually only cater to the snake oil audiophile. Otherwise, they only care what the average consumer wants. The average consumer wants loudly compressed digital audio that:
a) sounds great (to them) on a Bluetooth speaker or a phone
b) is cheap
c) is convenient to consume

The average consumer can't tell the difference between a well-mastered recording and a terrible one. A lot of that has to do with the low fidelity of their audio systems which mask mastering errors.

So, there's a large difference between what *should* happen and what *does* happen. The sad reality is that a lot of music is mastered over 0dBfs and clips horrendously. That makes it very uncomfortable to listen to, at least for me. This means we have two choices, either:
1) Design our audio systems for the best sound possible given what the record industry produces; or
2) Demand the record industry changes its nefarious ways to cater to our ideals.

Given the market power of ASR readers in comparison to the average consumer, I think you'll be waiting a very long time if you're expecting number 2 to happen. So, the most practical course of action is to go for option 1, which unfortunately includes having to mitigate intersample overs in our audio systems.

You don't necessarily need to buy a DAC that mitigates intersample overs internally as long as the DAC has adjustable output volume. In all my audio systems, wherever possible, I set the DAC to at least -3dB to mitigate intersample overs on the vast majority of CD recordings. Eg the Topping E30 II Lite in pre-amp mode. I also scan all downloads and CD rips using the Foobar2000 True Peak Scanner plugin to set the replaygain tags appropriately to avoid intersample overs when streaming to Chromecast and Airplay devices etc.

 

[NTK]

This means we have two choices, either:
1) Design our audio systems for the best sound possible given what the record industry produces; or
2) Demand the record industry changes its nefarious ways to cater to our ideals.

Fortunately, streaming platforms are doing this for us. They are normalizing loudness (by default), so the average listener will be listening to loudness normalized materials, which is taking away the incentive to participate in the loudness war.

For example:

Loudness normalization on Spotify - Spotify

support.spotify.com

 

[MRC01]

... The sad reality is that a lot of music is mastered over 0dBfs and clips horrendously. That makes it very uncomfortable to listen to, at least for me. ...

True. I stopped listening to new pop/rock and a lot of jazz for just that reason - it sounds so terrible and artificial it just hurts to listen to it. But classical music is not so affected, it's outside the blast radius of the loudness wars, thank goodness.

... You don't necessarily need to buy a DAC that mitigates intersample overs internally as long as the DAC has adjustable output volume. In all my audio systems, wherever possible, I set the DAC to at least -3dB to mitigate intersample overs on the vast majority of CD recordings. ...

This avoids clipping but the recordings still sound terribly artificial and fatiguing due to the heavy dynamic range compression.

... The average consumer wants loudly compressed digital audio that:
a) sounds great (to them) on a Bluetooth speaker or a phone
b) is cheap
c) is convenient to consume
The average consumer can't tell the difference between a well-mastered recording and a terrible one. A lot of that has to do with the low fidelity of their audio systems which mask mastering errors.

And that they can listen to on crappy earbuds in a noisy environment.
All this is true, but the best solution is to apply all that processing during playback. This would give those listeners everything they want or need, while preserving the music at its source, rather than crushing it. At the very least, apply the processing only to lossy streams like MP3 but not to lossless streams (or downloads, or physical discs) at CD or higher quality.

 

[Hayabusa]

The record industry doesn't care what audiophiles want, unless they can sell the product at a premium, and even then they usually only cater to the snake oil audiophile. Otherwise, they only care what the average consumer wants. The average consumer wants loudly compressed digital audio that:
a) sounds great (to them) on a Bluetooth speaker or a phone
b) is cheap
c) is convenient to consume

The average consumer can't tell the difference between a well-mastered recording and a terrible one. A lot of that has to do with the low fidelity of their audio systems which mask mastering errors.

So, there's a large difference between what *should* happen and what *does* happen. The sad reality is that a lot of music is mastered over 0dBfs and clips horrendously. That makes it very uncomfortable to listen to, at least for me. This means we have two choices, either:
1) Design our audio systems for the best sound possible given what the record industry produces; or
2) Demand the record industry changes its nefarious ways to cater to our ideals.

I guess 2 is the only option. Option 1 will not help you, the recording is already sounding bad, with or without correct handling of inter-sample overs.

 

[markstein]

Fortunately, streaming platforms are doing this for us. They are normalizing loudness (by default), so the average listener will be listening to loudness normalized materials, which is taking away the incentive to participate in the loudness war.

For example:

Loudness normalization on Spotify - Spotify

support.spotify.com

View attachment 504249

Unfortunately not available on Qobuz, so I terminated my test period and switched back to Tidal.

 

[Mike123]

This means we have two choices, either:
1) Design our audio systems for the best sound possible given what the record industry produces; or
2) Demand the record industry changes its nefarious ways to cater to our ideals.

There's a third option. Listen to properly mastered music. Fortunately, very much of it was released before 1990 that it's impossible to listen to it for a lifetime. I hardly listen to modern compressed music. It's optimized for listening outside with headphones. It's impossible to listen to it at home on a high-quality, powerful system.

 

[markstein]

There's a third option. Listen to properly mastered music. Fortunately, very much of it was released before 1990 that it's impossible to listen to it for a lifetime. I hardly listen to modern compressed music. It's optimized for listening outside with headphones. It's impossible to listen to it at home on a high-quality, powerful system.

Abandoning certain music (the artistic part) at all because of bad mastering (the technical part) would be ridiculous.

If there is a way to make badly mastered music listenable in better quality, we should use it.

 

[RexrothPigeon]

Abandoning certain music (the artistic part) at all because of bad mastering (the technical part) would be ridiculous.

Yep, it's the kind of thing that helps me understand why so many people deride audiophiles

 

[melowman]

It is funny and astounding to see, again, after all what’s been discussed here, people associate ISP overs with digital loudness.