Revisiting the TEAC UD-501 DAC (2013): THD(+N), DSD output, Sweeps, Jitter, 1/10 Decade Multitone 32. (And on SSD transition, Corsair MX500 SSD, and Hi-Fi+ closes comments...)

After Bennet/Dtmer Hk's comment on this previous post, I thought it might be interesting to revisit my longtime friend, the TEAC UD-501 DAC that I bought new when it came out in 2013. My preview and measurements (Part 2 PCM, Part 3 DSD) are still online of course.

I still remember doing those measurements in my previous home - it feels so long ago! ;-)

These days with the E1DA Cosmos ADC and RME ADI-2 Pro FS available for measurements, let's take a trip down memory lane at what has been over the years a reference DAC for me. I think the TEAC UD-501 is a special device that ushered in for many audiophiles a cost-effective, high resolution, well-built, reliable, low jitter, asynchronous USB audio interface which I suspect has influenced the design and performance of other DACs over the years.

Spec-wise, it was one of the first to catch my interest with the ability to handle PCM 384kHz and DSD128. The high quality metal case looks serious. It's based on a dual-mono TI/Burr-Brown PCM1795 configuration with dual toroidal linear power supplies. There are a number of PCM and DSD filters including the ability to turn off the 8x oversampling and allow "NOS" (Non-OverSampling) mode. The quad JRC MUSES8920 opamps they used were also another "feature" talking point back then. (Quite a complete specs page here for the TEAC.)

So, first let's look at what I'm getting with the 1kHz tone for THD(+N), XLR out, from the DAC fed by my Raspberry Pi 4 "Touch" streamer into the E1DA Cosmos ADC:

At 0dBFS, XLR output is measured at 4.0Vrms. Unweighted and directly into the ADC without notching of the fundamental, I'm seeing a THD+N of -108dB on account of the relatively strong 2nd and 3rd harmonics of almost equal amplitude. When I attenuate the 1kHz signal to -3dBFS, the THD+N improves to -111dB.

Every ADC is a bit different and each measurement will also vary unless tightly calibrated. So let's do the same test with the RME ADI-2 Pro FS. Again, let's use the same Raspberry Pi 4 streamer, and with mono "M/S" processing as well:

The resemblances when using the E1DA Cosmos ADC and RME ADI-2 Pro FS are not hard to see. There is a pair of low-level sidebands around the 1kHz signal. We can zoom in for a good look at those (captured here with the Cosmos ADC):

Exactly +/-120Hz sidebands, of low amplitude below -130dBFS. This is likely related to the 60Hz mains hum originating from the linear power supply.

In each of the measurements above, the 2nd and 3rd harmonics predominate and this is the main limiting factor with the measured THD. As for the higher order harmonics, at 0dBFS, the 9th declares itself quite strongly, and at -3dFS, the 7th and 5th harmonics become more prominent.

Likely due to the relatively better time-domain performance of the Cosmos ADC, we can see that the "skirt" at the base of that 1kHz tone is narrower than the RME's.

At 0dBFS, we're seeing THD+N around -108dB using both ADCs. Lowering the output level to -3dBFS improves the THD but THD+N improves a bit more with the E1DA Cosmos ADC as it maintains a lower noise floor.

In the last few weeks, we've been looking at DSD a bit more. Let's have a peek at the -0.1dBFS 1kHz tone encoded in DSD (with SoX-DSD SDM-8 setting). While there are some changes with the different analogue FIR options (1-4), let's just use FIR1 - both DSD64 and DSD128 FFTs:

Note that the output level is lower in DSD by about -6dB using this FIR filter, this varies depends on which setting is chosen. There's a bit more 2nd harmonic with DSD128, dropping the THD(+N). Noise level looks about the same though which is good. The "N" is measured at -109.5dB with DSD64 and -109.0dB with DSD128.

As you can see, peak ultrasonic noise with DSD64 is up at -60dB, 70kHz. This is better with DSD128, noise going up to -66dB at 135kHz; I trust this would be outside of the bandwidth for the majority of amplifiers. The pattern of the noise shaped ultrasonic content with FIR1 is actually quite similar to the RME with "DSD Direct" and a weak filter.

For comparisons, let's look at the RightMark results at 16/44.1 along with the RME ADI-2 Pro FS R Black Edition, and Topping D90SE (best performing PCM DAC at this point in history, 4V default setting):

Filter for each DAC set for "sharp" or equivalent on particular model.

As usual, there's not much to see here at "standard" CD-resolution. Even back in 2013, the TEAC UD-501 performed well and 16/44.1 can be reproduced with no difficulties. This is to be expected since this DAC is capable of better-than-CD resolution.

I think it's not unreasonable to argue that if all you're listening to is bit-perfect CD-resolution material, don't care to perform DSP or apply volume normalization in 24/32-bits, and already have a decent DAC, then there is probably no point spending more money on upgrades.

The more challenging job is reproducing hi-res audio, and this is where objectively, DACs will show more variability. Here's the TEAC's 24/96 results compared with the Topping D90SE, the summary:

The TEAC DAC can play up to DSD128 (of course these days many DACs can go to DSD512 or even DSD1024) so you see that I've included a comparison at DSD64 and DSD128 with the Topping D90SE as well. Clearly, the Topping is significantly more resolving than the old TEAC with PCM 24/96 playback. The difference is alas much less obvious with DSD playback. While the Topping's ESS ES9038 Pro chip can do a great job with minimizing harmonic distortion, the DSD intermodulation results and noise level are actually not any better than the TEAC. As I said previously, I don't think the current ESS chips are great for DSD playback even though they perform very cleanly with PCM data.

Here is the usual collection of graphs for comparison:

Let's now have a look at the J-Test using the very low jitter E1DA Cosmos ADC:

While not "perfect" (compare this to the Topping D90SE for example), there are some low-level (-130 to -140dB) sidebands likely due to 60Hz-multiple power-related spuriae and other low-level random noise. Notice that all the noise anomalies are below the 16th-bit LSB jitter modulation pulse in the 16-bit J-Test. BTW, I love how well the Cosmos ADC performs capturing the J-Test - nice, flat, very low noise floor (I think even better than the APx555)!

As I have said for years, jitter has been a selling point used by advertisers, companies, and audiophile media mouthpieces, a perennial "bogeyman" but really has not been a problem for years - especially with asynchronous USB. (Not that jitter is all that audible even when high.)

We can use REW for frequency and generator level sweeps:

Above are some frequency sweeps from 20Hz-20kHz at 0/-3/-6/-12dBFS. The main take-home message is that 3rd and 2nd harmonics are the strongest at the upper output levels. As we get lower in level, the odd 3/5/7/9th harmonics become relatively higher. Overall, the relative harmonic amounts remains predictably constant through the audible spectrum at each output level which is good.

And here is the Harmonic Distortion vs. Generator Level graph:

While there's a bit of fluctuation between -25 to -15dBFS, generally the 3rd harmonic is most prevalent across the output levels when I measured the DAC's L channel in stereo mode on the Cosmos ADC. Lowest THD+N appears to be around -4dBFS.

We can look at linearity of the output using that data above:

Excellent, with <0.1dB error all the way to -115dB (19-bits).

Finally, a look at the "1/10 Decade Multitone 32" PCM 24/96 on the TEAC UD-501:

I've put the cursor at -116dB, approximately the level above which the FFT is free from any distortions/noise. Likewise we can look at DSD64 playback encoded with SoX-DSD:

Basically PCM and DSD64 are quite similar and I'm being picky with where I'm putting "the line". Both PCM 24/96 and DSD64 versions have about the same 19-bit distortion/noise-free range of around 115dB or better. On the DSD64 version, we can see the noise creeping up on the right side as expected. While not shown, DSD128 would be similar but without the ultrasonic rise in noise.

There she is on the test bench - TEAC UD-501... Cosmos ADC on top, RME ADI-2 Pro FS behind.

Summary & Comments...

The results here look consistent with expectations for a DAC that's approaching a decade old now with relatively comparable PCM and DSD abilities. The TI/BB PCM1795 DAC datasheet quotes a dynamic range of 123dB and THD+N of 0.0005% (-106dB). My results show 120dB dynamic range and THD+N of 0.00038% (-108dB) at 0dBFS which improves to better-than THD+N -110dB if we back off the amplitude a few dBs down to -3dBFS. This looks to be within or exceed the datasheet expectations; perhaps we're looking at the dual-mono (2 x PCM1795) configuration helping out.

I suspect the numerical results could be better if I notched out the 1kHz fundamental and that might explain the discrepancy in THD(+N) compared to an AP measurement like John Yang's post here showing SINAD 113-114. Of course there will also be variations between devices and who knows if hardware versions changed without notice during the production of this machine. This one I bought was one of the first units shipped here in Canada. The specs sheet only listed the XLR output as 115dB SNR and 0.0015%/-96.5dB distortion which appear to be conservative numbers!

Compared to ASR's measurement of the T+A DAC8 done a year back (based on the Hi-Fi News & RR report, this DAC uses a "quad" 4 x PCM1795 DAC arrangement), I agree with the general sentiment on the thread that the TEAC UD-501 appears to measure better than the T+A 8 which was also a contemporary DAC from 2013. The T+A 8 is more expensive device back in the day although it did have more features including preamp volume control, remote and other resampling settings.

After all these years, my well "broken in" TEAC UD-501 from 2013 still does a great job. I had a listen to some Alannah Myles (1989, DR13), 2021 hi-res Pink Floyd The Wall (same as the Immersion box 2011 remaster), and the recent No Time To Die (2021, DR10) soundtrack the other night with this DAC and it sounded great; nice dynamics, soundstage, tonality, "smooth" upper end, etc. While the headphone output at 100mW into 32Ω might not be strong enough for low-sensitivity cans, it sounds very good as well using my Drop+HiFiMan HE-4XX at modest volume. While the TI/BB PCM1795 chip may not be considered current "state of the art", I think TEAC did an excellent job with the execution of this DAC.

These days, with all the apparent hype around multibit R-2R DACs, it's good to be aware that the TI/BB "Advanced Segment" chip DACs are actually a type of hybrid design - "segmented" into different techniques for how the conversion is handled. The most significant bits (I think top 6 bits assuming this is similar to the TI/BB PCM1794) are converted in a "multibit" fashion and the lower bits (26 bits of a 32-bit input for the PCM1795 I presume) are converted through a multi-level sigma-delta modulator which can achieve better low-level linearity at lower cost of implementation. As far as I'm aware, current ESS and AKM DACs internally are pure multi-bit sigma-delta designs.

Despite the hoopla about architectures and debates within the audiophile community between "multibit" versus SDM DACs, I have yet to see any blind test showing that "Golden Ears" can truly tell the difference. Yeah, I know, we see all kinds of forum comments about "multibit" being awesome and that R-R2 ladder DACs are somehow able to produce a more "real" sound. I would love to see this belief tested in a controlled, blinded fashion once we take away knowledge of $$$ and brand names! If anything, high performance R-2R DACs are more expensive to produce (hence "bragging rights"?) and one must be aware of potentially higher distortion and linearity errors. Also, some multibit DACs (like the Holo Audio Spring 3) do not incorporate oversampling which IMO is desirable as this is more accurate than NOS when playing 44.1/48kHz material. Of course you could turn off the TEAC UD-501's filtering and hear the "stair-stepped" NOS mode for yourself.

BTW, here's the NOS output from this DAC (1kHz 0dBFS 16/44.1, captured with E1DA Cosmos ADC):

With low-pass filtering at 20kHz, notice that the THD and THD+N are almost exactly the same whether using oversampling or NOS mode. Objective testing like this does not suggest NOS should sound "bad" in general. What we know though is that there is some roll-off in upper frequency response if we were to run a sweep, and the "imaging" artifacts were not in the original signal or performance.

Barring any controlled listening tests to prove otherwise, the technical performance of high-resolution DACs these days is simply excellent and representative of a "mature" product class. While we can see the incremental improvements in the PCM performance over the last 8 years between the TEAC UD-501 to the Topping D90SE with high-res signals (going from THD+N of -110dB to better-than -120dB for example), I trust that there's no need to get too excited given human perceptual limits.

It's actually nice revisiting a device after a few years using updated test equipment and software (eg. latest REW). When a device is new, it gets in the spotlight, people write reviews or make videos claiming how the device is so much better than the previous generations. We might even get a measurement or two. But once the generation passes, we're often left with the impression that this or that device is now "obsolete" or substantially pales in comparison to the latest and greatest. While that might still be the case for the latest cutting-edge technologies (the newest VR headsets for example might have even better resolution and more responsive framerate), I think it's fair to say that audio DAC sound quality isn't improving by leaps and bounds. Features may continue to expand of course but these evolutionary steps don't result in substantially more "accurate" sound quality.

At a time in history where planned obsolescence is common and so much of our consumer goods are built to be disposable, I think it's important to be responsible about material goods in a world with 7.75 billion people. We're all striving for a better quality of life, and ultimately, resources are limited on Starship Earth.

With DACs, IMO we've long reached the point of "diminishing returns" by the time we're spending even >$150 for basic digital-to-analogue conversion duties (the Topping D10s is a nice example I think). Even more so when we consider that still the vast majority of digital music remain as 16-bits and 44.1kHz resolution - 98dB dynamic range, 22.05kHz frequency response is all you're going to get - this is simply an unremarkable level of performance for any decent DAC these days!

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

For those of you who have upgraded to Windows 11 with AMD processors, make sure to get the most recent updates. With more intensive applications, the speed difference is quite noticeable.

This past week I grabbed another 2TB SATA-III SSD drive, the Crucial MX500 3D NAND:

Technically we're looking at Micron TLC 256Gb 64-layer 3D NAND storage which should have more endurance than QLC drives, controlled by a 6Gb/s Silicon Motion SM2258. This one has hardware AES-256 encryption which is good for work-related files.

Released back in 2018, there has been time for prices to drop nicely (<US$170 currently for 2TB, around US$350 4TB). Maximum power utilization is 6W but usually in brief bursts, and typically <110mW when not in use as opposed to something like 6-12W with 3.5" hard drives during read/write and 3-6W idle even with efficient models these days. Quite substantial power savings with SSD. Of course you also have the other benefits of SSDs - cool-running, silent, no mechanical vibrations, and no spin-up/wake-up latency.

The benchmark results look pretty decent using my AMD Ryzen 9 3900X CPU. Not the fastest SSD nor at the level of an M.2 NVMe solid-state drive of course (see below). I was curious if turning on encryption using BitLocker changed performance. As you can see on the right, there is a small "hit" to the performance especially with the small, low thread count, random access (Q1T1) test. Not bad at all and certainly the performance of the drive "feels" nice and fast in daily use.

For comparison, here's my 2019 model ADATA XPG SX8200 Pro 1TB M.2 NVMe which I use as the boot drive using CrystalMark 8 on the same computer:

Slowly but surely, the transition from mechanical hard drives to solid-state storage continues. The Workstation computer which is what I use to write these blog posts and edit the images with is now completely SSD-based (finally removed those last 1TB WD Green and Seagate Barracuda hard drives that have been in use since 2009!). The only hard drives I have left now are for my media archives (music, movies, photos) on the Server computer or backups on an external hard drive enclosure.

At some point, I'm sure to switch out my 12TB Seagate EXOS Enterprise and 6TB WD Gold drives that currently house the digital media. At present, an 8TB Samsung 870 QVO is going for US$750. Still a hefty price but I think if we give it another 5 years, 12TB and 16TB SSDs will be available and I look forward to the solid-state transition being complete at reasonable prices.

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

In other news, I see that Hi-Fi+ has joined The Absolute Sound in revamping their website. And in so doing, gone are the Disqus reader comments. Another audiophile website with no direct reader feedback mechanism. Hi-Fi+ has been an interesting one in that of the "mainstream" magazines in English, they have not been shy about "reviewing" very expensive cables. They have been bold (or crazy) enough to put out stuff like cable "buyer's guides". Over the years, I've been curious about the feedback on their reviews of (multi)kilobuck kables and it's interesting to see the arguments between believers and naysayers (generally I stay out of these even when tempted to say something ;-).

In the last year or two, the reader comments to cable "reviews" on Hi-Fi+ have actually become muted, much less argumentative or sarcastic. I will miss those days where cable-believers tell non-believers to go back to their Sansui sound systems ;-). I suspect this diminished level of conflict is a result of dwindling interest rather than increasing acceptance for beliefs in these products. I presume many "non-believers" have just moved on. This apathy is a good thing I think. (As usual, here's my collection of cable measurements/articles.)

Magazines like TAS and Hi-Fi+ exist IMO as glossy "infomercials" for the Industry. Within these reviews they claim that brands we've likely never heard of somehow caused a "splash" at some point or other with their supposedly fantastic products or genius designers. Most of this is basically fantasy I think to please advertisers and the real "product" is turning the readership onto what's available for sale. An obvious attempt to capture the fantasy of those audiophiles who still have faith in snake-oil and potions; including computer products like switches, supposed ethernet "filters", etc.

The other day, I was listening to Douglas Coupland interviewed on CBC (local Vancouverite and Generation X: Tales for an Accelerated Culture turned 30 this year). In the interview, he showed optimism that the Internet actually on the whole makes us smarter people. While we may have a multitude of examples to show otherwise, I actually think that in audiophile-world, this may be the case. Over the last decade, I think the discussions among audiophiles have improved. Back in the '90s and early '00s, the level of audiophile-discourse was maddeningly filled with "fake news" dictated by a few magazines. Voodoo, cults and related snake-oil beliefs were the norm on forums I used to visit. While there's still a good amount of that, at least it doesn't seem that believers in such products/theories are as forceful with their expression of The Faith. I think the Internet has opened up the opportunity for voices to demonstrate reason and question against the large amount of advertising when conditions are open for feedback (the reason they're closing discussion off). The fact that many audiophile websites and YouTubers close off comments are indicative of these being places where opinions are being peddled in lieu of facts. In time, I have faith and hope that the most truthful ideas will capture the greatest mindshare. We each have a part to play as catalysts for change wherever we find ourselves...

By the way TAS and HiFi+, you might want to try looking at how images are rendered on various mobile browsers with your site update! I see that aspect ratios are not being respected in various articles and those images look distorted and ugly. Not at all complementary to your advertisers/reviewed products.

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

I heard Jeremy Dutcher and Yo-Yo Ma's beautiful World Music collaboration and reimagining of the Mi'kmaq's "Honour Song" off Notes For The Future the other day:

Hope you're all doing well, dear audiophiles, and enjoying the music! (Regardless of what year your DAC came out. ;-)