As we continue to evaluate the S.M.S.L. DO100 DAC, let's spend time today delving into the resolution that this machine provides as a PCM DAC (see Part I last week for filters and jitter performance). Given that PCM is the de facto digital modulation scheme, this post will form the heart of DAC performance evaluations. We'll leave DSD to Part III along with discussions of subjective performance.
With PCM data, this DAC will accept up to 32/768. Let's run some tests as per usual to see how this device compares to others I've looked at on this blog. For most of the tests, I will use my Raspberry Pi 4 "Touch" (USB) or a Logitech Squeezebox Touch (S/PDIF) as digital source. The E1DA Cosmos ADC will be used with or without the Cosmos APU to measure. For some of the REW-based tests, I will perform a simple loopback that looks like this:
Intel NUC → SMSL DO100 DAC → [Cosmos APU] → Cosmos ADC → Intel NUC
As usual, all cables are generic RCA when testing unbalanced, for XLR measurements 6' Monoprice Stage Rights were used.
Unless otherwise noted, the DO100 will be set-up to use "FL6" Linear Steep filter.
Without further ado, let's jump to it...
I. RightMark Comparisons
Although the RightMark Audio Analyzer battery has not been updated in awhile, I think as an initial test of devices, this software still provides a useful synopsis of performance plus allows me to build up a database of results. It's a convenient way to express numerical scores and graphic overlays so long as the tests are done using the same conditions.
As expressed many time, although modern DACs these days have tended to exceed standard CD resolution, let's make sure this DAC does the job well at 16/44.1:
|Note: Topping D10s / D10 Balanced with stock firmware when purchased, not linear phase steep digital filter versions which I'm using more recently.|
As you can see, we're looking at the RCA and XLR outputs of the SMSL DO100 to the left and comparing the results with a number of other DACs. We have DAC chip representatives here from ESS, AKM (the RME set at +19dBu 6.9Vrms), TI/Burr-Brown (the TEAC) and a custom R-2R in the form of the Cayin RU6.
Looking at the numbers and graphs, it's quite obvious that a 16-bit signal is no challenge at all for hi-res DACs. The only exception is the Cayin RU6 which as we've previously discussed is NOT a hi-res capable device (not even a full resolution 16/44.1 device).
Feel free to do some volume-controlled, preferably blinded listening trials on any of these DACs other than the Cayin. I think you'll quickly learn that it would be very difficult (if not impossible) to tell them apart at least with 16/44.1 material as suggested by the data!
Moving on to proper high-resolution testing, here's the summary table at 24-bits, 96kHz:
|Note that the RME ADI-2 Pro FS R Black Edition was set to +19dBu (6.9Vrms) output compared to ~4Vrms with the other XLR DACs. This could give it some dynamic range advantage. The RCA outputs are at around 2Vrms.|
As usual, we see that the RCA output is noisier and can pick up anomalies such as the 60Hz hum and what looks like a 3rd harmonic at 180Hz.
Within the limits of the E1DA Cosmos ADC used for the measurements here, notice that the Topping D90SE remains the "king" of measurements still. However, the SMSL DO100 is literally "spitting distance" away for low noise and distortion.
My TEAC UD-501 which is almost a decade old now provides an interesting historical comparison documenting the improvements in USB DAC technology in the last decade. While certainly things have objectively improved, it's important to look at these results and consider the implications within the limits of human hearing. Beyond features that might be exciting and even "sexy" with newer products, older generation and vintage gear can and do sound fantastic given the maturity of audio technology as a whole over the last few decades.
I don't have much 192kHz music content in my library so it's hard to say what value measurements at 24/192 have. Regardless, let's have a quick peek to make sure there are no issues:
Again, we're looking at a level of performance very close to the Topping D90SE (at a fraction of the price of course). If we take away potentially some subjective and objective hype, consider seriously whether the data suggests that the Topping D10 Balanced is all that different from the D90SE. How truly "perfect" do we think a DAC needs to be?
More often than not, the data-driven, logical answer will give us a sense of the truth when it comes to actual differences in the physical world. The difference between the physical objective reality that we can measure and the subjective psychological impressions is the playground of advertisers and marketing people. And within that territory also the poles between utilitarian value and luxury.
If 192kHJz sample rate is not enough, let's go even higher to 384kHz (this is the limit of the E1DA Cosmos ADC):
For completeness, let's run a few 24/96 measurements over the S/PDIF inputs fed by my old Logitech Squeezebox Touch. I believe some of you out there might still prefer these inputs (I agree, there are reasons to do so including compatibility):
As you can see, despite different inputs - USB versus S/PDIF Coaxial and TosLink - done a few days apart and with different devices (the USB input was through a Raspberry Pi 4 "Touch" and the S/PDIF measurements from Logitech Touch), the results look very boring and similar. ;-)
I know, some people insist that they can hear "clear" differences between USB and S/PDIF, between Coaxial and TosLink, or even different digital cables. Other than slight timing variation which I suspect is the cause of that difference between USB and the S/PDIF frequency response (most of it >20kHz), there's simply no difference. Yeah, TosLink will provide galvanic isolation if for some reason you have a noisy system or run into ground loops.
Time and again with DAC measurements and listening to various digital inputs, I have failed to notice any difference once one applies a modicum of "control" to the listening test. It does not take a DBT to realize that differences are at best subtle and not worth getting excited about. I suppose it's possible that some DACs may have audible differences between digital inputs. However, I suspect that if this is indeed the case, then there must be something very significant with the way the DAC behaves when processing the digital data from the different inputs; this is probably not a good thing!
II. High Resolution Dynamic Range and THD+N
With the E1DA Cosmos APU + ADC, let's have a closer look at the dynamic range:
|These are the "raw" FFTs from a -60dB 1kHz test signal. Add 60dB back to the results for 0dBFS values.|
Looking deep into the noise floor we can see the superior noise-rejection ability of the balanced output. Nonetheless, the RCA output is quite impressive, still achieving an overall 123dB of dynamic range compared to over 125dB with the XLR.
This should be more than adequate for human hearing at almost 21-bits of dynamic range! ;-)
As for THD+N, let's send some 32-bit, 96kHz highest resolution 1kHz data to the DAC for a peek:
Very nice, we're looking at THD+N of better than -119dB RCA output, and around -122dB XLR (SINAD 122dB). Very impressive for a DAC around this price range!
Previously, I had documented that there could be variation in the THD depending on the sampling rate used (for example, the Topping D90SE had higher harmonic distortion >96kHz). Using the Raspberry Pi 4 "Touch", let's measure some 24-bit 1kHz 0dBFS signals at various samplerates:
|At this level of resolution, even the difference between 24-bit playback and the 32-bit generator in REW (the above THD+N graphs) can be measurable.|
Nice, around THD+N -120dB across the board regardless of sample rate. This is more consistent than the Topping D90SE last year suggesting that S.M.S.L. might have paid more attention to the THD compensation parameters across sample rates.
I saw this comment about THD on the SMSL M500 having high 3rd order harmonics when the L and R signals are asymmetrical. Let's check with the DO100 - here's L channel playing a 0dBFS 1kHz tone and R channel silent:
III. Stepped Frequency & Amplitude Tests
Of course, looking at 1kHz tones is limited so we should check out other frequencies and amplitudes just in case there are issues.
To start, let's look at a frequency step at a few output levels:
Other than at 0dBFS where we see a bit of frequency-dependent variation among the relative harmonic levels, the harmonic "structure" generally remain predictable across the audible spectrum typically with the 3rd order predominating. As output level decreases, harmonic distortions decrease and eventually get buried in the noise floor if we were to measure below -12dBFS.
Looking further down into amplitude levels with a 1kHz tone, we can examine the harmonics, calculate THD, and consider linearity; let's go down to -130dBFS using REW's level step:
We see that the harmonics rise above noise floor between -50 to -60dBFS. For the most part, odd order harmonics predominate - the 3rd, followed by the 5th - depending on the output level. The harmonic distortion should be of little to no consequence since THD barely rises above -90dB (0.003%) at any of the output levels above -50dBFS.
Linearity is excellent with 0.5dB deviation below -125dBFS.
With the recent update to REW, let's look at the DIN IMD (250Hz/8kHz, 4:1) & TD+N vs. output level step graphs:
Good symmetry between the two channels. Like the THD graphs above, we see that IMD rises from the noise floor as the level goes up, in this case starting about -45 to -50dBFS. The THD "bump" never rises above -75dB (0.018%), highest down around -40dBFS output level. Again, nothing to get audibly concerned about I think.
IV. Multitone Tests
Let's get a reading for the Triple-Tone Test which I've used typically in my amplifier measurements. This consists of tones at 48Hz, 960Hz, and 5472Hz (1:1:1) spread across the audible frequency, each one adding their own harmonic distortions and interacting to give intermodulation:
With RCA TD+N around -112dB and XLR at around -113dB this is pretty good. Note that the king of this test remains the Topping D90SE with a TD+N around -120dB. I would already be very happy to see an amp achieve -100dB on this test; just goes to show the very high relative fidelity of DACs compared to other components in the audio chain.
Let's have a peek at the "1/10 Decade Multitone 32". Simultaneous dual channel 512k-FFT at 96kHz captured:
|Both left and right channels overlaid.|
Other than that 180Hz tone picked up in the RCA output which may be idiosyncratic to my test set-up, results are very similar to the Topping D90SE with around -130dB distortion-free range in the XLR output.
V. Conclusions, Part II
Simply put, these are highly impressive results from the S.M.S.L. DO100 DAC at a street price of less than US$250 currently.
High resolution playback devices are quite common these days thanks to the forward march of technology over the years. What I'm seeing (and hearing) I believe is easily comparable to the best DACs on the market today regardless of price. To achieve THD+N of better than -120dB is in itself impressive and the rest of the testing including RightMark battery, IMD, level/frequency step graphs, are all in line with expectations. The fact that THD+N remained excellent across sample rates measured up to 352.8kHz is an improvement over the degradation I saw previously with the Topping D90SE (which I hope has been resolved with the latest firmware).
Of course, I'm defining "best DAC" here based on the concept of high fidelity (which can be objectively defined) rather than a subjective preference which may or may not lean towards a certain "taste" in sound. I bet that there will be folks using words like "clinical", or "cold", or "hi-fi sounding" or even "unengaging" when they listen to very high fidelity products like this. At times, I suspect there could be a conscious desire to downplay a device because of the price point with an intent to elevate megabuck gear (advertising-driven reviewers tend to do this all the time). Yeah, I think that's just how the game is played and everybody has the right to a subjective impression/opinion. As usual, once the blind test starts, more likely than not, that's when the truth is revealed about the nature of opinions especially with DACs of very high performance!
[Have a look at the blind test using recordings of 16/44.1 devices done a few years ago. How much more difficult to differentiate hi-res DACs these days!?]
I trust that the array of tests done already is quite complete. While we can dream of other kinds of test signals, the fact that this DAC has performed so well across a number of tests suggests that there's low likelihood of having missed some significant sonic issue. We are not quite done with this DAC yet though! In Part III, let's have a look at DSD playback, spend time around subjective impressions, and I'll make sure to get you an AMPT recording for listening comparisons.
While I really hate to keep talking about MQA, as a comment about contemporary issues, let's just be mindful that companies have been co-opting environmental concerns in their advertising. Sometimes this could be appropriate if we are looking at genuinely better and more environmentally friendly advances; however in the case of MQA, I think basically everything they've been doing to promote the company have been questionable. I believe this is simply the result of their foundation being flawed to begin with!
Notice their focus on "environmental impact" in the last year. They also have pages on this on their main website. Recently as discussed on Audiophile Style we have this graphic which was released over Instagram:
Here's my comment on the site in response to that graphic to put some numbers into this claim:
Co-opting environmental concerns like this is truly despicable by the company.
Let's look at the numbers here.
365days x 6hrs/d = 2,190 hours
24/192 2.0 stream (9,216kbps) compressed FLAC assuming conservative 30% compression = 1.125MB/s x 0.7 x 60 = 47.25MB/minute = 2,835MB/h
Therefore per year = 2,835MB/h x 2190h = 6,208,650MB = 6,209GB = 6.2TB/year streamed
Compare this to MQA - 24/48 2.0 stream (2,304kbps) which is 1/4 - and let's assume again 30% FLAC lossless compressed efficiency like above, so we're looking at 1.55TB/year streamed.
A conservative difference of less than 4.7TB streamed according to them = 5 return flights London/Berlin = 9 car journeys London/Manchester = 19 trees planted?
Can someone confirm that this is what the source "Carbon calculator" actually says is the environmental impact of sending 4.7TB across the Internet over 365 days?
Obviously even this calculation is over-estimating things...
1. Who listens to just 24/192 music!? ;-) Vast majority 16/44.1 out there anyways.
2. 30% FLAC compression too low for most good classical/jazz recordings, probably closer to 40-50% in real life. Also, MQA processing actually makes the efficiency of FLAC go down so we might still be looking at 30% compression with MQA, but 45% with the original 24/192.
And obviously switching to MQA has costs - environmental and otherwise:
1. The streaming service will need to pre-process all the native hi-res data to MQA, embed the cryptographic signature, and presumably there's implementation of the Ultimaco hardware infrastructure. All this needs energy and manufacturing. Who knows if each streaming service will need to duplicate this for their set of music files.
2. The consumers are encouraged to ditch otherwise perfectly good DACs for MQA-capable ones - increased trash? more manufacturing carbon footprint, etc.
3. The extra processing to decode MQA whether in extra cycles each time played back in software on your computer or implemented in the DAC hardware itself accumulates in unnecessary excess energy.
All this so we can listen to lossy "hi-res" audio? MQA truly is getting very desperate...
Chris Connaker also makes an important point that since the network infrastructure is always on 24/7, what actual excess energy is spent to send those 4.7TBs over a year? I bet it's negligible even on a mass scale.
Audiophile-related or otherwise, I think it's important for companies to not act like the way MQA has over the last 8 years or so! Sure, at an intellectual level, the lossy encoding/compression technique where they use their "origami" to stuff ultrasonic content into the lower bits in a 24-bit container is interesting, and embedding some cryptography into the stream might be seen as useful by some. But that's all this technology really is!
Environmental concerns are important, but this is a sickeningly obvious example of "greenwashing" in the audio streaming space. If they honestly cared about streaming audio's carbon footprint, they should lead by example and immediately pull TIDAL's license to use MQA in the HiFi+ tier, prepare to close their business, send Mike Jbara and their top PR people to lobby for lossy AAC/MP3/OGG streaming as the only solutions we need for good enough audio quality.
While they're at it, give the money back to investors instead of continuing to accrue losses so maybe the funds can truly be utilized with entrepreneurs who can do something environmentally beneficial.
I like GoldenSound's suggestion to lodge a complaint against this "award-winning" British company:
I genuinely hope there's nothing else to say about MQA on this blog other than maybe a celebratory note when this madness ends. I would recommend anyone who subscribes to TIDAL HiFi+ to switch over to just the HiFi stream and save yourself 50% while experiencing I believe negligible sound quality difference (might even sound better with genuine 16/44.1 lossless albums). This will also be good for the environment since your streaming bitrate will immediately drop by 30+% so over time you too can feel good about driving to Manchester or flying to Berlin from London return.
All the best audiophiles... Hope you're enjoying the music!