Does lowering DAC volume compress music?
Status
Not open for further replies.
Mar 28, 2019 at 10:46 PM Thread Starter Post #1 of 15

JustASnack

New Head-Fier
Joined
Jan 22, 2019
Posts
20
Likes
10
Location
Malta
My hypothesis is this:
Let's say that my DAC normally outputs a signal which can reach a maximum voltage of 3 volts. This way, I understand that the DAC 'represents' the signal's dynamic range with a bandwidth of 3 volts.
Now I lower the volume of the DAC (either through hardware or software) such that the signal can reach a maximum voltage of 2 volts and thus has only 2 volts of bandwidth to represent the signal's dynamic range.
Ignoring the dynamic range of the signal lost to, perhaps, reducing the signal's bit depth if it is reduced through software, have I effectively compressed the signal, since the highest and lowest parts of the signal have been forced closer together? Does this make sense? Or does it not matter since the signal is being sent to an amplifier anyways?
 
Last edited:
Mar 29, 2019 at 3:55 AM Post #2 of 15
My hypothesis is this:
Let's say that my DAC normally outputs a signal which can reach a maximum voltage of 3 volts. This way, I understand that the DAC 'represents' the signal's dynamic range with a bandwidth of 3 volts.
Now I lower the volume of the DAC (either through hardware or software) such that the signal can reach a maximum voltage of 2 volts and thus has only 2 volts of bandwidth to represent the signal's dynamic range.
Ignoring the dynamic range of the signal lost to, perhaps, reducing the signal's bit depth if it is reduced through software, have I effectively compressed the signal, since the highest and lowest parts of the signal have been forced closer together? Does this make sense? Or does it not matter since the signal is being sent to an amplifier anyways?

If you've reduced the output of the DAC by 1 volt, then you've reduced the level of the entire signal, both the highest and lowest parts of the signal. The relationship between the highest and lowest parts of the signal has been maintained, they have not been forced closer together and therefore you haven't "effectively compressed the signal".

G
 
Mar 29, 2019 at 5:29 AM Post #3 of 15
If you've reduced the output of the DAC by 1 volt, then you've reduced the level of the entire signal, both the highest and lowest parts of the signal. The relationship between the highest and lowest parts of the signal has been maintained, they have not been forced closer together and therefore you haven't "effectively compressed the signal".

G
Let's say for simplicity's sake that, without adjusting the DAC's volume, if you were to cut out the loudest section of a song, you would get a signal of 2Vrms. Likewise, let's say that the quietest section of the song is a couple moments of complete silence, such that if you were to cut the silent section of the song out, you would get a signal of 0Vrms.
Now, reduce the output volume of the DAC, and the loudest section now boasts a signal strength of 1.5Vrms, but the lowest section of the signal is still 0Vrms, since it can't really be made quieter.
Again please correct me if I'm wrong, but it seems that the relationship between the highest and lowest part of the signal has changed when affecting output volume, no?
 
Mar 29, 2019 at 6:17 AM Post #4 of 15
Again please correct me if I'm wrong, but it seems that the relationship between the highest and lowest part of the signal has changed when affecting output volume, no?

No, because zero volts is not a low level signal, it's no signal at all, it's complete silence which is the lack of any signal. A very quiet section is still a signal, just a very low signal and the relationship of that very low signal and the highest signal is maintained. The relationship between no signal (0V) and the peak level has obviously changed but changing that relationship is just lowering the output level, not audio compression, which reduces the range between the lowest level signal and the highest by ONLY reducing the level of the peaks.

G
 
Last edited:
Mar 29, 2019 at 6:39 AM Post #5 of 15
Not all DAC adjust volume the same way - on certain (and mostly older) design the DAC adjusts volume by adjusting the bit-depth of the digital data - so if you start with a 16bit digital audio data, then adjusting the DAC's digital volume cause the DAC to cut off a few bits at the end - you might ended up with a 14bit or worst (depends on how much you are lowering the volume) digital data and thus you are no longer getting the bit-perfect audio signal when it is converted to analog output. That is what many usually referred as the music being compressed with DAC's digital volume control. A lot of more modern DAC internally resample the digital data to higher bit-depth (i.e. ESS's DAC usually resamples data to 32bits) so even with digital volume control, you will still get the the full signal as the DAC only cuts off the extra empty bits (thinks of it as a space holder) the resasmpling process added in but not the original 16bits music data. You will have to lower the digital volume control to very very low before it will affect the original 16bit data.
 
Last edited:
Mar 29, 2019 at 7:03 AM Post #6 of 15
Let's say for simplicity's sake that, without adjusting the DAC's volume, if you were to cut out the loudest section of a song, you would get a signal of 2Vrms. Likewise, let's say that the quietest section of the song is a couple moments of complete silence, such that if you were to cut the silent section of the song out, you would get a signal of 0Vrms.
Now, reduce the output volume of the DAC, and the loudest section now boasts a signal strength of 1.5Vrms, but the lowest section of the signal is still 0Vrms, since it can't really be made quieter.
Again please correct me if I'm wrong, but it seems that the relationship between the highest and lowest part of the signal has changed when affecting output volume, no?
problem 1:
no sound isn't a level of sound. I get your idea, except that it relates to nothing in real life. not for a DAC, not for sounds in the air. the concept of silence is in fact a lot of noises, just a lot quieter than the music. so the only concept where your rational works is the concept of Signal to Noise Ratio(SNR). your playback system will have noises, even the DAC alone will generate some, and when you lower the voltage output to reduce the listening level, some noises will go down along with the music, but some will not. instead they will remain just a noisy as ever. so the difference in amplitude between the music and the noise will effectively be smaller.

problem2:
but for the music, in practice a song will pretty much never be silent, when it is, the song typically hasn't started or has already ended. if you look at the silent parts within a song, you'll notice that they're usually only 60 or maybe 70dB below full scale signal. I'm sure there is somewhere a song with no signal within the song, but I wouldn't count that a the norm because it really is not.
so just with that fact, let's look at your idea. a DAC will have let's say 100dB of dynamic range and SNR(on sigma delta DACs it's usually the same value). you have a song that effectively uses 60dB of dynamic and that's what you'll get at the output. it's not like the DAC with 100dB of dynamic will somehow stretch the song so that it now becomes a song with 100dB of dynamic. the dynamic range of the DAC is what it will use with all songs. maybe the terms are misleading because dynamic is in them?
now if we take the case of lowering the voltage output(AKA listening level) by half. so instead of a rather typical 2V, the DAC will now output a full scale signal as 1V. that's only -6dB. so you'd need to reduce the voltage by a good deal more for that reduction+ the 60dB of our previous song, to amount to more than 100dB that the DAC can handle. and even when we reach that limit, what will happen is that the quieter signals will be lost, even then the DAC will not dynamically compress the music. the song will effectively have lower dynamic range, but because the quiet parts are gone. it's something pretty specific.
also modern DACs may have a specif way to handle volume attenuation that doesn't risk crippling the signal so nothing would actually happen beside having quieter music.

problem3: if you wish to consider that actual silence or at least the quietest signal a playback system and a room can let you perceive, is part of the song, then it would mean that you'd need to play the song at a specific listening level to get the "correct" experience and "dynamic". but what would that be? I have no idea and don't know of any such information on a record. and if noises from your street become louder, should you then increase the listening level in proportion to maintain the "silence" at the right level relatively to the music? or the opposite, if you play the music louder than it should be, you'd get a silence that's too quiet relatively to the all dynamic of the song ^_^. this is more of a philosophical question than anything usable but I get that you'd ask yourself that, because I did the same a few years ago :wink:
 
Mar 29, 2019 at 7:09 AM Post #7 of 15
... so if you start with a 16bit digital audio data, then adjusting the DAC's digital volume cause the DAC to cut off a few bits at the end - you might ended up with a 14bit or worst (depends on how much you are lowering the volume) digital data and thus you are no longer getting the bit-perfect audio signal when it is converted to analog output. That is what many usually referred as the music being compressed with DAC's digital volume control.

But that is not "compression", it's "data truncation" which is pretty much the exact opposite of compression! Data truncation removes the least significant bits, the lowest level signals. Compression on the other hand does the opposite, it reduces ONLY the peak levels and leaves the lower and lowest level signals completely unchanged.

G
 
Mar 29, 2019 at 12:17 PM Post #8 of 15
Castleofargh provides a very satisfying answer! Thanks for your input, though I still have one more question (I apologize for being such a nuisance) :)

you'd need to reduce the voltage by a good deal more for that reduction+ the 60dB of our previous song, to amount to more than 100dB that the DAC can handle. and even when we reach that limit, what will happen is that the quieter signals will be lost, even then the DAC will not dynamically compress the music. the song will effectively have lower dynamic range, but because the quiet parts are gone. it's something pretty specific.
Let's say that we choose to lower the output volume of the DAC so much that the quieter bits of the song are infact lost, and as you said, the song's dynamic range is decreased in doing so. Would these quiet bits be lost below the noise floor of the DAC? I'm assuming that if so, then listening to the music in this state would be rather noisy, assuming I have enough power to amplify the signal to be able to hear it at a decent volume. Is this correct?
 
Mar 29, 2019 at 12:31 PM Post #9 of 15
Castleofargh provides a very satisfying answer! Thanks for your input, though I still have one more question (I apologize for being such a nuisance) :)


Let's say that we choose to lower the output volume of the DAC so much that the quieter bits of the song are infact lost, and as you said, the song's dynamic range is decreased in doing so. Would these quiet bits be lost below the noise floor of the DAC? I'm assuming that if so, then listening to the music in this state would be rather noisy, assuming I have enough power to amplify the signal to be able to hear it at a decent volume. Is this correct?
by attenuating like crazy in the digital domain and then applying a massive gain at the amp to compensate for it, you could indeed reach such extremity where you're hearing the quantization noise of the DAC very clearly. or just whatever noise floor is loudest, because at this point the amp is boosting everything a lot. you'd end up with a really horrible SNR.
 
Mar 29, 2019 at 12:32 PM Post #10 of 15
But that is not "compression", it's "data truncation" which is pretty much the exact opposite of compression! Data truncation removes the least significant bits, the lowest level signals. Compression on the other hand does the opposite, it reduces ONLY the peak levels and leaves the lower and lowest level signals completely unchanged.

G

Read this: http://www.esstech.com/files/3014/4095/4308/digital-vs-analog-volume-control.pdf

I am referring the idea that what most people is seeing as the output is being "compressed" against the fixed noise floor when using digital volume control, but not when using analog volume control where the noise floor will decrease when the volume is lower.
 
Last edited:
Mar 30, 2019 at 6:27 AM Post #11 of 15
Read this: http://www.esstech.com/files/3014/4095/4308/digital-vs-analog-volume-control.pdf
I am referring the idea that what most people is seeing as the output is being "compressed" against the fixed noise floor when using digital volume control, but not when using analog volume control where the noise floor will decrease when the volume is lower.

There's two problems here:

Firstly (and again): If we lower the volume, in the digital or analogue domains, we lower the volume of the entire signal equally. So if we reduce the volume by say 10dB, then the peaks (highest level points) in the signal are reduced by 10dB, the lowest level parts of the signal are reduced by 10dB and so is ALL the rest of the signal in between these two extremes. If we lower the volume sufficiently so that the lowest level parts of the signal are now below the (digital or analogue) noise floor, then obviously we've reduced the signal to noise ratio (SNR) but this is NOT compression! Compression ONLY reduces the peaks, it therefore cannot reduce the lowest level parts of the signal to below the noise floor and it does NOT affect anything in between the lowest level parts of the signal and the peaks.

Secondly (and unrelated to the fact it has nothing to do with compression anyway), the document you posted is, in practice, mostly irrelevant/not applicable. It's just audiophile marketing that employs one of the typical, fallacious audiophile marketing tactics: Circumstance/Situation "X" is bad because it causes such and such a problem, you therefore need to buy our product because it creates circumstance/situation "Y" that avoids circumstance/situation "X". It's a fallacy (false marketing) when circumstance "X" and it's associated problem don't actually exist in practice! In this particular case, the ESS document starts with a 16bit DAC chip (circumstance "X") and correctly illustrates the problem when lowering the digital volume. The ESS product avoids/solves this problem by using a 32bit DAC chip (circumstance "Y"). However, as 24bit DAC chips have a trade price of around $1 - $2, they're standard even in cheap DACs and AFAIK, none of the chip manufacturers even make 16bit DAC chips any more (for music/audio applications), so when does circumstance "X" ever actually exist?

G
 
Last edited:
Mar 30, 2019 at 6:37 AM Post #12 of 15
If a dac plays with no volume, did it really play.
 
Mar 30, 2019 at 9:03 AM Post #13 of 15
There's two problems here:

Firstly (and again): If we lower the volume, in the digital or analogue domains, we lower the volume of the entire signal equally. So if we reduce the volume by say 10dB, then the peaks (highest level points) in the signal are reduced by 10dB, the lowest level parts of the signal are reduced by 10dB and so is ALL the rest of the signal in between these two extremes. If we lower the volume sufficiently so that the lowest level parts of the signal are now below the (digital or analogue) noise floor, then obviously we've reduced the signal to noise ratio (SNR) but this is NOT compression! Compression ONLY reduces the peaks, it therefore cannot reduce the lowest level parts of the signal to below the noise floor and it does NOT affect anything in between the lowest level parts of the signal and the peaks.

Secondly (and unrelated to the fact it has nothing to do with compression anyway), the document you posted is, in practice, mostly irrelevant/not applicable. It's just audiophile marketing that employs one of the typical, fallacious audiophile marketing tactics: Circumstance/Situation "X" is bad because it causes such and such a problem, you therefore need to buy our product because it creates circumstance/situation "Y" that avoids circumstance/situation "X". It's a fallacy (false marketing) when circumstance "X" and it's associated problem don't actually exist in practice! In this particular case, the ESS document starts with a 16bit DAC chip (circumstance "X") and correctly illustrates the problem when lowering the digital volume. The ESS product avoids/solves this problem by using a 32bit DAC chip (circumstance "Y"). However, as 24bit DAC chips have a trade price of around $1 - $2, they're standard even in cheap DACs and AFAIK, none of the chip manufacturers even make 16bit DAC chips any more (for music/audio applications), so when does circumstance "X" ever actually exist?

G

To your first point - I see only issue of semantics.

To your second point - you lose me there so I am not going to attempt to answer.
 
Mar 30, 2019 at 10:32 AM Post #14 of 15
To your first point - I see only issue of semantics.

If we compress a signal we end up with a significantly different signal than if we just lower it's volume/level, so how can it only be an issue of semantics?

G
 
Mar 30, 2019 at 10:51 AM Post #15 of 15
Last edited:
Status
Not open for further replies.

Users who are viewing this thread

Back
Top