PEACE clipping meter question!

[Robbo99999]

@Robbo99999 I don't think lowering your windows volume will help you here. Once a sample has reached beyond 0dbfs, any information above is lost – in other words, it clipped. the result is that the clipped part is sent to conversion as 0dbfs, or as the highest maximum digital value. I believe the information lost has already happened by the time it's sent to the windows mixer, so lowering the volume won't help to regain that lost information.

I did the following experiment: I took a song which had many clipped samples (reconstructed peak at about +1.5 dbfs), and made a loopback recording into audacity twice. First time at 100% volume, second time at 80%. I then time aligned the recordings and normalized both to 0 db. Then I inverted one recording, and mixed them together to show the difference. the result is a perfect null – which means, the samples clipped before reaching the mixer, and lowering the volume didn't "reveal" any information that exist above 0dbfs.

Just as an extra thought, could you redo the experiment but this time use a larger negative preamp in Equaliser APO instead of using Windows Volume to counter the Reconstructed Peaks over 0dBFS? Because running a larger negative preamp in Equaliser APO is not exactly the same as running a lower Windows Volume (I think).

 

[MC_RME]

Another one to consider:

https://sound.stackexchange.com/questions/24981/why-is-my-cutting-equalizer-increasing-output-level

 

[Robbo99999]

Another one to consider:

https://sound.stackexchange.com/questions/24981/why-is-my-cutting-equalizer-increasing-output-level

Well I do use a High Pass Filter in PEACE to cut out the lowest bass frequencies....in a bid to decrease potential bass distortion & to increase clarity in the rest of the range. Is that what you're getting at? I could do some experiments with & without High Pass to see if makes a difference to PEACE clipping, however one of my previous posts (https://www.audiosciencereview.com/...ace-clipping-meter-question.15071/post-474193) shows a correlation between intersample overs in the track & PEACE clipping, so I'd be surprised if it was due to my High Pass Filter.

 

[Robbo99999]

Another one to consider:

https://sound.stackexchange.com/questions/24981/why-is-my-cutting-equalizer-increasing-output-level

Wow, well I did that little experiment turning off the High Pass Filter, and what do you know, no clipping signal lighting up in that Supermassive Black Hole track! How the jolly roger does one explain that!!??
EDIT: you can see my High Pass Filter in the PEACE graph in post #1 of this thread.

 

[Sgt. Ear Ache]

I've been using this recently too. The -6 dB limit on pre-amp and the lack of ability to copy and paste (or directly load) a preset in commonly found formats (like whats found on autoeq/oratory github) from are my two biggest gripes with it.

Luckily for myself, I really only have one primary set of headphones I bother to EQ so I just had to manually enter Oratory's profile for them and I'm pretty much done.

 

[Fluffy]

I'm not exactly sure of how inters ample overs should hurt the analog signal. They are basically a theoretical extrapolation of the signal – meaning, it doesn't matter how much the signal could have gone over the digital maximum, the end result is that the dac receives a value of 0dbfs for those samples.

This is my reasoning – digital audio is represented by assigning a value for each sample, that represents a voltage level. a bit depth of 16 bit can represent 65536 different voltage values, meaning +/-32768. So a sample at a positive 0 dbfs will have a positive value of 32768, which is the maximum allowed value. There is a specific voltage level assigned to that digital value. Let's say it's exactly 2V. any sample that reaches 0dbfs therefore reaches exactly 2V. if the dac receives a signal that should surpass the value of positive 32768, let's say double that at positive 65536 to represent a sample that is at +6dbfs – the dac cannot produce that level, and it is truncated down to value of 32768, and produced as 2V. meaning, that sample is clipped. if there is an audio file that some part of it exceeds 0dbfs, meaning gone beyond the value of positive 32768, it will all be truncated down to 32768 and clipped. the result of this clipping is that all those samples are produced at 2V – so the resulting signal will look like a straight line.

Now, if you take that audio file and lower its volume by 6 db before going to the dac, what happens in the case of the windows volume mixer is that all those samples at over 0dbfs are automatically truncated down to 0dbfs, and then sent to the dac. So the dac receives a signal which is at a lower value than its maximum positive 32768. But, all those samples where truncated to the same level of -6dbfs, or a digital positive value of 16384, which correspond to voltage level of 1V. so the resulting signal at those samples will still look like a straight line, though at 1V instead of 2V.

The end result is the same looking signal with the part of it that clipped looking like a straight line of identical voltages for that duration. The only difference is one signal is lower in amplitude. So why should these two signals sound any different?


One can conceive a simple solution for intersample overs – normalizing the audio file to be under 0dbfs. What you've seen in the examples of albums that don't have reconstructed peaks above 0dbfs is exactly that – the mastering engineer probably normalized the track so that its maximum value will be below 0dbfs, insuring no clipping would take place as a result of digital values exceeding that level. because this is a digital signal, it means you would lose a tiny beat of dynamic range, but that’s totally worth it.

 

[Robbo99999]

I'm not exactly sure of how inters ample overs should hurt the analog signal. They are basically a theoretical extrapolation of the signal – meaning, it doesn't matter how much the signal could have gone over the digital maximum, the end result is that the dac receives a value of 0dbfs for those samples.

This is my reasoning – digital audio is represented by assigning a value for each sample, that represents a voltage level. a bit depth of 16 bit can represent 65536 different voltage values, meaning +/-32768. So a sample at a positive 0 dbfs will have a positive value of 32768, which is the maximum allowed value. There is a specific voltage level assigned to that digital value. Let's say it's exactly 2V. any sample that reaches 0dbfs therefore reaches exactly 2V. if the dac receives a signal that should surpass the value of positive 32768, let's say double that at positive 65536 to represent a sample that is at +6dbfs – the dac cannot produce that level, and it is truncated down to value of 32768, and produced as 2V. meaning, that sample is clipped. if there is an audio file that some part of it exceeds 0dbfs, meaning gone beyond the value of positive 32768, it will all be truncated down to 32768 and clipped. the result of this clipping is that all those samples are produced at 2V – so the resulting signal will look like a straight line.

Now, if you take that audio file and lower its volume by 6 db before going to the dac, what happens in the case of the windows volume mixer is that all those samples at over 0dbfs are automatically truncated down to 0dbfs, and then sent to the dac. So the dac receives a signal which is at a lower value than its maximum positive 32768. But, all those samples where truncated to the same level of -6dbfs, or a digital positive value of 16384, which correspond to voltage level of 1V. so the resulting signal at those samples will still look like a straight line, though at 1V instead of 2V.

The end result is the same looking signal with the part of it that clipped looking like a straight line of identical voltages for that duration. The only difference is one signal is lower in amplitude. So why should these two signals sound any different?


One can conceive a simple solution for intersample overs – normalizing the audio file to be under 0dbfs. What you've seen in the examples of albums that don't have reconstructed peaks above 0dbfs is exactly that – the mastering engineer probably normalized the track so that its maximum value will be below 0dbfs, insuring no clipping would take place as a result of digital values exceeding that level. because this is a digital signal, it means you would lose a tiny beat of dynamic range, but that’s totally worth it.

I see your points, but it's established practice that intersample overs can be dealt with by lowering Windows Volume, and this is coming from 2 different people that are engineers in DAC producing companies.....JDS Labs and Topping (as well as other established members on this site), so it can't all be FUD, there's gotta be something in it.....in fact the JDS Labs guy even measured it as increased distortion as opposed to just theorising.

I don't know if you can redo your loopback comparison test but this time using the negative preamp in Equaliser APO rather than using the main Windows Volume Control? I just want to test that theory re intersample overs. Also, is it possible that the loopback test that you're doing doesn't have "enough resolution" to determine the differences? Just theorising possibilities.

 

[Fluffy]

Wow, well I did that little experiment turning off the High Pass Filter, and what do you know, no clipping signal lighting up in that Supermassive Black Hole track! How the jolly roger does one explain that!!??
EDIT: you can see my High Pass Filter in the PEACE graph in post #1 of this thread.

Surprising, yet interesting. Actually this makes a lot of sense in the context of that specific song, because it has bass content all the way to 10hz. So filtering out that lower register can result in eliminating low frequency oscillation in the signal that prevent it from clipping. for example, if the combined frequencies above 20 hz would send the sample over positive 0dbs, and there is a 10z component that at that moment has a negative amplitude, they cancel each other.

I didn't think much of your sub bass filter, but now I think you should get rid of that. it makes sense in the analog realm to get rid of very low frequencies nearing DC, to avoid damaging amps and speakers. But in the digital realm, all signal is limited by the unmovable 0dbfs limit – so sub-bass frequencies can only cause so much problems before they hit that wall.

 

[Fluffy]

I don't know if you can redo your loopback comparison test but this time using the negative preamp in Equaliser APO rather than using the main Windows Volume Control? I just want to test that theory re intersample overs. Also, is it possible that the loopback test that you're doing doesn't have "enough resolution" to determine the differences? Just theorising possibilities.

I did it again, this time with EqAPO providing -3db or pre-amp. Same exact result – complete null. And resolution here is not the issue, but how this loopback recording is dealing with samples above 0dbfs. If the recording takes the signal that is sent to the dac (and I have every reason to believe it is), then it is reliable in this scenario.

I see your points, but it's established practice that intersample overs can be dealt with by lowering Windows Volume, and this is coming from 2 different people that are engineers in DAC producing companies.....JDS Labs and Topping (as well as other established members on this site), so it can't all be FUD, there's gotta be something in it.....in fact the JDS Labs guy even measured it as increased distortion as opposed to just theorising.

You mentioned that people were talking about it. do you have any links with proofs to support that idea? And even if someone showed distortion at some test, I would still like to know what is the theoretical basis to that conclusion.

 

[Robbo99999]

Surprising, yet interesting. Actually this makes a lot of sense in the context of that specific song, because it has bass content all the way to 10hz. So filtering out that lower register can result in eliminating low frequency oscillation in the signal that prevent it from clipping. for example, if the combined frequencies above 20 hz would send the sample over positive 0dbs, and there is a 10z component that at that moment has a negative amplitude, they cancel each other.

I didn't think much of your sub bass filter, but now I think you should get rid of that. it makes sense in the analog realm to get rid of very low frequencies nearing DC, to avoid damaging amps and speakers. But in the digital realm, all signal is limited by the unmovable 0dbfs limit – so sub-bass frequencies can only cause so much problems before they hit that wall.

Hmm, I understand the words and can loosley see the relationship you talk about in the first paragraph, but I don't know enough about sound waves, etc to know if what you say is actually correct, some of what you say goes against my logic cutting frequencies below a certain point isn't going to change the amplitude of frequencies above because I just imagine in the digital realm that they would still be "as described". Maybe someone else can chime in on this with some links or examples. It's true though that the High Pass Filter I have put in place is somehow tripping the PEACE clipping meter, but I don't know if it's just a bug or what/how it's actually creating that effect....you've certainly put across a (to me) possibly plausible theory, but I can't latch onto it 100% just yet without some more understanding/proof.

 

[Robbo99999]

I did it again, this time with EqAPO providing -3db or pre-amp. Same exact result – complete null. And resolution here is not the issue, but how this loopback recording is dealing with samples above 0dbfs. If the recording takes the signal that is sent to the dac (and I have every reason to believe it is), then it is reliable in this scenario.


You mentioned that people were talking about it. do you have any links with proofs to support that idea? And even if someone showed distortion at some test, I would still like to know what is the theoretical basis to that conclusion.

For your first paragraph, what DAC have you got? Maybe you got a multibit R2R DAC that is able to resolve intersample overs in all situations and therefore it's not being negatively affected by intersample overs & therefore you would see no difference in loopback.

For your second paragraph, I'll try to find those links now for you.

EDIT: here's links to JDS Labs guy: https://www.head-fi.org/threads/o2-amp-odac.616331/page-249#post-11106527
and here's the links to various times JohnYang from Topping has mentioned it:
https://www.audiosciencereview.com/...intersample&c[users]=JohnYang1997&o=relevance

 

[dasdoing]

HP filter causing clipping first thing came into my mind is the phase shift.
So I googled that and https://www.soundonsound.com/sound-advice/q-why-do-my-mixes-clip-when-i-apply-high-pass-filter

I personaly discussed the use of FIR filters instead of EQ for headphones in another topic. I think it makes more sense using linear phase filters since the added phase of the correction with EQ filter wont correct nothing

 

[maverickronin]

I have a meter program called Orban loudness meter. It has an option to show reconstructed peak, and in compressed music it shows peaks above 0dbfs, though the file itself never goes beyond 0dbfs according to audacity:
View attachment 76353

Thanks for the link. Pretty cool program.

Not Unicode friendly though. Which track was #4 again?

 

[Robbo99999]

HP filter causing clipping first thing came into my mind is the phase shift.
So I googled that and https://www.soundonsound.com/sound-advice/q-why-do-my-mixes-clip-when-i-apply-high-pass-filter

I personaly discussed the use of FIR filters instead of EQ for headphones in another topic. I think it makes more sense using linear phase filters since the added phase of the correction with EQ filter wont correct nothing

Interesting idea. I looked at your link there, he describes two mechanisms for the increase in level, first one is phase shift and the second one is just the plain old reason that sometimes a cut will create a peak just to the right of the cut which is seen in the Total EQ Curve. However, that second point re Total EQ Curve, I don't have a visible peak to the right of my High Pass Filter in the Total EQ Curve (I made sure of that when I implemented it)....you can see that in the Total EQ Curve in PEACE in my first post, one of the graphs. It still could be the Phase Shift thing you mention, but the problem is I don't fully understand that, lol! If it was Phase Shift, how would I implement a High Pass that doesn't create "phase shift issue that's creating boosting of level".....if indeed that really is the mechanism of the cause?

 

[dasdoing]

Interesting idea. I looked at your link there, he describes two mechanisms for the increase in level, first one is phase shift and the second one is just the plain old reason that sometimes a cut will create a peak just to the right of the cut which is seen in the Total EQ Curve. However, that second point re Total EQ Curve, I don't have a visible peak to the right of my High Pass Filter in the Total EQ Curve (I made sure of that when I implemented it)....you can see that in the Total EQ Curve in PEACE in my first post, one of the graphs. It still could be the Phase Shift thing you mention, but the problem is I don't fully understand that, lol! If it was Phase Shift, how would I implement a High Pass that doesn't create "phase shift issue that's creating boosting of level".....if indeed that really is the cause?

I don't realy understand it either ,
but I THINK it could be explained this way:
let's say we have 2 frequencies in the original at 10Hz and 15Hz. they are not actualy in phase in the track. now the phase shift of the filter could have an effect of actualy phase aligning them (the phase shift is a curve); so in the original they don't add, but after phase aligning they add.

the same about two signals that cancel in the original. they cancel because the phase is exactly the opposite. now when you phase shift both in the same direction they don't cancel anymore

 

[Robbo99999]

I don't realy understand it either ,
but I THINK it could be explained this way:
let's say we have 2 frequencies in the original at 10Hz and 15Hz. they are not actualy in phase in the track. now the phase shift of the filter could have an effect of actualy phase aligning them (the phase shift is a curve); so in the original they don't add, but after phase aligning they add.

the same about two signals that cancel in the original. they cancel because the phase is exactly the opposite. now when you phase shift both in the same direction they don't cancel anymore

Ha, well yeah, that definitely makes two of us that don't totally understand phase! Yep, I've heard of phase aligning to be important during mixing to ensure that sounds aren't cancelled out, etc, so you're essentially using that same idea to the High Pass Filter.....where you say that the High Pass Filter is affecting Phase on a curve rather than in an absolute step wise fashion....thereby that creates (in the "transition phase" of the curve) a series of altered phase relative to each other, and it's in that "transition period" that misalignments of phase could occur that result in boosting or nulling of some frequencies?? Just trying to piece together my very incomplete and likely partly erroneous understanding of phase!

Would welcome any Phase gurus to comment on use of High Pass Filter and the negative effects of using one, along the lines of our discussion.

 

[Fluffy]

For your first paragraph, what DAC have you got? Maybe you got a multibit R2R DAC that is able to resolve intersample overs in all situations and therefore it's not being negatively affected by intersample overs & therefore you would see no difference in loopback.

If you are not familiar with the loopback feature of Audicity, I'll tell you that it happens completely "in the box". It captures what audio is being sent to the dac, but it has nothing to do with the dac itself. It's useful because it lets you see what your system is sending the dac after all effects and mixing took place.

Theoretically I could test what actual analog signal is coming out of the dac to see if there is any intersample over issues, but that would involve analog signals and ADCs and the result would be a lot dirtier and maybe even less revealing. I don't have the kind of measurement equipment Amir has to go to such low resolutions.

As for why low-cut creates clipping, I concocted this test:

I took a complex signal (two unrelated frequencies), and add a small bump that just pushes the signal to clipping at a specific point. then I added the key ingredient, which is asymmetric distortion – basically, the positive goes over 0dbfs while the negative doesn't. this is actually a common type of distortion used in tubes. After that, I mixed in a 5hz sinewave, and aligned the signals such that the a negative peak of the low tone will coincide with the clipping samples of the complex signal.

So this is how the signal looks with will all of them mixed together:



And now if I add a sharp high pass filter at 20hz, we get this:



Clipping!

 

[dasdoing]

where you say that the High Pass Filter is affecting Phase on a curve rather than in an absolute step wise fashion....thereby that creates (in the "transition phase" of the curve) a series of altered phase relative to each other, and it's in that "transition period" that misalignments of phase could occur that result in boosting or nulling of some frequencies??

that's how I understand it, yea.

phase is not that dificult to understand. it's just delay with another scale (realtive to the wave instead of time).
understanding what it does to the sound it the hard part lol

 

[Robbo99999]

If you are not familiar with the loopback feature of Audicity, I'll tell you that it happens completely "in the box". It captures what audio is being sent to the dac, but it has nothing to do with the dac itself. It's useful because it lets you see what your system is sending the dac after all effects and mixing took place.

Theoretically I could test what actual analog signal is coming out of the dac to see if there is any intersample over issues, but that would involve analog signals and ADCs and the result would be a lot dirtier and maybe even less revealing. I don't have the kind of measurement equipment Amir has to go to such low resolutions.

As for why low-cut creates clipping, I concocted this test:

I took a complex signal (two unrelated frequencies), and add a small bump that just pushes the signal to clipping at a specific point. then I added the key ingredient, which is asymmetric distortion – basically, the positive goes over 0dbfs while the negative doesn't. this is actually a common type of distortion used in tubes. After that, I mixed in a 5hz sinewave, and aligned the signals such that the a negative peak of the low tone will coincide with the clipping samples of the complex signal.

So this is how the signal looks with will all of them mixed together:
View attachment 76411


And now if I add a sharp high pass filter at 20hz, we get this:
View attachment 76412


Clipping!

Ah, ok, your Audacity software is not sending it to the DAC for loopback, ok. In that case, I think I heard that when using WASAPI (& maybe ASIO) that they are not negatively affected by intersample overs but DACS can be.....so if you're using WASAPI / ASIO as part of your loopback test (& it's not going to the DAC like you said), then that might be the reason for you seeing no difference in your experiments when you decrease windows volume. I'm a bit hazy on this, but sure I remember audioscience members saying WASAPI not affected by intersample overs.

Sorry though, I don't understand your other experiment re High Pass Filter, you might have to expand on that.

 

[Robbo99999]

that's how I understand it, yea.

phase is not that dificult to understand. it's just delay with another scale (realtive to the wave instead of time).
understanding what it does to the sound it the hard part lol

If that is indeed the case, how do you implement a High Pass Filter without incurring phase issues that seem to be affecting the levels?