So I got another DAC here at home - purchased through usual retail channels of course :-).
The Topping DX3 PRO (~US$220) is a device meant for those who want wired (S/PDIF TosLink, Coaxial, USB) as well as flexible Bluetooth 5.0 wireless input. There are 2 versions of the DX3 PRO out there - the earlier version reviewed/measured at Audio Science Review by Amir and the newer one which I have here with LDAC support (Sony-developed "high resolution" Bluetooth CODEC, also measured by WolfX-700) released in late 2019.
At some point, I figure it would be interesting to compare how the various Bluetooth codecs perform since this device can accept the lowest common denominator SBC, plus AAC, aptX, aptX-LL (Low Latency), aptX-HD, and LDAC audio - about as broad a range as I've seen among Bluetooth DACs (here's a good primer on these acronyms).
For today, let's examine the device's standard DAC performance and run it through my typical procedure with a few new measurement variants to see how it performs.
I. The BoxHere's a top view of the box with the antenna sticking up - notice the "Hi-Res Audio" stickers - one for the standard hi-res DAC, another for the aptX-HD and LDAC wireless hi-res.
It's a pretty nice solid metal box. Plastic remote with a good clicky feel to the buttons. Here's a picture of the remote with the rear of the DAC:
The DAC uses dual AKM "VelvetSound" AK4493 chips for conversion (spec of -113dB THD+N / 126dB DR). XMOS XU208 USB controller to handle up to DSD512 (DSD256 DOP) and PCM 768kHz. Qualcomm CSR8675 Bluetooth chip to support the various audio codecs. Remember that Qualcomm purchased CSR, the makers of aptX back in 2015 so it makes sense that this chipset supports the range of aptX flavors. Headphone out amplification uses the TI TPA6120A2 (spec -112.5dB THD) with 0dB and +9dB gain settings available. The spec says up to 700mW into 32Ω load. Given the price, this is a capable DAC.
The older V1 model used the TI OPA2140 opamp to drive the headphones.
Similar to the Topping D10 previously measured, the manual comes with some measurements already printed for you - nice to see, audiophiles need not be phobic over numbers and charts :-)
|Nice OLED display - easy to read volume control setting. Topping D10 on top.|
As usual, let's start with the "microscopic" tests. Right off the bat, we see that this DAC provides us with a number of digital filter options. To be honest, while it's nice to have choices, unless I'm purposely fooling around with stuff like the custom intermediate filtering, I'll listen with standard linear phase, steep filters. Typically these provide best flat frequency extension and the linear phase setting will not add phase shift/group delays in the time domain.
II. Oscilloscope Waveforms, Digital Filters, Headphone Out
For the DX3 PRO, this means filter setting F-1 (labeled as just "F1" in my diagrams) would be my focus for the majority of these measurements and when listening. So with the DAC set at F-1, this is what a 1kHz 0dBFS sine wave and -3dBFS square waves look like off the rear RCA outs (connected by USB from Raspberry Pi 3 B+ running Volumio):
As you can see, the output is a standard 2.1Vrms at 0dBFS. Channel balance is excellent with the 2 channels precisely overlaid in the sine waveform. For the square waves, as expected, with the filter set at linear phase, pre- and post-ringing are symmetrical.
So, for filter F-1 setting, the impulse response looks like this:
Note that "absolute polarity" is maintained. Here's the "Digital Filter Composite" (DFC) consisting of an overlay of various signals that can help us understand the digital filter better in the frequency domain, based on discussions with Jürgen Reis. This is an extension of what Stereophile has been doing:
Looks great; for context, compare this to the AudioQuest Dragonflies. The filter is reasonably steep with minimal imaging artifact from the 19 & 20kHz tones or wideband white noise. Already we can see that IMD would be quite low. Furthermore, this filter has enough intersample overload protection to prevent anomalies with the 0dBFS white noise signal. Certainly an excellent performing digital filter among the DACs I've evaluated over the years.
This is of course just filter F-1... The DAC provides 6 settings total, here are my descriptions rather than using the AKM or Topping language which can be a bit vague:
F-1 = Linear Phase, Sharp Roll-Off
F-2 = Linear Phase, Slow Roll-Off
F-3 (Default) = Minimum Phase, Sharp Roll-Off
F-4 = Minimum Phase, Slow Roll-Off
F-5 = "Super Slow" Roll-Off - essentially non-oversampling
F-6 = "Low Dispersion", "Short Delay" - Intermediate Phase
I didn't bother testing them all with DFCs, but here are the impulse responses for F-2 to F-6:
For the most part, the filters are as per the descriptions. Of interest, the F-5 setting is useful if you want to listen for a NOS-like sound. I'm not even sure why they call it "Super Slow Roll-Off" - it's essentially unfiltered! And the F-6 "Low Dispersion" is an intermediate phase setting with moderate steepness.
Since F-3 is the default and I assume many people will just listen with that setting, here's the DFC graph:
While the steep roll-off is the same as F-1 above, notice that it doesn't handle "intersample overs" as well and we can see this showing up with the 0dBFS wideband white noise signal.
As I said above, for those who love the sound of Non-Oversampling DACs, Audio Note's silly misnomered "1x oversampling", stair-stepped waveforms, or the "euphonic distortion" that this might bring with copious ultrasonic imaging/"aliasing" artifacts, check out setting F-5 and this "beauty" of a DFC graph and stair-stepped 1kHz sine wave:
The headphone output is indeed quite powerful. It's capable of pushing 4.6Vrms into a 30Ω load at the +9dB gain setting (filter setting F-1):
So that's a good 700mW as per spec. The one disappointment however which the WolfX-700 measurement/review found was that the output impedance unfortunately is rather high:
That voltage drop from Open to a 30Ω load (1kHz sine) calculates out to an output impedance of 10.4Ω. The specs sheet rated the headphone output impedance as "<10Ω". Yeah, this definitely could be better and suggests that one should use higher impedance headphones to reduce variation in frequency response (80+Ω). The previous DX3 PRO V1 was superior with a measured output impedance of 1Ω. Remember guys and gals, just like amplifiers with higher output impedance, ears/brains are forgiving and sticking a 32Ω IEM into this DAC still can sound good depending on the headphone even if technically suboptimal.
III. RightMark noise and distortion resultsOkay, let's now run the DAC through the battery of tests in RightMark Audio Analyser 6.4.5 Pro. As usual, the standard measurement chain I use these days is with the RME ADC:
Topping DX3 PRO V2/LDAC --> Generic 6' RCA cable --> RME ADI-2 Pro FS ADC --> 6' Generic USB --> Microsoft Surface 3 Pro computer for analysisFor these measurements, I've set the filter to F-1. As I said above, I like "orthodox" linear phase, steep roll-off filters for listening and measurements.
Other than listening for a few hours before measuring (maybe 4 hours total) to make sure all features worked, no "burn-in" is needed. Over the years using audio electronics, I've certainly heard "warm up" change sound quality, "burn out" of components (smoked some amps over the years), and "break down" resulting in distorted sound... But other than maybe mechanical changes over time like speakers settling in, I have never heard "burn-in" actually improving sound quality over hours and hours of use.
Starting with standard resolution, let's just make sure there are no surprises with the DAC - here's the summary results with a few other comparison DACs:
Not bad at all for the Topping DX3 PRO, with RCA output. As has been the case for years, standard CD-resolution reproduction is no problem with essentially any decent DAC. This is consistent with the blind test results last year as well.
Performance is essentially the same at 16/44.1 compared to the inexpensive Topping D10 (currently ~$US90) although marginally better THD and IMD+N can be seen. The TEAC UD-501 was my reference DAC for many year purchased in 2013. The Oppo UDP-205 has since become my daily use reference DAC for the main system since I got it in 2018. For "fun", I've included the venerable PonoPlayer that I got back in 2015 as well as the inexpensive SMSL iDEA (the equivalent Sabaj Da2 can be had for <US$60 these days).
Here are some graphs for more details:
Notice that the PonoPlayer rolls off the highs due to its slow-roll filter at standard resolutions similar to the minimum-phase Ayre "Listen" filter. The noise level graph shows that the TEAC UD-501 and Oppo UDP-205 have a little bit of 60Hz mains hum (absent if you use the XLR balanced output). Notice that the Topping DX3 Pro does not have hum with its little switching power supply.
The Topping DX3 Pro IMD+N suggests a device with very low non-linear distortion across the audible spectrum.
It's with the 24-bit high-resolution tests that we typically see more measurable variations between otherwise good DACs. Here's a summary with the same selection of devices:
This is where we see the DX3 Pro pulling out ahead of the D10 in resolution. In fact, for RCA output, the dual AKM AK4493's are essentially as quiet as the Oppo UDP-205's ES9038Pro RCA out although the Oppo ultimately is superior when it comes to insanely low distortion and of course has the option for balanced XLR output.
Here are some graphs:
Left / Right channel balance is excellent across the spectrum. At most +/-0.01dB difference between channels which could just be representative of variation with the interconnects and ADC used.
Very impressive results from this little DAC! In terms of distortion and noise level, the Topping D10 is around the performance of the TEAC UD-501 while the Topping DX3 Pro is approximately at the level of the Oppo UDP-205 RCA out.
For completeness, let's look at the measurements at 192kHz. In reality, I have few albums at this samplerate and honestly don't have any special desire to waste storage space on these (regardless of whether storage is "cheap"). Summary chart with the same devices:
Again, excellent performance from the DX3 Pro. A couple of graphs for frequency response and noise level:
I have measurements up to 24/384 and the results look good. But seriously, who cares? Anyone have albums at 384kHz? :-)
However, let me reiterate something from the 384kHz data. Here's the noise floor of the DX3 Pro alone keeping it nice and clean:
The rise in ultrasonic noise isn't terribly high even into 192kHz and is a reflection of the combination of DAC and ADC's modulator noise. But notice just how clean this is for the stereo RCA output! No noise peaks at all. And this is all using the stock switching power supply. Remember this when you hear about folks online spending money on often very expensive linear power supplies for their DACs, computers, and streaming devices claiming that the power supply somehow has special abilities to keep noise level low! Unless these audiophiles/companies can prove such a thing, it could all be nonsense and in fact, one could be introducing noise into the playback system while wasting money and making the system less power efficient.
|Time to try out DSD mode...|
Remember that DSD playback can have lots of ultrasonic noise due to the use of strong noise shaping to keep noise level low in the audible frequencies (thanks to 1-bit quantization). The Topping DX3 Pro provides the option of 39kHz (default) or 76kHz low-pass filtering. I would recommend staying with the default for typical DSD64/128 playback, the 76kHz setting would be fine for DSD256/512 but I doubt any of us have many of those recordings either!
Anyhow, here are summary results of the 24/96 RightMark test signal converted to DSD64/128 to show resolution and noise level differences. Low-pass filter kept at default of 39kHz:
Looks good. No surprises really, and highly suggestive of "transparent" hi-res PCM to DSD conversion. Remember for the distortion and noise measurements, RightMark is calculating from 20Hz to 20kHz. As such, DSD64 is capable of good "high resolution" within the audible spectrum. It's in the ultrasonic frequencies where we start detecting noise.
Here are the comparison graphs:
Again, this is all rather typical of DSD. DSD64 performs similar to 24/44.1 with the noise rapidly rising after 20kHz. DSD128 is more akin to 24/88.2-level performance (we can start to see a little excess noise at 40+kHz) and if we were to continue that comparison, DSD256 will have similar performance to 24/176.4 which is why if you wanted to use that samplerate, you could increase the DSD low-pass filter setting up to 76kHz.
IV. Stepped Sine TestsI haven't done this previously with my DAC measurements but it's worth having a look for comparisons. The newer beta versions of Room EQ Wizard (currently 5.20 beta 55) have the ability to use the stepped sine facility to measure harmonic distortion across frequency and output levels. As usual, amazing work from John Mulcahy!
All tests performed at 96kHz with 131k point FFT using the RME ADI-2 Pro FS as my standard set-up. Again, the DAC has been maintained at filter F-1 (linear phase, steep roll-off).
For the Topping DX3 Pro, here's what the frequency stepped sine distortion graph looks like:
Across the frequencies, for the most part, we see a pattern of harmonics where lower order (2nd and 3rd) components generally predominate over the higher ones. In principle, this should be good for nice clean hi-fi sound quality.
We can compare this to the Topping D10 and its ESS Sabre ES9018K2M DAC:
Overall THD of the Topping D10 is higher than the DX3 Pro, mainly because of the higher 2nd and 3rd harmonics. The higher order harmonics are very effectively suppressed at -3dBFS as you can see. Notice that the 3rd harmonic rises above the 2nd above 2kHz and becomes the predominant harmonic in the high frequencies.
In REW, we can also perform an amplitude level stepped sine measurement:
The top half is the FFT of a 950Hz tone at the end of the sweep with the fundamental signal at 0dBFS. I'll speak more about the choice of 950Hz instead of 1kHz below.
The bottom half is the stepped sine measurement as the fundamental 950Hz tone is gradually increased from -120dBFS to 0dBFS measured every 3dB (4 sample average). We can visually inspect the linearity of the amplitude steps; notice that I have overlaid a very fine red line through it. Linearity looks excellent even down to -110dB. Beyond that, we can see slight deviation from the red line as the signal encroaches into the noise floor. At the bottom of the graph, we see a measurement of THD, 2nd and 3rd harmonics. As you can imagine, there are limits to how low we can measure these levels based on the ADC system and noise. We can see that above -15dBFS, there is a rise in the 3rd harmonic which is also reflected in a slightly rising THD above -10dBFS.
Here's the frequency stepped sine data reformatted relative to the 950Hz fundamental:
As you can see, through most of the amplitude range, THD is primarily consisting of the 2nd and 4th harmonics until -13dBFS where the 3rd harmonic arises. Above -7dBFS there's a bit of "saturation" resulting in overall THD rise.
Again, for comparison, here are the same amplitude level stepped sine graphs using the Topping D10:
If we look closely at the thin red line overlaid on the fundamental, we see that it deviates from linear below -100dBFS; not as good as the DX3 Pro as expected, but realistically, good linearity below -90dBFS is more for bonus points rather than "essential" resolution unless one's soundroom is located in an anechoic chamber. :-)
And if we plot the harmonics relative to the fundamental 950Hz sine wave:
Quite a different picture emerges compared to the DX3 Pro with this DAC showing more fluctuations in the relative levels of harmonics depending on amplitude!
While the DX3 Pro and its AKM DAC has a more predictable pattern of 2nd and 4th order harmonics through much of the amplitude range, with the D10's ESS DAC, notice that the harmonics vary quite a bit more relative to each other and the 3rd harmonic became most prominent between -47 to -30dBFS. While I certainly would not claim that such differences are significantly audible, it's interesting to see the different "signatures" the DACs leave. These are some of the subtle differences between ESS and AKM DAC chips we would not notice when looking just at a single THD(+N) number. I know some audiophiles claim to have preferences for or may dislike the sound of certain types/brands of DACs (so-called "Sabre glare" for example which I can't say I've personally had concerns with).
Also, it is known that some ESS-based DACs have the THD/IMD "hump" as discussed here; not really an issue with the D10 as far as I can see other than the elevated 3rd harmonic between -47 and -30dBFS. Maybe a very slight bump from -39 to -17dBFS.
Note the accentuated proximal sidebands on either side of the fundamental. These side bands correlate with the 229.7Hz square wave embedded in the lowest bit of the 16-bit J-Test.
The low-level jitter-related anomalies are easier to see with the 24-bit test:
This pattern of J-Test would not be audible due to extremely low levels (only around -130dB peak) and the proximity to the fundamental frequency would mask audibility further.
Yes, jitter performance can be improved with this DAC. No, things like UpTone's USB Regen would not improve the situation because the jitter is intrinsic to the clock inside the DX3 Pro itself, not dependent on the computer/streamer - feel free to prove me wrong!
Since USB is not the only wired input, here are the jitter results from the S/PDIF interfaces - coaxial and TosLink. I used my old Squeezebox Touch S/PDIF output which is not known for low jitter as the source:
As you can see, asynchronous USB simply performs better compared to S/PDIF for most devices when it comes to jitter.
Remember that typically, coaxial S/PDIF has higher bandwidth than TosLink which requires conversion of electrical to an optical signal and back. The benefit of TosLink of course is galvanic isolation which may or may not be needed (typically unnecessary). As you can see above, generally, coaxial S/PDIF provides better jitter performance than TosLink which is more evident when you compare the 24-bit results. In any event, with the fundamental signal at -6dBFS, the jitter-affected sidebands are still below -120dBFS. Remember, the S/PDIF source will have an effect, so a more accurate device than the Squeezebox Touch should be cleaner than the FFTs above. Also, some DACs like the Oppo UDP-205 will also achieve intrinsically lower jitter using its S/PDIF inputs.
Pssst... Want to see almost "perfect" J-Test results with the Topping DX3 Pro?
Beautiful. This is the J-Test result when the signal is upconverted to DSD128. DSD64 is basically just as nice looking. I suppose if one is aiming for "perfection", upsampling PCM to DSD128+ with a high quality algorithm (like HQPlayer) likely will get you very "accurate" results. You'll need a powerful playback computer or server (eg. converting through Roon Core) though.
VI. Subjective Listening Impressions & ConclusionsOkay, let's end off with some subjective impressions, listening comments and summary.
It sounds good. As a traditional USB DAC which is how I measured the DX3 Pro here, it's all plug-n-play to the Linux-based Raspberry Pi 3 streamer. The Topping USB driver for Windows with ASIO has been stable on my laptop. The Bluetooth streaming from my smartphone works well also and with a bit of fiddling around on my Huawei P30 Pro with Android 10, I was able to select which Bluetooth codec music was being transmitted through using the Developer Options. I can easily cycle in realtime between SBC, AAC, aptX variants, and LDAC. I was suitably impressed by the sound of LDAC at >900kbps "Best Quality" setting which was the default for my phone with up to 24/96 target quality.
Up to now, most of my listening has been with the headphone output to my Sennheiser HD800. The Sennheiser's high nominal impedance of 300Ω is a great match. Since I measured the RCA out mainly, see WolfX's measurements for headphone out performance (I might do my own measurements at a later time). Fidelity is excellent, consistent with the measurements. A fellow blog reader David S. recommended some Monte Montgomery albums and I recently got his live CD New & Approved (2003, DR7) which is one of David's "desert island" picks. I agree that this is a great album for music and sound (still wish the studio engineers laid off the compression of course :-). Energetic performances, remarkably detailed sound of guitar and percussion! I love the audience participation on "Tug of War", and the cover of Hall & Oates' "Sara Smile" is great. The DX3 Pro provided plenty of power, volume can be pushed up to deafening levels with the Sennheiser. Excellent rendering of the frequency range from deep bass to limits of upper frequencies, nice transients on the percussion and realistic rendering of vocals.
The other album from Montgomery I enjoyed was the studio effort Mirror (1999, DR9). A great showcase of the Austin, TX guitarist. As an aside, I listened to this album with my 1MORE QuadDriver IEMs which are only rated as 32Ω and they still sound good even if not optimal and certainly not "reference" quality like the Sennheiser HD800. The 1MOREs have an accentuated bass response which could work well with some music. One pairing I found to be excellent was the DX3 Pro with my old AKG Q701! Plenty of power to drive these 62Ω 'phones and the resulting sound was very pleasant with adequate bass strength; I have found lesser headphone amps to sound anemic with these.
I have not needed to use the +9dB gain setting as my headphones are all sensitive enough to not need so much power (nor do I listen to music too loud!). Good to know the extra "oomph" is there if needed!
So, bottom line... The Topping DX3 Pro is a great little DAC for not too much money (currently ~US$220). The higher-end AKM DACs are right up there in terms of sound quality if accuracy, high-fidelity and "transparency" is what you're after (no romantic/sentimental tube-DAC sound here, friends!). I really like the convenience of the remote for control of settings and general turning ON/OFF from standby. The OLED display is a nice size with good functional details.
The main "limitation" with this device is the higher headphone output impedance (~10Ω). Othewise, while I would have been happier with lower jitter measurements, it's not audible anyways. It'll be cool if Topping came out with a V3 that addresses the output impedance, cleaned up jitter a little more, maybe go with a 1/4" headphone jack to be even more "pro" - that would be "perfect".
Needless to say, using this as a Bluetooth DAC with its powerful headphone amplifier is practical, flexible, and something I will very much enjoy in regular use in the days ahead streaming from my cell phone, tablet, and laptop!
As usual, great to see technological progress continuing to provide more ways to enjoy music, maintain fantastic sound quality, and at an unapologetically great price point for meaningful "hi-fi end" gear catered to all music lovers!
This supposedly "strong" noise was at least part of his justification for the UpTone USB Regen back in 2015. After all these years, I have yet to see the company publish any results to show their device suppressing the 8kHz noise or demonstrate any benefit for jitter. All we have are lots of words and some subjective testimony as typical in the audiophile world.
Without objective evidence, I fail to see why anyone would pay money to hang all kinds of doohickeys off their USB ports whether the Regen, or AudioQuest's JitterBug, or the iFi USB iPurifier variants.
In other news...
A friend texted me when he saw that Apple is going to be introducing "Spatial Audio" updates to the AirPods Pro with the ability to simulate 5.1/7.1/Atmos content. You can see the WWDC 2020 Keynote video here - fast forward to 43:20. It looks like it will incorporate much of what Redscape is capable of for head-tracking wirelessly using the AirPods Pro's internal accelerometer along with DSP to simulate the multichannel "surround" effect.
To synchronize the soundstage with motion of the mobile device and keep the sound "anchored" seems like a neat trick. I look forward to giving this a try at some point although I have some suspicions about the accuracy of the head and device tracking - hopefully it's "good enough" as a first implementation. Also, remember that the AirPods Pro's H1 chip can only handle the AAC codec and I suspect have limited DSP power otherwise. Therefore, I assume the vast majority of the surround processing is being done by the mobile device. This could be a lot of work for the iPhone or iPad handling the HRTF processing, encoding in AAC for Bluetooth transmission and achieving minimal latency in the process if you're watching a movie for example. Hearing is believing and I hope to be impressed by Apple knowing their usual slick "magical" hype in the promo presentations.
As I said a year ago (see Multichannel and the Audiophile), getting the "masses of headphone users interested (in multichannel) with HRTF-based DSP-reconstructed binaural playback" would be very useful to getting us more surround content ahead.
I see that Tidal is already streaming Atmos-encoded content (currently something like 40 albums). Now this I can support, not that nonsense called MQA even though it's a lossy Dolby Digital+ stream. It would be very interesting if this Apple AirPods Pro update is compatible and significantly improves the experience of streaming this type of content.
Recently, I listened to one of the first Atmos-encoded music Blu-rays - INXS' Kick 30th Anniversary (released 2017). I love this album having enjoyed it when it came out in my high school days but I must admit the results were mixed. Love the bass-heavy spatially-expanded "Guns In The Sky" at the start. Also enjoyed the "Never Tear Us Apart" mix. "Need You Tonight" seemed a bit too rear-channel accentuated for my taste as if most of the music were coming from behind. And "Mystify" just sounded too harsh and dated. Regardless, I'm a big surround music fan so I'd love to have more multichannel albums coming, Atmos or otherwise.
Have a great summer everyone! I'll be on vacation for a good part of July. Free time permitting, I might post some "Summer Musings" as I try to catch some R&R on a family road trip...
Stay safe especially to the American friends. I'm really quite proud of how British Columbia maintained communications and managed the pandemic with great results to show from the measured, calm, science-based, more objective, unifying, rational approach thus far.
Hope you're enjoy the music!