MEASUREMENTS: Oppo UDP-205 DSD Playback (DSD64-DSD512), and PCM 768kHz. Plus CES2019 coverage thoughts...

Alright guys and gals, it has been months since I completed the set of measurements of the Oppo UDP-205 last year (here, here, and here).

In early December, David M wondered how the Oppo performed as a DSD player as I had neglected to measure that.

Over the years, I have measured DSD output performance but remember that this is a little bit of a pain :-). To obtain some results for comparison with PCM in RightMark, what I typically do is take the test signals (originating in 24-bit PCM), convert to DSD using software like Weiss Saracon, and then play back the DSD file through the DAC into the ADC which of course takes that analogue output resampled back into PCM for analysis. Doing this understandably adds other variables to the measurement system which should still be minuscule. Over the years, I have looked at things like PCM-to-DSD converters (here and here) to demonstrate that the conversion programs do have an effect as one would expect with different resampling and modulation algorithms.

You can look over the previous results I obtained with DSD playback of devices like the ASUS Essence One (with DSD Kit), TEAC UD-501, PonoPlayer, Oppo BDP-105, and Oppo Sonica DAC for reference. Note that over the years, I have changed the measurement ADC device and technique a bit so the numbers will vary but I believe comparisons can still be made. Also, I've always compared the DSD with direct PCM output side-by-side so that in itself would provide a frame of reference.

For today's measurements, the testing set-up looks like this:

Intel i7-3770K computer → 16' generic USB cable → Oppo UDP-205 → generic 6' XLR balanced cable → RME ADI-2 Pro FS ADC → generic USB → Windows 10 measurement laptop

Notice that I'll be using the reasonably powerful Intel i7-3770K computer described recently to perform playback duties. As you can see, I've done nothing all that special with the machine, and we're looking at a 16' AmazonBasics USB 2.0 cable connecting the computer to the Oppo. The Oppo UDP-205 was updated with the latest "public beta" December 2018 firmware, RME ADI-2 Pro likewise updated to recent version 185/90.

Remember that the UDP-205 is capable of up to DSD512 (and up to 768kHz PCM). Alas, I could only do off-line PCM-to-DSD conversion with Weiss Saracon 1.61 up to DSD128. Using JRiver 24 and TASCAM Hi-Res Editor, I can convert to DSD256. I don't know if there are programs that will do offline conversion to those huge DSD512 files, so I used JRiver's "DSP Studio" to convert PCM to DSD512 in realtime on the Intel i7-3770K CPU which worked well for the measurements today.

For playback, I'll be use JRiver 24.0.72 x64 with Oppo's native ASIO drivers and "Bitstreaming" set to DSD like this:

No DSP Studio setting used of course for the pre-encoded DSD files. For realtime PCM-to-DSD512, all I did was turn on the "Output Encoding" to 8xDSD:

I. RightMark Results:

Here are the results taking a 24/96 test signal and converting it to DSD64/128/256 or realtime conversion in JRiver to DSD512, played back on the Oppo UDP-205 over the XLR balanced output:

To the left are results from the unaltered PCM test signal. As typical for the Oppo UDP-205, the results are excellent with noise level and dynamic range around 20-bits of resolution and remarkably low distortion.

To the right of the PCM column are the various DSD test results using the test signal converted to DSD. The RightMark test only calculates the result within the audible spectrum (20Hz-20kHz). Notice what's interesting about the noise level is that there is actually a relative increase in noise as we increased the DSD samplerate. There is particularly an increase in the noise floor at DSD256, then the noise level decreases slightly into DSD512.

Notice that the higher noise level DSD256 results were found with both off-line conversions using JRiver 24 and TASCAM Hi-Res Editor. JRiver looks to be slightly better than the TASCAM for distortion results. Clearly the difference is so small that I cannot imagine this would be audible.

In order to appreciate the ultrasonic noise and see the relative noise floor differences, here are my usual composite graphs:

DSD512 JRiver RT refers to "realtime" / on-the-fly conversion of PCM to DSD512 using JRiver 24 and DSP Studio.

What's very obvious is the increasing ultrasonic noise level with DSD64 rising rapidly from around 25kHz. Remember that this is typical of DSD64 where high-order noise-shaped modulation is needed to keep the noise floor low within the audible frequencies. The JRiver conversion slightly rolls off the high frequency response compared to the TASCAM DSD256 and Saracon DSD64/128 conversions.

As you can see, the noise level within the audible frequencies, particularly at DSD256 was the highest. Though not shown here, testing using other software (including HQPlayer which we might talk about another time), the higher noise level at DSD256 was consistently reproducible on the Oppo.

To have a better look beyond 48kHz, let's have a peek using a 24/384 PCM test signal:

Again, remember that these numerical results are reflective of the 20Hz-20kHz frequencies.

We can see the difference in frequency response and noise floor going from DSD64 to 128 to 256 and 512 quite well here:

Notice particularly that DSD64 allows for a relatively flat noise floor until about 25kHz (as shown with the 24/96 test above as well). DSD128 is good to about twice that - over 40kHz. With the 24/384 signal, both DSD64 and DSD128 achieved about the same noise level within the audible spectrum.

II. Jitter anyone?

For completeness, let's have a peek at the J-Test results with DSD. Remember that the J-Test was not constructed for DSD, it's at least a convenient comparison to look for anomalies using a strong primary signal. Here are the 16-bit and 24-bit J-Test results at various DSD samplerates:

No worries. Looks pretty good across the board. Over the years I have not seen any issue with jitter and DSD playback (just like there are no issues with asynchronous USB playback of PCM in general). Notice the increase in the noise floor with DSD256 is particularly noticeable. The realtime conversion of PCM to DSD512 using JRiver 24 still looks good although notice a little burst of noise just below 14kHz in the 16-bit J-Test and a single pair of sidebands about +/- 1kHz on either side of the 12kHz primary signal in the 24-bit test but way down below -120dBFS.

III. Quick look at PCM 768kHz...

For completeness, since we're measuring all the way to DSD512, let's just have a peek at PCM at the UDP-205's maximum 768kHz samplerate. Note that this is a bit complicated because I really don't have any software that handles 768kHz other than again realtime upsampling (with JRiver 24 or HQPlayer). So that's basically what I did... RightMark can measure up to 24/384, so taking that test signal, I had JRiver's DSP Studio upsample to 24/768 for playback on the UDP-205, and measured the result on the RME ADI-2 Pro FS in 24/384:

Not bad at all. A very slight increase in noise floor by doubling the sample rate. A possibility is that JRiver's upsampling algorithm may have added a little noise. Distortion results essentially no different between the two sample rates. The slight difference in stereo crosstalk is inconsequential in real-world audio reproduction.

Here are the graphs:

And here's the 16-bit and 24-bit J-Tests when upsampling to 768kHz in realtime with the Windows 10, Intel i7-3770K machine using JRiver 24:

No problem.

IV. Conclusions...

There you are, some measurements of the Oppo UDP-205 as DSD playback machine and ultra-high 768kHz PCM sample rate ("2xDXD").

Bottom line: subjectively, DSD sounds great and objectively I have no concerns. Remember that while we can use these tests to verify that there are no major issues, this is mostly academic but can be useful for future comparisons if we look at other devices that can handle these speeds... Just know that DSD playback on the Oppo UDP-205 is fine. :-)

Considering that 99% of all the DSD music I have is DSD64 (2.8224MHz, typically SACD sourced), the results are excellent with minimal distortion and we can see the usual high frequency noise from noise shaping being utilized beyond ~25kHz. Personally, I still prefer 24/96 PCM as a better system with more playback compatibility and opportunities for DSP processing.

I have less than a handful of DSD128 (5.6MHz, or "Double-DSD") material and these sound great on playback. I have 1 album in DSD256 (11.29MHz, or "Quad-DSD") which is basically a demo/sampler. Again this one album sounds great despite the objectively measurable higher noise level with DSD256 playback. To be honest, I have zero interest in owning any DSD512 (22.58MHz, "Octuple-DSD") material in my music library - it just doesn't make any sense wasting so much storage space!

Considering how many years higher samplerate DSD playback ability has been around (many DACs have been able to handle DSD128+ since 2013), I think the paucity of DSD128/256 material is telling.

While there could be measurable differences because of the encoding software used (Weiss Saracon, JRiver, TASCAM), the interesting finding in these measurements I think is that DSD256 has higher noise level in the audible spectrum than DSD64 or DSD128 with the Oppo; about +6-9dB higher. Though different SDM (Sigma-Delta Modulation) algorithms will be applied in the PCM-to-DSD conversion depending on the software (ie. Saracon is using its default 8th order cascaded resonators with feedback loop [CRFB] setting), I believe the resulting higher noise level that I measured is consistent and independent of otherwise accurate conversion algorithms. Regardless, we are still looking at an overall noise floor in the audible spectrum below -105dB which is still superior to 16-bit CD resolution.

I did a similar test in mid-2017 with the Oppo Sonica DAC up to DSD256. Back then, I already saw that the DSD256 noise level was higher as well (at about the same -105dB level). Remember that the Sonica DAC also uses the ESS ES9038Pro DAC chip so this finding appears consistent between the two machines.

As usual, no worries about jitter... At.All.

Remember folks, let's not get crazy with the tech specs like claiming that DSD512 is "better" than DSD128 or even that DSD1024 (what is that - "Sexdectuple-DSD"?) somehow has magical sonic properties. Taking the above comments into account, I would argue that the "sweet spot" is at DSD128; less data wastage than DSD256+, excellent low noise level in the audible frequencies, and ultrasonic noise pushed beyond 40kHz. Assuming one's amplifier doesn't have issues dealing with the >40kHz noise ramp (up to -70dBFS using Saracon), DSD128 is more than good enough for human consumption. About equivalent to PCM 24/96 which as far as I am aware, surpasses human hearing limitations.

After having listened and tested this stuff myself over the years (BTW, this is the first time I've published DSD512 and PCM 768kHz results for any device), I find there's little to get excited about despite some folks seeming to be ecstatic about bigger and bigger sample rate numbers... Fun to play with upsampling to DSD256+ if you have the CPU capability (or ability to offload to GPU if you use HQPlayer) of course, but let's not get too giddy :-).

As I discussed back in 2013, my perception is that the higher one goes with DSD, the cleaner the signal with less ultrasonic noise one gets, and the more it just sounds like PCM! If I want to hear DSD's "euphonic" effect often described as being "more like analogue", I think sticking with DSD64 might be more pleasing. Once we get to DSD256 and DSD512, the signal just looks/sounds clean like PCM. You might not have heard this perspective from DSD advocates who seem to think "bigger is better". Let's zoom down to the level of the waveforms to illustrate what happens as we increase the DSD samplerate:

PCM vs. DSD waveforms from Oppo UDP-205. Analogue output from DAC captured at 24/384 with RME ADI-2 Pro FS. The PCM is played back with good reconstruction filter of course.

See what I mean? DSD64 with its higher noise level could simply sound subjectively "more analogue" because of the distortion introduced into the signal and the potential effects this may have going through one's audio system. YMMV. As usual, if you have a DAC capable of high DSD sample rates, go have a listen and experiment for yourself.

For completeness, PCM 768kHz playback measured and sounded just fine off the Oppo UDP-205. Other than folks wanting to do some upsampling with their own filters, there's really no point to go to such extremes. IMO, it's ludicrous to seriously believe that a consumer would have use for music at 384kHz (DXD) or 768kHz sample rates; that would be a colossal waste of storage with no audible benefit at all. While I might have faith that perhaps somebody, somewhere can benefit from 24/96 (hence I don't mind keeping some hi-res 24/96 in my archives), the idea of going beyond even 192kHz and perceiving benefits I think is rather fanciful.

One final note about the JRiver 24 realtime conversion using my i7-3770K machine. I was rather impressed that the program is quite efficient even converting 24/96 PCM to DSD512. CPU utilization was typically <20%. Not bad. Here's a shot of Windows 10's CPU utilization performing playback of 24/96 to DSD512 in realtime:

One observation I made while doing these DSD512 measurements is that on my machine based on an older ASrock Z77 Extreme4 motherboard circa 2012, the USB 2.0 ports were better than the USB 3.0 ports for connecting to the Oppo UDP-205. I could hear and measure subtle errors during realtime PCM --> DSD512 playback in JRiver with the USB 3.0 ports which were absent when plugged into the USB 2.0. Realtime PCM --> DSD256 playback was no problem with either the USB2 or USB3 ports. Also, you might need to increase the JRiver or ASIO buffer size for DSD512. Errors could be a combination of the USB hardware, device drivers or the CPU load... Which is why it would have been nice to do off-line conversion to DSD512 instead of realtime processing and see if these were primarily USB throughput issues or the CPU just not able to keep up with the realtime demands every once awhile.

Remember that sometimes episodic playback errors can be due to latency issues, so make sure to double check with LatencyMon if you are having troubles. Do not use the old DPC Latency Checker which is only good up to Windows 7 (Windows 10 will just give you a constant stream of 1000μs). My machine was fine:

Interrupt-to-process latency over 9 minutes while surfing the web. Occasional bursts up to few hundred μs but typically <100μs.

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

Well, another CES (2019) has come and gone.

As noted over the years, not just here but elsewhere, CES is a place for companies to show off innovation and drum up excitement in new technologies. I think traditional 2-channel "high end" audio has little to offer in general in this regard given the technological maturity. As much as new products can be exciting to hear about, it's also what has "disappeared" from the show that's just as telling in terms of the direction the wind is blowing... Here are a few observations.

1. The "Hi-Res Audio" dog and pony show apparently is a "no show" this year. Remember that in CES2018, it looked like a number of companies collaborated on the "Hi-Res Audio Pavilion" and in CES2017 there was likewise all kinds of hub bub around hi-res following the excitement of sorts since 2015 when the PonoPlayer made its debut at the CES, supposedly ushering in the high definition audio movement.

Perhaps the music industry is finally realizing that hi-res offers little for consumers because a.) bigger containers in and of themselves do not sound "better", and b.) the Industry has not been serious about producing better masters to provide a better sounding hi-res version of their music that can potentially benefit from the technology. Of course all this is old information for readers here and have been spoken of for years by people like Mark Waldrep.

To make matters worse, the audiophile press isn't doing anybody any favours by not being critical of the nonsense, and even adding to the empty rhetoric. For example, consider this recent TAS article on the "hi-res" Hamilton Broadway cast recording. As one of the commenters noted, this is a dynamically compressed DR9 recording - silly to think that noise levels and "grain" have improved by going to 24-bits if one actually listened to this in a proper comparison!

Isn't it remarkable that in the video/TV world, the idea of "high dynamic range" which we talked about two weeks ago is literally a "must have" these days on high-end TVs and everyone is out to show off the benefits this can bring? Yet, the audio world doesn't seem to recognize that this might be somewhat important for "Hi-Res Audio"?

Hey music industry: give the consumers well produced and mastered albums first with low noise levels and higher dynamic range, then consider if the album is worthy of a larger, high-resolution "container". What we need are the production sensibilities of music from the late-80's/early-90's using modern high-res digital techniques. Unless this is done, "hi-res" is essentially worthless. IMO, Mr. Market's already making the call on the Industry's bluff.

2. Cause for some celebration is the impending arrival of Qobuz into the lossless streaming marketplace. While I'm not a streaming service kind of guy at the moment, competition is good! The "upgrade" from Spotify, Amazon Music, and Apple Music is primarily having lossless full 16/44.1 streaming - not "hi-res". I believe Qobuz is more catered to classical and jazz recordings which are the appropriate genres for focusing on albums which can potentially benefit from high-resolution playback. It's good to see that Qobuz is streaming real lossless hi-res streams in FLAC rather than that partially-lossy-cryptographically-marked-pseudo-high-resolution system called MQA championed by Tidal "Master".

Though it's only January, I nominate J.V. Serinus for the "Greatest Shill of 2019 Award" for his work on this article. Particularly this fantastic line:

"While Tidal's potential rival, the French company Qobuz, is currently beta-testing its own hi-rez streaming platform for Android, iOS, Apple, and Windows devices, prior to its official launch in the US, Qobuz is limited to 24/192 resolution. MQA's folding and unfolding technology, as seen on the Android phone held by MQA's Ken Forsythe, enables Tidal to stream Masters titles of up to 24/352.8 or 384 resolution."

Bravo! Now that's surely some top-shelf "shilling" from Mr. Serinus. Amazing supporting role to Mr. Forsythe as well. (Ahem... A little worried about Qobuz splitting the few subscribers who care about high-res streaming and dumping MQA? How many serious, knowledgeable audiophiles still care about "big numbers" like 24/352.8 as if that means anything!?)

But seriously, dear readers, I obviously do not need to explain the issues in that quote to you. What's so disturbing is the obvious underestimation of the audiophile consumer's intellect from both Mr. Serinus and Forsythe to publish such words as if the shallowness isn't patently obvious! Then again, perhaps this is sadly all the depth the audiophile press can muster...

Speaking of inflated numbers, it looks like Tidal (only place to stream MQA) is being investigated for inaccurate streaming figures (original complaint back in earlier 2018). Of course, Tidal has been under suspicion of inflating numbers for awhile, including the number of subscribers back in 2017.

As a reminder to how important streaming is these days, consider that this week's Billboard #1 album only sold 823 "traditional copies" while credited with 83 million streams!

3. As usual, questionable products are still trotted out at CES. AudioQuest with yet more "Mythical" speaker cables (starting at a mere US$4500 and going much higher)? Regardless of one's stance on expensive cables, I suspect that everyone can agree on the name "Zero Technology". "Rubber-cork-rubber" innovative "vibration control" devices you say? Truly "increcable"! Of course the company needs a slogan/philosophy like "Trust what you hear... your perceptions are unique to you". Mr. Serinus apparently was impressed in any event.

4. The NAD Master M10 does look like a nice device. Power-packed with Dirac DSP, HDMI connectivity, gigabit ethernet, wireless, AptX Bluetooth, 100W nCore amp, flexible DAC, etc... That's what a modern space-constrained "audiophile" device looks like I think. Alas, it is missing a USB-B input for a computer/external streamer to plug into unfortunately. Also, it would be nice if BluOS could stream as a UPnP/DLNA endpoint device (doesn't look like the company is interested in implementing this) for flexibility. Of course, the inclusion of MQA by NAD is forgivable :-). It just ticks off another check box that some audiophiles might be looking for.

Otherwise, not much else new or different in audiophilia. I see for example that Technics has updated their venerable and familiar line of turntables with the SL1200/SL1210MK7 at US$1200 coming out in the summer. I do wonder if the recent rise of analogue audio (vinyl and turntables) might be reaching some kind of plateau or steady state as CD sales have already declined dramatically, digital downloads having fallen, and streaming has its place in the order of things.

5. For me, the most interesting reasonably priced audio gadget I might grab - Audio-Technica's ATH-M50xBT (~US$200). Well, can't say I "need" yet another pair of headphones :-). But if I wanted to, I'd certainly consider grabbing these. The wireless Bluetooth headphone market is maturing in the mainstream and the choices are starting to blossom. Nice to see the inclusion of both aptX and AAC codecs with these headphones - I wonder why not just good 'ol MP3!? Since I'm in no hurry, I will wait and see the feedback on these.

6. Industry pushing for 8K TVs now? Moving away from audio for a bit, remember how I said in December that in time, I believe other technologies will be sold to consumers based on subjective characteristics like how the audiophile world has become once the core function of achieving fidelity to satisfy human perception matured. I believe it begins when the specifications start exceeding the "need" for the consumer. At some point, the "big numbers", whether it's how many "K" screen pixel resolution, or hundreds of kHz PCM, or tens of MHz DSD audio will no longer meaningfully matter. When that day comes, companies will have to differentiate themselves based on features or appeal to "non-utilitarian" factors to garner interest which includes the psychological "joys" of ownership of luxury products. My belief is that 4K will be with us for a long time for 2D video. Beyond 4K/2160P, while there can be objective benefits, it would require rather special situations for the benefits to be meaningful for consumers.

Remember in my examination of 1080P Blu-Ray vs. UHD Blu-Ray about a year back, unless movies truly originated from high-resolution sources (70mm film, hi-res digital captures), and are actually being transferred to 4K digital intermediates, you might as well get the 1080P. While some will disagree, IMO, forget 35mm film transfers (for example here and here) as worthwhile material for 4K resolution other than to archive the material. There's only so much definition needed for film grain :-). Thank goodness that at least with video, they had the good sense of promoting high dynamic range as discussed above with better contrasts, improved gradients, and a wider color gamut which ultimately can still make a 35mm transfer worthwhile on the 4K disk even if resolution is not the main draw.

It is good to see that the technology enthusiast / "geek" reporters are for the most part realistic and articulate well the unlikely need for 8K TV resolutions any time soon. Great job! I think it would be nice for the audiophile press to learn to be a little more critical. It's important to have honest reporting to confront the hype that inevitably comes from the Industry's advertising departments. Oooohhh, look, there's now an 8K Association to promote this technology! Perhaps the video equivalent to the CEA/CTA for Hi-Res Audio?

Finally, looking over the CES 2019 "best of" posts, I think the Savvy Smart Mirror looks like a cool addition that will show up in fancy homes in the next decade. Imagine brushing teeth, flossing and shaving while checking out the latest news. Assuming it doesn't record sounds and video of what happens in the bathroom, I'll take one! :-)

Enjoy the tech everyone!