SMSL DO100 Review (Stereo DAC)

[staticV3]

@sarumbear @Robbo99999 @restorer-john
Because PCM filters came up during the discussion:

Here's a real good look at the ES9038Q2M's PCM filters, including the "hidden" filter preprogrammed into the programmable slot:


Edit: full-res version

 

[RichB]

Of course the numbers are 'incredibly good' because the entire top of the audible spectrum is chopped off, along with its harmonics.

Default filter 1 is circa -12dB down @ 20kHz according to your AP plots... Not sure when that became remotely HiFi. Like, not ever. Certainly not in the digital era, or the open reel era, or even the cassette era.

Filter 3 is clearly the one you should be using for tests.

The last D/A converter you reviewed had the best filter response you've ever measured (the Chord) and yet this thing is getting top marks with arguably one of the worst default filters I've ever seen. Go figure. Where is the consistency?


Hardly. The default filter response (you used for testing) is poor and you know it.

Benchmark uses custom linear-phase filter that is not as steep as Chord but looks good:

I just bought a used one.

- Rich

 

[PeteL]

Benchmark uses custom linear-phase filter that is not as steep as Chord but looks good:

View attachment 208488
I just bought a used one.

- Rich

about 87 dB of attenuation is OK I guess but not great. There will be some imaging unlinearities as well. I am one to think we may put a little too much importance in these small flaws, it's out of our hearing range, but if you are going to go and design your own I feel there should be an objective improvement on what's available in the DAC chip itself? Not saying its a bad purchase, Benchmark has an outstanding reputation for their DACS, but it does feel like an exemple of over engineering. Unless maybe their whole conversion chain is custom and not from an off the shelf DAC?

 

[staticV3]

about 87 dB of attenuation is OK I guess but not great.

-43dB to -56dB at Nyquist is 13dB of attenuation.

 

[PeteL]

-43dB to -56dB at Nyquist is 13dB of attenuation.

If you want, I was talking of the total attenuation of the filter, not at Nyquist. I already said that there would be imaging so it's implied that it's not brick walled at Nyquist Nitpicking a bit?

 

[Garrincha]

@sarumbear @Robbo99999 @restorer-john
Because PCM filters came up during the discussion:

Here's a real good look at the ES9038Q2M's PCM filters, including the "hidden" filter preprogrammed into the programmable slot:
View attachment 208485

Edit: full-res version

With "hidden" you refer to the "reserved"? Because this one really appears to be almost perfect, only less than 1dB down at 20 kHz and almost 100dB down at Nyquist. If this filter is a given one with the chip, why not everbody is using it, also the manufactorers as a default?

 

[staticV3]

Understood. Please forgive my mistake.

 

[respice finem]

facts. Who needs them?

Not what I meant. But, the fact that nobody's hearing is perfect, is just that - a fact. And believing to hear everything is - a belief...
Being an objectivist, I acknowledge my own limitations.

 

[staticV3]

With "hidden" you refer to the "reserved"?

Yes.

If this filter is a given one with the chip, why not everbody is using it, also the manufactorers as a default?

ESS advertize one freely programmable filter slot.
I guess just a fraction of manufacturers realize that ESS put a unique filter in there from the factory.
And just a fraction of those make an effort to make that filter accessible to the end user.

 

[PeteL]

Yes.


ESS advertize one freely programmable filter slot.
I guess just a fraction of manufacturers realize that ESS put a unique filter in there from the factory.
And just a fraction of those make an effort to make that filter accessible to the end user.

That is indeed the best filter, and it does seem to me that there gotta be a different reason than just manufacturers not realizing it's there. DACs takes many months of development, normally you would lift every stones to try to get optimal performance. May I ask how you got these graphs? I think there gotta be some payback. Maybe it require to add some memory, maybe the latency is impractical, I don't know, strange.

 

[amirm]

I always thought of "threshold of hearing" as a SPL metric. may you tell us the math behind this 15 dB below threshold?

It is a simple conversion. Assume the peaks of the tone are at 120 dBSPL, then subtract the distortion from it and see where it lands. In other words, the math assumes amplification with reference playback (peak) of 120 dBSPL (sometimes I use 115 dBSPL). If distortion is -130 dB, then it is -10 dBSPL.

 

[DrZingo]

I see, so not Threshold of hearing but threshold of proven transparency for houses built with the collaboration of top gun acoustical Engineers, out alone in a field with no windows or other electronics, listening in the dark.

Even then. Doesn't this sentence sound contradictory?

"CDs, with undithered audio, have a maximum dynamic range of 96 dB (16 bits). Studies, such as the one conducted by Meyer and Moran, have shown 96 dB of dynamic range is transparent for any normal listening conditions.
120 dB (20 bits) of dynamic range is required for transparency. Anything below -120 dBFS is inaudible. "

Yes, it does. I suppose it has to with what you define as normal listening conditions. The threshold of hearing a multiband stimulus (rustling leaves, a whisper) is usually regarded as 0 dB SPL around 1 kHz. For a pure sine wave it's higher (3 dB SPL) as fewer hair cells are stimulated. When operating a jackhammer you are subjected to about 100 dB SPL so that's roughly the range of the undithered 16-bit format (and a reasonable dynamic range for most conditions and people, I would say). With dithering and noise shaping you gain several dB more, equivalent to about 20 bit - I would contend that this format is more than good enough for human sound perception.

 

[amirm]

Someone asked me about coil whine while the unit is powered but off. I checked for this and mine doesn't make any noise. But I know some who have bought the unit and heard it.

 

[HansHolland]

Yep, that's right, JBL 308p Mkii speakers. I'll generally set the volume control on the back of the speakers to the loudest I'd want to listen when the DAC is at 0dB, sometimes I kick it up another notch or two on the back of the speakers but most of the time I'll be using the preamp in the E30 DAC to control the volume - generally -6dB to 0dB, but sometimes down to -10dB.

just like me, we are able to set the gain of what's up after the DAC. For us that measurement we are talking about is not that interesting (I don't expect much difference between 0 dB and -10 dB (yes, 10 dB)).

For those who do not have that option, it is possible that their gain is way too high and than they have to attenuate a lot with the volume control of the DAC.

 

[PeteL]

Yes, it does. I suppose it has to with what you define as normal listening conditions. The threshold of hearing a multiband stimulus (rustling leaves, a whisper) is usually regarded as 0 dB SPL around 1 kHz. For a pure sine wave it's higher (3 dB SPL) as fewer hair cells are stimulated. When operating a jackhammer you are subjected to about 100 dB SPL so that's roughly the range of the undithered 16-bit format (and a reasonable dynamic range for most conditions and people, I would say). With dithering and noise shaping you gain several dB more, approaching 20 bit - I would contend that this format is more than good enough for human sound perception.

I think it is too but this is implying optimal gain structure. I think that many of these studies are a bit old, A CD player, hooked to an analog preamp for volume, to amp and speakers. I understand the aim to get more dynamic range nowadays. With very sensitive Iems, that block tons of outside noise and large amount of digital attenuation, sometime some margin needed for EQ, combined with many use cases where the analog gain is high and uncontrollable, this direct comparison with real world levels is not accurate. A lot of these DR dBs are lost in the chain. Bottom line is that there are situations (rare) where the hiss become audible in IEMs but anyway, I am totally OK with the 16 bits format but I am not that good at detecting low level noise and if I don't hear it it's good enough for me.

 

[cheapmessiah]

Ive been really interested for the measurements of this DAC, i suspected it was going to ace amirs tests as SMSL already made stellar performing DACs, and this one packs all the functinalities most people will ever need, and it edges the E50 with the addition of BT for a lower price and it even has internal PSU.

I believe this is the DAC to get at the moment, since once you factor in cables, the D10B isnt that much cheaper than this one, and the feature set makes it really atractive.

What a time to be an audio enthusiast.

 

[staticV3]

May I ask how you got these graphs? I think there gotta be some payback.

I measured the filters myself.
The DAC was an E1DA 9038S G3, which lets you choose between all eight filters via the Tweak9038 Android App. See here
Measurement setup was a Cosmos ADC in Stereo at 96KHz sample rate.
65K FFT in REW using Flat-Top window and a ridiculous amount of AVGes (like 12-16K)
White noise was generated in Adobe Audition, exported as 32/44.1 WAV, then played back via a smartphone running UAPP.

 

[PeteL]

I measured the filters myself.
The DAC was an E1DA 9038S G3, which lets you choose between all eight filters via the Tweak9038 Android App.
Measurement setup was a Cosmos ADC in Stereo at 96KHz sample rate.
65K FFT in REW using Flat-Top window and a ridiculous amount of AVGes (like 12-16K)
White noise was generated in Adobe Audition, exported as 32/44.1 WAV, then played back via a smartphone running UAPP.

Thanks.

 

[Labjr]

Ive been really interested for the measurements of this DAC, i suspected it was going to ace amirs tests as SMSL already made stellar performing DACs, and this one packs all the functinalities most people will ever need, and it edges the E50 with the addition of BT for a lower price and it even has internal PSU.

I believe this is the DAC to get at the moment, since once you factor in cables, the D10B isnt that much cheaper than this one, and the feature set makes it really atractive.

What a time to be an audio enthusiast.

I prefer XLR jacks on the SMSL. E50 has ES9068AS and XU218 chips for...cough! cough! M*A. I said it!

 

[pablolie]

If it had a good headphone output, it'd be perfect!