Digital domain volume control

[georgehifi]

Hi, very much a novice in this area. I have a Cary 303/200 CDP it has PMD200 hdcd and PCM1704 dac.

What I would like to know the digital domain VC goes from 0 to 64 on the readout, all sound great if it's used between say 56 to 64 on the VC.
The other day we used it on some Parrasound JC1's and Wilson Alexia's, because this setup is much more efficient we had to use the VC at 48 much lower and it sounded very flat (non dynamic and maybe rolled off at both ends).
Is there a some sort of bench test I could do that some give a reading/indication of at what reduced volume setting when "Bit Stripping" starts to happen, as we were all sure this was what was happening. As when I reduced the gain of it's output stage so the VC could be used at 60 all the goodness came back.

Cheers George

 

[georgehifi]

Judging by the overwhelming response on this one, is it in the too hard basket???

Cheers George

 

[rfbrw]

"Bit Stripping" - Its maths, Jim, but not as we know it.

Judging by the overwhelming response on this one, is it in the too hard basket???

Cheers George

Given that one cannot make something smaller without actually making it smaller perhaps you can explain what "Bit Stripping" entails.

 

[georgehifi]

I don't know either Ray, from what I've read around "bit stripping" happens with non up sampled convertor Redbook dacs that have digital domain volume controls before the da conversion???? And when it happens you get 14 bit, 12bit resolution or even 10bit the lower the volume becomes??

And this was the only thing we could put it down to, because after we lowered the gain of the output stage so then we could use the digital volume almost at full for the same listening level, everything was great again.

Cheers George

 

[Pan]

I don't know how the math is done in that box but if the sound collapses at -12dB from a digital level change I'd say something is very wrong.

Remember that the only time you use all 16 bits on redbook format is when you play at full level. When you listen to dynamic material most of the information lives below the 13bit level and all sound pass thru LSB regularly.

It makes more sense to think about digital audio as the level you have and some noise. If you can not hear the noise you don't suffer from restricted dynamic range or "resolution".

The parts you mention are fine stuff but that is not enough of course.

 

[randytsuch]

Here is ESS's explanation of digital volume control.

Digital volume control

I think this explains what is happening.

Randy

 

[rfbrw]

Ummm, some idea of how many bits are active relative to the position of the digital volume control ? A peak meter type thingy mapped to the wordlength of the dac in question is what you need.

 

[georgehifi]

Thanks for that Randy, I think?

So it's noise that was doing what we all heard????

1: Lack of dynamic swing
2: Seemingly rolled off at both ends?????

Cheers George

 

[randytsuch]

Thanks for that Randy, I think?

So it's noise that was doing what we all heard????

1: Lack of dynamic swing
2: Seemingly rolled off at both ends?????

Cheers George

That's what ESS thinks.

In my simple mind, I always figured if you incorporated digital volume control without doing what ESS does (adding more bits), then you would lose bits, and information.

If my simplistic thoughts are correct, then the question becomes what does this loss of information sound like?
Noise is probably the simplest way to describe it, but I can believe you could hear it as a loss of dynamics and roll off.

Guy I know used to use a stepped attenuator for volume control, but then used digital volume control for fine settings. But he was careful to keep the digital volume at over 90% or something like that, to stay away from what you noticed.

Randy

 

[abraxalito]

It would be useful to know how much signal you're getting when VC=48 compared to when VC=64. Otherwise there's nothing much to go on. Do you have an AC voltmeter and a test CD you could use to check?

 

[georgehifi]

With a 1khz 0dbf sine wave I got at VC 64 =1.871vac
With the same track but at VC 48=.113vac

Cheers George

 

[abraxalito]

Ah thanks, that's very helpful. So yes that's a difference of 24dB, pretty significant. So you're losing 4 bits - in effect, with a digital volume control, you're listening to 12bit DACs.

 

[georgehifi]

Great now were getting somewhere thank's very much.

So 4 bits lost from 16bit Redbook, and it was very noticeable.

How did you work out 4 bits lost??

Cheers George

 

[abraxalito]

Step 1 - divide the output at max (VC=64) by the output at VC=48. 1.871/0.113 gives 16.5.

Step 2 - take the log (base 10) of 16.5 - this gives 1.22

Step 3 - multiply 1.22 by 20 - gives 24dB

Each bit on a DAC is half the next-door higher bit. DACs are binary. So this means 1bit = 6dB.

Step 4 - divide 24dB by 6dB, result 4bits.

 

[georgehifi]

Your the man Abraxalito.

As you know the PCM1704 dac is a 24bit dac, would the same "bit stripping" be happening when we play Reference Recordings 24bit 96 cd's with those same VC settings or maybe worse??

Cheers George

 

[abraxalito]

No matter what the source material (digital tape, DAW) word width, what's on CD is always 16bits. So yep, you'll be losing the same number of bits no matter what was used to make the CD.

Its an interesting test for a DAC chip to see how well its dynamics holds up (subjectively) when its given fewer bits to work with. I suspect the PCM1704 isn't a particularly good example (based on your listening here and other people who don't much rate it). I'd be interested to know if the PCM63 fares better under the same test.

 

[abraxalito]

Inspired by your experience George, I have now tried this (24dB digital attenuation) on my homebrewed TDA1387 DAC. Since I have no digital volume control I took a track and attenuated by 24dB using Audacity, saving it on my FLAC TFcard player. Pleased to report, apart from adding a bit more noise, it doesn't seem to lose dynamics very much. Slightly less bloom at the very low freqs but soundstage stays nice and deep.

I suggest you upgrade your DAC chips

 

[hollowman]

Some other issue... in-DAC digital attenuation

I suggest you upgrade your DAC chips

The 1704 is a very fine DAC, and much superior to 1387/1545A.
There must be some other issue in George's setup.
I also had (have) "mysterious" volume control issues with in-DAC digital attenuation, but OPPOSITE to what George reports:
http://www.diyaudio.com/forums/digital-line-level/240045-dac-digital-volume-level-mystery.html
Bottom line for me: For both my portable DAPs (CS4398 in Colorfly CK4) and Asus computer sound-card (PCM1792 in Xonar ST), I have the SOFTWARE volume setting at roughly 66% (so 33% below max).

 

[rfbrw]

No matter what the source material (digital tape, DAW) word width, what's on CD is always 16bits. So yep, you'll be losing the same number of bits no matter what was used to make the CD.

It is not that simple. In this case the PCM1704 is preceded by a digital filter. Assuming the filter is set to output 24 bits any bits lost in attenuation would be out of 24 not 16.

 

[abraxalito]

The 24bit word length gives an advantage in that dither at any change of gain will be at a lower level. But even with 24bits going into the DAC, its still the 4 MSBs which are being lost so I can't see how the digital filter's presence changes that.