Saturday 14 August 2021

MEASUREMENTS / REVIEW: Topping D10 Balanced (D10B) DAC. Simple, no-nonsense USB DAC, excellent balanced sound quality.

Last month, I published on the Topping D10s DAC which performed excellently for such a small, USB-powered device. Fantastic price to boot!

Today, let's have a look at the newest sibling in this line of DACs released in July 2021 - the Topping D10 Balanced (I'll just call it D10B for short). As you can see in the image above, the contents in the box are similar to the Topping D10s, with manual (including measurements), pamphlet with various other Topping products listed, generic USB A-B cable, and for this model a couple of TSR-to-XLR male-male adaptors.

At <US$150, this is certainly still inexpensive. And as the name suggests, this baby is capable of balanced analogue output. Balanced transmission provides improved common-mode noise rejection and this should result in quieter analogue output including rejection of potential issues like mains hum in the signal.

I bought this DAC through the usual retail channels; no relationship with the company.

As you can see, I picked the silver model this time. Here in North America, most of our electronics tends to be black in color although companies like Schiit feature quite a bit of silver. I heard that over in Asia and Europe, silver and gold are popular colors as well for hi-fi gear. Notice that the front fascia is less marked up than the D10s; no text about DSD, 384kHz, etc. When turned on, there's a tasteful amber OLED display that's not too bright showing sample rate and whether PCM/DSD being processed.

The build is good. Weight is 335g on my scale without adaptors, about the same as its single-ended siblings. I noticed the enclosure is different from the D10s in that it's a single body piece rather than top-bottom halves. Inside, the DAC chip is the same ESS Tech ES9038Q2M low-power unit as the D10s. I'm not sure if there's still a socketed opamp for folks who care to play with opamp-rolling. The main difference you'll notice is on the back:

Instead of the single-ended RCA output, we now have dual TRS balanced ports. As you can see I'm showing the included TRS-XLR adaptors. Due to the size of these adaptors, it can look a bit funny with both attached:

Looks a bit like the souped-up car in this post from a few years back on audiophile mods. ;-)

Make sure to provide some support (like put something on top) if you have heavy XLR cables so the whole DAC unit doesn't get tipped up. As per the other D10-family DACs, this unit is powered with the single USB input. I measured current usage at <250mA so this should be easily powered by most standard ports (mobile device USB ports probably not enough current). The device remained cool after hours of use. Also like the other D10 devices, there's only S/PDIF coaxial and TosLink outputs

While not recommended by Topping, and in fact they have a warning:

Here at the Musings, we're not afraid to try :-):

Topping D10B with typical TS-RCA adaptor - NOT recommended.

Pictured with some standard mono 1/4" TS-to-RCA adaptors. Obviously, if you're just going to be using single-ended output, don't waste money on the D10B, just get the D10s. And heed Topping's warning above. For now, let's just say that I have not destroyed the DAC by doing this, but I agree with Topping that this is definitely NOT recommended.

Okay then folks, as usual, I listened to the DAC for a few evenings and wrote down some listening impressions before test bench measurements. As a product review, IMO, it's better to get down to the objective measurements first...

A. Objective Performance

Here's the D10B sitting on my test bench. Top left, you can see the Raspberry Pi 3 B+ "Touch" I used as playback device for most of these tests. Since the TRS-XLR adaptors were included, I just used those, however, feel free to get something like the yellow TRS-to-XLR cable to connect the DAC to your pre-amp. This will at least look better than the big dual-exhaust-XLR adaptors sticking out from behind!

As usual, the measurements here will primarily be performed with the RME ADI-2 Pro FS as the ADC used to capture test signals and my Intel NUC computer for data processing. XLR cables will be generic 6' lengths (similar to the Monoprice Stage Right described before).

I. Oscilloscope Tracings and Filter Performance

Since the oscilloscope doesn't accept balanced input, with the unbalanced TS-RCA adaptor, we can see that this DAC provides clean-looking 1kHz sine and bandlimited square wave tracings with excellent stereo channel overlay. We see that the sine wave is around 2.13V. This means that the balanced output is twice this and I measured it at 4.29Vrms (+14.9dBu).

Looking at the bandlimited square wave "ringing" pattern, we can surmise that this device uses a minimum phase filter. Here's the impulse response with a 16/44.1 signal:

Yup, minimum phase setting, and given the length of the ringing we should see a relatively steep filter. Polarity is maintained like the D10s (and unlike the original D10, not that we'd generally hear this beyond test signals). Let's have a look at this using my Digital Filter Composite graph (DFC, as per discussions with Juergen Reis) in the frequency domain:

Interesting how Topping has changed the digital filter setting with each of the D10-family of devices! The original D10 used a linear filter in many ways similar to this one but without the noticeable overloading using the 0dBFS white noise signal. The D10s used a steep linear phase "apodizing"-type filter with attenuation by 20kHz, unfortunately it showed some mild audio band rippling. And now, we see a minimum phase filter with this D10B which could be one of the standard settings on the DAC chip.

While ideally I would have preferred a non-overloading, linear phase filter as a default setting, the way it is currently is subjectively fine.

II. RightMark comparative DAC performance

Okay then, let's line up and compare performances of a few DACs, including each of the Topping D10 family:

Notice that for comparison, I've included the Oppo UDP-205 which uses the higher spec ESS ES9038Pro DAC chip with XLR output, and also the Topping DX3 Pro which while sporting single-ended RCA out, uses dual AKM AK4493 chips.

Some graphs with more details:

As you know, 16/44.1 isn't a particularly challenging performance level these days so most hi-res DACs tend to perform well on these tests.

We see some variations in the frequency response between the DACs but notice the zoomed-in dB scale and just how tiny the differences really are. Furthermore, most of the difference is >10kHz which will be even more difficult to hear.

Let's now step up to hi-res...

Folks, this is literally a parade of very high resolution devices sporting huge dynamic ranges beyond what's really needed when paired with even some of the best amplifiers and speakers. Fantastic low crosstalk, and tiny distortion figures!

For this set of comparisons I've added both the Oppo UDP-205 and RME ADI-2 Pro FS R Black Edition (+24dBu output level, maximum resolution) into the mix; both with balanced outputs. Topping DX3 Pro has RCA single-ended output.

Again, we see the variation in frequency response zoomed in. Balanced DACs are less prone to 60Hz mains hum and typically achieve lower noise levels. Slight variation in crosstalk but nothing really to get excited about. Higher IMD+N with the original Topping D10.

These days, I don't particularly care for 192kHz music streaming/downloads, but I will use this samplerate for upsampling, so let's make sure the DAC performs well.

Everything numerically looks great among the field of high performance DACs and due to the limitation of the RightMark software, unfortunately I can't plot the crosstalk and IMD+N as I usually do at 24/96 to compare. However, I can show that the two channels of the D10B are well matched in performance:

For completeness, some DSD test results using PCM 24/96 --> DSD64/128 conversions of the test signal using JRiver. As usual, the conversion system is lossy so DSD results will not be as good as PCM but should show good hi-res performance in the audible frequencies.

Looking great. As usual, there's quite a bit of ultrasonic noise especially with the DSD64 playback. Otherwise, below 20kHz, the noise shaping used to handle the high 1-bit quantization noise does a good job.


Let's have a look at the detailed 1kHz 0dBFS THD+N FFT. For the record, this is done at 24/96 samplerate for both playback and ADC capture, 128k-points FFT using Blackman-Harris 7 windowing, 4 samples averaged, log scale - I'll standardize on these settings.

Nice. Basically, this DAC is hitting the limits of my RME ADI-2 Pro FS ADC-based test system around THD+N of -113dB. Audio Science Review's measurements using the Audio Precision APx555 show that the DAC can go down to -118dB which is fantastic!

Peeping at the results, we can see that with my unit here, the right channel overall has less noise (the N component) but slightly more harmonic distortion with the 3rd harmonic around -115dBFS. The left channel harmonics are all around or below -120dBFS - beautiful!

For context, as a test of "real world" audibility in a blind test, I was not able to show that listeners were able to identify a preference for THD+N below -75dB with real music in home speaker and headphone systems (as per the informal THD+N test last year). Personally, any DAC with THD+N better than -100dB is already likely to be many times "cleaner" than needed for real-world audiophile listening!

For completeness, as pictured above, I tried the TS-RCA to convert the balanced output to a single-ended signal. While the oscilloscope tracings above look fine, have a look at what happened using the high resolution THD+N graph:

Clearly, there is a significant increase in distortion if we use a simple converter adaptor that ties the sleeve and ring connections. Notice that I showed the -3dBFS FFT just to make sure the extra distortions are not due to clipping (basically no difference between 0dBFS and -3dBFS). Realize though that a THD+N of around -85dB actually still sounds good even if far from optimal for this DAC. Consider that this is still a better THD+N than the SMSL SA300 desktop amplifier I use. Again, a reminder of just how much resolution DACs have these days compared to the rest of the audio reproduction chain!

Yes. Follow the Topping warning and avoid simple balanced-to-single-ended converters like this.

For testing, I ran the Topping D10B with a single-ended adaptor for 3 days or so continuously on my desktop system as above. Not great performance compared to a D10s or even D10, but no damage done at least from short-term use like this.

IV. Jitter

Looking beautiful! Nice, clean squiggles with no significant sidebands typically indicative of data jitter. Maybe a pair of tiny sidebands below -150dB. "Skirt" at the base of the 11kHz and 12kHz primary frequencies small.

I noticed in the ASR measurements, Amir mention that the 24th-bit jitter modulation signal can be seen on his J-Test graph. I had a look using my measurement gear as well:

Since the low-level signal is a 250Hz square wave, we can see the 250Hz spike as well as it's odd 3rd harmonic at 750Hz, there might even be a small 5th harmonic at 1250Hz visible with the eye of faith. Cool, the noise level is low enough from the DAC and measurement ADC to show this.

(Needless to say, this 24-bit jitter modulation pulse would be impossible to hear with human ears and speaks to the power of what can be demonstrated objectively as an audio reviewer without needing to spend a lot of money on measurement gear! Although admittedly AP gear is cool, as a pragmatic guy, unless I'm sure it's going to show me something audibly significant I can't already demonstrate with a bit of creativity if necessary, or know for sure that I can get value on the investment with good returns, I simply don't see the need to use something like that.)

V. Frequency and Amplitude Sweeps / Stepped Measurements

Finally, let's just make sure we don't see any irregularities across the frequencies and amplitude levels.

Very nice. Basically no variation of concern across the frequencies from 0dBFS to -12dBFS. At lower amplitudes, distortion will get further buried under the noise floor.

And if we look at distortion as a function of generator level for a 1kHz sine signal:

I thought it was interesting to show the variation between the two channels. At this level of detail, given that DAC chips themselves will differ slightly in performance, we can see a kind of "fingerprint" that will be unique with each device. As output level goes above +10dBu (~2.5Vrms), the third harmonic tends to rise and predominate.

Finally, with the generator level data, we can have a peek at output level linearity:

Simply beautiful. With the -2.19dB as the 0 point on the graph, even at -120dBFS, the error is <0.4dB and much of this is likely due to a limitation of the RME ADC rather than the Topping DAC. No point showing the right channel because it's just as "perfect" (actually, it even looks a little better!).

B. Subjective Performance

For size comparison, the image above gives you an idea of how small the silver Topping D10 Balanced is compared to the RME ADI-2 Pro FS R Black Edition, and the TEAC UD-501 below that.

I spent most of a week listening to my system with the D10 Balanced DAC to see if I can come up with some impressions of the sound that this little wonder "makes". As usual, I made sure to listen to a varied list of songs across genres. For example, on the screen above, we see the cover for A.R. Rahman's 2009 soundtrack to Slumdog Millionaire. I loved the flick when it first came out and it was a real treat over the summer holidays enjoying it again this time with the kids old enough to appreciate it as well! The music is a great complement to the vibrant visuals of the movie. Catchy pop ("Paper Planes"), electronic dance, urban rap ("Gansta Blues"), merging beautifully with sentimental instrumentals ("Latika's Theme"), ethnic-inspired instrumentation ("Mausam & Escape"), and the Bollywood ending dance theme ("Jai Ho"). The Topping D10B didn't miss a beat and rendered the album with fantastic dynamics, deep bass, and precise sense of pacing (BTW folks, let's leave the audiophile term "PRaT" back in the '70s where it belongs - days when turntables were the predominant playback devices with all their timing anomalies!). I love the delicacy of "Mausam & Escape" and its beautiful sitar work by Asad Khan intermingled with the more intense synths. With the excellent low noise floor, this DAC has no trouble rendering every nook and cranny of the complex multi-tracked audio layers.

Sometimes audiophiles talk about gear rendering spatial depth in the music. For example, I was quite impressed by both the sense of depth and width in the soundstage of Bryan Adams' Unplugged (1997). On a track like "Summer of '69", there's a nice layering with Adams' voice recessed, flanked by guitars on the sides sounding a bit closer to the listener, with background singers to the side and behind Adams, and the cheers of the crowd "surrounding" the listener to the far corners and behind. In comparison, Eric Clapton's MTV Unplugged, while one of my favourite live recordings has great side-to-side soundstaging but less depth (eg. "Tears In Heaven"). The precision of each guitar strum, pluck, backing vocals, crowd noises again sound fantastic through the D10B along with the expected sense of realism and "liveness" one would expect from acoustic performances such as these. With live recordings we can judge whether the timbral accuracy sounds "right". Some of the recorded guitar sounds can be a bit harsh when close-mic'ed and I consider this simply a reflection of what's on the recording itself, not an indictment against the hardware (and in fact, I would not want it any other way by coloring or "taming" what's embedded in the audio data). I have not noticed any excess sibilance in female vocals using this DAC through my set-ups.

Local Vancouver-based musician Yu Su's album Yellow River Blue (2021) was a treat to listen to the other night. This is electronic music of the "chillout" variety. Asian-inspired instrumentation with some old-skool synth sounds reminiscent of the '80s interspersed. A good variation with each track which is good. I often find music like this to be too repetitive, not so with this album. As expected, no problem at all with the amount of detail from the D10B with excellent lows and synthetic surround effects (check out "Gleam").

For the audiophiles in a sentimental mood or on a rainy Sunday afternoon, have a listen to Richard Hawley's Coles Corner (2005). I'm sure you'll be reminded of the sounds and lyrics of decades past. Many of the songs evoke melancholia (like the title track). It's a well produced album and while with an average DR8, this album is not going to require the DAC to perform any extreme gymnastics, a DAC like the D10B has no trouble with the detailed, warm male vocals, accompanied by various instrumental layers. "Darlin' Wait For Me" sounds like something that would be completely at home in one of the late great Roy Orbison's albums.

Billie Eilish released her sophomore album Happier Than Ever (2021) just recently and I spent an evening digging into this release; let's spend a little more time talking about this one.

Not bad at all! Good to see her progression as an artist with songs touching on relationships, media expectations, privilege, body image, etc. Sixteen tracks, ranging from slower ballads like "Getting Older", "My Future", and "Halley's Comet" compared to the heavier beats of "I Didn't Change My Number", "Billie Bossa Nova", "Everybody Dies", and the rather intense "Oxytocin"; listen for any excessive bass bloat in your listening room on these latter tracks. On a well set up system, you should be able to appreciate the nicely done "surround" effect on "Oxytocin".

While soundstage is synthetic, the hall-reverb "air" at the start of "Goldwing" is quite nice. Eilish's voice is featured front-and-center on most of the tracks with good separation of the vocals from the rest of the music, creating an impression of the voice floating in front of the listener even during congested passages and with heavy bass.

While the average dynamic range is far from "natural" (DR5), I would like to present the award for "Best Use of Clipping" as an artistic choice to the title track "Happier Than Ever". Check this out:

Yes, from 3:00 onwards, it sounds just like it looks - harsh, distorted, severely dynamically compressed. But it works! This is because that second half of the song intentionally uses the dysphonic sound quality as a counterbalance to the sweet, dreamy, sentimental first half as she sings emotionally dissonant lyrics. That latter half effectively creates a cathartic emotional release that finally synchronizes the lyrics (distressed), music (harsh), and emotional tone (angry) as it ends. Intentionally not "beautiful" sounding, but artistic and well done!

See, while rare, Archimago is OK with dynamic range compression sometimes if it fits the artistic intent. ;-) Production folks and mastering engineers who simply turn the volume up to "11" across the board for the sake of being "loud" are simply ruining the sound.

One more thing, while listening to "Halley's Comet", notice the subtleties in that first 30 seconds - kids making noise in the distance just over the relatively high noise floor. These are like the "microdynamic" details audiophiles often talk about and can be easily missed if ambient noise is high or the system lacks resolution. You should easily hear the low-level "contrast" between the kids and the rest of the sounds with good clarity and separation of space in your soundroom (and very easy to hear with good headphones).

Although by no means is this Billie Eilish recording "high-resolution", I hope this is a useful "audiophile" sound analysis of what I'm sure will be a popular album if you're over at a friend's place and want to get a comparative sense of the system quality.

Wonderful evening listening sessions with the Topping D10 Balanced. Seriously guys and gals, this is one honest sounding high-fidelity DAC which reproduces the music beautifully if you have a balanced sound system to take advantage of this level of resolution.

C. Summary

No need to beat around the bush, the objective and subjective qualities of this little DAC makes the Topping D10 Balanced a fantastic deal if you're in the market for a simple no-frills balanced device with S/PDIF coaxial and TosLink outputs. Apart from USB2.0, no other inputs, no wireless, no control mechanism to change filter settings. At <US$150 (due to stock and price fluctuations, double check on Aoshida), I really think this level of performance is simply going to be very very hard to beat, highlighting the level of resolution the little ES9038Q2M can accomplish! Heck, this line of D10 devices are inexpensive enough to give as "stocking-stuffers" for friends and family who might be curious to hear what "high fidelity" sounds like in their systems assuming they use USB out.

More likely than not, even at this price point, the quality of this little balanced DAC I bet surpasses the performance of pretty well any other piece of gear (preamp, amp, definitely speakers/headphones) one might have in the system when it comes to low noise, low distortion, high dynamic range, temporal precision, etc. The sound will also easily surpass the fidelity of other much more expensive DACs out there.

The only technical criticism I would make would be wishing that Topping kept the digital filter linear phase as a default. This is a small thing and software upsampling during playback is easy to do these days which is pretty well the only purpose for the 352.8/384kHz samplerate this DAC supports (like using Roon's settings). Between the two, I prefer the D10B's filter setting than the D10s and its subtle frequency response rippling so if Topping is to issue a change to the filter setting (firmware update possible?), I'd encourage a change to the D10s first.

If I were to use this DAC as the primary playback device in my sound room, I would definitely make sure to buy some 1/4" TRS-to-XLR male-to-male cables instead of using the nozzle-shaped converters they come with for esthetic and weight balance reasons!

I thinking I'll probably end up using the combination of the Topping D10s and the D10 Balanced as signal generators in my PC-based testing system if I'm not using the RME's own DAC. For single-ended operation, the D10s will provide clean output below -110dB THD+N at +8dBu, while the D10B can achieve -113dB THD+N with its +15dBu balanced output. Not bad for a home hobbyist measurement rig. When measuring amplifiers or preamplifiers, this level of resolution will be plenty. In fact, I would be impressed with any reasonably priced amplifier capable of achieving -100dB THD+N already. I honestly cannot imagine "needing" anything more for enjoyment of sound until the day I build my own anechoic chamber ;-). And I better do this soon before I hit the next decade of life since ears are not like fine wine and do not get better with age!

I said in the D10s review that I believed it's already a "perceptibly perfect" DAC. With the benefits of a demonstrably lower noise floor, the D10 Balanced is even more perceptibly perfect. I personally would have no problem enjoying an inexpensive, low-power DAC like this in my sound room with zero regrets that I'm somehow "missing out" on some gnostic "magical" sound quality.

I hope you're enjoying the music and the toys, dear audiophiles.


  1. Hi Archi

    Nice report. Great. In the last 2 years, I am impressed with every DAC or Headphone Amp or Preamp, Topping is able to bring to the market, and especially for the price. No gimmick, just clear design language concerning the technical realization. Highly respective.


    1. Yeah indeed Juergen,
      The value for the money and just overall performance is impressive for sure! As a company, business people always talk about "price, quality service" - certainly price and quality great. I've not had to deal with service but I am glad to see that the software drivers and updates look good. Hope consumers are experiencing reasonable service if there are warranty issues

      BTW, for this box the warranty says "One month guarantee replacement. One year free maintenance."

      Hope you're having a great summer.

  2. Yu Su's album is US$ 13 on band camp and is only 40 minutes .... no way!!

    1. Hey El Oscense,
      Yeah, I agree that's a bit excessive on Bandcamp. I was listening to it on Apple Music, also available on Amazon Music; presumably other places as well.

      I think it's interesting to think about the low cost of streaming these days and how this will feed back on what prices artists and labels can charge when doing downloads. Likewise, I think this will have impact on potential units sold in the physical formats; although "collector item" type packaging may dissociate in price since the buyer isn't primarily paying for the music itself!

  3. Hi Archimago,

    Regarding the DSD tests, for example in your DX3 Pro V2 review:
    There is almost no difference in PCM and DSD noise floor, right?
    Then on the UDP-205:
    Everything with DSD is significantly noisier with something like 105dB DNR, and I was able to get 111dB DNR on a 16 years old Sound Blaster using the old and non-TOTL CS4382, recording by another sound card with the TI PCM4220 ADC:

    I suppose you were using the same DSD encoders with same settings in these tests right? The only difference is the DX3 Pro V2 has an AKM chip and the others are ESS. What do you think?

    mansr has compiled a special version of SoX with DSD support:

    For example, when using the sdm-8 modulator, <20kHz noise floor should be low enough to not bottleneck anything, you may give it a try.

    1. Thanks Bennet for the very thorough post and the interesting observation!

      Yeah, there are definitely some nuances to be mindful of in the DSD conversion and test results. Most of the time over the years, I have "standardized" on using JRiver 24 PCM --> DSD which is a number of years back now and might not represent the same quality as today's JRiver 28 version.

      To be honest, I have probably not spent enough time on thinking about the DSD results as much as I should. Most of the time, I just look at the numbers and graphs, make sure I'm at least seeing "hi-res" results, and move on.

      Maybe there's something there with ESS vs. AKM (noting that AKM has the "DSD Direct" mode). However if we look at the RME ADI-2 Pro FS R BE results based on AKM:

      Notice the DSD noise floor is similar to what I'm seeing with this Topping D10B and not as good as the DX3 Pro even though the RME has a menu item to do "DSD Direct".

      Hmmm, I wonder if it could be the dual AKM chips in the DX3 Pro that might be giving an advantage especially seen in these DSD tests? Very interesting if so!

      I've actually been wanting to re-standardize the DSD portion of testing for a bit. Part of why I haven't I think is because the DSD market looks pretty dead with little demand. It has been quite awhile since I've bought a new SACD in fact.

      Since you mentioned Mansr's SoX with DSD, maybe I should go with the SoX conversion as the new standard which would be good since it'll be free of commercial interests! Any recommendations? Do you think it'd be best if I went with the sdm-8 modulator?

    2. Thanks Archimago,

      If you read the link I previously posted, mansr answered several questions about the settings. In digital domain I cannot find any anomaly in encoding and decoding the standard RMAA sequences with sdm-8, and it is a known issue that RMAA's stereo crosstalk signal is clipped unless "32 bit (float)" is selected in RMAA's drop down list. The clipped parts can be easily identified in spectrograms, for example, when compared to the files generated with "32 bit (int)" and "24 bit" in the drop down list.

      These clipping won't affect the test result since RMAA only evaluate crosstalk and the distortion caused by clipping is irrelevant. So it should be totally fine to use the unattenuated "32 bit (float)" file as the input format to SoX. Yes the command-line SoX tool is integer-based, but the clipping won't affect how RMAA interprets the crosstalk figures.

      For example, I used this command to encode the 32-bit float 352.8kHz RMAA signal and it seemed to work fine at least in digital domain. With appropriate ultrasonic filtering, RMAA showed over 133dB DNR, 0.00000% THD and so on and it is unlikely to bottleneck analog performance.

      sox -V input.wav output.dsf rate -v 2822400 sdm -f sdm-8

      The available modulators are clans-4 to clans-8 and sdm-4 to sdm-8, their digital domain performances are known, the unknown parts are how they actually perform in different DAC chips/boxes. Perhaps you can try them out and see which one works best/worst with which DAC.

    3. Excellent, thanks Bennet,
      Got SoX-DSD last night and started examining some of the RightMark test signals. As you noted, the RM signals clip with some signals up to +2.8dB true peak with intersample overs inevitably when resampling/transcoding to DSD.

      Starting some conversions now with signals at 16/44.1 (converted to 24-bit and applying attenuation which I assume any knowledgeable label would do when putting CD-quality into SACD!) and high-res with 32-bit floating point to feed the encoder with highest resolution. Currently planning to add a -3dB attenuation across the board so as to prevent clipping in the DSD. Probably not unreasonable since Scarlet Book calls for -6dB (even though many/most SACDs did not follow this), so I'll split the difference at -3dB.

      Let's see how this goes. Will have a peek at the clans/sdm-4/8 differences. Probably will form the basis of future DSD testing of DACs here!

      Thanks for the comments and suggestions. ;-)

    4. Thanks for looking into it Archimago,

      My initial worry is that manually attenuating the reference RMAA files, depends on the used methodology and RMAA settings (for example, in RMAA's General --> Normalize amplitude of test signals before analysis), will change the analysis figures even within the reference files themselves. Your proposal is perfectly fine, but readers may curious about how the reference files look like after adjustment, so there would be no mystery if the results of the reference files are also included in the reviews. Perhaps, the differences are caused by different output voltages between PCM and DSD? It would also be interesting.

      Yes, I realized you reviewed some DSD encoders years ago, so the SoX results are nothing more than adding another one in your database.

      Also, regarding operations between SoX and RMAA, there are cases that SoX may generate PCM files which are unreadable by RMAA. I made a tool to solve this issue:
      The readme file described the issue, as well as the issues related to 0dBFS+ sample values.

    5. Thanks Bennet,
      Nice work on the software and will see if oldsCool is needed once I start trying some of the tests. Yeah, the manipulation will keep resolution of the test files as high as possible and should be a good way to examine the effects of the different modulators plus allow more deliberate look at relative abilities of DACs and DSD playback. Could be useful for folks doing PCM --> DSD upconversions I suppose to be aware of relative resolution abilities.

      Other than folks playing SACD rips, I really wonder these days if many people still bother with DSD downloads. I see that for example is still around and selling something like Patricia Barber's latest Clique! in DSD1024 for CAD$80, which is just a conversion of DXD PCM source (which is CAD$42)! Not a good deal IMO. That's gotta be a microscopic market...

  4. Hi great review. Thanks!

    I think you should also test their top of the line DAC, the Topping 90SE, which has broken all measurement records to date. You may contact its designer, the 24-year old John Yang, who goes by JohnYang1997 on the Audio Science Review forum for samples...

  5. Hey Archimago, the D90SE is an MQA DAC so I bet you won't buy it. However, if you are going to try another Topping product, D30 Pro is a plausible candidate, and it uses the latest Cirrus Logic flagship CS43198 with Direct DSD support. Yes AKM is in trouble due to the fire, so why not try something new?

    1. Hi David and Bennet,
      Yeah, I try to avoid MQA the best I can. Dollars spent on that hardware and software just means some taxes going towards the MQA company - bad for the audiophile ethos.

      Yeah both the 90SE and D30 Pro look like great options with insane levels of resolution.

      Wow. Did not know that John Yang is 24 years old! I think I read some of his comments on the D10B on the boards which was why I wanted to try out the unbalanced conversion for myself. ;-) Maybe time for the oldskool audiophile press to find some new folks like this to talk about instead of the usual near-retirement audio designers.

    2. I'm running a D30 Pro in the kitchen, together with a Hypex Ncore power amp and a pair of Elac wall-mounted speakers. As far as I can tell, it sounds great and works just fine. The only disadvantage is that it requires some more ugly cables than my previous integrated amp with built-in DAC.

      As always, great stuff Archimago! We are truly spoilt for choice these days.

  6. I saw that you tested this balanced DAC with an TRS-RCA adapter. Topping emphatically does NOT support this configuration for the balanced D10B and voids all warranties and any damages are your own to fix. They recommend instead that you get the D10s DAC, which was designed for RCA outputs. Same price, I believe, as the d10B.

    1. Hey David,
      Yup, it was because I read this from Topping that made me want to try for myself. ;-)

      As per the text, I reiterated a number of times that this is a bad idea and worsened distortion even though I haven't damaged things so far.

      The way I see it, like it or not, there will be people doing this, perhaps accidentally so I think it's worth knowing what happens. Obviously if I destroyed the DAC in the process, I would have let everyone know to be very careful!

  7. BTW, someone from the ASR forum by the name @maty (from Spain) COPIED your TRS-RCA THDN graph and text, (word for word!) and claimed it was him who performed the tests!!!

    1. Hmmm. That ain't cool... Hope folks correct that!

    2. I did ... a few posts afterwards.

  8. Hi Archimago!

    "Currently planning to add a -3dB attenuation across the board so as to prevent clipping in the DSD. Probably not unreasonable since Scarlet Book calls for -6dB (even though many/most SACDs did not follow this), so I'll split the difference at -3dB."

    Do you know that there is a smart VST plugin that can change the volume of the audio file in 6 dB steps in a completely lossless way? By "completely lossless" I mean literally: at if you take a 16 bit audio file (file A) and convert its bitrate to 24 (or 32), you can attenuate it by -6 dB, save it in 24 or 32 bit resolution (file B). Then you can open it, increase back the volume by +6 dB, and save it in 16 bit resolution with no dithering applied (file C). If you then bit-compare files A and C, they will be identical (bit perfect).

    The VST plugin I am talking about is free. It's called AirWindows BitShiftGain:
    from a guy called Chris.

    Knowing how you always strive to perfection, I thought you would want to know about this plugin.

    I use it in my VST-chain in Foobar, when I do digital room correction (when I listen through speakers) or when I apply crossfeed (when I listen through headphones). Both processes result in overclipping if you don't attenuate the waveform, so I use this BitShiftGain to reduce the volume of the incoming audio by -6dB before further processing. I mean if you have to attenuate the signal anyway, why not do it losslessly?

    1. Hi fgk,

      While it is easy to perform lossless volume adjustment in digital domain, for example, if you look at the "Digital Filter Composite" graphs in Archimago's various reviews, the resulting spectrum plots with 4dB attenuation applied on the waveform did not have their noise floor reduced proportionally by 4dB because it is limited by analog performance. The same applies to bit-shift (~6.0206dB/bit) as well. No, I don't think it is an audibility issue, but for measurement, it matters.


      Regarding the DSD tests, as long as the generated DSF files are usable, they can still be sent to the DACs, as a form of stress test. Perhaps some DACs may exhibit issues with these test signal, but that's also a way to differentiate technical competence from one manufacturer to another. Also, DSD encoders like SoX and others already have their level lowered by 6dB upon input, take a look at mansr's website:
      Fig.6 One period of a sine wave (blue) and its DSD representation (red)
      One can see that the sine wave has an amplitude of +/-0.5 but the samples are +/-1.

      The issue with RMAA's crosstalk signal is that the sample values themselves are already above 0dBFS even without any frequency domain manipulation (SRC, filtering, EQ...). Which means all references files, except the 32-bit float ones, are already clipped in the first place, like this:

      I wonder why no one talked about this for so many years, likely because it has no impact on the measurement results themselves. I also emailed the RMAA developers and reported issues like the 88.2kHz tests have botched frequency response, I got an acknowledgment, but no bug fix.

      Sorry about the verbose posts, no intent to flood the comment section, but there are something that you, Archimago, will take care of (DSD tests, showing ripples in digital filters, impulse response, polarity...).

    2. Thanks for the discussion fgk and Bennet.

      Yeah fgk, the level change will maintain the full resolution of the original test signal. Good to know about this VST "bit shift" plugin. Certainly a "perfectionist" way of getting the job done ;-).

      Thanks for that clarification Bennet on the -6dB already in the DSD encoding. Hmmm, maybe I should just leave the clipping and not worry about any extra attenuation. I'll give it a try both ways since I have both sets of files available and see if there's a difference. I would hope that when mastering engineers convert PCM to DSD, especially the "hot" modern PCM, they would be mindful of clipping and would manage these with that extra -3dB or so. Will talk about this when I give the SoX test files a go...

  9. "Balanced DACs are less prone to 60Hz mains hum and typically achieve lower noise levels."
    This is true, but we can put numbers to the assertion from your graphs: the D10B has a 60Hz peak at -144dB whereas the D10s is a WHOLE 3dB WORSE at -141dB ...
    I managed to start a bit of a fight over on ASR when I dared suggest that, for the majority of domestic applications, balanced connections are simply irrelevant as the improvement they offer lies below the threshold of perception. The results you report here aren't going to make me lose any sleep ;-).

    IF you've got a ground loop, then balanced connections may be an easier way to solve the problem than testing all your cables and connectors to find the stray impedance that's causing it. But if you don't have a ground loop and can't hear any mains hum then you won't be getting any _perceptible_ improvement by 'upgrading' to a balanced DAC and amp. Chunky XLR connectors look more impressive, but in most cases they're just audio jewelry.

    Don't tell anyone, but I wasn't that impressed by Eilish's first album (heresy!). But her second is a lot better, largely (IMO) because her brother Finneas has improved his production skills markedly in the past couple of years. Of course he won a couple of Grammies for his drab production of When We Fall, so what do I know @_@.

    1. Right Charles don't loose sleep over this!

      Depending on the device, set-up, shielding of cables, etc. you will see varying amounts of 50/60Hz hum and certainly it gets ugly with ground loops of course.

      Even with a really ugly set-up like the one I used when testing the JitterBug FMJ (see part B "USB 3.0 HUB CONNECTED TO COMPUTER"):

      If that was a balanced DAC instead of the AudioEngine D3, there would be no noise anomalies and resistance to that 60Hz hum.

      Another example I think of "Good Enough". For the vast majority of us, single-ended/unbalanced is completely adequate unless you actually hear noise that can be pin-pointed to the RCA output. Using XLR at home is icing on the cake and may not add anything of value. Obviously in complex audio systems like in the studio, balanced would likely be essential!

      Yeah, I'm OK with that 1st Billie Eilish album as well, and agree that I enjoy this 2nd one more!

  10. Umm yes measurements aside i just got the d10b and i hope that there is something wrong with my unit. It is easy piercingly shrill. And not in headfi superbest terms. I mean just so harsh sounding. I mean in a omg right out of the box what the hell is wrong with this thing way. Certain sounds like cymbals the word s have something beyond sibilance. Multiple amps cables headphones just to be sure. It's going back. Its unlistenable even at low volumes

    1. Hmmm. Interesting larry.

      Definitely sounds like something is wrong. As a purely USB-powered device, be careful if your set-up has any complexity to it which might create ground loops or noise. Usually the single-ended D10s is more susceptible but I have seen it with the D10 Balanced as well.

      Curious, what computer / streamer are you using? And what pre-amp or amp are you connecting this to? Would be interesting if others have issues with those up/downstream devices.

  11. I purchased the D10Bal because of your great review. I am very pleased with the DAC. Simple, clean, sounds great. It is one of the best frugal purchases I have made Using my plain old Mogami Gold balanced cables to my powered Heavenly Soundworks 517 speakers I am really impressed. I am very interesting in your upcoming review if their “pro” DAC. I will be hard pressed to change this DAC out.

    1. Glad to hear Bob,
      I like no-nonsense devices like these. Simple single USB connector and just the analogue outs. Small. Enough capabilities to handle hi-res DSP playback which we optimize on the software side!