REVIEW: S.M.S.L. DO100 DAC [Part III - DSD, Subjective, AMPT Recording, and Conclusions]. And a very close look at the Topping D10 Balanced THD(+N) vs. Level graph.

As you can see, the SMSL DO100 displays DSD samplerate in the form of DSD64 (64 x 44.1kHz = 2.8MHz), DSD128 (5.6MHz), DSD256 (11.3MHz) and DSD512 (22.6MHz).

Okay audiophile friends, let's finish off the trilogy of posts on the S.M.S.L. DO100 DAC with some final measurements around DSD performance, let's talk about subjective listening impressions, and as usual, let me provide a direct recording from the XLR output for comparison with others I have done (as discussed here).

We'll then put this all together as final concluding thoughts on this device.

I. DSD Performance

Over the years, DSD (1-bit PDM, SACD and such) has been promoted as an alternate modulation system for audiophiles yet you've probably seen precious little testing done! While I think the hype around DSD has faded out since the last resurgence attempt in the mid-2010's, I have tried to change this in my reviews with the incorporation of SoX-DSD encoded test signals these days with in-depth device posts, as discussed last week. The hope is that we might be able to detect DAC limitations not just between PCM and DSD but also within the family of DSD sample rates; as I noted before, just because a device might be advertised as being able to play DSD512 doesn't mean it can do it just as well as DSD128 or DSD256!

As a convenient "battery" which allows us to look at a number of characteristics, let's start with the RightMark as I usually do with PCM testing. Since there is typically quite a bit of noise with DSD, let's just look at side-by-side comparisons of 24/192 DSD-encoded signals at various rates with the native PCM playback; this will show us the noise up to 96kHz. XLR output was measured for best performance:

The synthetic RightMark test signal originated as a 24-bit PCM so the conversion to 1-bit DSD is to some degree "lossy". One does not expect exact measured resolution but it should be close. Indeed, if we look at the numbers above, the results are close with similar frequency response, dynamic range around -120dB(A) or even better, and likewise very low distortion results.

Here's the usual composite summary graph:

For DSD graphs, due to the high ultrasonic noise, sometimes a simple noise graph will be misaligned. This can be realigned in the -60dB Dynamic Range graph while still showing the noise shaping effects.

As you can see, when comparing PCM and DSD performance, they all look equivalent except for the anticipated ultrasonic noise with DSD (which distorts the IMD+N sweep as well) and the PCM original data shows slightly better performance.

I must say that I am impressed at the performance of this little US$250 DAC as a DSD playback machine thus far!

Remember years ago SACD (DSD64) advertisements would claim an "extended frequency response up to 100 kHz"? As you can see over time with measurements, DSD64 does get very noisy starting just above 20kHz so we're certainly not seeing anything like a flat frequency response to 100kHz. Furthermore, most playback hardware will implement a filter around 30-50kHz for DSD64 playback which means that in reality, that idea of 100kHz frequency response was always just advertising hyperbole.

With the high resolution E1DA Cosmos APU, let's have a better look at the THD+N of a 1kHz signal at various samplerates compared to 24/96 PCM (as per the "standard" signal posted last week):

As suggested by the RightMark test, this DAC does a great job with DSD material, with DSD256 and DSD512 THD+N less than 1dB away from the PCM result. We have also seen last week with the Topping D10 Balanced that the DSD playback does seem to consistently show lower THD than the PCM at this output level.

Moving on, let's have a peek at the complex "1/10 Decade Multitone 32" in DSD:

That's rather beautiful! Both DSD64 and DSD256 capable of at least 125dB range between the peaks and any distortions/noise. Rising noise with DSD64 above 20kHz as expected.

Finally, for completeness, let's run a 24-bit J-Test signal at DSD256:

The little spikes at the bottom in the above image represent low-level random noise that fluctuates during the measurement.

I don't believe the J-Test when converted to DSD is valid anymore as a tool to evaluate jitter. Unlike PCM where each sample represents an output level and slight "wobble" can result in frequency-domain errors (which is the basis of the J-Test as per Dunn), DSD is a type of PDM (Pulse Density Modulation) where levels are determined relatively over time with numerous samples. Over the years of testing DSD, I have yet to see significant sidebands after DSD conversion; I suspect high jitter in DSD will look more like subtle noise and amplitude level changes than structured sidebands as we see in the PCM J-Test.

I think this is good enough of an examination of DSD playback with the DO100. I must say that as an ESS DAC, I like what I'm seeing here. Unlike AKM DACs where there's a "DSD Direct" option, these ESS "HyperStream" DACs will reprocess the 1-bit data internally into multibit-SDM. Sonically, I don't really think this makes much of a difference and on a device like this which appears to be implemented very well, the results look very accurate.

II. Subjective Impressions

Size-wise, this is a relatively small DAC - smaller than the Raspberry Pi "Touch" streamer in width. Notice BTW that I'm playing the album as 48kHz. As an old 1986 analogue recording, the Top Gun OST was released in 1999 on SACD and the rip did not suggest anything more than 16/48 resolution was needed. 

Okay then, with the objective content out of the way, let's talk about how this device sounds subjectively with some music.

As per usual testing procedure, I listened to this DAC in my sound room for a couple of evenings for some initial impressions prior to sticking it on the measurement test bench.

It's Top Gun summer blockbuster time. As a child of the '80s, there is clearly no way the original Top Gun OST (1986, DR11) will be supplanted by any sequel soundtrack! The new Top Gun: Maverick OST (2022, DR8) has reprised a number of the tracks including Harold Faltermeyer's "Anthem" and Kenny Loggins' "Danger Zone". Much of the music consists of soundtrack instrumentals (with many tracks giving Lady Gaga scoring credits) and there are not as many songs and no great ballads here like Berlin's "Take My Breath Away", or even secondary pop hits like Loggins' "Playing With The Boys".

Lady Gaga's "Hold My Hand" is clearly the attempt here at creating a hit single and possibly targeting the Oscars Best Original Song category. I'll need to listen to this a few more times to determine if I really like this song. To me it doesn't have the "sweetness" of '80s pop, and there's an awkwardness here suggesting that Gaga's just trying a little too hard... OneRepublic's "I Ain't Worried" sounds alright, much like most generic pop I hear these days and just not memorable.

Regardless, the SMSL DO100 handled the live Miles Teller cover of "Great Balls of Fire" conveying good vitality, plenty of resolving ability to differentiate the numerous voices and incidental audience noises. Good width and depth to the soundstage. The classic Faltermeyer "Main Titles (You've Been Called Back to Top Gun)" theme sounds beautiful with variations throughout the score - evoking wonderful memories from decades back (I'm reminded of the first time I watched Top Gun on a Laser Disc and experienced Dolby Pro Logic, no AC3 back then, in a friend's home theater), revisited now in the 21st Century. 

Moving on from blockbuster soundtracks, I've been enjoying the social activism, protest rock of Midnight Oil's Resist (2022, DR5). Yeah, it's very much compressed sound but the material is good, including catchy melodies ("Lost at Sea"). Important lyrics to ponder upon in tracks like the urgency of "Rising Seas", the slower meditative "We Resist", and hopeful inspiration of "We Are Not Afraid". Since the first time I heard Diesel and Dust (1987, DR14), I've been impressed by the band's sincerity and passion. Although I have not followed the band over the years with every release, I know some critics have suggested that this new album is some of their strongest material in years.

No problem with full frequency extension through this DAC (one would not expect any issues given the frequency response!). Despite the restricted dynamic range of the album, the songs come through "loud and clear" for the most part. Although there's a significant amount of peak limiting applied, audible distortions were not obtrusive except very obviously on "Nobody's Child" where the cymbals and high-hats sounded harsh, unnatural. Also the louder, complex portions on tracks like "Reef" (around 4:00) is a reminder that this is obviously not "audiophile" sound quality.

With the death of longtime bassist/vocalist Bones Hillman in 2020, it looks like the current tour will be the last for Midnight Oil. Check 'em out if you can. (I regret that I missed their stop here in Vancouver on June 1st.)

Alright, how about some "audiophile approved" tracks?

There are a number of demo discs from audiophile "high end" brands and Burmester released a series of "Reference" discs a couple decades back. There's also this newer one shown above - Burmester Selection Volume 1 (2017, DR10). Of course as an audiophile demo disc, one cannot go wrong when Nils Lofgren's "Keith Don't Go" is one of the tracks in this compilation. ;-)

Track 1 Scarlatti's "Sonata in C Major, K.406" is a nice piano workout for your system. Excellent attack on the keys with realistic sound envelope. Nice sense of space with the piano sounding a bit further back behind the plane between the speakers. Good articulation and dynamic variation as well. This is followed by Al Di Meola's rendition of the Beatles' "And I Love Her" on guitar (amazing guitarist) which sounded awesome - the realism of the guitar was goosebumps-inducing. 

The techno-beat female-vocals of Norwegian Kari Bremnes' "Spor (2015 Version)" has a nice vibe and I had no idea what it was about until I saw this translation for "Birds". Tight bass, excellent clarity and presence of the vocals. Mckinley Black's "My Diamond Mine" is one of those excellent acoustic, broad soundstage recordings with singer front and center, guitar and bass spread out behind the singer. A typical well-recorded Stockfisch track.

While admittedly I don't listen to these audiophile demo albums usually, it's good to have a few of these in the streaming library when friends and family come by for a listen to what "hi-fidelity" sounds like. Needless to say, the S.M.S.L. DO100 reproduced the tracks wonderfully and shows its ability to just play back material like this in a "transparent" fashion.

I've been a fan of the Estonian composer Arvo Pärt's since I was turned on to the fantastic recording Tabula Rasa more than a decade back. During the COVID-19 pandemic, the Chor Des Bayerischen Rundfunks (Bavarian Radio Choir) went back to work to record many of the tracks on Pärt's Miserere (2021, DR12); a liturgical composition which was penned back in 1992. Pärt has been credited with bringing sacred/spiritual works back to the mainstream audience and this is another rewarding example. Much of this work is minimalistic ("Versus III"), other times dramatic and dynamic ("Dies irae" - Day of Wrath).

The recording itself sounds excellent through the DO100 on my system. Choral voices are smooth and clear with the ambiance of the venue (BR Studio 1 for many of the tracks) nicely captured and envelops the listener in the sweet spot. Beyond the standout multipart "Miserere", there are a few other choral works including "Tribute to Caesar" and "Which Was the Son of..."; the latter starting off this recording and seeds the contemplative atmosphere.

After a long day at work, sometimes it's nice to get into the "man cave", turn on some music like this and "float" into the spiritual ether...

For a taste of Pärt's music to those who are not already familiar with this composer, consider the compilation The Very Best of Arvo Pärt (2010, 2CDs, DR13). There's some great stuff here including excerpts from the aforementioned Tabula Rasa, parts of Magnificat Antiphons, and versions of Fratres.

Again, excellent sound quality from the DO100 with these acoustic recordings. Strings reproduced with natural timbre (lovely performance by the Estonian National Symphony Orchestra on "Summa"), superb separation of the multiple vocal parts on "Cantate Domino", and smooth, gentle overlay of voices and organ on "Beatus Petronius" (composed for 2 church organs and 2 choirs).

Of interest, contrast "Beatus Petronius" with the dynamic power of "Statuit ei Dominus" which is a piece also commissioned for the 600th anniversary of the Basilica San Petronio in Bologna; both can be found on Adam's Lament (2012, DR14).

Let's end off with one more album I had a listen to while evaluating the other night. How about a live concert soundtrack? I ripped George Michael's Live In London (2009, DR10) Blu-Ray awhile back with both the 24/48 2.0 PCM, and also decent 5.1 DTS-HDMA multichannel versions. While I would highly recommend multichannel for live recordings to better reproduce the energy and ambiance in the performance venue, the 2.0 version, played over the S.M.S.L. DO100 certainly does a great job reproducing Michael's interactions with the crowd and backing vocalists on "Faith". Good pace and bass on the electropop "Fastlove". The live version of "A Different Corner" is perhaps my favourite recording of George's amazing voice.

While the measurements show excellent DSD performance, I only spent maybe 30 minutes browsing through some SACD rips. It has been years since I listened to Arne Domnérus' Jazz At The Pawnshop, or Ennio Morricone's The Mission. As discussed last week in the comments, I don't think DSD64 is completely transparent but it does have a sweetness to the sound that I think works well with analogue recordings like these which were never really "high-res" to begin with, yet still compelling and as good as I've heard these albums played.

The SMSL DO100 is very very much "perceptible perfect" as far as I can tell. It just plays the audio data it's fed without editorializing.

While it's all well and good to hear descriptions about sound through the "ears" of others and look at graphs of the technical abilities, I'll also leave it to you to listen to an actual recording from this device:

S.M.S.L. DO100 (2 x ESS ES9038Q2M chips, XLR) AMPT

As per the AMPT protocol, the file above is the standardized recording created by connecting the SMSL DO100 to the RME ADI-2 Pro FS ADC. The 44.1kHz playback has been recorded as 24/96 hi-res, volume equalized. Note that the 6 seconds or so of silence at the beginning of the AMPT is intentional so you can have a listen to the level of the very low DAC noise floor. As usual, I am assuming that your playback system is also uncolored, and hi-res capable in order to reproduce a reasonable facsimile.

III. Final Conclusions

Nice sounding pairing: S.M.S.L. DO100 connected directly into Hypex nCore NC252MP amp. To my front Paradigm Signature Reference S8 v.3 speakers (not shown).

I am very impressed by this little SMSL DO100 DAC; simply excellent, near state-of-the-art objective and IMO sonically transparent performance at around US$250 which I believe will hold its own against basically anything out there regardless of price fidelity-wise.

Objectively, while I'm sure we will see even more resolving DACs in the future at this price point, I think we have to be very honest with ourselves in questioning just how much more resolution is needed. With 7 filters to choose from, essentially perfect time-domain performance with at worst microscopic jitter through all inputs (USB, S/PDIF coaxial, Toslink), true hi-res performance even down to THD+N below -120dB at 0dBFS and dynamic range of 125dB over the XLR output, uniform performance across all PCM sample rates, and excellent low-noise, low distortion DSD playback, I really have nothing to complain about. I suppose the only nit-pick I can propose is that they could improve the intersample overload characteristics of those digital filters.

In terms of "look and feel", well, it's a small black metal box which feels robust and has a useful remote control mainly to turn ON/OFF, select inputs and adjust volume. LED brightness setting is nice and should be standard these days for darker rooms. Also nice to have both RCA and XLR output (not balanced TRS like on Topping D10 Balanced or their E50 that might require an adaptor). In my sound room, while I will more likely use the remote, I do prefer the large control knob like this on the front over push button controls, looking a bit more elegant and cleaner. I did not run into any control issues or playback problems with S/PDIF when fed the digital out from my Squeezebox Touch, or a Raspberry Pi + JustBoom Digi HAT. As discussed in Part I, perhaps the DPLL setting could come in useful if you have very jittery S/PDIF sources like TVs or older CD transports.

Comparing the objective performance of this DAC with others, while we can say that the Topping D90SE with its ESS ES9038Pro DAC can provide even higher dynamic range and lower distortion (in PCM at least), we again need to question whether the improvements are necessary. I would strongly suggest that they are not audible.

Unless over the last few months Topping updated the D90SE firmware to address inconsistencies in distortion level (ie. I found higher distortions >96kHz playback) and improved the DSD noise level, this little DAC performed with none of those issues.

Arguably, with the release of the Topping D90LE (~US$800) which incorporates processing like "Tube Simulation" that increases significant amounts of harmonic distortion, we have reached the point of recognition even among hardware manufacturers that we are near the end of the pursuit for "high fidelity" alone in modern DACs. Perhaps it is also time to "scientifically" start giving the end-user control over their "euphonic distortion" preferences as well.

Maybe in the near future, the goal of digital audio is not just ubiquitous hi-res/hi-fi playback, but also to make tube preamps obsolete. Without purchasing any other piece of equipment, the DAC then might be able to satisfy the desires of both the "high-fidelity enthusiast" and the "euphonophilic" audiophile. ;-)

----------------------

There was an interesting discussion in the last blog post's comments section from KSTR about glitches that one might see in the 6-bit SDM of the Topping D10 Balanced (ESS ES9038Q2M) with noise modulation and the potential for "ripples" related to output levels.

Previously I had published the THD vs. Frequency graph using typically 1dB stepped level 1kHz sine:

Notice that the harmonics start to rise from about -25dBFS up as highlighted in the graph above. The relative amount of distortion looks chaotic at this resolution. The result above was from back in August 2021 before some significant upgrades more recently.

What if we zoom into those output levels and increase the resolution by an order of magnitude using the E1DA Cosmos APU + Cosmos ADC; taking 0.1dB steps?

Interesting! With this amount of resolution (both from the hardware of the APU notch + low-noise ADC, and software 0.1dB steps), we can now see that there is a structure to the distortion amounts with variable sinusoidal periodicity which I had not appreciated before.

I assume then this is the effect of the DAC chip's internal SDM architecture. Of course, I cannot exclude that to some extent, the Cosmos ADC's ES9822Pro has a role to play (I doubt it as I have already checked this elsewhere).

No need to be alarmed since the harmonic levels remain very low compared to the overall noise level and calculated THD+N (20Hz-20kHz bandwidth) is down below -110dB! This is simply a microscopic level of analysis afforded by my current measurement set-up and I would not claim that this is audible. I can apply this same method to AKM DACs and have a peek as well another time and likewise put the SMSL DO100 on the bench again and see if this same pattern can be replicated with another ES9038Q2M DAC. [Further explored here.]

Hope you're enjoying the music, dear audiophiles. I think we are well in the "golden age" of affordable ultra high-fidelity with digital audio conversion boxes. It's summer - time for some R&R in the near future...

Addendum - July 9, 2022:

I noticed the comment by bobfa on Audiophile Style about the popping sound when first turning on the SMSL DO100 connected to Raspberry Pi (maybe all Linux-based devices?).

Looks like this may have to do with the DO100 going into DSD512 mode when turned on and when you play something (likely PCM and almost certainly not DSD512!), that initiation/transition creates that noise.

S.M.S.L. - can you look into this? Maybe when starting up, putting it into PCM mode might not result in such an issue - might need a firmware update? I've noticed this using both Ropieee and Volumio on both the Raspberry Pi 4 and 3 B+.

For the time being, folks running into this issue can put the DAC into UAC1.1 mode. You'll be limited to 24/96 maximum resolution and no DSD; but no popping noise.