Topping D10s USB DAC and Bridge Review

[Veri]

MPC-HC is such a good player.. has Meier crossfeed too. Among other things.

 

[Sprint]

Not that I think the differences are audible but just wanted to point out there are trade-offs among filter length (delay), ripple, steepness and attenuation. Also, if one wants to use software upsampling anyway it can be done for free using something like the SoX plugin for foobar2000. Even MPC-HC can do some implicit auto upsampling with the built-in (libsamplerate?) resampler with more than decent results.

Is the filter ripple bad and affects sound quality?

 

[bennetng]

Is the filter ripple bad and affects sound quality?

If my answer is "yes" then I need to provide some blind test results so I would not say this. The whole thing started from someone claimed some minor imaging up to ~0.55fs is wrong and driven by marketing:
https://www.audiosciencereview.com/...r-mirror-images-short-study.20318/post-671014
For more details read the subsequent replies from other members as well.
I just presented my counter argument about this, because this minor imaging greatly reduced the ripple without using an additional DSP chip to create a "mathematically correct" filter with low ripple and flat response up to 20kHz, which adds costs and latency.

 

[AnalogSteph]

Is the filter ripple bad and affects sound quality?

Maybe. It's certainly higher than any of the examples given by Julian Dunn in his 1998 paper.
Ripple of +/-0.075 dB at a period of ~3.25 kHz in a linear phase (FIR) filter would give pre- and post-echoes spaced at 0.3 ms at a relative amplitude of 20log(10^(0.075/20)-1) dB - 3 dB = -44.2 dB. While the post-echo would definitely be masked, I wouldn't be 100% sure about the pre-echo. The psychoacoustics department may be able to come up with some thresholds depending on displacement.

In any case, it's getting a little close for comfort. We're not talking about anything 70, 80 or even 90 dB down, at which point audibility would almost certainly be out of the question either way. For -80 dB we'd need periodic passband ripple of no more than +/-0.0012 dB. Now very few ADCs have any lower than 0.001 dB and a lot 0.005 dB only (the best I've seen is the PCM4220 "classic" filter, +/- 0.00015 dB), so you probably won't need a whole lot less in a DAC either. Anything lower than +/-0.01 dB I would not be particularly concerned about.

Running the filter faster will reduce echo displacement correspondingly, making things more and more likely to be masked. By the time we're up to 4X speed (period ~13 kHz), it should be pretty much a non-issue. This is why I tend to be relatively unconcerned with the filters in several Cirrus Logic converters that merely show 1 or 1.5 periods of ripple across the entire passband, even if those are reaching +/-0.02 dB.

You can get away with far more passband ripple in a minimum phase (IIR) filter, which will produce a post-echo only, easily masked by the auditory system.

In any case, in this day and age, we should not have to worry about filter ripple in a pure playback device. About the only price to pay for low ripple is filter group delay, and that's just not a major concern in a pure playback application as it will rarely exceed 1 ms. Worst-case for the ES9038Q2M is 35/fs.

 

[bennetng]

Strong ripples are not uncommon, but just uncommon on newer and discrete DACs. RMAA uses a rather relaxed standard (40Hz to 15kHz) so what the summary showed are mostly likely ripples rather than extremely hi/low end roll off.

 

[AnalogSteph]

Ah, the bad old days of consumer (computer) audio. A +/-0.1 dB ripple spec used to be quite common back in the '90s.

For kicks, here's some performance data for a few typical 8X oversampling digital filter chips as used in CD players ca. 1989-1993:
NPC SM5840A/B: ripple 0.125 dB +/- 0.03 dB, stopband atten 55 dB
Yamaha YM3433(B): ripple +/-0.002 dB, stopband atten >70 dB (and mostly, >75 dB)
Yamaha YM3434: ripple +/-0.0001 dB, stopband atten >100 dB
NPC SM5803A & SM5813A (& SM5843 "sharp"): ripple +/-0.000025 dB, stopband atten >110 dB (FIR filters are given as "153 + 19 + 17 order" - am I right in calculating my group delay as 76 samples at 2fs + 9 samples at 4fs + 8 samples at 8fs = 41.25/fs?)
Sony CXD1144: ripple +/-0.00001 dB, stopband atten >120 dB (but only up to 150 kHz = 3.4fs, then there's two "ears" around 4fs peaking at -70 dB)
Sony CXD1244: ripple +/-0.00001 dB, stopband atten >100 dB

 

[bennetng]

"Ears" are not too bad when compared to DSD64 noise cluster, but yes, 4x is a bit low.
Some filters you listed are quite impressive! For CD players, there are at least several hundred milliseconds of delay due to disc/optical pickup operation anyway, so no big deal.

 

[pedrob]

I have in the past. Not much to look at there that is significant.

Interesting. What is the best USB->S/PDIF bridge?

 

[jannek]

Interesting. What is the best USB->S/PDIF bridge?

You can only test it with an external dac. And we paid attention to that as well. Coax output is better than AP's. The toslink output is also good enough.

I think with a quality level like this there's no need to look for better. The remaining jitter is easily compensated by modern and decent DACs.
Apart from that the Audiophileo was tested quite good, if I remember correctly.

 

[hotdogsandwich]

Any reason I can’t use this with a pair of RCA splitters to drive a pair of active speakers and a sub? Will the output be halved or will the opamp increase voltage? I apologize in advance if this is a dumb question, I am generally mistrustful of electrons and doubt their motives.

 

[Sprint]

Any reason I can’t use this with a pair of RCA splitters to drive a pair of active speakers and a sub? Will the output be halved or will the opamp increase voltage? I apologize in advance if this is a dumb question, I am generally mistrustful of electrons and doubt their motives.

I am doing this but digitally. I use the SPIDF into my Genelecs active monitors (using SPIDF AES cable) and use the toslink output to miniDSP 2x4HD DDRC24 which feeds into my 2 SVS subwoofers. Works very well.

 

[somebodyelse]

Any reason I can’t use this with a pair of RCA splitters to drive a pair of active speakers and a sub? Will the output be halved or will the opamp increase voltage? I apologize in advance if this is a dumb question, I am generally mistrustful of electrons and doubt their motives.

I'm guessing you're talking about analog output not the digital. It will probably be fine. The output voltage will stay the same, but the D10s will have to provide more current. This shouldn't be a problem unless your speakers or sub have an unusually low input impedance, say <10k Ohms.

 

[hotdogsandwich]

I'm guessing you're talking about analog output not the digital. It will probably be fine. The output voltage will stay the same, but the D10s will have to provide more current. This shouldn't be a problem unless your speakers or sub have an unusually low input impedance, say <10k Ohms.

Yes, I meant analog out. My speakers, are KRK G4’s and the sub is an older Atlantic Tecnology with RCA input. I went ahead and tried it, everything seems to be working ok — no smoke pouring out of anything.

 

[hotdogsandwich]

I am doing this but digitally. I use the SPIDF into my Genelecs active monitors (using SPIDF AES cable) and use the toslink output to miniDSP 2x4HD DDRC24 which feeds into my 2 SVS subwoofers. Works very well.

Cool, just starting to read up on the miniDSP, I think i’ve convinced myself I need one so that would be handy.

 

[Sprint]

Cool, just starting to read up on the miniDSP, I think i’ve convinced myself I need one so that would be handy.

Definitely it is worth it. My Yamaha AVR did not do a good job for the subs. After running MiniDSP DDRC24, I did not even know that my SVS subs could bring more juice. The bass became more prominent, precise, tight and it is fun. Really worth it.

 

[sterkoff]

Curious as a novice to the following: Could I connect this to my pc for the USB connection and then optical out from the D10s to a Geshelli Labs Enog2 Pro while taking advantage of the display to show what is the current sample rate? That would make for a nice way to add some information to DACs that do not provide a screen if I am to understand correctly. Plus I have a few amps at my computer desk and would still use the d10s as well (and sell my Khadas Tone Board which already has this job of being my ESS dac).

 

[AnalogSteph]

I don't see why it shouldn't work. Seems worth a shot.

 

[sterkoff]

I don't see why it shouldn't work. Seems worth a shot.

Then I have just one more follow-up question that I honestly haven't thought about before. I have only had the option of optical out from my motherboard but the D10s would now allow for a coaxial digital connection. Any advantage or disadvantage to either type (assuming to a Geshelli Labs Enog2 Pro?) Edit: Or a link to an objective article on this topic?

 

[Veri]

Then I have just one more follow-up question that I honestly haven't thought about before. I have only had the option of optical out from my motherboard but the D10s would now allow for a coaxial digital connection. Any advantage or disadvantage to either type (assuming to a Geshelli Labs Enog2 Pro?) Edit: Or a link to an objective article on this topic?

Optical can break ground loops if you have any. Coaxical is otherwise generally better received inside the DAC because good optical receivers are expensive and not as available since COVID. Then again these low levels of jitter are proven not to be audible though (-120dB is audible threshold).

 

[somebodyelse]

I think the coaxial on the D10s is transformer isolated, which would also break ground loops. Check it though - I'm going by memory of seeing a transformer near the coax outputon the PCB photos.