Hello Everyone,
This is a review and details measurements of the SMSL D200 DAC.
It was kindly sent to me by
Aoshidaudio.com
View attachment 475835
SMSL D200 DAC - Presentation
This is a nice looking device with only one knob on the front. It reads PCM (up to 768kHz) and DSD (up to 11MHz or 22MHz depending on DoP or DSD direct).
It features a ROHM BD34352EKV DAC that offers two interesting options, implemented in this SMSL D200:
- A classic Oversampling Filter selection : Sharp, Slow or None
- A much less classic oversampling rate selection for the Delta-Sigma modulator : 16x or 32x
As for the rest of the technical description and features, I invite you to read to go to the
SMSL site.
The specs are the below:
View attachment 475842
On the back we get:
- XLR and RCA out
- USB in
- SPDIF (Coax and optical) in
- Clock input too, for external synchronisation
- Bluetooth for LDAC (24bit/96kHz), APTX/HD,SBC,AAC
View attachment 475838
Note that the analog outputs offer two settings:
- Preamp volume control, with output up to 5.2Vrms (XLR) or 2.5Vrms (RCA) at 0dBFS.
- Fixed output (no volume control), with output of 4Vrmr (XLR) or 2Vrm (RCA) at 0dBFS.
Ok, time has come for measurements
SMSL D200 - Measurements (USB in - XLR Out)
As opposed to
CD Players that I like to review, I'm not used to measure these high resolutions DACs. So I'll follow the lead of
@amirm, as he did recently for the
Fosi Q6, and will try to concur to his way of measuring, so you get pretty much the same information.
All measurements performed with an E1DA Cosmos ADCiso (grade O) set to mono mode (SNR >129dBA), and the Cosmos Scaler (100kohms from unbalanced input).
I did not succeed to completely get rid of some power supply related leakage, but it stayed very low (<-135dBr).
I upgraded to latest firmware but I did not see any differences in performances before and after.
The D200 outputs a high 5Vrms from the balanced output at 2 times less from RCA. The channel imbalance is a very low 0.01dB, which is very good.
This is the XLR output (Variable output -2dB) with 1kHz @0dBFS (so that is 4Vrms output):
View attachment 479803
Similar to the AP of
@amirm, this is from 24bits/44.1kHz sampling rate, with 32k FFT and 4 averages (I can't do 3).
The two channels are shown but only one is evaluated in the dashboard, the other one has a THD+N of -115.7dB, so that's the same. It's a little better than the specs (-114dB) and the same as what
SMSL exhibits on their website with couple of views from an Audio Precision.
By the way, this is a view of the same but with RCA outputs, since SMSL shows them too:
View attachment 475851
I measured precisely the same performances as them, which is good news. Also, due to the negligible variation between RCA and XLR, I'll continue this review from XLR outputs only.
These performances, on SINAD perspective, would put this DAC in the middle of the "Excellent" raking of
@amirm.
[BEGIN EDIT]
@Rja4000 suggested I increased the gain at the input of the Cosmos Scaler in fear I'd bit hitting the noise floor of the Cosmos ADC, when measuring the DR. And indeed I got 2dB better DR when doing so.
So, on the
Dynamic Range side, I got the below:
View attachment 480305
This is from 997Hz test tone with a CCIR weighting curve, as the AES now mandates. So, we get a
DR of 119dB. This is very good. For the sake of comparison with measurements performed with the old
A-weighted curve, I get
123.6dB, which nearly exactly what SMSL reports with 123.4dB
.
[END EDIT]
Anyways for me who's used to measure CD players, that's crazy good, since this is 20dB better than CDA
Next is the bandwidth (48kHz sampling rate):
View attachment 475858
These two measurements concur with the published specs of the
Rohm BD34352EKV DAC. No surprise here.
From what I can measure, the IMD performance is good:
View attachment 475859
And the jitter test is nailed:
View attachment 479811
I forgot the Multitone test (48kHz sampling rate):
View attachment 476015
Despite that spike at 150Hz (an interaction from mains (50Hz)), we have 20bits or more of distortion free range.
EDIT: The below is from the WAV file that
@amirm shared some time ago, and is the multitone test from the Audio Precision (192kHz sampling rate). So that helps to compare with the other reviews here. Same results, basically:
View attachment 479809
For compatibility with Stereophile measurements, here is the 50 Hz spectra under 640 ohm load (not 600, sorry, I can't), Linear Frequency Scale:
View attachment 479817
This is good but we have a little more low level random noise with low frequencies tones, as we've seen especially with the Multitone test. This is nevertheless very good.
Oh yes, and the THD+N ratio vs level:
View attachment 475861
This is an excellent result, really. We see a some distortion at low and high frequencies but the THD+N ranges from -111dBr to -102dBr meaning a very low level of ultrasonic noise (since distorsion and noise are accounted for up to 90kHz in that test).
And here comes the interesting par of being able to change the oversampling factor of the delta-sigma modulator. Maybe you wonder what is "Sound Color #3" I put on the graph. Well, that was how SMLS named the different settings of the ΔΣ Modulator (from Sound Color #1 to #4) in their original firmware. After I updated the firmware, the different setup are aligned with the doc of the Rohm DAC, and the SOUND COLOR shows the below:
- LOW OSR + HPC ON
- LOW OSR + HPC OFF
- HIGH OSR + HPC ON
- LOW OSR + HPC OFF
Let me decipher all of that:
- OSR stands for OversSampling Rate, and my guess is LOW is for 16x and HIGH is for 32x oversampling rate of the ΔΣ Modulator.
- HPC stands for High Precision Calculation, and this is related to the mathematical precision of the computation, draining more current when in HIGH mode.
You can find all of that explained in the datasheet of the
Rohm BD34352EKV DAC, and they refer to it as "Sound Settings".
Let me further describe what that is. Look at the block diagram of the DAC:
View attachment 475867
You can see the oversampling filter, that can be set to sharp or slow, and that is the one adding more samples in between original samples, and performing the anti-imaging function, by filtering everything above Fs/2.
Next, there is the Delta-Sigma modulator that reduces the bit-depth to the one required for the final conversion. This is done again with an oversampling of a set order (generally 4, or more). To reduce the PCM stream from 24 bits, to 4 or 5 bits (I don't know here) required by the Delta-Sigma converter, there's a need to keep safe the original SNR. Since we loose roughly 6dB every bit we drop, it is necessary to compensate that loss by a combination of oversampling and noise shaping. These two mechanisms together allow the engineers to reject the quantization noise (when reducing the bit depth) far away in frequency.
And voilà, the idea of Rohm, offered to us to play with in the SMSL D200, is to reject even further the quantization noise of the noise shaper. Of course, this require more computing power, and has some drawbacks too, as I've seen and will show you.
The below shows a 1kHz sine @0dBFS played either with LOW or HIGH oversampling of the ΔΣ Modulator:
View attachment 475868
This linear frequency scale view goes up to 80kHz. You can see the advantage of the low ultrasonic noise with a higher oversampling rate. The green trace is with the ΔΣ Modulator oversampling rate set at LOW. You can appreciate the building of the noise that is rejected here in this mode, by the noise shaper.
This is the reason why the THD+N vs freq measurement was so good despite me including all noise up to 90kHz, like
@amirm usually do. But with the ΔΣ Modulator oversampling set to low, I assure you the performance would have been much lower, because of all the noise rejected from 35kHz by the noise shaper.
Ok, you thought this is a little too complex? So let me add a bit more
Have a look at the very interesting and busy graph below, based on white noise long time averaging:
View attachment 475869
This is an even wider view, up to 300kHz. I played with the FIR Oversampling Filter (Sharp and Low) and the ΔΣ Modulator oversampling rate too.
With that view, you can appreciate that:
- The slow filter (LOW) is fully active at only 44kHz and exhibits ringing, ie it filters less of the aliases of the conversion
- The sharp filter (SHARP) is fully active ad 24kHz and filters way better the aliases.
- The effect of decreasing the oversampling rate of the ΔΣ Modulator generates noise being built as soon as 35kHz and going up to -60dB at 165kHz.
The dark blue trace is the SHARP FIR and the HIGH ΔΣ Modulator oversampling rate, and I like it
Note that there is a "NO FILTER" option, but there's a problem with this mode, at least with my early device. I saw a significant decrease of level and high distortion. There's probably something wrong, but at least I saw it indeed does not filter at all, plenty of unwanted aliases are created out of the audio band, meaning it mimics a Non Oversampling DAC (which is bad, don't use that), but since there's still a ΔΣ Modulator that must oversample, I see the effect of that oversampling but I can't understand it. So as long as I'm not 100% sure of what I measure, I'll refrain to publish, but anyways don't go NOS mode with a DAC, please pretty please.
SMSL D200 - Measurements (SPDIF in - XLR Out)
I briefly checked the quality of the SPDIF optical input (16bits only), to find excellent performances. But I saw additional low level random noise at the foot of the fundamental, when feeding it with 16bits stream from a CD Player. It shows better itself when performing a jitter test (16bits/44.1kHz):
View attachment 475873
This would have been a perfect trace without the little random noise at the foot of the main tone. It is negligible below -120dBr, though.
EDIT: I just did the same test from a Pioneer PD-D9 and the result id better:
View attachment 487116
The noise is almost at the foot of the fundamental. I don't why it was higher with the SMSL PL150, but I'll check that when time allows (I suspect a problem with the Coax cable I used with the SMSL).
And that being said, this is a very high precision DAC anyways.
Best proof is one of my favorite measurement that Stereophile was often using as a proof of low noise DAC. It is from an
undithered 997Hz sine at -90.31dBFS. With 16bits, the signal should appear (on a scope) as the 3DC levels of the smallest symmetrical sign magnitude digital signal
View attachment 475871
Without surprise, this is one of the best trace I've seen, nearly 0 noise comes in the way, it is very close to the digital data.
Theoretically we should get a square at this lowest level of the PCM 16bits format, and we don't because of the limited bandwidth (Gibbs phenomenon). The ringing tells us that the reconstruction filter is symmetrical.
By the way, since I'm here, at fixed output level, that is 4Vrms XLR (-2dB compared to max possible), let's see how this DAC resists my usual intersamples over test:
Intersample-overs tests
Bandwidth of the THD+N measurements is 20Hz - 96kHz | 5512.5 Hz sine,
Peak = +0.69dBFS | 7350 Hz sine,
Peak = +1.25dBFS | 11025 Hz sine,
Peak = +3.0dBFS |
| D200 from PL150 (CD player) | -68.5dB | -56.9dB | -39.1dB |
| SMSL PL150 | -52.1dB | -30.8dB | -19.2dB |
Well, we get some headroom but not much. If I go to "preamp mode" and lower the volume by a lot more (eg -6db) I don't get more headroom, and that is because the volume control is not done in digital domain, prior to conversion, but post conversion by an IC that attenuates the output voltage (digitally calculated attenuation).
Volume Control
Following a conversation started
here about the volume knob of the D200, I edit the review to add couple of information, from the feedback of
@Panelhead and
@Roland68.
The volume control is provided, post DAC, by a digitally controlled analog attenuation chip (NJU72315). This means there should be a difference in measured performances between send an attenuated digital signal to the DAC, or attenuating by the same level via the front volume know.
And indeed, this is a comparison between the two for a -6dBFS output (D200 set at 5Vrms output, meaning it will be reduce to 2.5Vrms):
View attachment 495916
The digitally attenuated volume decreased the SINAD by nearly 3dB, while it remained the same when attenuating with the Volume knob
And now the same test but at -12dB attenuation:
View attachment 495917
The difference is now significative. In tems of SINAD and therefore ENOB, we loose more than 1bit of resolution when decreasing the output digitally, compared to using the volume knob. It is good to see that a -12dB, you get the exact same performance (SINAD >115dB) than at full scale.
At -24dB, there is a now a significant difference in SINAD of near 10dB:
View attachment 495918
Digital attenuation (left) gets us to below CD Audio resolution, while volume attenuation by the same -24dB keeps us way above CDA, on a SINAD perspective (97dB vs 106.3dB).
In addition, I did not see any trace of channel imbalance from 0dB to -32dB (the two channels matched at 0.00dB).
I also thought about another way to understand the performance of the volume control, as I wanted to compare to more standard digital attenuation pre-conversion.
For those, I used a Topping D50III, kind of a reference, which offers digital volume control. And I compared with the D200 at different levels of attenuation: I sent a constant 1kHz @0dBFS, and then attenuated the output using their respective volume controls. I then captured key elements of performances, the THD+N, THD and ENOB:
As you can see, if the Topping does better than the D200 in pure performances above -10dB, the D200 leads from -10dB and below.
This means that the D200 is recommended if your plan to use it as a preamp. Its volume control is not as advanced as the one of an RME ADI-2 DAC for instance, but it is nice to see that you can keep a high level of precision even with significant attenuation. It is also good to confirm that digital volume attenuation, pre-conversion, can be beaten by a well-designed volume IC.
All the above means that the volume control of the D200 is very precise and ensures minimum loss. It is to be preferred if you need to attenuate the output of this DAC.
Note that the DAC will recall the last attenuation at start up.
Conclusions
Sorry, I know it's a lot of information to digest, it was for me too and it's not finished!
But from what I gathered so far, I think this DAC is a high performing one, and the possibility to play with a selection of FIR filters and ΔΣ Modulator oversampling rate is cool, even if I can't hear a difference.
And for those of you worried about a noise shaper generating too much noise, too close from the audio band... well, you have a winner here!
@Scytales if you read me
The little random noise I've seen from a CD Player, as a source, is of no concern.
I hope you enjoyed the review and have a nice weekend!
