Topping D90SE desktop balanced DAC

In October 2020, amidst the continuing world-wide virus-induced disaster, a separate disaster for the high-fidelity community struck. The Asahi Kasei Microdevices’ semiconductor fabrication plant in Nobeoka, Japan, burnt down. It was where just about all the best AKM DAC (and other) chips were made. Now, some high-end high-fidelity manufacturers go their own way entirely, with custom-programmed Field Programmable Gate Array processors and R2R resistor ladders. But the others tend to build around existing DAC chips. And the two brands of DAC chips most highly regarded have been those from ESS (the Sabre range) and AKM (with its Velvet Sound series).

Both brands were well-represented in much of the best digital audio out there. And then the AKM factory caught fire.

Companies with quality products based on AKM chips were left scrambling. Some had plenty of stock in their warehouses, while others had to do some redesigning, typically to use ESS chips rather than AKM ones. AKM has since resumed production, but now there are quite a few products with versions using ESS rather than AKM chips.

Topping D90SE desktop balanced DAC

Which brings us to the Topping D90SE DAC. It appeared because the Topping D90 MQA employed the AKM AK4499 DAC chips. You can still purchase the latter device, but here were having a very close look at the Topping D90SE.

In short

  • The Topping D90SE is an incredibly high-performance DAC with balanced output via XLR with additional Bluetooth functionality
  • Based around the ESS Sabre ESS9038PRO digital to analogue converter chip (datasheet here). This is an eight channel DAC (with four channels in this implementation dedicated to each of the two stereo channels)
  • Supports PCM up to 32 bits and 768kHz sampling
  • Supports DSD64, DSD128, DSD256 and DSD512 via USB and IIS. Supports DSD1024 via IIS input
  • Supports MQA streams
  • Bluetooth 5.0 playback, with support for SBC, AAC, aptX, aptX HD, aptX LL and LADC codecs
  • Seven user selectable D/A filters. No NOS option
  • Front panel display showing sample rate, audio format, input and output
  • 1 x stereo XLR balanced outputs, 1 x stereo RCA single-ended outputs, can switch to one, the other or both
  • 1 x USB-B input, 1 x coax digital audio input, 1 x optical digital audio input, 1 x AES-EBU digital audio input, 1 x IIS input
  • Fixed output and variable output modes, switchable
  • Internal power supply with IEC socket
  • IR remote control included
  • Available in black or silver finish
  • 222mm wide by 46mm tall by 180mm deep
  • 1074 grams
  • The Topping D90SE is practical, versatile and delivers astoundingly high performance. Given all this, the selling price makes it a real bargain.
  • Price: $1599
  • Available at fine high fidelity retail outlets, and direct from distributor’s retail division (Topping D90SE)
Topping D90SE desktop balanced DAC

    More about the Topping D90SE

    The switch from AKM to ESS would have been easier for Topping than some other companies because it already produced a number of products based around ESS silicon. And this unit has been well received, especially by those who look to measured performance as a mark of quality. We will definitely return to that.

    A quick word on the inputs: both optical and coaxial S/PDIF inputs are provided, along with a USB Type-B for use with a computer. When connected to a computer that way, the Topping D90SE acts as a USB Class 2.0 Audio device. Less familiar to regular high fidelity folk could be the other two inputs. AES/EBU is an S/PDIF-like standard that runs over balanced XLR connections. This is commonly used in recording studios. The IIS connection uses a HDMI socket, but don’t go plugging your TV or Blu-ray into it. The signal is entirely different. It is designed to connect to devices that output digital audio in a format often used internally in digital audio devices. So it isn’t constrained to pre-existing USB standards or other digital audio standards. Apparently the Topping D90SE will decode DSD1024 (that’s 16x DSD!) if you have an IIS-equipped output device.

    The D90SE is controlled by three front-panel buttons, but more conveniently by the included IR remote control. This is the same remote that comes with several of Topping’s DACs, including some which are quite modestly priced. It works well.

    Topping D90SE desktop balanced DAC rear

    Signal support

    Regarding the fixed and variable output modes, the same outputs – XLR and RCA – are used, but the whole unit can be set (defaults to) DAC mode where it acts as a fixed level source device, or pre-amp mode with adjustable volume. So you could just plug it into a power amp and plug several digital components into it, switching between them and controlling the level with the remote control. I used it only as a DAC in this review, almost exclusively fed by a computer plugged into its USB-B input. It worked fine with Windows with the downloaded driver (if you want to use DSD, you’ll need the Windows driver), and with a Mac without any need for a driver. The Mac reported that the DAC supports all sampling frequencies which were multiples of 44.1kHz and 48kHz, right up to 768kHz.

    I tested it on both the Mac and a Windows computer at all those sampling rates, and all worked perfectly. I also tested DSD64, DSD128, DSD256 and even DSD512. All but DSD512 worked on the Mac, and all including DSD512 worked with the Windows computer. I’ve never been able to get DSD512 to work on a Mac. I think – I’m not certain – that’s because the JRiver software I use employs DoP (DSD over PCM, which means DSD disguised as PCM, so that it can be carried by the standard protocols) on the Mac. With the Windows version you can switch to direct DSD (if you’re using the ASIO interface).

    So, if you’re a Mac music person with a substantial collection of DSD512 material (I suspect there’s not one single person in the entire world who fits that description, primarily due to the almost complete absence of DSD512 material), maybe you should see if some other software can deliver the content. I’m guessing not.

    Topping D90SE desktop balanced DAC remote control

    For anyone else, if you have PCM or DSD digital audio content, the Topping D90SE is going to handle it … with ease. Including MQA. I ran through my collection of MQA test tracks, and the DAC happily unfolded them variously to 96kHz, 192kHz or 352.8KHz as appropriate. At least it placed “MQA” in the top-right corner of the display and showed the relevant sampling frequency. The display is light blue on black, so it can’t show coloured dots to indicate the MQA mode. Instead “MQA” indicates, well, standard MQA, while “MQA.” (note the full stop) indicates original artist or producer approval of the source. According to the unit’s instructions, “OFS” is another indicator for MQA which confirms that the player is being delivered “the final unfold of the MQA file”. I didn’t see that indicator with any of the eighteen test tracks I had available. Maybe that’s for tracks capable of unfolding to more than 700KHz.

    I had a little look at the datasheet for the ESS9038PRO DAC chip used in this unit. ESS Tech says that it “is the world’s highest performance 32-bit solution designed for Audiophile and Studio equipment applications such as SACD players, Blu-ray players, digital preamplifiers, A/V receivers, studio consoles and digital audio workstations.” That’s a bold claim, but no-one seems to disagree. The specs are impeccable. In mono mode (ie. all eight channels grouped in to handle just one channel), the dynamic range is claimed to be 140 decibels. The theoretically perfect noise floor with 24-bit audio is at -144dB, so that’s pretty incredible. THD plus noise is rated – the conditions aren’t stated – at -122dB. Keep that figure in mind when we get to some measurements below.

    Listening with the Topping D90SE

    Equipment (System 1):

    Equipment (System 2):

    Equipment (System 3):

    Regarding System 3, I’ll be reviewing the Topping headphone amp in coming weeks, and likewise for the Topping PA3s. Clearly the D90SE and the A90 are companion devices. Not to say they must be used together, but they are quite the match. When you place the A90 atop the D90SE, the XLR inputs on the former align perfectly with the XLR outputs of the latter. Topping have short XLR interconnects designed specifically for this arrangement.

    Topping D90SE desktop balanced DAC

    At which point I have little to say about the sound. The fact is, the Topping D90SE did so close to a perfect job of decoding digital audio, regardless of format, to analogue, any variance from perfection was well beyond my ability to detect it.

    It did it quickly too. There were no lost notes at the start of the track due to a delay in latching onto sampling frequency. And it did it without fuss. There were no noises or clicks as formats switched. The unit was a delight to use and a delight to listen to.

    Which brings me to the main story:

    Measuring the Topping D90SE DAC

    Over the better part of the past twenty years I’ve been gradually improving my analogue to digital converters. These are what I use to measure other gear, including DACs. Each upgrade has added several decibels of signal-to-noise capability. Presently I’m using an RME ADI-2 RS R Black Edition, out of Germany. Somewhat relevant note: this uses high-end AKM chips for digital to analogue and analogue to digital conversion. I bought mine only a few months after the fire.

    This is a truly great analogue to digital converter. But I suspect its limits may have been exceeded by the Topping D90SE. Don’t worry, I’ll borrow someone else’s work shortly to go beyond all that.

    First, before digging into (what I consider to be) the main functionality of the Topping D90SE, I’ll note that I checked the Bluetooth stuff as well. Yes, it does indeed work with SBC (as it must), aptX, aptX HD, aptX LL and LDAC. Furthermore, LDAC does indeed offer significant advantages over all the others, including an extended high frequency response. I’ll go into a lot more detail on this when I put up my review of the FiiO BTA30 Pro Bluetooth transceiver.

    So, let’s look at my measurements. First frequency response. Here we have it with 44.1kHz sampling:

    Topping D90SE desktop balanced DAC 44.1kHz frequency response

    Take your choice. Topping’s (ie. the default) is the cyan trace, “Fast roll-off minimum”. That takes the response solidly out to 20kHz, but “minimum” refers to timing, and in my opinion shifts things away from a proper unbiased representation of the signal. I won’t claim to hear a difference, but it’s clear that “Fast roll-off linear” is closer to reality. So I’d suggest changing the filter setting to F-5. That’s the … light puce is it? … trace.

    Now, we bump up to 96kHz sampling:

    Topping D90SE desktop balanced DAC 96kHz frequency response

    I’d stick with that F-5 light puce setting. That gets you to more than 40kHz at -0.2dB.

    With 192kHz sampling, the variances narrow, so why not just leave it on F-5? Minus 1dB at 84kHz seems okay to me:

    Topping D90SE desktop balanced DAC 192kHz frequency response

    Now, let’s look at noise. I did most of my measurements from the XLR output. I’ll have a look at RCA shortly. So, with 16-bit audio via XLR, the Topping D90SE consistently scored -97.8dBA. Which filter I set made pretty much no difference:

    Topping D90SE desktop balanced DAC 16-bit noise

    Which means: no audio noise.

    Things get better with 24-bit audio since it has a lower intrinsic level of noise and this DAC is able to deliver much of that. Here’s the graph using 24-bit, 192kHz audio:

    Topping D90SE desktop balanced DAC 24-bit noise

    Don’t worry about that gentle rise to the right. That’s a combination of the DAC and my ADC both shifting noise out into the ultrasonic, thereby reducing it in the audible band. The measurements amount to -121dB A-weighted. Now, I did a little bit of playing around with connections during this. For these tests I feed the signal via USB-B from a Microsoft Surface Pro. I sometimes have this plugged into a Surface Pro Dock which is itself plugged into mains power and Ethernet. The Surface Pro isn’t reticent about feeding noise from that connection into its USB connection. Mains breakthrough shows up as a 50 hertz spike, often with a bunch of harmonics. Noise from Ethernet is just a messy, higher level noise floor.

    Fact: many DACs let some of this through to the analogue outputs. Fact: I’ve even tested some in which enough has broken through (particularly the hash from Ethernet) to be clearly audible.

    Do you see any of this in that graph above? Nope, neither do I! I checked the various combinations of power and battery, Wi-Fi and Ethernet, and the graph was always the same. This is high quality stuff.

    A final graph: what about the RCA outputs? The frequency response performance was virtually the same, but there was some visible – although still totally inaudible – difference. Here’s the graph, in which you’ll see a little 50 hertz and some of its harmonics. Again, at -130dB or less, you aren’t going to hear any of that:

    Topping D90SE desktop balanced DAC XLR vs RCA noise

    Finally I measured total harmonic distortion with 16-bit signals at 0.00034% or less, while intermodulation distortion was at around 0.0035%. There are a lot of highly welcome zeros after the decimal points in those results.

    But even more with the 24-bit signal results. THD was 0.00009% or less. IMD was 0.00034% or less.

    A second set of measurements

    THD of 0.00009% is essentially out of this world. It’s simply amazing. But it isn’t what Topping claims. It claims around half that, at just 0.00005%.

    Well, my measurements unavoidably include the contributions of the device being measured, and the device doing the measurement, my RME ADC. That ADC is simply one of the finest in the world, but when you’re measuring a DAC that would seem to be also one of the finest in the world, it simply isn’t enough to extract all the technical performance.

    So here’s the Audio Science Review. This guy uses a $20,000+ laboratory-grade Audio Precision analyser for his tests. And he measures stuff that I don’t, such as low-bit linearity and jitter. Check it out. But here’s his conclusion:

    It is not every day that every barrier we have seen in DACs gets shattered to new levels. We are seeing extreme attention to every aspect of this design to squeeze this much performance out of this ESS DAC. Just when we though the race was over, here comes one more over achiever! Those of you who have made it a hobby to upgrade the latest and greatest, here is another one for you!


    Have I been too light on with listening impressions? Maybe. But if some reviewer claims that the Topping D90SE sounds inferior in some way to some other DAC, the burden of proof should be on that reviewer. How can it sound different when there is, for all realistic purposes, no noise, no frequency balance colouration, no distortion? What special magic can that reviewer detect that evades these measurements?

    Which means that in my opinion, if you’re after an essentially perfect DAC, I’d suggest that you check out the Topping D90SE.

    DacsDesktop headphone amps and dacsTopping