In the last few years, we've seen a resurgence in interest around R-2R DACs. Back in the day (like these old CD players for example), multibit ladder technology was the way to do the conversion until the rise of 1-bit and multilevel sigma delta modulation DACs became the norm.
High quality R-2R does require extra effort to make sure the resistor levels are precisely trimmed to achieve higher resolution. This is why hybrid techniques have also been implemented over the years such as the TI/Burr-Brown segmented architectures (like with my TEAC UD-501).
With this post, let's focus on the Cayin RU6 DAC (~US$250), a new USB-C dongle device (comes also with a USB-A adaptor) which aims to provide high performance headphone playback of PCM up to 24/384 and will accept DSD256 by converting the 1-bit SDM to PCM internally (no native DSD conversion).
Thanks to my buddy AudioPhil for lending this device for testing!
|Front (top) of device with OLED screen (currently off) below the shiny glass. The other end of the stick has a USB-C connector. 3 buttons: MODE, "-", and "+".|
|Underside. Notice the slight triangular edge.|
The device is a good looking 6.5cm x 2.5cm x 1.4cm stick, all black with pleasant orange/goldish lettering, gold connectors, and a useful 1" 128x64 pixel OLED screen. The aluminum case feels sturdy and at 28g, this is nice and light for mobile use. No internal rattling when you shake it. There are two headphone outputs, a standard 3.5mm single-ended phono and a balanced "Pentaconn" 4.4mm phono (JEIT RC-8141C standard). Note that this does not have fully balanced circuitry internally, but provides more power if your headphone needs it. Unfortunately, I don't have a 4.4mm adaptor so I'll just stick with listening to and testing the standard 3.5mm out.
Cayin's specs: 3.5mm single-ended output = 0.5Ω output impedance / 138mW into 32Ω, 4.4mm balanced output = 1Ω / 213mW into 32Ω. This should provide enough power for most headphones. Obviously if you need more power for low sensitivity headphones, consider a desktop amp for non-mobile purposes.
Most of the time, the OLED screen is off until you press a button to change the volume. If you press the MODE button, it shows a display with some details:
|Shown with 1MORE Quad Driver IEM.|
As you can see, there's the samplerate at 96kHz currently (using upsampling on USB Audio Player Pro). The smaller text on top tells us that it's currently at 100% hardware volume, "LdB" means it's at "Low" gain, and we're in "NOS" mode. For convenience, when listening with the phone, I've typically left the DAC hardware volume at 100% and used the phone volume buttons. However for best fidelity, it's good to leave the phone volume at 100% so there's no software attenuation applied and use the hardware volume buttons. The volume setting is thankfully saved between uses. There can be a delay and occasional clicks/noise when changing volume due to the relay mechanism; nothing I found concerning.
When showing the display screen, if we then hold down the MODE button for a couple seconds, we go into the settings screen:
Press the MODE button to advance through the 3 options, and "-" to toggle the options. This is obviously how we select the "High" vs. "Low" gain. You can also change whether playback is "OS" (Oversampling) or "NOS" (Non-OverSampling) mode. "Backlight" determines how long the display screen stays on when the MODE button is pressed.
Okay, as usual, let's start talking about some findings and the objective performance, then subjective impressions/opinions afterward...
I. Oscilloscope, Impulse Response, Digital Filter, Output Impedance, USB power draw
|Yellow = Left channel. Blue = Right channel.|
II. RightMark PCM
Raspberry Pi 4 "Touch" --> Cayin RU6 USB DAC --> 3' 3.5mm Phono-to-RCA --> RCA-to-XLR --> RME ADI-2 Pro FS ADC --> Intel NUC for data capture
III. RightMark DSD & Ultrasonic Noise
|As expected, we see the high frequency attenuation due to NOS.|
The DSD64 noise level shows the characteristic rise in ultrasonic frequencies. Let's see if the DSD --> PCM process attenuates any of that ultrasonic content resulting from noise shaping. This time, using the E1DA Cosmos ADC running at 384kHz (using FlexASIO in REW) to give us a full 192kHz bandwidth:
Indeed, we can see the typical ultrasonic noise from DSD conversion except for DSD256 which introduces its own anomalies on this DAC like the pair of sidebands around 1kHz, irregular ultrasonic noise, and higher THD. This is also our first look at a THD+N number as well - let's turn to that now in native PCM playback...
IV. 1kHz THD+N
Now that you've seen the 1kHz signal played back through DSD as a converted 24/96 signal, let's have a peek at native PCM playback at 16/44.1 which of course is still the most common, "standard" resolution with actual music. Testing at 16/44.1 is also important to see the difference that NOS makes.
First, let's start with the oversampling (OS) setting:
Let's add some further complexity to the signal and have a peek at a few multi-tone tests - first up the Triple Tone TD+N (24/96) which I typically use with amplifiers:
|Note: Normally I turn on the 20Hz/20kHz filter, forgot to this time when capturing the image. When on, the TD+N result improved to -59dB.|
As you can see, that's quite a bit of harmonic and intermodulation distortion products! Again, TD+N results in the mid-50's like this is power amplifier domain, not DACs or headphone amps usually.
For comparison, here's the old Geek Out V2. This is how a "typical", high-resolution USB dongle DAC performs on this test:
And here is the 1/10 Decade Multitone 32 with the RU6:
Finally, let's have a look at the J-Test results from this little DAC/Amp. Again 100% hardware volume, "Low Gain" setting:
|For reference, the LSB 16-bit jitter modulation content should be below -120dB.|
|For 24-bit test, the jitter-modulation pulse should be buried below the noise floor with a clean hi-res, jitter-free DAC.|
Clearly these are not clean J-Test FFTs. OS mode cleans up some of the anomalies but this is quite different from most jitter-free asynchronous USB DACs. The noise floor is clearly elevated with only the 24-bit "OS" J-test relatively spared from high-level sidebands although there's a "skirt" below the 12kHz primary signal suggesting some low-level jitter.
VII. Subjective & AMPT Recordings
As usual, I listened to the DAC/headphone amp before running measurements and then after the measurements to make sure I'm not biased by knowing the results when first listening.
For most of my listening, I used a combination of the 1MORE Quad Driver IEM, the Drop + HifiMan HE-4XX, and Sennheiser HD800. I found the Cayin to have enough power for all of these headphones. Even with the AKG Q701 and my modded Dekoni Blue, the volume is fine with "High gain" activated although without too much extra overhead if music is recorded a bit on the soft side. Situations like this would benefit from the higher power 4.4mm Pentaconn "balanced" output which I did not test.
I was told that there has been at least 100 hours of "burn in" done by the time I borrowed the unit (not that I necessarily believe this makes a difference with electrical gear). I spent a few hours listening to some "audiophile" stuff I have used over the years like tracks from the Dali CD Volume 3 and some familiar music like Ella Fitzerald's Pure Ella: Ella Sings Gershwin with Ellis Larken on piano (you can see a list of some of my test material like Jennifer Warnes, 1812, Roger Waters, and such, in my Soundroom post from 2020).
Subjectively, I thought this is quite a good sounding DAC/headphone amp with a more "meaty" presentation especially in NOS mode with some caveats. Midrange does sound a little accentuated which I think is a pleasant effect from NOS DACs in general (probably because of that higher frequency attenuation with 44.1kHz material).
I had a listen to The Weeknd's (aka Abel Tesfaye) recent album Dawn FM (2022, DR6) through these and quite enjoyed the pop tunes and modern production values. Fun hearing Jim Carrey doing his faux-DJ cameo, and interesting interlude by Quincy Jones ("A Tale by Quincy"). Yeah, this is typical Weeknd material. Catchy melodies, clean disco-like rhythms appropriately meant for Top-40 stations (I like "Gasoline"). There's certainly more than a passing nod to '80s synthpop (for example check out "How Do I Make You Love Me?"). Some '70s-like funk on "Sacrifice". Nice rendering of the sound through the Cayin RU6 paired with the Sennheiser HD800. The HD800 can sound too bright with some amps if there's any kind of tipped up treble, no excessive harshness with this pairing.
Acoustic tracks like The Modern Jazz Quartet's live Blues At Carnegie Hall (1966, MoFi release, DR12) sounded appropriately natural with a good "soundstage" and "air". Soundstage through headphones of course isn't like listening to a good pair of speakers in a treated sound room. Detail retrieval is good. Nice transients on the percussion. Nuances like the audience applause and cheers are well defined. The annoying stuff in live recordings like audience coughing ("Blues Milanese") just as well defined.
Bass response is excellent, dependent more on your headphones of course. In the picture above, I'm listening to the Les Miserables Live! (2010) performance through my modded Dekoni Blue headphones. As you can see on the screen, high gain ("HdB") was needed for these relatively low sensitivity planar magnetics (based on the Fostex T50RP Mk III). Tracks like "The Second Attack" had cannons and gun special effects that came through with plenty of bass heft (and that's with MP3 from the smartphone).
There's a noticeable physical warmth to the casing over time with use, but nothing concerning or excessive whatsoever.
Putting on my "critical listening hat" for a bit, while usually not a problem, I noticed occasionally audible distortions with lower impedance headphones. Here's a concrete example I came across while rummaging through some music on my phone: with the closed-back lowish impedance AKG K371 headphones (measured here) in a quiet room, turn the Cayin to Low Gain, 90% volume (not too loud), NOS mode. Listen to the first few seconds of track 2 "A Million Dreams" on The Greatest Showman (2017, DR8) soundtrack. With this combination, I noticed during the lower-level piano intro, a subtle "haze"/"hash" has been introduced into the sound. Imagine something like the top end distortion of low bit-rate MP3 but transposed down an octave or two. It's not as strong with "OS" mode but still clearly not as clean as with a higher resolution DAC+amp (like say the Topping D10s with the Drop + THX AAA 789 headphone amp). I suspect this is an example where the non-harmonic distortions we see in the measurements surpassed the threshold for detection and I remember thinking "that sounds different...".
I can imagine that some listeners might interpret the distortions as "details" and believe that this is somehow a result of extra content "retrieved" because of NOS, or that this "layer" of sound represents extra "depth". While some might actually like this, I just found it distracting once I noticed it. Thankfully this did not happen often and I was certainly able to enjoy the music with this DAC/amp even knowing the results from the objective tests. Again, I'm happy to take the "minority report" and state that I don't particularly find multibit R-2R +/- NOS DACs that I've heard as being particularly "natural" or "real" sounding as some people seem to claim.
As we have discussed over time, "euphonic distortion" is always a subjective possibility depending on the listener so long as the effect is subtle and doesn't cross the threshold of being unpleasantly dissonant. Where that threshold lies and how we each interpret such anomalies would be idiosyncratic.
Okay dear audiophiles, I'll leave it to you... Here are the hi-res 24/96 "AMPT" recordings captured from this DAC in NOS and OS modes:
Embedded in the NOS recording are the slight interchannel temporal shift with the right channel delayed ever-so-slightly as discussed in Part I, NOS-type -3dB high frequency roll-off, along with some squarish stair-stepped waveforms. Some of these differences will be evident when you A/B compare through a high-resolution system. Note that since this is captured by connecting the headphone output to the 9kΩ input of the RME ADI-2 Pro FS ADC as per standard procedure, distortions introduced by low-impedance headphones which might strain the amplifier will not come through in these AMPT recordings; this is meant as a reflection of DAC performance.
|Cayin splash screen when first powered up.|
Overall, the physical design of the Cayin RU6 R-2R USB DAC/headphone amp feels robust and convenient for portability, nice to see USB-C in new DAC designs, the 1" OLED screen looks good with useful information like sample rate, high/low gain, and volume level. It's interesting to see Cayin discussing some tech details about this DAC/amp especially the resistor-array volume control, that it's not a fully balanced design, and admission that "measurement is not the strong suit of R-2R technologies"; clearly true!
Indeed, we're not looking at a high resolution DAC/headphone amp here. This is why I mentioned at the top the inclusion of those little "Hi-Res Audio" stickers - maybe that's why they also didn't slap them on at the factory. Other than the fact that this device can accept >16-bit data, and can reproduce >22.05kHz frequencies, the noise and dynamic range deficits are very much sub-hi-res. As Mark Waldrep (Dr. AIX) mentioned a number of years back, the JAS "Hi-Res" logo can be meaningless, and this is a good example.
The way I see it is this. This mobile USB dongle DAC/headphone amp is currently a unique product riding on a kind of "revival" (some might call it "hype"?) using R-2R conversion among manufacturers looking to differentiate themselves from mainstream, already high performance devices, based typically on sigma-delta chips. The Cayin RU6 does a number of things technically well including adequate power for most headphones, very good channel balance, low output impedance (0.5Ω), equivalent PCM-DSD playback, and hassle-free compatibility among portable USB devices I plugged it into.
However, the OS digital filter is weak, jitter test results are far from impressive, and that slight right-left temporal delay in NOS mode is unusual. Most significantly, this is not a "high fidelity" or "high resolution" DAC in that the objective resolution actually does not approach even clean 16-bit conversion and the distortion amounts are clearly higher than the norm. In fact, the resolution doesn't even beat the multibit chip DACs from CD players of the 1980's! Cayin advertises that this uses "0.1% resistors" - better precision appears to be needed for true 16-bit, never mind "hi-res" performance, and that doesn't come without a hefty price!**
Having said this, I think we can subclassify audiophile gear into 2 gross categories based on intent. On the one side we have "high fidelity" gear. Machines that perform with an eye towards "accuracy" and transparency. Admittedly, the goal of "accuracy" is more in line with my philosophy and we've talked about this before. Then there are devices that purposely take a different path and along the way, instead of aiming for "transparency", will show themselves to have "character" and potentially "euphonic" to some. Cayin must clearly be aiming for this latter goal with the RU6 DAC/headphone amp. Users have commented that this little DAC has an "analogue" sound. As I hinted at in the text above, the results actually look very much like modest power amplifiers; in fact, the Triple-Tone TD+N with all its harmonic and intermodulation distortions reminded me of the Melody Onix tube amp measured last year.
We have met other devices "with character" on this blog. I would consider the Pass Amp Camp Amp and First Watt SIT-2 amplifiers as examples. So long as we understand the potential pitfalls and benefits within the audio systems we build (in this case mainly the headphones we intend to pair the DAC/amp with), have fun! Like with human temperaments and personalities, we're all free to love, hate, or even be indifferent to the character idiosyncrasies.
At the time of this publication, I've seen one other set of measurements on the Chinese site L7AudioLab with the typical suite of Audio Precision data which looks consistent with what I've found. Obviously there are a host of subjective-only reviews (here, here, here, and "No" this is not likely an "end game" dongle if you want high fidelity). Generally this DAC has been reviewed favourably. As usual, make sure to do your own research; at-home trials and a good return policy help tremendously when we're trying to decide whether to keep devices "with character" for long-term use. ;-)
Thanks AudioPhil for the opportunity to listen and test this device out!
** As an aside, but still important discussion, I think there's more than a reasonable argument to be made that using purely discrete R-2R architecture for the goal of high-resolution "24-bit" DACs should be viewed as an anachronism. It's obviously an antiquated way of doing things demanding extra cost for what usually ends up being lower technical performance. Obviously, thought and expense have been put into the design of the Cayin RU6 with its double circuit boards, 96-piece high spec resistors, and 9-segment resistor/relay for volume control, so I appreciate what's being done here.
Even if anachronistic, this does not mean hobbyists can't or shouldn't enjoy these things! Vinyl LPs, reel-to-reel, cassette tapes, 8-track, tube amps, vintage audio gear can all sound subjectively great even if not particularly "high fidelity" by modern standards and can hold their monetary and psychological value.
Heck, one might even view this little Cayin DAC/headphone amp as a mobile DAC with tube amp distortions thrown in!
When I previewed this article with AudioPhil, we discussed the idea of finding balance between objectivism and subjectivism as human beings with our own unique qualities, limitations, and peculiarities.
Objectively, IMO DACs can be developed quite easily these days with distortions well below audibility using off-the-shelf chips; that is no longer an engineering challenge for manufacturers. The future (and perhaps present) may actually be a challenge for companies to see if intentional imperfections might be the target for products they want to make. While I personally am not keen on it philosophically (and would not call these products "high fidelity"), I accept and support that a significant percentage of audiophiles might be interested in such distortions that create an impression/illusion of "harmonic fullness" and "soundstage depth" like this one. An insightful audiophile who is interested in creating a "euphonic" playback system which is not strictly "hi-fi" is just as fair a philosophy to subscribe to!
It's disheartening when objectivism becomes dogmatic, indignant and automatically label devices that measure technically "poorly" with words like "garbage", or "crap". I obviously don't think there's any need to be abrasive like this. While I believe there are some very expensive but poorly made devices out there, this is certainly not universal. What's important is that manufacturers and reviewers be clear with letting buyers know when a product is not technically accurate, and that the device intentionally caters to a certain type of subjective listener preference. Having owned a number of high resolution mobile DACs already, I can say that personally the RU6 provided a more interesting listening experience than yet another AudioQuest DragonFly, even if not "transparent" nor necessarily something I would spend too much money on.
My suspicion is that for DAC+headphone amps like these aimed at the the intimate headphone-listening experience, a moderate distortion range - something like Triple Tone TD+N around -50dB (~0.3%) to -70dB (~0.03%) - can add intentional "euphonic" properties without being objectionable for most listeners. When distortion level is better than -75dB (<0.02%) TD+N with a flat 20Hz-to-20kHz frequency response, some might use the terms "hi-fi", "sterile", or "clinical" for this kind of "clean" sound (to be clear, this is what I prefer). This is consistent with our blind test results here back in 2020 exploring THD. Obviously there are many variables to dissect including frequency response (eg. that -3dB high frequency dip from NOS could be less fatiguing) and temporal characteristics will also have an effect. Likewise, each person will have their own thresholds of audibility depending on which headphones are used, so there will be a range that manufacturers can aim for.
Designers like Nelson Pass are already aware of certain listeners' preferences for adding nonlinear distortions - here's an interesting discussion of his H2 Harmonic Generator. Ethan Winer also wrote a short article about distortion being potentially euphonic in Sound-On-Sound back in 2006. This is something that can be studied empirically with listening tests and statistical probabilities applied if we want. "Unmeasurable" voodoo and pseudoscientific "quantum" explanations offered by certain unsavory companies simply need not be entertained, and should be resisted.
My feeling/hope is that the future of audiophilia is not the dichotomous tribalism of "objectivists" vs. "subjectivists", but that combination in the "Rational Audiophile". Hobbyists who can see value on both sides, understand the technology and more importantly themselves (audiophile psychology), as such can speak clearly about his/her desires. Relative immunity from irrational hype and "snake oil" comes with the objective analysis but objectivists must watch out for a tendency to lose touch with art and emotions. The love of music regardless will require a subjective embrace but undisciplined purely-subjective "Golden Ears" reviews often accentuate hype, irrational fears and countless false claims. I believe we each must find our honest balance.
Until next time, I wish you all wonderful listening and great enjoyment of the music!
PS: Meat Loaf - anyone with such a name clearly must have been an interesting "character". RIP.
Discussion on Topping D90SE DAC and -60dBFS signal...
As per the comments by Bennet/Dtmer Hk, there are some posts by Miska on ASR about variation in noise floor with the Topping D90SE using -60dBFS input signals. As I mentioned in my comment response, I actually don't have the D90SE at this point to retest, but I do have some -60dBFS measurements which are simply part of the RightMark suite used for calculation of dynamic range of the DAC.