DAC under 1500USD with top performance volume control

[unpluggged]

Do you mean this is both theoretically and also practically, repeatedly measurably wrong? Or do you just disagree?

With the "wrong" portion of my reply I refer to the statement that analog control maximizes the SNR, and then I explained why this is not true. Because after you attenuate your DAC output, you still have to amplify it, and that's where amplifier's own noise is added, which is usually tens of decibels higher than the unattenuated DAC noise, especially with active speakers.

Now there is the issue of level mismatch when your source output level is too high compared to your amp's sensitivity for effective digital volume control, and this is solved by using fixed passive attenuators that don't degrade the signal.

And besides SNR, analog volume control usually introduces much more noticeable artifacts like channel imbalance (even with relay-based controls due to resistor tolerances), distortion, scratching, switching noise and pulses.

With my ADI-2 Pro FS R I usually listen to music at volume levels between −16 to 0 dB (at +4 dBu ref level), so converter's SNR does not fall below 103 dB. And again, I don't really care about it because I can't hear its noise floor, which in any case is overwhelmed by the noise floors of my speakers and my environment.

 

[Sokel]

Digital VC can be as good as it gets as long as there's a bulletproof way NOT to ever forget it's setting.
NOT the case in many of those as we read all the time.
That's my only concern.

 

[mdsimon2]

guys, is it possible to submit a request to Amir to run some measurements and clarify this? the answer to this question could basically save a lot of people a lot of money. If we could for example compare measurements of RME dac maxed out to a Schiit Freya or Pre90 vs RME dac volume adjusted. If they are matched in output level, wouldn't that settle this discussion?

The primary difference between digital volume control and analog volume control is noise when the volume is attenuated. Analog volume control theoretically should have the advantage because it attenuates both noise and main signal thus maintaining the SNR while digital volume control only attenuates the main signal. However, as mentioned elsewhere in this thread the analog preamp will add some noise of its own and high performance DACs are so low noise these days that the difference in system SNR is minuscule even in the best cases for analog volume control.

Fortunately it is rather is easy to model system SNR (i.e. SNR at the amplifier output terminals) as a function of volume. I gave some examples of this in another thread here -> #274 and here -> #323 that may be helpful. The short story is that analog volume control makes a huge difference when you are starting from a sub par DAC (like in an AVR) but if you start with a typical high performance DAC then it doesn't make a difference and can actually make things worse.

It is also a good idea to have an understanding of what residual noise levels you can hear with your system. I did some listening tests with an unpadded, rather high sensitivity tweeter (~92 dB) here -> #28 and found that a residual noise level in the low-100 uV range is noiseless to me and you can achieve this with rather middling equipment. Of course if you are using super high sensitivity horns that value may be different for you.

One final thing, I really don't think the Topping Pre90 belongs in the discussion of high performance preamps. The low input impedance is a joke and will cause elevated distortion in many DACs.

Michael

 

[restorer-john]

With the "wrong" portion of my reply I refer to the statement that analog control maximizes the SNR, and then I explained why this is not true. Because after you attenuate your DAC output, you still have to amplify it, and that's where amplifier's own noise is added, which is usually tens of decibels higher than the unattenuated DAC noise, especially with active speakers.

Now there is the issue of level mismatch when your source output level is too high compared to your amp's sensitivity for effective digital volume control, and this is solved by using fixed passive attenuators that don't degrade the signal.

And besides SNR, analog volume control usually introduces much more noticeable artifacts like channel imbalance (even with relay-based controls due to resistor tolerances), distortion, scratching, switching noise and pulses.

My goodness, you really are digging yourself further into the hole of misunderstanding aren't you?

@MaxwellsEq is correct.

And we have all these people making assumptions on exactly how the gain blocks, attenuation and amplification is implemented in their systems when they really have no understanding of the actual topology of each device. People calling volume controls 'digital' when they are digitally controlled analogue chip based attenuators, resistor ladder or VCAs. Others are digital, prior to the D/A itself. Many digitally controlled volume ICs use combinations of attenuation and active gain internally. Stating that analogue controls introduce clicks and noise when they realistically don't- it is the digital versions where you hear the clicks and noise.

Many vintage high performance preamplifiers of the past (and some from today) use 4 gang volume pots, using different value tapers for the front end (before the gain stages) and back end (buffer/out stage). That way, they attenuate both the signal AND the entire preamplifier's noise as the control is turned down. This is done as 99% of listening is done within a range of volume which is nowhere near full output and S/N is preserved.

Yamaha:

Vintage 1970s Pioneer:

Yamaha TOTL preamplifier with a rated <1.5uV residual and a 106dB S/N...

In short, I see a heap of young ASR members messing around arguing about preserving S/N through randomly connected components with no idea about levels, measured residual noise, gain structures, impedances and ensuring they are not massively under utilizing the available gain and consequently missing out on what their systems are capable of.

Remember, all this is completely dependent on the source content and you will see enormous ranges of levels. I have discs where the average level is over 20db lower than others and you need to have sufficient reserve gain to be able to listen to all your content at the level you want and not leave a bunch of output on the table.

 

[Daverz]

A Topping E70 should do what you want. Amir even provided plot of SINAD vs. output level.

Topping E70 Stereo DAC Review

This is a review and detailed measurements of the Topping E70 stereo USB DAC with Bluetooth. It was sent to me by the company and costs US $349. I really like the new white multi-segment display with its large volume level indicator. Response is excellent as you rotate the encoder. A select...

www.audiosciencereview.com

 

[dshreter]

My goodness, you really are digging yourself further into the hole of misunderstanding aren't you?

@MaxwellsEq is correct.

And we have all these people making assumptions on exactly how the gain blocks, attenuation and amplification is implemented in their systems when they really have no understanding of the actual topology of each device. People calling volume controls 'digital' when they are digitally controlled analogue chip based attenuators, resistor ladder or VCAs. Others are digital, prior to the D/A itself. Many digitally controlled volume ICs use combinations of attenuation and active gain internally. Stating that analogue controls introduce clicks and noise when they realistically don't- it is the digital versions where you hear the clicks and noise.

Many vintage high performance preamplifiers of the past (and some from today) use 4 gang volume pots, using different value tapers for the front end (before the gain stages) and back end (buffer/out stage). That way, they attenuate both the signal AND the entire preamplifier's noise as the control is turned down. This is done as 99% of listening is done within a range of volume which is nowhere near full output and S/N is preserved.

Yamaha:
View attachment 256582

Vintage 1970s Pioneer:
View attachment 256583

Yamaha TOTL preamplifier with a rated <1.5uV residual and a 106dB S/N...

View attachment 256585

In short, I see a heap of young ASR members messing around arguing about preserving S/N through randomly connected components with no idea about levels, measured residual noise, gain structures, impedances and ensuring they are not massively under utilizing the available gain and consequently missing out on what their systems are capable of.

Remember, all this is completely dependent on the source content and you will see enormous ranges of levels. I have discs where the average level is over 20db lower than others and you need to have sufficient reserve gain to be able to listen to all your content at the level you want and not leave a bunch of output on the table.

I don't think it's a good idea to look at the matter in absolute terms. I don't question at all that there are very high-quality analogue volume attenuators, and there is quite an art in design and construction of them.

But there are also loads of poor-quality ones as well, and I have my own backlog of devices with scratchy volume pots and with channel imbalance. I had one of these Naim pre-amps which had terrible channel imbalance when it wasn't turned up a lot and it's apparently a common issue with the design. In hindsight I may have been an audiophool buying Naim, but regardless it's not like it's easy for a consumer to research the volume potentiometer used in the thing they buy.

Attenuation pre-DAC is something that can be done inexpensive devices at high quality. If someone is only spending $300 on electronics, it might be better advice do attenuate in the digital domain. If buying a high end pre-amp, the same advice might not apply.

 

[restorer-john]

Attenuation pre-DAC is something that can be done inexpensive devices at high quality. If someone is only spending $300 on electronics, it might be better advice do attenuate in the digital domain. If buying a high end pre-amp, the same advice might not apply.

I agree with that sensible advice.

 

[unpluggged]

you need to have sufficient reserve gain to be able to listen to all your content at the level you want and not leave a bunch of output on the table.

Thank you for your invaluable advice. That's where the autoref comes into play in my setup.

Concerning the rest of your condescending rant, allow me the privilege not to react.

 

[Wunderphones]

What you really want is a digital volume control, but one that's implemented by way of a 64-bit DSP, rather than the DAC chip.

 

[antcollinet]

Setting the DAC at 0 and using a volume control afterwards maximises your signal to noise ratio* and minimises any degradation caused by the way the digital values are manipulated in order to "reduce volume"**
* When you turn down the analogue volume control you also turn down the noise in the preceding elements
** Some DAC digital volume controls are better than others in terms of how/whether they reduce bit depth

In practice, assuming you don't turn down the DAC volume very much, you may not notice any degradation given that very little real world music uses all 16 bits anyway. But if you are serious about trying to make the domestic playback chain the best it can be, this is the correct approach...

It's better than that - almost all DACs now process in 24 bit (or higher) - any 16 bit material is upsampled to 24 bit before the volume control. That means you've got -48dB of volume control before you lose any of those 16 bits. Even more, as you turn down the 24 bit volume, the 16 bit noise floor** of the source material is also turned down until it vanishes into the noise floor of the 24 bit output.

If your source material is 24 bits - it still doesn't matter. You can still use -48dB of volume control before going below 16 bits. Try telling the difference between 16 bit and 24 bit material when you are listning at signal levels 1/250th of full scale.

The only area you lose out on compared to post DAC analogue volume control is in the noise generated in the analogue output electronics of the DAC itself. This is not attenuated. But since you are using a DAC with inaudible levels of noise in any case (you are, aren't you? ), this really doesn't matter.


**both quantization noise, and noise introduced at the recording stage.

 

[Willem]

I once again want to point to the volume control of the RME ADI-2. Its auto reference level functionality combines the accuracy of a digital volume control with a stepped analogue attenuation to maintain the best S/N ratio. It does not get any better.

 

[Ken Tajalli]

It's better than that - almost all DACs now process in 24 bit (or higher) - any 16 bit material is upsampled to 24 bit before the volume control. That means you've got -48dB of volume control before you lose any of those 16 bits. Even more, as you turn down the 24 bit volume, the 16 bit noise floor** of the source material is also turned down until it vanishes into the noise floor of the 24 bit output.

If your source material is 24 bits - it still doesn't matter. You can still use -48dB of volume control before going below 16 bits. Try telling the difference between 16 bit and 24 bit material when you are listning at signal levels 1/250th of full scale.

The only area you lose out on compared to post DAC analogue volume control is in the noise generated in the analogue output electronics of the DAC itself. This is not attenuated. But since you are using a DAC with inaudible levels of noise in any case (you are, aren't you? ), this really doesn't matter.


**both quantization noise, and noise introduced at the recording stage.

Do DACs upsample to 24bits?

 

[antcollinet]

Do DACs upsample to 24bits?

They have to. Othewise the output voltage would be at -48dB. They may not do it intelligently (Just pad the LS 8 Bits with 0's) but it doesn't matter.

 

[Ken Tajalli]

They have to. Othewise the output voltage would be at -48dB. They may not do it intelligently (Just pad the LS 8 Bits with 0's) but it doesn't matter.

But that is not upsampling, nor will it affect the volume control as you described, unless I didn't get it right.
Padding is a given. True upsamling may add 3-4dB of dynamic range, so no more than 17bits.
At any rate, in essence, I agree with you. All this talk about noise is irrelevant. Modern few hundred dollar dacs, have noise floors so low, even at -20dbfs, they are still better than many poweramps.
And if you listen at -20dBFS, then the room background noise is going is still higher than even the poweramps, so noise is not an issue.
I doubt anyone listening to a 12bit dynamic range of even a good recording at quiet levels, would be able to tell it from 24bit!
Sure, if I was listening at 110dB on a fine system, maybe I could tell a full 16bit from 12 (doubtful), but not at home and quiet levels.
I think the real question may be somewhere in dithering or noise shaping or both

 

[antcollinet]

But that is not upsampling, nor will it affect the volume control as you described, unless I didn't get it right.

Depens on your definition of upsampling, and you didn't get it right wrt volume control . 16 bit PCM goes from −32,768 to 32,767 That has to be upscaled (at least by padding) by a 24 bit DAC which needs to present values in the range -8388608 to 8388607 to the DAC chip.

So if it is padding, the 16 bit number is effectivley multiplied by 256, and we have a 16 bit signal but in a 24 bit range. Albeit with steps of 256 - so still only 16 bit resolution, and occupying the most significant 16 bits of the range (Bits 23 to 8).

Now we do digital volume control, lets say with 1dB steps. So for each 1dB volume reduction the dac must multiply the values by 0.891. Each time it does this we still have the full 16 bits of resolution, as you can understand if you consider each 6db volume reduction.

Each 6 steps of 1 DB is the same as dividing by two. Which is the same as shifting the number right by one bit. So with -6dB we now have the same 16 bit number but occupying bits 22 to 7 of the 24 bit range, and steps of 128, with the MSB=0. This includes all the noise of the 16 bit signal which has also now been divided by two, relative to the 24bit full scale.

-12 DB occupies bits 21 to 5 - and so on, until you are at -48 db where you have the original 16 bit signal occupying bits 15 to 0.

 

[Ken Tajalli]

Isn't that what another member referred to as "Volume control in the DSP" ?
I get what you are saying, this why I said, the question is in dithering and/or noise shaping - my knowledge of this is not huge, but applying digital volume control within DSP requires a bit more than just a simple number crunching.
Never the less, thanx for your explanation.

 

[antcollinet]

Isn't that what another member referred to as "Volume control in the DSP" ?
I get what you are saying, this why I said, the question is in dithering and/or noise shaping - my knowledge of this is not huge, but applying digital volume control within DSP requires a bit more than just a simple number crunching.

Well to implement digital volume control you have to at least be able to multiply - which I guess is DSP (at it's most basic). The point of my argument is that it doesn't have to be any more than that - if you are working on 24bit PCM - for it to be effectively transparent.

Sure - if you are working with 24 bit material that has been dithered, and you don't want to lose that noise shaping, then you may want to make it more complex to "re-dither" the attenuated signal. But I'm going to suggest that with 24bit, -144dB of quantisation noise is probably going to be inaudble, shaped or not.

 

[tensor9]

Depens on your definition of upsampling, and you didn't get it right wrt volume control . 16 bit PCM goes from −32,768 to 32,767 That has to be upscaled (at least by padding) by a 24 bit DAC which needs to present values in the range -8388608 to 8388607 to the DAC chip.

So if it is padding, the 16 bit number is effectivley multiplied by 256, and we have a 16 bit signal but in a 24 bit range. Albeit with steps of 256 - so still only 16 bit resolution, and occupying the most significant 16 bits of the range (Bits 23 to 8).

Now we do digital volume control, lets say with 1dB steps. So for each 1dB volume reduction the dac must multiply the values by 0.891. Each time it does this we still have the full 16 bits of resolution, as you can understand if you consider each 6db volume reduction.

Each 6 steps of 1 DB is the same as dividing by two. Which is the same as shifting the number right by one bit. So with -6dB we now have the same 16 bit number but occupying bits 22 to 7 of the 24 bit range, and steps of 128, with the MSB=0. This includes all the noise of the 16 bit signal which has also now been divided by two, relative to the 24bit full scale.

-12 DB occupies bits 21 to 5 - and so on, until you are at -48 db where you have the original 16 bit signal occupying bits 15 to 0.

Excellent post, although I’m not following your bit ranges.

But to underscore your general point, the 9038 Pro is a 32 but DAC, so that means we’re extra golden with digital attention even if a 24 bit signal?

I really want to see measurements of SINAD in 5 or 10 dB increments, in volume control, from 0 down to the floor of some DAC. Tedious test but it would put these discussions to rest. Let’s just see the data. After all, that’s almost all that’s left to distinguish dacs these days: SINAD at lower volume settings. With todays, DAC’s output, who one needs level gain anymore? It’s all about attenuation. And who listens to their DAC at full volume?? These low volume data are arguably more important than full volume.

@amirm the people have spoken.

I will donate to ASR if I can get you to do this test.

 

[antcollinet]

Excellent post, although I’m not following your bit ranges.

But to underscore your general point, the 9038 Pro is a 32 but DAC, so that means we’re extra golden with digital attention even if a 24 bit signal?

I really want to see measurements of SINAD in 5 or 10 dB increments, in volume control, from 0 down to the floor of some DAC. Tedious test but it would put these discussions to rest. Let’s just see the data. After all, that’s almost all that’s left to distinguish dacs these days: SINAD at lower volume settings. With todays, DAC’s output, who one needs level gain anymore? It’s all about attenuation. And who listens to their DAC at full volume?? These low volume data are arguably more important than full volume.

@amirm the people have spoken.

I will donate to ASR if I can get you to do this test.

Bits are by convention numbered from 0 for the lsb to n-1 for the msb. So from 15 to 0 for a 16 bit number.

 

[dwkdnvr]

But to underscore your general point, the 9038 Pro is a 32 but DAC, so that means we’re extra golden with digital attention even if a 24 bit signal?

No, 24 vs 32 bit is irrelevant for volume control. Even 24 bit gives us a theoretical capacity of 144dB or 20+ dB of headroom compared to the actual performance of the best measuring DACs. Those bottom 3 bits will be buried in the noise.
32 bit resolution IS important for internal calculations and DSP (and some even go to 64 bit), but for output 24 bit is sufficient.