Saturday 12 September 2020

MEASUREMENTS: RME ADI-2 Pro FS R Black Edition as ADC.


About a month ago, I posted this preview of the RME ADI-2 Pro FS R Black Edition AD/DA converter. It has been a wonderful month of simply listening to this device without a care around performing measurements. For listening, I have it set up as per the image in my sound room with the Raspberry Pi "Touch" running RoPieee previously discussed.

So far, from a subjective perspective, it has been a joy to listen to music with this machine.

Alas, as a "more objective" audiophile, it came time to have a look at the performance for this device to check if I'm missing any significant limitations. Over the years, I've discussed limitations to human hearing as well as the idea that many of our audio electronics these days have surpassed human auditory perception such that while measurements of DACs or amplifiers can show limitations, often these are beyond the resolution thresholds of human hearing/perception. Remember that while not all measurable anomalies need to be chased after when the goal is simply to be able to enjoy music, technical "ideals" are useful for devices one may want to use as a "reference".

So to start, today let's just focus on the ADC side of this RME device. Remember that the RME ADI-2 Pro FS R Black Edition (again, let me just call it the "FS R BE" for short) is an evolutionary step from the ADI-2 Pro FS ("Pro FS") that I started using 2 years ago. Consider this article as building on some of what we already know about the Pro FS as an ADC from 2018 with requisite repeated tests comparing the two machines and using today's software.

Over the years on this blog, I have upgraded the ADC used in order to improve the quality of measurements published. For today, let's just look at the last 3 devices I have been using - the Focusrite Forte, ADI-2 Pro FS, and ADI-2 Pro FS R BE now (in my previous article, I also included the old Creative E-MU 0404USB in the mix):


Remember that compared to the RME machines, the Focusrite is limited to 192kHz (the RME can go up to PCM 768kHz and DSD256). Again, if we think about human hearing and music, this isn't a big deal as in reality there isn't much in the ultrasonics anyway.

In terms of specifications, the "FS R BE" uses the same ADC chip as the "Pro FS" - the 4-channel AKM AK5574 in 4-to-2 mode (as per specs: dynamic range 124dB, -1dBFS THD+N -114dB). So right from the start, I am expecting good performance yet I am not expecting much difference between the two RME boxes unless the company further tweaked the surrounding circuitry significantly. The main thing is to make sure the new box doesn't have any noticeable anomaly that could impact my tests in the future when used for measurements.

I used the shipped firmware version on the "FS R BE" for these tests which is USB 238/DSP 98. The "Pro FS" is using the latest firmware online which is 213/98. As with firmware upgradable devices, sometimes we will see small differences with updates over time.

I. The Noise Floor

First, let's start with a look at the noise floor since it is from this foundation that everything else arises. ADCs of course should be connected to a source and the reference I've used over the last few years has been the Oppo UDP-205. As an ESS ES9038 Pro DAC, the Oppo measures well and the balanced XLR output is silent. Let's plug the Oppo into the new "FS R BE" and compare with the "Pro FS" playing 24-bit digital silence through the balanced output. Room EQ Wizard's (REW 5.20 Beta 60 currently) real-time FFT analyzer these days provides an excellent tool!

I've set the RME input to +13dBu reference at 0dBFS as a convenient level for many of my measurements (remember that the RME can be set to 4/13/19/24 dBu options). A typical 2.1Vrms line level output is equal to 8.7dBu and the 13dBu option provides overhead.


As you can see (and as expected), the two RME devices are excellent with extremely low noise, consistent floor. Notice in cyan I have the "FS R BE" running off the 12V rail from the TalentCell 12V/8300mAh / 9V/11000mAh lithium battery. While clearly the switching stock power supply works well without any real noise issue, the battery was able to suppress even that tiny bit of 60Hz hum highlighted in yellow. This also speaks well of the Oppo UDP-205 for not passing along any mains hum through the balanced output.

The above results were obtained with the left channel. It's good to know if the right channel is just as quiet when testing gear!


Yup, beautiful, the 2 channels of the "FS R BE" are just as quiet. BTW, you can see the smoothing (1/48-octave), averaging (4) and windowing settings for the FFT graph to the right I'm using here. At these extremely low levels, smoothing and averaging help visualize the noise floor so we can pick out the really small deviations like that microscopic 60Hz hum in the previous graph.

In comparison we can look at my older ADC, the Focusrite Forte, both channels, running off the Microsoft Surface Pro laptop's USB (the laptop of course is also running off its battery):


As you can see, calibrated for the same 0dBFS input level, the Focusrite doesn't achieve as quiet noise level as the RME ADI-2 Pro devices. Also, as I had pointed out many times when I was using the Focusrite Forte for measurements until earlier 2018, the ADC has an idle tone at ~37kHz which mars the otherwise flat noise profile to 50kHz. Of interest is that the tone is stronger in the right channel. Here's an example of why it's nice to check if both channels are functionally equivalent.

A few years ago, I measured the Tascam UH-7000 ADC/DAC and found that the power supply added distortion as the device warmed up over time. Since then, I've checked to make sure there are no issues with my other ADCs:


No problem here with the "FS R BE" over 4 hours of warm-up. In fact, not even a hint of increased noise or spurious tones over time while you can feel the box getting warmer. This is even with the stock switching power supply (notice the subtle 60Hz and 120Hz bumps down in the noise floor around -150dB).

II. Frequency Response

We know that the analogue output frequency response of the Oppo UDP-205 is pretty well "ruler flat" with the official Oppo specs as -3dB by 160kHz (presumably running at 384/768kHz). Due to various limitations, at best I was able to measure the Oppo at 192kHz which of course means up to 96kHz bandwidth. Let's compare the frequency response obtained with the "Pro FS" and the new "FS R BE":


As you can see both devices show an equivalent, very flat, frequency response sweep coming out of the Oppo set to "Linear Phase Fast" filter. There was a slight <0.05dB difference in amplitude recorded between the two machines. Notice that the whole Y-axis only represents 4dBs in the graph above. On both machines we see a little rise in level up at 80kHz (only +0.14dB above 1kHz level) before we hit the bandwidth limits.

Just as DACs have filter options, the RME ADC can be set to various low-pass filters also. "Sharp" linear phase filter was used above and allows a bit more extended frequency response and has been my standard setting over the years. The ADI-2 Pro FS devices actually have 4 filter options; here are the resulting frequency responses obtained by the various settings with the 192kHz sweep:


Notice that while there are 4 filter options to choose from, we only see 2 curves - basically both "Sharp" and both "Slow" settings overlap precisely. The "SD" designation stands for "Short Delay" and refers to minimum phase versions of the filters which in itself does not change frequency response but would result in very slight group-delay distortion. At 192kHz, the "Slow" settings provide flat frequency response into 25kHz and the roll-off might be preferred with music recording.

By the way, you can see the variation in filter options on page 84 of the manual with curves shown for 44.1 and 96kHz samplerates as well.

III. A look at Distortion

Using the REW "Frequency Step" function, we can now look at the distortion across the audible frequencies from the Oppo using the ADI-2 Pro FS R BE. The Oppo is playing the step sweep at -3dBFS (XLR output level at 12.5dBu). Unless otherwise specified, all sweeps done at 24/96, 64k FFT length:


As you can imagine, to obtain high resolution results, the DAC and ADC must function in tandem. Since we know the Oppo UDP-205 performs very well achieving low harmonic distortions, this suggests that likewise the RME ADC is pulling its weight. Notice that I have put the cursor at 1kHz and we can see the THD+N of the Oppo measures at -111.9dB (-3dBFS signal).

Here's a high-resolution FFT of 0dBFS 1kHz playback comparing the "FS R BE" and the "Pro FS" (24/96, 128k-point FFT):


Realtime measurements will obviously show small differences. Notice how close the THD+N results are; -113.1dB with the "FS R BE" and -113.0dB with the "Pro FS". As you can see, the Oppo UDP-205 has some non-harmonic sidebands around the 1kHz signal (power related?) easily demonstrated with both RME devices. Here and elsewhere, I've noticed that the "FS R BE" captures the base of the fundamental tone a little cleaner than the "Pro FS". When we do tests like the Dunn J-Test, this smearing of the base suggests low-level timing inaccuracies; perhaps this slight difference comparing the ADCs is also reflective of improved ADC time resolution?

Let's now have a look at the "Level Step" test in REW derived from increasing DAC amplitude level. Note that I use 950Hz in these instead of 1kHz because occasionally with some USB gear we could run into 8kHz PHY noise which would be difficult to differentiate from the 8th harmonic if I were to use 1kHz. We know that the Oppo UDP-205 is a very linear DAC - essentially "perfect" down to -120dB based on Audio Precision measurements:


I've overlaid a thin red line over the "Fundamental" amplitude slope from -120 to 0dBFS (+14.7dBu). The ADC was able to capture essentially perfect linearity - as assessed by the lack of visual deviation from that red line at least! We can reformat that and get a more precise plot of linearity:


In comparison, we can look at the equivalent linearity test with the Topping DX3 Pro (V2) and its comparatively lower resolution RCA output as we get below -100dB generator level.


Finally, we can examine the Level Step graph from the Oppo UDP-205 closer to look at the individual harmonics and compare levels between the "FS R BE" and "Pro FS":



We're generally seeing equivalent results between the 2 RME devices and very much representative of inter-test/inter-device differences. As you can see, by -40dBFS generator level, harmonics start poking up above the noise floor and we can see a bit of fluctuation with various order harmonics taking the lead. Remember that different DACs will have different patterns and typically I've seen AKM DACs show a less chaotic harmonic profile even if ESS chips tend to have lower THD (I discussed this briefly back in June comparing the AKM Topping DX3 Pro vs. ESS Topping D10).

IV. RightMark result comparisons

RightMark Audio Analyzer (current version 6.4.5 Pro) has been a staple in my DAC tests over the years as a simple test suite that provides a table of summary results and allows quick comparison between devices. Unfortunately it hasn't been updated in awhile (since 2018) and has various stability and graphing issues especially at 24/192+. Nonetheless, the "core" tests have been reliable and the software works well at 24/96.

Here are the results between the 2 RME ADCs measuring the Oppo UDP-205 and Topping DX3 Pro (V2) at 24/96 and notice I've also included 16/44.1 results with the Topping. For the Topping measurements, I'm playing the signal through a Raspberry Pi 3B "Touch" using Volumio software in this case off a USB stick as an example of what a typical playback source might look like for me:



As you can see, numerically the summary results are basically equivalent for the two ADC's measuring the same devices. As usual, when we look at ES9038Pro DACs, it's impressive how small the % THD and IMD distortions are! The graphs are basically equivalent between the two ADCs although we can see that the "Pro FS" was able to discern lower IMD+N sweep results. We are looking at very small overall differences down between -108 to -110dB so I won't be obsessing over this of course.

As a side note, I must say that I am still very impressed at the performance of the little Topping DX3 Pro (V2)! IMO with any blind test using actual music and controlled for volume, I would have a hard time believing that human listeners would be able to discern a difference between the Topping and Oppo. Even the frequency response is only 0.15dB difference between the two up at 20kHz. Note that the Topping uses RCA outputs compared with the Oppo's XLR cables which would explain the higher (but still very low!) stereo crosstalk. Despite being "handicapped" by the RCA single-ended output, notice the very low noise level and absence of 60Hz hum even with the stock Topping switching power supply. Again, I don't understand why audiophiles bother spending money on linear power supplies for stuff like DACs; sometimes these power supplies can cost even more money than the DAC itself!

For "fun", let's have a look at the 16/44.1 graphs for the Topping DX3 Pro:


We're looking at almost exact overlays. This is simply a reflection of how these days, standard PCM 16/44.1 or CD-resolution - which is still by far the most common digital music format of course - can be reproduced essentially perfectly by any good DAC, even if it's only around US$200. Likewise, 16/44 resolution gear/signals should pose no challenge to hi-res ADCs when running measurements and achieving highly accurate, and reproducible results.

V. Conclusions

I'm finding excellent performance from the RME ADI-2 Pro FS R Black Edition used as an ADC. Noise level is very low even with the stock switching power supply across both channels, only slightly bested with the lithium battery pack if one cared about the tiny 60Hz hum way down at -140dBFS or thereabouts. There is no evidence of change in noise level as the device warmed up over 4 hours. Noise level for the two channels are equally quiet. As a professional-level tool used in a studio, clearly this device will capture your audio at a very high resolution.

As an audiophile doing vinyl rips, this is way more accuracy than needed (remember, I'm a guy who's happy with 16/48 vinyl rips these days)! Then again, if you're a perfectionist, I don't think you could reasonably ask for any more than this. Furthermore, for those who are into DSD, the fact that this machine can do up to DSD256 recordings is a feature you may care about. I have made a few recordings with DSD using SoundIt! Pro that sound excellent although objectively I've not analysed them since I would then have to convert back to PCM for testing.

The real value of these ADI-2 Pro machines for me and why I want an ADC of this quality is of course because I could use it for my measurements. As a hobbyist, although I don't have a calibrated high resolution "pure" signal to check the absolute ADC distortion figures, using the Oppo UDP-205 as a clean audio source with its known high-fidelity performance, the results achieved are close to the published values by those using formal calibrated test gear (here and here). Looks like I'm missing at most the lowest 3dB compared to an Audio Precision for THD+N for such an already high resolution DAC.

In this regard, I'm happy to say that both the ADI-2 Pro FS and now the ADI-2 Pro FS R BE are excellent and produce equivalent results. Although not significant, looking over the minutiae of my comparison between the two ADCs, I'd say the new "FS R BE" is marginally quieter with lower overall noise floor but did seem to report slightly higher 3rd harmonic distortion amounts. As mentioned at the outset, one should not expect much difference between the two as they are both based on the AKM AK5574 converter. The consistent results from the two devices made years apart speaks well for RME in terms of quality control among their products.

As alluded to in my Preview post, the real updates for the "FS R BE" are the addition of the remote control and an improvement of the DAC to the AK4493 converter (along with the Black Edition esthetics which matches my listening space better!). The former means I can enjoy using this device in my main soundroom conveniently from the sofa. As for the latter, we'll look at DAC performance in an upcoming post and see what RME has done to improve digital-to-analogue playback resolution...

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

Happy September everyone. I hope you stay safe whether it's back-to-work or back-to-school, or just chillin' in retired life.

Perhaps I'm missing something, but I don't see much happening or contentious in the audiophile world recently to comment on as world affairs and business continues to be hampered through the pandemic among other stresses recently.

Regardless... Hope you're enjoying the music...

PS: Wow. Smoke from the fires in the US West Coast has really traveled up north to us here in Vancouver! Here's a shot from the porch of my house over the neighborhood at noon on September 12:


It gets hazy during fire season here over the years but I don't think I've ever seen it this severe... We're #2 worst in the world according to this ranking just below Portland today.

12 comments:

  1. Hi Archimago. Great blog. I am also very happy with the ADI-2 Pro FS R BE as an ADC, not mainly because of it’s great measurement, but mainly because of it’s “sound”, as it does sound very transparent. Juergen

    ReplyDelete
    Replies
    1. Thanks for the comment Juergen,
      Yeah, great little device with excellent sound. And the technical support and firmware updates from the company over the years have been excellent...

      Take care and have a great week!

      Delete
  2. Glad you noticed the new linearity plot for stepped level measurements, I added it after I saw your overlaid line estimating linearity last time you did a review :)

    As an aside, the ADI-2 Pro is also very good for making impedance measurements with the headphone output, very low output impedance and plenty of power.

    ReplyDelete
    Replies
    1. Nice John,
      As usual excellent work! You're a hard working man as I've followed along with the beta releases. What can I say but "Thanks". And of course a donation coming your way! ;-)

      Delete
    2. How would someone go about taking an impedance measurement with the ADI-2 Pro?

      Delete
  3. I'm right in between the two big fires in Oregon, visibility has been about 100' for the last 5 days, saw a glimpse of the sun for the first time today. Cool damp weather at least has allowed the firest to stall out, things are looking much better.

    ReplyDelete
    Replies
    1. Oh man, take care of yourself.

      Today being the 15th, the sky is still overcast, smells of smoke outside, bad enough that at least yesterday the postal people weren't doing mail deliveries due to respiratory risks.

      Looks like Portland still ranks at the top today followed by Vancouver and Seattle with the dubious distinction of "worst in the world". At least here we have rain coming by end of the week so hopefully we might see the sun again by weekend.

      Delete
  4. If we could just get one or two of these storms in the gulf up on the west coast...trying to stay away from any of the new cycles as it is all depressing pretty much.

    I was reading the sales figures off of Analog Planet about Vinyl and CD sales and Covid has had a big impact. The cost of vinyl is also a factor and without going out browsing the used bins is out. I am bumbed about Joe Rogan's $100 million deal with Spotify. There are no musicians making that kind of money with their streams. Seems wrong to me to have a talker being funded my musicians who can actually DO Something, often quite well.

    ReplyDelete
    Replies
    1. Hey Jim,
      I know what you mean man... The news cycle is depressing. With the way things are looking and smelling here in Vancouver, it literally feels like a weight of depression pushing down on the West Coast. I feel bad for those in Cali/Oregon/Washington living near these fires.

      Saw the news item about vinyl as well.

      LOL Rogan and his $100M deal. This is the culture of the day I suppose. I guess there is value in this kind of stuff even if I don't quite get it. Remember that Howard Stern signed onto SiriusXM for multi-millions as well.

      Delete
  5. Archimago, I've been reading your blog for a long time. I really enjoy the mix of technical insights, scientific measurements, and a relaxed attitude!

    Lately I have been looking more and more at "studio gear" and I'm thinking of buying an FS R BE, as it seems so versatile. I have seen some complaints that it runs really hot - is that something you have noticed?

    ReplyDelete
    Replies
    1. Hi Freddie,
      Yeah, it does run warm, I wouldn't call it "hot" exactly... Tube amps are "hot", but this feels warm to touch after an hour and certainly won't be burning any fingers. I would give it some ventilation however and as per the 4-hour noise measurements above, I have not noticed any issues with function or sound quality.

      In my experience, good quality "studio gear" is really excellent "bang for buck" even if audiophiles and the audiophile press don't cover these products as much. Pro guys want function, reliability and of course good sound generally have no patience for the audiophile esoterica and fringe stuff like crazy cables with good reason. Important I think for the audiophile world to wrap their minds around this because at the end of the day, when it comes to sound quality, I believe audiophiles and pro audio should not be worlds apart in understanding what makes a difference and what doesn't!

      Anyhow, have fun Freddie!

      Delete
  6. Thanks, Arch
    With specs like these, one hardly needs an Audio Precision analyzer for real-world measurements.
    I picked up a Motu M2 for my own purposes: should do me fine for RMA and REW measurements of my vintage (and not so vintage) equipment. And subjectively, it sounds very good as a headphone amp.
    Keep up the good work: hope your air stays breathable.
    Phil

    ReplyDelete