Software volume control - does this affect sound quality in a negative way?

[snapcrackle]

Compared to using an analog volume control from a pre-amp, or in fact the volume control built into a DAC chip, can software volume control be equally good as the aforementioned?

For example, using the 'software' volume control in an Auralic Mini, or Volumio etc etc?

 

[DVDdoug]

Technically, you do loose resolution when you reduce the volume. i.e. If you reduce the volume of 16-bit audio by -6dB (50%) you are only "using" 15-bits. There are also "rounding errors", but they are insignificant.

But you don't hear any loss of quality - Imagine if you reduce the volume down to where you are only using a few bits... You can't hear anything so you can't hear any loss of quality. Or if the level is very-low so you can barely hear it, you won't hear any noise or distortion which will be below the level of the signal.

The danger would be if you re-amplify it -And, something similar happens with analog... As you reduce the (analog) signal you reduce the signal-to-noise ratio (for any noise in the next-stage electronics). But if the noise isn't bad (or isn't audible) you don't hear any loss of quality unless you re-amplify which amplifies the signal and noise together, making the noise more noticeable.

 

[Tangband]

Compared to using an analog volume control from a pre-amp, or in fact the volume control built into a DAC chip, can software volume control be equally good as the aforementioned?

For example, using the 'software' volume control in an Auralic Mini, or Volumio etc etc?

Some digital volume controls in software are almost transparent with no apparent loss of soundquality. Others are not so good.
Conversion with dithering can sometimes be done from 16 bit and up to 32 or 24 bit. Doing this, you have a lot of headroom when using volume control in digital domain.

 

[nikosidis]

In most cases it would be better with digital volume control. Preamp is a filter and might colour the sound or mask it.

 

[mdsimon2]

With digital volume control there are two issues to worry about, these will be the same whether you are using digital volume control implemented on a DAC chip or software volume control.

First issue is quantization noise, in practice this is rarely an issue these days as volume calculations are typically done at 24 bit or better so you can attenuate 16 bit data by 48 dB before you even have the possibility of an issue.

The second issue is that a digital volume control only attenuates the signal and does not attenuate noise. This means that when you attenuate you are effectively reducing your signal to noise ratio. However in practice if you cannot hear any noise from your listening position with nothing playing then attenuating really doesn't lose anything as your ambient noise floor is higher than your system noise floor. This is an area where it pays to have a DAC with high SNR so you can ensure a low noise floor.

In contrast to a digital volume control an analog volume control will attenuate both signal and noise, preserving the signal to noise ratio as you attenuate. Of course your effective SNR will still be limited by the analog noise floor.

Michael

 

[Chester]

https://www.audiosciencereview.com/...truth-about-dsp-volume-control-in-roon.23552/

May be of interest OP. My understanding is the same applies to all well implemented digital volume. I’m no expert though.

 

[MCH]

I wish someone with the knowledge drafts a "complete guide to volume control" covering all these recurring questions.
BTW, the document linked in post 6 seems to contradict, at least partially, that digital volume control does not attenuate (digital) noise, or were you referring to the noise out of a DAC? Aaaarrr... there are so many combinations... software volume, dac volume, pre Amp volume, noise floor of the signal, noise floor of the system..... Good that nowadays you can get very good results even if you don't really understand what you are doing (in my case at least)

 

[nikosidis]

I wish someone with the knowledge drafts a "complete guide to volume control" covering all these recurring questions.
BTW, the document linked in post 6 seems to contradict, at least partially, that digital volume control does not attenuate (digital) noise, or were you referring to the noise out of a DAC? Aaaarrr... there are so many combinations... software volume, dac volume, pre Amp volume, noise floor of the signal, noise floor of the system..... Good that nowadays you can get very good results even if you don't really understand what you are doing (in my case at least)

I agree. For none tech.. people this is hard to understand. I'm part of that group.
I still kind of understand it and can try to describe it in a way normal people can understand.

If you are using a passive or digital volume control you will have the voltage that come inn to work with.
Standard RCA connection is 2V. That will be max possible gain or volume (volume or gain is the same thing).
For XLR it is 4V. That is 6db more and should be plenty for any setup.
In many situations 2V is not enough to play real loud and you would need active preamp. before the power amp.
When you lower the volume on any passive volume control, digital or analogue the signal to noise ratio will go down and introduce noise. To explain signal to noise we are talking about dynamic range. Not about resolution.
It is not that the music will be distorted or unclear.
The way to solve the signal to noise problem with 16bit audio using a digital volume is to upsample the signal to 24 or 32-bit so you have a huge amount of headroom. This should be done in software on a computer or from the hardware itself. That way there will not be any noise when you lower the volume.
If you have a DAC or some other source with digital volume go close to the speakers.
Without any signal just go from low to max gain and listen if you hear any change of noise.

Sure, there are not always a digital volume does all things as it should. Just like most other components or speakers.
If done right it should be the best possible solution for digital audio.
Early in the digital age when it was not possible to upsample the signal it was for sure a limitation.
Today it is not.

 

[voodooless]

I wish someone with the knowledge drafts a "complete guide to volume control" covering all these recurring questions.
BTW, the document linked in post 6 seems to contradict, at least partially, that digital volume control does not attenuate (digital) noise, or were you referring to the noise out of a DAC? Aaaarrr... there are so many combinations... software volume, dac volume, pre Amp volume, noise floor of the signal, noise floor of the system..... Good that nowadays you can get very good results even if you don't really understand what you are doing (in my case at least)

These tests are purely done on the digital signal. In both cases, Roon DSP and Xmos, the noise was kept below -160. That is just about at the analog noise floor of SOTA DACs, so I would not expect any audible issues here.

Besides. You need a bit of added noise to reconstruct the missing resolution using dither. The lowest bit of a 24-bit file is noise anyway. With a bit of luck, you'll get 19 really usable bits. And then most audio is still only 16 bits. You can attenuate it about 48 dB before you lose any resolution anyway. That is quite a lot.

The rule to thumb with a decently implemented digital volume control is as was already stated by others: if you play a silence at full volume, and you hear nothing, you're most probably good.

 

[mdsimon2]

I wish someone with the knowledge drafts a "complete guide to volume control" covering all these recurring questions.
BTW, the document linked in post 6 seems to contradict, at least partially, that digital volume control does not attenuate (digital) noise, or were you referring to the noise out of a DAC? Aaaarrr... there are so many combinations... software volume, dac volume, pre Amp volume, noise floor of the signal, noise floor of the system..... Good that nowadays you can get very good results even if you don't really understand what you are doing (in my case at least)

I agree that it was hard for me to wrap my head around the various volume control issues. You hear things like "digital volume control is perfect because it is done in 32 bit" which is true in some aspects but not others.

The other issue that makes things difficult is that noise metrics for various equipment are quoted at very different levels, for example DAC SNR might be quoted at 2 V, 4 V or something even higher, at ASR amplifier SNR is measured at 5 W in to 4 ohms and full power, it makes understanding how these numbers come together somewhat difficult.

A few things helped me understand it better. One was playing around with digital volume control in a similar manner to that Roon volume control white paper and measuring noise digitally. This helped me understand that quantization noise is not really an issue, certainly a non-issue compared to the overall DAC / amplifier noise floor. The other was measuring actual noise level at my amplifier outputs with various DACs as this gave me a real feel for the ever constant noise floor of my system. And finally building a spreadsheet to calculate system noise levels based on DAC SNR, amplifier SNR, amplifier gain and amplifier power rating.

For example here is a plot from the spreadsheet showing system SNR as a function of volume position with 16 bit input data. Modeled parameters here are DAC SNR of 110 dB at 2 V full output (0 dBFS), amplifier SNR of 100 dB at 5 W in to 4 ohm, amplifier gain of 25.6 dB and full amplifier power of 125 W in to 8 ohm / 250 W in to 4 ohm. For reference this roughly models a MOTU Ultralite Mk5 and a Hypex NC252MP amplifier.

A few things to note, the plot doesn't go up to 0 dB because the amplifier will clip at volume positions above -2 dB. The high end of the graph is rather flat as the noise floor is limited by 16 bit input data at that point and not the DAC / amplifier noise floor.

For reference here is the same modeled system with 24 bit input data, as you can see although we are able to do better than 16 bit performance (96 dB SNR) we do not actually reach 24 bit performance (144 dB SNR) because we are limited by the DAC / amplifier noise floor.

Michael

 

[Vincent Kars]

Technically, you do loose resolution when you reduce the volume. i.e. If you reduce the volume of 16-bit audio by -6dB (50%) you are only "using" 15-bits. There are also "rounding errors", but they are insignificant.

True, but only if you are using a 16 bit DAC.
With a DAC with a 24 register you can chop of 48 dB and still have the entire 16 bit sample unaltered.

16 bit register
MSB LSB
1111111111111111
0000000011111111

24 bit register
111111111111111100000000
000000001111111111111111

 

[MCH]

I agree that it was hard for me to wrap my head around the various volume control issues. You hear things like "digital volume control is perfect because it is done in 32 bit" which is true in some aspects but not others.

The other issue that makes things difficult is that noise metrics for various equipment are quoted at very different levels, for example DAC SNR might be quoted at 2 V, 4 V or something even higher, at ASR amplifier SNR is measured at 5 W in to 4 ohms and full power, it makes understanding how these numbers come together somewhat difficult.

A few things helped me understand it better. One was playing around with digital volume control in a similar manner to that Roon volume control white paper and measuring noise digitally. This helped me understand that quantization noise is not really an issue, certainly a non-issue compared to the overall DAC / amplifier noise floor. The other was measuring actual noise level at my amplifier outputs with various DACs as this gave me a real feel for the ever constant noise floor of my system. And finally building a spreadsheet to calculate system noise levels based on DAC SNR, amplifier SNR, amplifier gain and amplifier power rating.

For example here is a plot from the spreadsheet showing system SNR as a function of volume position with 16 bit input data. Modeled parameters here are DAC SNR of 110 dB at 2 V full output (0 dBFS), amplifier SNR of 100 dB at 5 W in to 4 ohm, amplifier gain of 25.6 dB and full amplifier power of 125 W in to 8 ohm / 250 W in to 4 ohm. For reference this roughly models a MOTU Ultralite Mk5 and a Hypex NC252MP amplifier.

View attachment 190329

A few things to note, the plot doesn't go up to 0 dB because the amplifier will clip at volume positions above -2 dB. The high end of the graph is rather flat as the noise floor is limited by 16 bit input data at that point and not the DAC / amplifier noise floor.

For reference here is the same modeled system with 24 bit input data, as you can see although we are able to do better than 16 bit performance (96 dB SNR) we do not actually reach 24 bit performance (144 dB SNR) because we are limited by the DAC / amplifier noise floor.

View attachment 190330

Michael

Very helpful explanation!
Fortunately i don't have a problem with noise, my approach, being conscious of my limited understanding, was to listen to johnyang and buy the DAC with the better SNR that i could afford (topping D30pro) and use it as preamp.
My power amp is precisely a NC252MP (boxem almost literally is next door from where I live). I would imagine, seeing your graphs and th d30pro review graph, that I will be getting to the 96db snr (most of my music is red book) pretty low in terms of attenuation (at around -30db?) what seems quite decent even having a fixed -3db because of room correction. Am I missing something or can I sleep in peace now?
Do I understand well that setting the volume at the source (moodeaudio) would not make any difference? (I am not going to do it anyways but I am just curious)

 

[mdsimon2]

Very helpful explanation!
Fortunately i don't have a problem with noise, my approach, being conscious of my limited understanding, was to listen to johnyang and buy the DAC with the better SNR that i could afford (topping D30pro) and use it as preamp.
My power amp is precisely a NC252MP (boxem almost literally is next door from where I live). I would imagine, seeing your graphs and th d30pro review graph, that I will be getting to the 96db snr (most of my music is red book) pretty low in terms of attenuation (at around -30db?) what seems quite decent even having a fixed -3db because of room correction. Am I missing something or can I sleep in peace now?
Do I understand well that setting the volume at the source (moodeaudio) would not make any difference? (I am not going to do it anyways but I am just curious)

Here are the plots for a DAC with 128.5 dB SNR at 4.1 V, they look really good and it is no surprise you do not have any noise issues. There is no practical difference whether you attenuate in software or the DAC.

16 bit input

24 bit input

You are definitely amplifier limited (which IMO is a good thing) and an analog volume control will not help at all. Here is a comparison with a good analog volume control with 126 dB SNR at 2 V (1 uV residual noise), analog is orange and digital is blue.

Michael

 

[MCH]

Wow, thanks for the personalised graphs. I doubt that any of the music i listen to reaches even 10 bits dynamic range, but the peace of mind these graphs give is priceless!!

 

[threni]

I wish someone with the knowledge drafts a "complete guide to volume control" covering all these recurring questions.
BTW, the document linked in post 6 seems to contradict, at least partially, that digital volume control does not attenuate (digital) noise, or were you referring to the noise out of a DAC? Aaaarrr... there are so many combinations... software volume, dac volume, pre Amp volume, noise floor of the signal, noise floor of the system..... Good that nowadays you can get very good results even if you don't really understand what you are doing (in my case at least)

There's a lot of topics that's true for. In fact, all of them. Loads of really good articles, statements, quotes, descriptions of every aspect of audio, hopelessly sprinkled across thousands of threads, most of which also contain contrary statements from the great unwashed. I'm not sure what the answer is but it would be great if it were possible to crowdsource a single, curated, moderated article on all interested topics here. If it were a book, I'd buy it! Some answers involve the reader getting a list of books about sub-woofers. Absolutely nobody is going to read all that unless they're literally going to start building them!

 

[Tangband]

In theory a digital volume control seems to be , in many cases , good enough to be transparent . But there are many exeptions .

The digital software volume control in Windows mixer sucks and is really bad .
The sample rate conversion in Windows is also very badly executed .
You really have to listen in each case to hear if the digital volume control is good enough for you .

One example of a product that sounds really good is Yamaha wxc50 if you turn of the pre amp mode .
Using the preamp mode and digital volume control on spdif out from Yamaha, you can hear a small decrease in fidelity and for me the Yamaha wxc50 is not good enough as a volume regulator to my Genelecs .

In contrast, my Mac with xmos USB bridge is slightly better sounding. Everyone can hear the difference with the Genelecs , but the difference is much smaller using an external dac with less good passive loudspeakers . One can then get the false impression that the volume control in the Yamaha is transparent .

Using Mac with an USB bridge, in the xmos meny you can choose between 16 bit or 24 bit and regulate the volume . I cant hear any difference in the sound in this case with 24 bit regulating the volume with the Genelecs.

 

[snapcrackle]

Thanks everyone - I've been away for the Weekend so have a lot of catch up reading (and understanding) to soak in here.

 

[snapcrackle]

I'm going to add a link from the Auralic forum which mentions software volume control as a result of a technology implemented in Lumin called Leedh which is advertised to be a 'Lossless Digital Volume Control'

https://community.auralic.com/t/lumin-added-new-volume-control-algorithm/6554

Xuanqian responds as below (copied and pasted from the link) ... implying there is a major problem with digital volume controls ....

The algorithm won’t fix digital volume control major problem.
Let’s say we have a DAC which has dynamic range (DNR) of 120dB. It means you DAC has a 1uV noise floor from the DAC chip itself and the all the analog circuits after that comparing to a maximum of 1V output signal level; this makes 120dB DNR.
When you apply digital volume control, the signal level will drop but the noise floor will probably not drop, or at least will not drop accordingly. If you apply a -40dB volume control, your music signal level will drop to 0.01V, but the noise floor won’t drop to 0.01uV because it is way beyond any of current component’s limitation. Let’s say the noise floor drop to 0.1uV (which I will be very impressed), you will lost 20dB dynamic range. In real world, you will lose way more than 20dB, that’s why digital volume control can make sound so dull when you turn the volume down, you just lost a lot of details in the music.
It is not to say analog type volume control is perfect, all resistors have Johnson noise thermometry it is temperature based noise from physical law. JNT is related to the resistor value because the higher value, the more heat it will create when current flow through. In a very high resolution, low noise audio system, JNT must be considered. However JNT can be very low and beyond audio system’s consideration if the entire system has low impedance design.

​

 

[voodooless]

The algorithm won’t fix digital volume control major problem.
Let’s say we have a DAC which has dynamic range (DNR) of 120dB. It means you DAC has a 1uV noise floor from the DAC chip itself and the all the analog circuits after that comparing to a maximum of 1V output signal level; this makes 120dB DNR.
When you apply digital volume control, the signal level will drop but the noise floor will probably not drop, or at least will not drop accordingly. If you apply a -40dB volume control, your music signal level will drop to 0.01V, but the noise floor won’t drop to 0.01uV because it is way beyond any of current component’s limitation.

Yeah, we already established all that.

Let’s say the noise floor drop to 0.1uV (which I will be very impressed), you will lost 20dB dynamic range. In real world, you will lose way more than 20dB, that’s why digital volume control can make sound so dull when you turn the volume down, you just lost a lot of details in the music.

120-20 is still 100 dB, more than CD quality. Even at 40 dB, you'll still be left with 80dB DNR. Vinyl will do 70 dB on a good day, and people don't complain about dull sound either.

Besides, you actually lose not that much dynamic range with a proper implementation: dithering will restore a lot of it at the frequencies where it really counts.

 

[freemansteve]

Need to be careful here.
You can have a digital control (pulsed knob, remote etc) that can affect volume in the digital domain or in the analogue domain.
I asked Cambridge Audio about this for their Evo, and they said the volume is internally analogue. They may well have been making it up though! I know other kit does volume in DSPs.