MEASUREMENTS: Oppo Sonica DAC (ESS Sabre ES9038Pro based device)

My Raspberry Pi 3 "Touch" on top of the Oppo Sonica DAC - notice the DAC can handle up to PCM 768kHz (and DSD512 / 22.4MHz).

A few weeks ago, I published the preview of the new Oppo Sonica DAC and noted some issues I ran into with unusual jitter and digital filter measurements. Thankfully, Oppo was able to correct the issues within a week with the beta firmware and I see that the official firmware incorporating the improvements is now out - Sonica-33-0511 on the support website (May 15, 2017 date).

With it becoming official, I figure it was time for me to run and post results of objective tests on the device. This should give us a taste of both the capabilities of this Oppo Sonica DAC as well as look at what the "latest and greatest" ESS Technology ES9038Pro 8-channel DAC chipset is capable of... Although I have not seen it specified anywhere, my assumption is that the stereo configuration in this device ties 4 of the DAC channels for each of the stereo pair.

I. Intro & Set-up

I've already discussed the external appearance and control mechanism of this DAC previously. There are a number of subjective reviews recently so I refer you to those writings if you want more information on subjective matters, the control app, "look and feel", etc. (see here, here, here, here).

I've posted the picture above previously but I wanted to remind everyone of the physical connectors back there. I'm most interested for this post in showing the objective performance with standard digital inputs - USB and S/PDIF. Along the way we'll compare the performance with some of the better DACs I've used over the last few years and I'll show results from DLNA streaming over WiFi and the quality of the AUX In connection for passing analogue through.

The measurement chain, unless noted otherwise looks like this for USB2.0 input:

Pi Touch --> 6' USB --> Oppo Sonica DAC --> 6' RCA / XLR --> Focusrite Forte ADC --> 6' USB --> Win 10 laptop

When I'm measuring through the S/PDIF input:

Microsoft Surface Pro 3 (i5) --> 6' USB --> CM6631A USB-to-SPDIF --> 6' TosLink / coaxial --> Oppo Sonica DAC --> 6' RCA --> Focusrite Forte --> 6' USB --> Win 10 laptop

If you're familiar with my posts here, the measurement chain is typical of other devices I've looked at. I typically will use the same USB2.0, TosLink, coaxial, RCA, XLR cables... For measurements, unless stated otherwise, they're all generic wires as I have found no evidence for significant improvement with more expensive wires in typical audio use.

II. Digital Oscilloscope, Digital Filter, Impulse Response

As usual, we start microscopically and have a peek at the waveforms from the device.

Here's the 1kHz 0dBFS 16/44 square wave with the digital oscilloscope - all volume settings "locked" at 100%:

Peak voltage from RCA is ~1.81V (1.3Vrms). We can see that the channel balance is truly excellent. When I see that kind of precision on the oscilloscope reading, it usually means very good performance is ahead. Notice that there's only post-ringing in the square wave transitions so this means it's using a minimum phase filter.

Indeed, Oppo has decided to use a "sharp" minimum phase filter with a moderate amount of post-ringing. Polarity is maintained with the initial deviation positive (as specified in my test signal).

And here is the "digital filter composite" based on discussions with Juergen Reis and similar to what Stereophile has been doing over the last few years:

Overall that's a good "DFC" plot of the 16/44 test signals showing good suppression of imaging - about -100dB suppression of the ~40kHz harmonics compared to the ~20kHz cluster. Small amount of overload seen with the 0dBFS wideband white noise but practically should not be audible. Remember that this is the improved digital filter setting from the updated firmware as discussed in my previous preview post.

III. RightMark Tests

Now let's run the DAC through the usual RightMark suite of test signals and see if we can get some comparisons of frequency response, dynamic range and distortion amounts.

16/44:
As usual, we start with the "standard resolution" test signal.

As you can see, we have a good selection of devices here for comparison. There's the Sonica DAC on the far left, then Oppo's BDP-105 fed by USB, TEAC UD-501, Logitech Transporter, the little SMSL iDEA DAC, and finally Logitech's Touch. Even a cursory examination of the data suggests that there's very little difference between the devices at 16/44!

As you can see in graph form, the differences at 16/44 for modern devices are trivial! You can determine this for yourself by doing a volume controlled blind test.

Not unexpectedly, the device which performed the poorest in this round-up was the Logitech Touch which came out in 2010 using a middle-of-the-road "consumer level" AKM4420 DAC in a rather feature-loaded package.

24/96:
As we turn to high-resolution performance, this is where the "spread" between devices widens a bit, but to be honest, there's no "bad" DAC here so we can't expect to see anything terrible.

Clearly we see the ESS Sabre reference DACs in the Sonica DAC (ES9038Pro) and BDP-105 (ES9018) perform extremely well... Then again, the dual TI/Burr Brown PCM1795s in the TEAC UD-501 are excellent.

If we're super critical, we can see that the TEAC UD-501 has a couple of noise spikes at 120 and 180Hz. The Oppo BDP-105 is remarkably good except for a little bit more noise below 60Hz compared to the Sonica DAC. Though not clearly demonstrated in the busy noise level graph, the Sonica DAC is remarkably quiet with no evidence of hum at 60Hz. Excellent, and speaks to a very quiet power supply.


24/192:
Finally, let's have a look at the devices that can handle 192kHz sample rate.

Like the 96kHz test, we see a similar picture for the noise floor. A couple noise spikes in the TEAC sticking out, slightly higher low frequency noise floor with the BDP-105... Nothing of concern at all IMO. Very impressive how the SMSL iDEA mini-USB DAC hangs in there with the other big boys!

XLR vs. RCA:
IMO, if you really want to squeeze out the absolute best from today's DACs in high resolution, make sure to use balanced cables...

Unfortunately, the Focusrite Forte cannot handle the high peak voltage of 3.5V from the Sonica DAC. As a result, I had to use the digital volume control to pull back the amplitude by 4dB to get the following results (let's just look at 24/96):

As you can see, I've highlighted the -4dB adjustment I had to make to the Sonica DAC and -3dB with the TEAC UD-501. These compensations obviously would hamper the noise floor results. As impressive as the BDP-105's noise level and dynamic range are, if we added another 4dB back into the Sonica DAC's results, we would be expecting close to 120dB dynamic range or about 20-bits of resolution. This is the highest resolution I have ever seen within the limitations of the hobbyist equipment I'm using. Along with this, we see vanishingly low distortion results with the Sonica DAC. Impressive.

Remember, despite the improvements over regular RCA, these differences are extremely small and it would be ridiculous for me to claim that I can hear any difference. This is of course unless there's something else happening like noise polluting the RCA playback chain to widen the difference.

IV. DSD64 / DSD128 / DSD256

Quickly, here are the results from the RightMark 24/192 PCM test converted to DSD64 and DSD128 using Weiss Saracon. I've also included a realtime upconversion to DSD256 using JRiver 22. All measured through the RCA output.

Alas, I was not able to achieve stutter-free DSD512 realtime upconversion with my Intel i5-6500 based PC. (Honestly, I see very little benefit to DSD512 as there is no actual content out there plus DSD256 already would achieve quality way beyond human hearing limitations!)

Remember, with conversion there will be small losses as seen in the numerical data. Most notable is the higher noise level with DSD64. Of interest, notice the realtime conversion to DSD256 using JRiver has a slightly higher noise floor than the Weiss Saracon conversion but because it is DSD256, the noise-shaping ultrasonic noise is pushed beyond the frequency limits of a 192kHz sampling ADC.

V. S/PDIF & DLNA Input

Although these days I would highly recommend that audiophiles use USB 2.0 or ethernet, for those who want or need to use S/PDIF, here are the results. As noted in the Intro, this is with the Surface 3 sending USB to a CM6631A USB to S/PDIF coaxial or TosLink converter:

In the summary and graphs, we also see the results from using the Sonica DAC as a DLNA streamer over WiFi. I had set up my music server computer in the other room with JRiver 22 as DLNA server.

Based on the objective data, the evidence suggests that USB remains the preferred interface for this DAC. There's a reduction in the dynamic range with coaxial and TosLink input and slightly worse THD and IMD distortions as well. Note that these difference are extremely small and I would not worry about a sonic difference; it's just that because I know there is a difference, I would recommend sticking with USB preferably.

One interesting finding I noticed when putting together these graphs is that there's a small 60Hz hum noticeable with TosLink and WiFi DLNA playback in the "Noise Level" graph, but not with USB and coaxial input. This was a reproducible finding. Of course, given that the noise is down at -130dB, it's an academic difference rather than of practical concern. For comparison, with the Logitech Touch analogue output, the 60Hz hum is up at -114dB.

For those wondering, yes, the Sonica DAC can also handle DSD through DoP over DLNA up to DSD256 using JRiver.

Sadly, the Sonica DAC does not currently provide gapless playback as a DLNA streamer whether of PCM or DSD data. Admittedly this is a significant limitation and could be a "deal breaker" for many who are aiming to use this device as a stand-alone DLNA streamer. It's rather unpleasant listening to a live concert with breaks between tracks!

VI. Jitter

Remember the old firmware had issues as reported in my previous "Preview" post. This has since been fixed. Here's what the 16-bit and 24-bit J-Tests look like with the various inputs:

As expected, USB output looks very clean and jitter-free. Using my Surface 3 USB-to-S/PDIF CM6631A converter, likewise, we see that jitter is not an issue at all with coaxial and TosLink inputs. Fantastic reminder that jitter is just not worth really worrying about - not only would it be inaudible with most decent modern equipment, but with a device like the Sonica DAC, it's even objectively essentially non-existent.

If you're wondering, I had a look at DLNA jitter over WiFi and there was nothing unusual at all. For fun, here's JRiver 22 converting the J-Test to DSD256 DoP realtime, streaming to the Sonica DAC over WiFi!

That looks rather beautifully jitter-free! Despite realtime computer processing sent over a wireless connection... I've heard of subjectivist audiophiles getting anxious about this kind of thing worsening sound quality and especially jitter; maybe they should try some disciplined listening tests before uttering such nonsense :-).

To be clear, I would not advocate transferring DSD or any kind of hi-res audio through WiFi unless you're absolutely certain that connection strength is reliable. Dropouts from buffer underruns don't sound good.

VII. AUX Input

Finally, I just wanted to test the quality of the AUX Audio Input for when you have an analogue signal you want to play back through the DAC acting as a preamp.

What I did was:
TEAC UD-501 DAC playing RightMark output --> 6' RCA --> Oppo Sonica DAC AUX input --> 6' RCA --> Focusrite Forte --> 6' USB --> Win 10 laptop

I can then compare the direct output from the TEAC UD-501 with what comes out from the Sonic DAC having been fed through the AUX input. I kept the amplitude at 100% to maintain full signal strength and least potential signal processing:

Hmmm. Not so good. The AUX In is able to pass through just slightly higher than 16-bit resolution. It looks like at best the signal is resampled to 48kHz. Obviously higher distortion and crosstalk. According to this review on Secrets of Home Theater & High Fidelity, the AUX In is resampled with the TI PCM1808 ADC at 24/48. No wonder I'm seeing these results... BTW, that review shows measurements from the old firmware with notable jitter anomalies; I left a comment about this and the moderator deleted my message!

Bottom line is simple. If you want to maintain the highest quality output from your device, best not to pass it through the AUX In and go straight to a higher quality pre-amp.

VIII. Conclusions

There you have it everyone, I hope a reasonably thorough look at what the Oppo Sonica DAC can do for you through a more objective lens... Remember that subjectively I thought this device sounded excellent as I had previously discussed. Over the weeks, I've listened to more albums through this device and also utilized room-correction DSP to improve the sound in my room. Absolutely top-notch levels of sonic fidelity.

Also, check out the manual online to have a closer look at the operations if you have further queries.

In summary then:
1. This is a very high performance stereo line-level DAC. No headphone output. No extra fancy processing provided; look elsewhere for MQA for example. And no selection for digital filters (at least as of this current firmware).

2. For best performance, use XLR output. This generally goes for all of the better DACs with the option of RCA vs. XLR. Balanced output allows for improved noise rejection especially good if you have the need for longer cables.

3. Fantastic performance especially highlighted by the USB 2.0 interface. Very impressive low noise floor, microscopic levels of distortion, irrelevant amount of jitter (no need for advertising hype claiming "femtoclocks" and the like). This is of course what we expect of the ES9038Pro DAC - the current "top of the line" Sabre DAC from ESS.

4. S/PDIF performance was good, although objectively the output was bested by USB and DLNA network streaming. I found the level of jitter rejection through the S/PDIF to be excellent with this device.

5. Yes, by all means, enjoy upsampling to PCM 768kHz and DSD512 with software like HQPlayer, JRiver, and Audirvana+. Unlikely we'll ever see actual native music at such extreme bitrates (storage-wasting and absolutely unnecessary for human consumption). Remember that you'll need quite a fast CPU or GPU when converting/upsampling to DSD512 in realtime to avoid dropouts!

6. I found the DLNA streaming feature worked well. Logging on to my network with the initial set-up was easy with the Sonica App (I used the Android version). Streaming over DLNA from JRiver 22, I had no troubles at all. The device was easily found on my network, responsive, and sounded great. Spotify and TIDAL streaming are supported through the app so you can play music without needing a server computer running on your home network.

7. Don't use the AUX In for high quality sources. The 24/48 digital resampling for the AUX Input likely will diminish the quality of whatever device you're connecting. Not ideal but OK if all you're doing is connecting a "standard resolution" device like a CD or MP3 player. Maybe next time, they can use a higher quality 24/96 ADC.

8. Wish list:

- GAPLESS PLAYBACK - this is the big one. It's 2017, I've been doing gapless over the Squeezebox / Logitech Media Server system for years. A real shame that the DLNA streaming feature cannot handle this. The workaround of course is to do something like stream to a Raspberry Pi connected to this DAC through USB.

- Selectable digital filters. Not essential but I think this would be a nice addition since the ES9038Pro is touted as having 7 settings in PCM to choose from. Having this available also provides at least the sense of a more complete high-performance package for the audio enthusiast since we are missing other common features like headphone out. Also, I wonder if there's any way to turn off the Sabre's filter to run in NOS mode for those who might want to do this in conjunction with software upsampling? Remember that the recently released Oppo UDP-205 is even capable of selecting filters, and that's a video player!

- More levels of OLED dimming. Currently there are only 3 levels - "High" (full), "Dim" (maybe around 50%), and "Off" (turns on for ~5 seconds with sample rate changes or starting playback, then turns off). Surely we can do a few more levels than just the single "Dim"?

- A way to turn ON the device with Sonica App? At this time you can use the App to turn the device off / put to sleep but there's no way to turn it on other than pressing the front power button. I don't know if the hardware will allow it, but this would be useful for convenience.

- Support for other streaming options? How about Roon RAAT, HQPlayer NAA, or even good 'ol Squeezebox SlimProto TCP? These protocols would also allow gapless playback and presumably more seamless PCM/DSD transmission with RAAT and NAA. Probably unlikely to happen with this device but having options like these incorporated into future network-aware DACs seem like a worthwhile move.

Although not discussed today, Bluetooth 4.1 streaming works well although it is lossy compressed audio sent through my Samsung Galaxy Note 5 phone. It might be interesting showing the results of the RightMark test over Bluetooth another time.

The question then is: "Are we there yet?" in regards to high-fidelity playback of digital audio such that there's really nothing more we can ask of our desire for pristine playback quality. I would say yes. In fact, given the limitations of human hearing, we've been "there" for years now.

As I said a few weeks back, if we think about audio reproduction rationally, there does come a time when we have achieved "high fidelity" to a level where it really is "good enough". In terms of technical accuracy, I don't believe you can really beat the quality of the Sonica DAC using USB playback these days.

So if we're "there" in terms of sound quality with DACs, what's next? The obvious answer to that is features. Sure, maybe the next generation of ESS Sabre DAC chips in a few years could be smaller, use less energy, cheaper, and even provide a few more dB's of dynamic range, but IMO we've reached the end of the road when it comes to any "need" for technical accuracy. I think consumers are already expecting features like network connectedness, filter choice selection, perhaps MQA for some, DLNA/Roon compatibility, Bluetooth connectivity, gapless playback... The Sonica DAC can do some of this already but missing others as noted in my "wish list". All these features will become the most important elements of the "complete package" to differentiate devices as audiophiles recognize that they can still choose between "different sounding" gear, but technically "better sounding" has clearly reached a point of grossly diminished returns (this can be easily proven with blind testing high quality DACs).

No doubt, the Sonica DAC is a well constructed device offering reference quality sound in 2017 at an affordable price of ~US$800. Though I think there are a few things Oppo could have done to complete the feature set, and present a more thorough enthusiast-oriented package, but at this price point, I have yet to see any competitors challenge with similar ES9038Pro-based devices.

RacingPHT wanted to have a look at the AP measurements from Oppo for the Sonica DAC - here they are. Enjoy...

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

The other week, I went with a buddy to visit Commercial Electronics here in Vancouver. Specifically, I wanted to see first hand and listen to these babies:

Those are of course the BeoLab 90 speakers referring to the 90th anniversary of B&O in 2015. They are an interesting integration ("marriage" perhaps) of modern audio technologies: multiple small but powerful Class D integrated amplifiers (touted as 8200W total), multiple DACs, digital crossovers and DSP processing for room compensation. Check out the detailed whitepaper PDF to read more.

How do they sound? Well, there's an impressive ability to adjust DSP parameters like the "beam control" to change apparent width of sound dispersion and define direction of the "sweet spot". I see it's hooked up with gigabit ethernet to the store's music server system:

Check that out... No fancy power cable needed!

I was told it can handle up to 24/192 input. I listened to a few "classic" audiophile tracks like Steely Dan's "Babylon Sisters", tracks off Beyoncé's Lemonade, a few 2L samples, yes, even Diana Krall and The Eagles :-). Unfortunately I didn't see a CD player connected to these speakers to play my own CD I brought along.

Very clear sound. Good definition and nice "sparkle" (I can see why some have commented on these sounding a bit "bright"). On the "wide" setting, the soundstage is appropriately broad, and there's an "omni" mode as well. As usual, it's difficult to evaluate sound without a familiar, better, and quieter sound room. What I heard certainly had great potential. At an asking price of >US$80k (I was quoted a price of CAD$115k here in Canada), this of course better sound decent! :-)

I had seen pictures of the BeoLab 90 on the internet over the last couple of years and wasn't all that impressed. However, in real life I actually thought they looked OK if not "interesting"... They were not as tall as I had originally imagined. Obviously one would have to consider how they would look in the context of the rest of one's home decor!

Have a great week ahead folks and hope you're all enjoying the music!