Topping D90 Balanced USB DAC Review

[Pluto]

For example, if analog volume is available it should be used

I'm sorry, but that is Internet forum nonsense. A well-designed digital volume control is absolutely harmless in the real world. One might criticise the “law” of the control but the same may be said whether analog or digital. Likewise, some insist on a rotating knob rather than up/down buttons but that too is a personal preference.

The important point is that, in the context of a domestic DAC at an attractive price, the digital approach to a volume control engine is to be preferred.

 

[curiouspeter]

I'm sorry, but that is Internet forum nonsense. A well-designed digital volume control is absolutely harmless in the real world. One might criticise the “law” of the control but the same may be said whether analog or digital. Likewise, some insist on a rotating knob rather than up/down buttons but that too is a personal preference.

The important point is that, in the context of a domestic DAC at an attractive price, the digital approach to a volume control engine is to be preferred.

I'm fine with digital control. However, it's nice to have a two-way volume control so that the knob and the Roon Remote are connected.

 

[rcstevensonaz]

I'm sorry, but that is Internet forum nonsense. A well-designed digital volume control is absolutely harmless in the real world. ...

I've often wondered: is this "internet" nonsense, or just dated information that is increasingly no longer applicable?

My understanding was that in older DACs (16 bit), volume control was done through bit robbing which reduced audio fidelity. By contrast, newer DACs (32+ bits) have sufficient processing headroom that volume adjustments can be made without the mathematical loss of precision that results in lower fidelity. Or, is my understanding also wrong?

 

[curiouspeter]

I've often wondered: is this "internet" nonsense, or just dated information that is increasingly no longer applicable?

My understanding was that in older DACs (16 bit), volume control was done through bit robbing which reduced audio fidelity. By contrast, newer DACs (32+ bits) have sufficient processing headroom that volume adjustments can be made without the mathematical loss of precision that results in lower fidelity. Or, is my understanding also wrong?

It also depends if the manufacturer is indeed implementing a 32-bit volume control. Or is it just doing a bad implementation as a freebie because it does not expect people to actually use it for volume control.

 

[Veri]

It also depends if the manufacturer is indeed implementing a 32-bit volume control. Or is it just doing a bad implementation as a freebie because it does not expect people to actually use it for volume control.

The volume control will pretty much always be within the DAC, which tends to be at high bit depth with modern chips and low chance of fuck ups.

 

[pLudio]

It also depends if the manufacturer is indeed implementing a 32-bit volume control. Or is it just doing a bad implementation as a freebie because it does not expect people to actually use it for volume control.

The AK4499 is a 32-bit DAC with 32-bit digital filters so I assume the volume control is also 32 bits. It's not the tricky part.

 

[curiouspeter]

The AK4499 is a 32-bit DAC with 32-bit digital filters so I assume the volume control is also 32 bits. It's not the tricky part.

If only it is done by the DAC. It can also be done in the XMOS controller.

https://www.audiosciencereview.com/...nced-mqa-dac-review.17419/page-74#post-660300

 

[curiouspeter]

Check if MQA decoding works with volume below 0 db. This usually tells you whether volume attenuation happens before or in the DAC.

 

[Tokyo_John]

Check if MQA decoding works with volume below 0 db. This usually tells you whether volume attenuation happens before or in the DAC.

I can confirm that MQA works below 0dB. The process chain for PCM in the AK4499 block diagram is:
DATT -> De-emphasis & Interpolator -> delta-sigma -> SRDAC

Asahi Keisei describes in the AK4499 data sheet that the chip has an 8-bit level control (0-255) at half dB increments with the highest setting 255=0 dB. The AK4499 ostensibly goes down to -127dB at level 1 (0=mute). The volume on the D90 goes down to -99dB. I imagine that this is because the display doesn't allow space for another digit, so they just stop at -99. It is more trouble for Topping to do something different than using the AK4499 attenuator, so I doubt they've done something different.

I looked for more info on how AK does the volume attenuation but the details were thin, and perhaps they aren’t forthcoming. Perhaps Topping themselves don’t really know how the AK4499 DATT works...and it seems that nobody here knows, either.

 

[Nicolaas]

I can confirm that MQA works below 0dB. The process chain for PCM in the AK4499 block diagram is:
DATT -> De-emphasis & Interpolator -> delta-sigma -> SRDAC

Asahi Keisei describes in the AK4499 data sheet that the chip has an 8-bit level control (0-255) at half dB increments with the highest setting 255=0 dB. The AK4499 ostensibly goes down to -127dB at level 1 (0=mute). The volume on the D90 goes down to -99dB. I imagine that this is because the display doesn't allow space for another digit, so they just stop at -99. It is more trouble for Topping to do something different than using the AK4499 attenuator, so I doubt they've done something different.

I looked for more info on how AK does the volume attenuation but the details were thin, and perhaps they aren’t forthcoming. Perhaps Topping themselves don’t really know how the AK4499 DATT works...and it seems that nobody here knows, either.

Maybe some scientific guessing here. Thanks for the info! Now what does a crappy 8 bit attenuator do to sound quality? Maybe neukrisz was right after all and we need to apologise...lt would be great to have some measurements about the influence of digital attenuation on sound quality. Does it add distortion to the signal or does it only decrease signal to noise ratio? At what level would this be noticed?

 

[Tokyo_John]

Maybe some scientific guessing here. Thanks for the info! Now what does a crappy 8 bit attenuator to sound quality? Maybe neukrisz was right after all and we need to apologise...lt would be great to have some measurements about the influence of digital attenuation on sound quality. Does it add distortion to the signal or does it only decrease signal to noise ratio? At what level would this be noticed?

My impression is that high quality re-sampling/interpolation in the amplitude domain (e.g., digital attenuation) is easy, unlike re-sampling/interpolation in the time domain (some people are actually paying almost 5 grand for the Chord M Scaler). I trust that Asahi Kasei knows what they're doing, and they wouldn't pollute their flagship DAC chip with a process that crudely strips away bits. Still, I would like to know more about what, exactly, they do in the 4499...

 

[mocenigo]

Maybe some scientific guessing here. Thanks for the info! Now what does a crappy 8 bit attenuator do to sound quality? Maybe neukrisz was right after all and we need to apologise...lt would be great to have some measurements about the influence of digital attenuation on sound quality. Does it add distortion to the signal or does it only decrease signal to noise ratio? At what level would this be noticed?

Why should it be crappy and why 8 bit? The fact that one multiplicand is an 8 bit value does not imply that everything is reduced to 8 bits of precision. You multiply a 16, 24 or 32 bit number (the PCM input) by this 8 bit number and get a 24, 32, or 40 bit number, which you can then divide by 256 and round to the next integer, for instance. This already gives a good result. Or apply dither before truncation, which is even better.

 

[alexj84nl]

I did not went through all 151 pages before typing this, but does anyone know if there is a successor of the D90 coming soon? My dealer (local hifi shop ;-) ) told me this but I cannot find any information on this.

 

[Nicolaas]

Why should it be crappy and why 8 bit? The fact that one multiplicand is an 8 bit value does not imply that everything is reduced to 8 bits of precision. You multiply a 16, 24 or 32 bit number (the PCM input) by this 8 bit number and get a 24, 32, or 40 bit number, which you can then divide by 256 and round to the next integer, for instance. This already gives a good result. Or apply dither before truncation, which is even better.

Yes I understand that there are ca 65 k PCM voltage levels with 16 bit PCM. Which are attenuated in 256 steps (8 bit).
I listen to background music on ca -40 dB. For serious listening I use ca -20 dB. Most of my music is in FLAC 16b - 44.1k.
Maybe it would be interesting to know the influence of these digital attenuation steps on sinad...

 

[Veri]

I did not went through all 151 pages before typing this, but does anyone know if there is a successor of the D90 coming soon? My dealer (local hifi shop ;-) ) told me this but I cannot find any information on this.

The D70/D70s/D90 all use AKM premium chips, but those will soon be unavailable due to factory fire incident.
Maybe this is Topping's upcoming DAC (ESS Sabre?)

 

[EdW]

Yes I understand that there are ca 65 k PCM voltage levels with 16 bit PCM. Which are attenuated in 256 steps (8 bit).
I listen to background music on ca -40 dB. For serious listening I use ca -20 dB. Most of my music is in FLAC 16b - 44.1k.
Maybe it would be interesting to know the influence of these digital attenuation steps on sinad...

I imagine the effect of the digital attenuator would be similar to running back up the IMD curve by the same number of dB. I.e if you set -20dB digital atten then that is the same as running the DAC with a -20dB signal. So your 120dB SINAD DAC becomes a 100dB DAC. Not too much of an issue if you have a 4V O/P DAC feeding a 4V sensitivity mono block this will will nicely scale. In fact you might wish to have the capability to overdrive the power amp by a dB or two so that you can reduce volume by a dB or two to allow for inter sample overs. A problem arises if you feed the DAC into a much more sensitive amp then you end up throwing away dynamic range. Topping’s answer buy the PRE90

 

[pLudio]

Yes I understand that there are ca 65 k PCM voltage levels with 16 bit PCM. Which are attenuated in 256 steps (8 bit).
I listen to background music on ca -40 dB. For serious listening I use ca -20 dB. Most of my music is in FLAC 16b - 44.1k.
Maybe it would be interesting to know the influence of these digital attenuation steps on sinad...

It's a 32-bit DAC so your 16-bit source material is usually the limiting factor.

 

[nimar]

Topping’s answer buy the PRE90

Or the A90, It's cheaper and apart from volume control looks to have all the exact same specs, which leads one to believe that it has much of the same moving (not moving) parts.

 

[EdW]

Or the A90, It's cheaper and apart from volume control looks to have all the exact same specs, which leads one to believe that it has much of the same moving (not moving) parts.

The missing thing on the a90 for a lazy slob like me is the remote

 

[nimar]

The missing thing on the a90 for a lazy slob like me is the remote

I set the a90 to my ideal listening level, and leave the DAC in preamp mode at 0db, but can subtract a few db If I want to turn it down. 95% of the time I am at 0db, I enjoy getting up and moving the dial if I want to get loud. And my wife can turn it down from my level, she doesn't care about the loss of SINAD that -5db of digital attenuation is going to cause.