Saturday 4 June 2022

REVIEW: S.M.S.L. DO100 DAC (dual ESS ES9038Q2M) [Part I - Overview, Filters & Jitter]. (And the likely continued contraction of MQA...)

Over the last number of years, we have seen a proliferation of DACs from Chinese manufacturers offering models of increasingly better resolution - higher fidelity - and greater feature set. These days, I am of the opinion that we have achieved way above and beyond the ability for human hearing to differentiate these machines unless the manufacturers purposely added some kind of "coloration" to the analogue output whether it be frequency response or changes to the noise and distortion levels (the "tube sound" being a prototypical example we may come across).

As you can see in the image above, we have the S.M.S.L. DO100 DAC (~US$250) in today for review. I bought this unit through retail channels and intend to either use it myself or give it away if I have no need for it.

For many devices these days, opening the manual will show us technical information like the frequency response of the filters in this example. I think this is a nice evolution among manufacturers recognizing that showing product performance is important to educated customers with graphs and technical details rather than mere prose which often provides little satisfaction for the consumer's intellect.

Today, let's start a multi-part look at this DAC; no need to hurry because I think the performance at this price point warrants serious consideration and the many ways and features one can gather performance metrics on (I don't promise to test everything of course!). Obviously it's not an expensive device, has a remote controller, has both unbalanced RCA and balanced XLR outputs, Bluetooth 5.0 input, capable of up to 32/768 PCM and DSD512 (1-bit, 22.6MHz) using USB 2.0, along with S/PDIF TosLink and Coaxial inputs for up to 24/192 and DoP DSD64. No money wasted on MQA licensing. Depending on what we find, this might be right for you...

I. Hardware Overview

As a DAC, this device internally uses dual ESS ES9038Q2M DAC chips. These Q2M chips have become quite popular with many mobile and desktop applications thanks to the fact that they are low-power chips yet provide excellent performance from the dual channels; each chip capable of stereo or mono operation. They are the "little brother" of the top-of-the-line ES9038Pro with 8 channels. And like the ES9038Pro, these Q2M's also can function in both voltage and current modes.

Just as the chips are like the little sibling to the top-of-the-line device, the way the SMSL DO100 is designed is reminiscent of the design of the Topping D90SE we talked about last year. The idea here is similar; both are driving the DACs in current mode, and both are using high quality TI OPA1612 opamps for current-voltage conversion (I/V C). A difference of course being that the DO100 only requires 4 sets instead of the 8 found in the D90SE.

I didn't open up the DAC, but the company's advertising material shows this which I'll take the liberty of reproducing here:

Based on the datasheet, these ES9038Q2M chips are rated as THD+N -120dB, and DR 129dB(A) when operating in differential current mode. Any DAC capable of this level of performance assuming everything else like linearity, frequency response, and jitter look good will already place the device in the upper echelon of the highest performance DACs out there.

Here's a further look at the DAC unit:

The box itself measures around 15x13x4cm, simple all-black aluminum with rounded corners. Not too many logos or text, and just the JAS "Hi-Res Audio" sticker on top. Weight of 1.25lbs. On the front there's an articulated knob that pushes in to select. The glass screen on the front will display output level (when we turn the knob for volume), samplerate, and also if it's playing DSD. Brightness is adjustable.

And a look at the rear:

That's quite compact with both gold-plated RCA and XLR outputs. Input over Bluetooth (Qualcomm QCC5125), coaxial/TosLink S/PDIF, as well as modern USB-C (XMOS U208-based). There's also a 3-prong grounded IEC; great if you want to try some high-end audiophile approved power cables. ;-)

Regarding the power supply, it's obviously integrated inside and is described as a low-noise linear design. When turned on, I'm seeing a continuous 5W on the Kill-A-Watt meter which drops to <1W when turned off / put into sleep mode.

Press and hold down on the control knob to turn on / off. You can also use the control knob to change menu items and click to either register a change or toggle between parameters - all quite intuitive. Parameters include which digital input, filter settings FL1-7, DPLL 1-9 jitter tolerance (higher values might be needed to increase S/PDIF tolerance for some devices like TVs, I think my default was 3), UAC1/2 (use UAC2 mode for modern hi-res USB 2.0 performance with DSD), and screen brightness 1-5.

[DPLL refers to the "Digital Phase Lock Loop" which is part of the "Jitter Eliminator" circuit. This parameter adjusts the bandwidth for the timing of the input signal in order to maintain a lock. With highly jittery sources like say an old CD transport, if you hear crackles or pops, you might need to push the value higher. It has been said that keeping this value as low as possible (ie. keeping at 1) could improve sound quality. I have not personally been able to verify this. I'm guessing this parameter is only meaningful for the S/PDIF inputs.]

Here's the remote control, labeled as an "S.M.S.L. RC-8C" which takes 2 AAA batteries. It's a typical IR unit with good range, no problems used in my sound room. Matte finish, rounded contours:

Power, volume (levels from 0-99) and input buttons would be the most utilized. FN will blank the screen, and the mute button works as expected; the display will show a blinking volume number to remind you of mute being on.

II. Oscilloscope & Digital Filter Options

Let's begin the measurements with a look at the RCA output into the digital oscilloscope:

Left and right channels are basically perfectly overlaid indicating excellent channel balance. This is with 2Vrms RCA output, with XLR output, voltage will double to 4Vrms.

Let's throw up some -3dBFS 1kHz 16/44.1 square waves:

Left channel (yellow) is a bandlimited 16/44.1 square, and on the right channel (cyan) we have a non-limited square wave showing the "ringing" due to the digital filter. This was done with filter FR6 which is the linear fast roll-off filter hence the symmetrical pre- and post-ringing at the edge transitions.

This of course leads us to discussing the digital filters. Because there's no MQA here, there's also no need to worry that the settings might change during use. Furthermore, there are no concerns that the DAC might end up using any of the degenerate low-quality MQA variants.

There are a total of 7 filters you can choose from and as per the manual, they are:

FL1 - Fixed minimum phase fast roll-off
FL2 - Linear slow roll-off
FL3 - Apodizing Apodization Filter [superfluous naming!]
FL4 - Minimum phase fast roll-off
FL5 - Minimum phase slow roll-off
FL6 - Linear fast roll-off
FL7 - Brickwall Mixing Filter

The manual shows this graph of the relative frequency responses:

That's a great start. However, we can look deeper into the filter options with the "Digital Filter Composite" (DFC) overlay graphs I've shown over the years (an extension of Juergen Reis' ideas).

Here are the first three "FL1-3" filter options:

As you can see, I've embedded the impulse response into the top right of each DFC plot for the first 3 filters.

And here are the last four filters - "FL4-7":

These basically look like the default filters described in the ES9038Q2M datasheet except for a change in the order being presented. Here's IMO the more logical order laid out in the datasheet along with the corresponding SMSL setting:

Linear Phase Fast Roll-Off = FL6
Linear Phase Slow Roll-Off = FL2
Minimum Phase Fast Roll-Off = FL4
Minimum Phase Slow Roll-Off = FL5
Apodizing Fast Roll-Off = FL3
Hybrid Fast Roll-Off (Intermediate Phase Fast) = FL1
Brick Wall = FL7

The ESS sequence makes sense because the first 4 are the "typical" options most people would probably use (linear/minimum and fast/slow roll-off), and then the last 3 gets us outside of the norm with apodization and intermediate phase options. Not sure why S.M.S.L. felt they needed to change the ordering.

Great to see that "absolute polarity" has been maintained as per the intentionally positive-pointing impulse responses.

Note that all of the filters showed intersample overload with the 0dBFS Wideband White Noise signal. This basically means that the filters do not provide a full +3dB overhead to cover most of the overload situations we might run into. This is quite common among DACs using the built-in chip filters. While these 7 filters are similar to the ES9038Pro as used in the Topping D90SE, the Topping/ES9038Pro is doing a better job with controlling intersample overload (as evidenced by the cleaner 0dBFS wideband white noise curves).

I didn't pay attention to which of the filters was the default (possibly FL1 which is pretty good with some characteristics like "goldilocks"). As a technically accurate audiophile, I like FL6 "Linear Fast Roll-Off" as being one with good frequency extension, sharp roll-off, minimal ultrasonic imaging, and linear phase (best temporal performance without phase shift).

As usual, filter preference can be a subjective or idiosyncratic choice; for example some people like NOS which would be technically poor for example...

III. Jitter

Let's move on to jitter.

I've left the DPLL setting as 3 on the SMSL DO100 DAC, the E1DA Cosmos ADC was set to capture at 32/192, using SpectraPlus-SC software at 1M-point FFT. Let's have a peek at the standard 16-bit (44.1kHz) and 24-bit (48kHz) J-Tests using USB input:

That looks excellent. Essentially ideal J-Test performance suggesting no jitter worth mentioning at all.

Let's have a look at the S/PDIF inputs - TosLink optical and Coaxial. As a standard over the years, I'm using my Squeezebox Touch's S/PDIF to feed the DAC:

I'm impressed by the excellent J-Test result from the TosLink and Coaxial S/PDIF inputs! 

Given how good it all looks above, while it has been awhile since I've shown this, let's accelerate the J-Test to 96kHz with the primary signal at 24kHz:

Wow, still excellent even with the S/PDIF interfaces! We can start barely making out a pair of sidebands at this accelerated sampling rate.

Okay, let's take it up another notch... Here's the J-Test accelerated even further to 192kHz samplerate, primary signal now up to 48kHz**:

Finally we see some notable sidebands around the primary signal. Impressive that the TosLink interface was able to maintain the 192kHz speed with no issues (Squeezebox Touch loaded with Enhanced Digital Output applet).

The lack of any meaningful difference between USB and S/PDIF on the J-Test is simply excellent performance. I played around with the DPLL setting and could not see any difference whether I used the default of 3 or the extremes at 1 and 9. (I did not see a difference either with the DPLL setting on the SMSL M100 Mk II measured awhile back.)

For comparison, I ran the extreme 24/192 J-Test with a few of my other DACs that I also know have reasonably low jitter:

The Topping D10 Balanced has a similar pair of +/-1kHz jitter sidebands although higher in level compared to the SMSL DO100. From previous testing we know that the Topping DX3 Pro tends to show a number of sidebands already at 44.1/48kHz and continues to do so at 192kHz. The RME ADI-2 Pro FS (AKM AK4490 DAC) pattern is clearly quite different with more of an irregular "skirt" at the base of the primary signal, suggesting varying amounts of low-level jitter. Regardless, the jitter sidebands are all below -120dB from the peak of the 48kHz primary signal which in itself would be inaudible.

Anyhow, all of this is consistent with the idea that jitter these days isn't really a problem with decent DACs - at all - regardless of what talking points some manufacturers and magazine writers want to claim they can hear! Talk is cheap - demand some kind of demonstration if people are going to insist on this and that being significant.

** To double check that running the J-Test at 192kHz, those sidebands I saw were not simply artifacts from the E1DA Cosmos ADC, I had a look at a 48kHz 24/192 tone without the J-Test 1kHz LSB square wave jitter modulation tone:

As you can see, without the stimulatory effect of the LSB modulation tone, we do not see those sidebands. Although the J-Test was created back in the days of testing AES/EBU and S/PDIF digital interfaces, we can still see jitter effects with USB digital transmission using this signal.


IV. Summary so far...

Clean, understated appearance with large sample rate display and DSD indicator.

Okay, wrapping up the first part on the S.M.S.L. DO100 DAC, we have here a device with excellent channel balance, and superb low-jitter performance whether it's USB or S/PDIF inputs. Certainly one of the best I have come across over the years.

The digital filter settings appear to be the standard options when looking at the ES9038Q2M data sheet. They're not perfect and all show a certain amount of intersample overloading which is common except for a few DACs, so feel free to use high quality software upsampling (Roon, HQPlayer) or apply some volume normalization like ReplayGain to attenuate excessively "hot" recordings. Another option is to just drop the output to "96" (max 99) using the volume control which will preserve overhead. To be honest, while we can discuss and show intersample overload technically, in practice it's not generally an issue with real music. The only time this might add to distortion in the sound quality is when we're playing excessively loud, likely distorted and clipped waveforms to begin with (these would usually not be considered "audiophile quality" masterings anyhow).

So far then, this is looking quite good!

Next week, in Part II, we'll delve deeper into the resolution of this DAC.


In other news, since I mentioned the Topping D90SE (US$900) above, I noticed that they've now released the Topping D90LE (US$800). Basically, you save about $100 by not getting MQA; money well saved IMO since the feature is worthless unless you subscribe to the $20/m TIDAL tier or still buy those questionable MQA-CDs from Asia.

Speaking of MQA-CDs in Asia, about a year back in May 2021, Bob Stuart wrote this article claiming around 600 MQA-CD titles have been released and linked to cdJapan as an outlet to search for these.

I was curious the other day, so had a look. Now in June 2022, in total we have 702 results, many limited editions, some no longer available. If we filter the list to show the available ones, there are only 347 available currently. The majority of those titles are reissued "classic" rock and jazz recorded on analogue magnetic tape with no "hi-res" provenance to begin with. Furthermore, it's rather distasteful/dishonest that these CDs are labeled "Hi-Res CD" when they are absolutely nothing of the sort; in fact quite the opposite! This must be some kind of "Newspeak" where "ignorance is strength" or something like that.

Obviously the CD market is nowhere as vibrant as it used to be, but even then, in the last year, it doesn't look like MQA-CD is growing in any significant way. The only CD player (other than the discontinued Oppo UDP-205) that I've seen here in North America that can handle MQA-CD is the Technics SA-C600; not exactly a large enough library worth buying an expensive player for! This, plus MQA only being available in TIDAL on their more expensive tier now I think all leads to the obvious conclusion that we're well along in a state of contraction for MQA unless I'm missing some other remarkable source of encoded content in the world.

At its peak in 2000, even HDCD had something like 5000 titles available! I highly doubt MQA will even approach that kind of number... Thank goodness.

On the hardware side, I hope this is the beginning of a trend towards off-loading "features" and logos that do not align with improving fidelity when being marketed to audiophiles as the next great thing. There was never any "need" for MQA to begin with for consumers (as summarized years ago) and this I suspect is not just the beginning of, but entering the last phase of a rapid denouement as something that was a foreseeably inevitable, money-wasting venture for those involved in MQA.


If you like your female "audiophile recording" vocalists, check out Vanessa Fernandez's I Want You (2019, DR12). A good mix of covers that I think will show off your audiophile systems. Also a good hybrid SACD if you're still collecting those!

All the best, dear audiophiles. I hope you're enjoying some nice music and experiencing the excellent fidelity of devices these days without wondering if you need to mortgage a home! ;-)


  1. Really interesting that Topping is marketing the Topping D90LE: makes the DAC a really super value. And it reveals some of the actual cost of MQA, for all those fanboys claiming it's "free".
    The Vanessa Fernandez album is a good one.

    1. Greetings Unknown,
      Yeah, good to see the D90LE!

      Saving $100 is certainly significant. While there might be some redesign to save costs (maybe a less powerful XMOS USB controller?), the question of how much MQA licensing played a part lingers. Looks like they've improved some things like jitter and added the "Sound Simulation" DSP like "Tube Simulation" which is cool - I'll take this over MQA any day. ;-)

      I hope it also addressed some of the issues I found in testing like increased harmonic distortion >96kHz samplerate and improved DSD noise level. You can bet I'll be looking into this with the SMSL DO100.

    2. Curious as to your findings =)

  2. Making sure I understand the inter sample overload characteristics of this DAC. If the volume control on the DAC is set to, let's say 96, then that would avoid the a possible overload because the volume control is done digitally before the conversion. Is that right? Or as an alternative, if using Roon or other similar software for volume control, one could set a volume limit to 96 for that device to similarly avoid a possible overload? In any normal listening situation, I don't think I'd come close to needing full output of the DAC and would use the DAC volume control.

    I haven't been able to find out in the manual or on SMSL's web site if both balanced and unbalanced outputs are active simultaneously. Can you confirm? Seems that this would be a good device to supply output over to balanced to my amplifier and unbalanced to a subwoofer.

    I like that they don't have a gaudy MQA sticker but would much prefer if the stickers were on the back. I have an SMSL M300 MKII with a blue finish in a second system and would really like to get rid of the HiDef Audio sticker. I tried to gently peel it off but it's seems like a strong glue. A sticker is better than a possible blemish but no visible sticker would be best.

    1. Since you mentioned RME in the review, I found the information that RME shared around their migration to ESS chips in the ADI-2 interesting. They mentioned the overload scenario with ESS chips and how they addressed it in the ADI-2. If I followed along correctly, they limited the input to avoid the scenario. The net result was a DAC with essentially the same measured performance as the AKM implementation. According to RME, the AKM chip handled that scenario more gracefully but at a cost of slightly lower dynamic range although those differences are measurably inaudible.

    2. Hi Doug,
      Thanks for the discussions on intersample overload and the situation here...

      Basically, yeah, we see that the better AKM DACs have a more graceful ability to withstand intersample overs. For example, have a look at the "Digital Filter F1" for the Topping DX3 Pro here:

      Which is the AKM AK4493 version of "Linear Fast Roll-Off" in this ESS DAC. Notice that it doesn't overload with that 0dB wideband white noise signal. Since ESS seems to pride themselves on the "best" specs like THD+N and DR, we can sacrifice a few of those dBs and achieve the same overhead as the AKM.

      Digital volume control on DACs like this one will attenuate the 24/32-bit data first, apply the filter, then feed into the DAC for conversion (you can see it in the datasheet block diagram that this is done first before processed in the HyperStream block).

      So that's the long way of saying... Yes. Turning the volume down to 96 will give the system enough overhead to not clip with the 0dB Wideband White Noise (it's actually somewhere between 95-96, but 96 is good enough).

      Alternatively, other ways of volume reduction will also do the job - giving it -3dB "Headroom management" in Roon DSP for example is easy.

      Use ReplayGain during playback to normalize track volumes like down to -18LUFS which I use, will also keep the volume in the "safe" zone. Discussed here:

      Regardless of what we do, remember that this is all just "perfectionistic obsessive-compulsive" stuff we as audiophiles engage in. ;-)

      Good quality "audiophile approved" recordings/masterings from places like MoFi, Analogue Productions, Reference Recordings, Soundkeeper Recordings, 2L, DCC, etc. with good dynamic range will not (in general) have those intersample overs anyways.

      Loud, excessively dynamically compressed modern stuff already comes "pre-distorted" so some overloading during playback will not substantially make things sound much worse IMO...

  3. Can the remote control adjust the balance (left - right)? I guess no, again...

  4. Should the voltages stay within 2V and 4V spec? Why go to 2.1?

    1. Hey 1984,
      Quite typical with DACs to see slight variation off from 2.0 or 4.0Vrms depending on the design as it doesn't really matter and there's no strict need to be exact (2.1V is 0.4dB higher than 2.0V).

      We saw the same thing with the Topping D90SE for example and the Topping D10s was 1.95Vrms.

      Others like the RME ADI-2 Pro will have adjustable output levels at 0dBFS (XLR) - +4dBu (1.23Vrms - often regarded as pro audio line level voltage), +13dBu (3.46V), +19dBu (6.9V), and +24dBu (12.3V); none of which 2 or 4V just to give you an idea of how much this can vary depending on the device.