RME Digiface USB's Internal DAC: PCM5102A - Sound

The AKM fire (AKM fire's still affecting us, years on!) affected the Digiface USB (and many other RME products...). I'm pretty sure the AK4432 was used in the previous design, but if not it, another mid-tier 2 channel variant. Anyway, it doesn't matter - I'm not talking about the 1/4" output of the first revision of the Digiface USB, I'm talking about the second.

When I first learned that RME chose to replace the AKM DAC in the Digiface USB with the PCM5102A, I was a bit disappointed... This was not my first but second time this chip was surreptitiously forced on me...
OK, not surreptitiously, but I ended up with it instead of the AKM DACs I was expecting. First time? Denon X3700H - the AKM fire ended availability of its DACs - two 8 channel'ers (it's a 5.2.4 receiver with additional Zone 2 (stereo)).

I'm not going to say that the digital sourced sound from the Denon is bad. It's not bad- but it could be better. My experience with it is actually what set my expectations for the Digiface USB. The X3700H's sound? Its frequency response sounds flat from 20Hz-20kHz (maybe it's a bit rolled off in the sub-bass), it's non-fatiguing. But microdetails are missed and imaging could be better (it's only slightly worse than being a non-issue). The described character comes through both the speaker outputs and headphone jack. These detrimental qualities are much more obvious with music, of course - for home theater content, I find performance to be more than good enough. Loud, complicated scenes with voices could, I'm sure, sometimes be more intelligible - but that's my only gripe. Since the receiver's primary (almost exclusive) purpose is movies/TV, I don't feel like I need to get anything better.

The X3700H did informed my opinion of the 5102A though - when my receiver was still new, to determine the quality of its internal DAC I did some testing. First, into the RCA analog inputs, I fed the output of an excellent DAC - I got noticeably better sound (through the fronts). I'd read up on the internals of the X3700H after buying it, and came to know that the volume control was handled by a pair of 8 channel volume control chips, a model commonly used in multi-channel home theater receivers. I don't remember the number anymore, but there's one spec I do - THD+n @ 1kHz: -100dB (...you can't have everything for $2k). Since Denon has some amplifier know-how, I decided to try feeding the power amps as directly as possible - it took me a bit to work out how without soldering to the board, but eventually, I found the way to use most of the channels by feeding signal to the preamp outs on the rear (you have to mess with some settings: the amp has to be receiving a signal which would have the channels you want to use enabled). Front L+R and Center always work, the two rears in 5.1 never work, and the other 4 channels will work when you have an Atmos signal or something to get those other four channels working (it depends how you have them configured, you'll know what I mean if you have a HT receiver). What doesn't work is switching into multi-channel stereo mode to get all channels going (don't ask me why!)

Anyway, I fed the DAC as directly to the amp as I could -probably through nothing but a digitally controlled switch (like a 3710), but that's it-, and the improvement was even bigger than the first (the one switching from the internal DAC to external through the RCA "CD" input {with "Pure Direct" mode enabled, btw}). Fed this way (optimally), the amp's sound was good. Good, not great: good. It had some of the same problems that I heard in the internal DAC - a somewhat lack of micro details and a lighter bottom end, but less apparent. I expected a bit more from Denon.. If you've ever used Dolby B noise reduction on a cassette, the type of lack of detail is similar to that, but the amount robbed is less (obviously, or that'd be a really bad amp!).

After the above description, you can imagine the expectation I had for the PCM5102A in my Digiface USB: not high. I expected that the negative qualities I noticed in the receiver would be present, albeit in (probably) smaller quantities.
I was wrong!

The PCM5102A in the Digiface sounds pretty good! If you told me it was a 5102A, I wouldn't have believed you... The sound is quite clear, much more so in the midrange/lower midrange, the bass is more present, and the detail doesn't sound immediately like it's lacking any detail! It is still missing some, especially when things get more complicated, and the sound is a bit fatiguing, actually. It's very different.

This is unexpected because with the PCM5102A, I/V is done internally, and there's another stage included as well -you can literally run a trace from the chip straight to the RCA jack! Of course you can add things between the DAC and output, but because you don't have to (and the entire I/V will be done internally), many of the decisions which can be made to tune the sound of audio, are simply not available. The output of the 5102A can be routed directly to an RCA jack, and when it is, TI says you're good to drive any line down to 1000 ohms.
Now, is performance at 1000 ohms? It's certainly possible... But so is anything.
I'm speculating here: a 1k ohm load being driven directly from the chip, I do think performance wouldn't be ideal. I think that starting somewhere in the 2-4kHz range, distortion will begin rising from -94dB, and quite steeply!

Most op-amps (especially Ti op-amps {this is a TI DAC}) give a THD+n plot from 20Hz to 20kHz at different impedances and gains. Usually it's 600 ohms and 2000 ohms with a gain of 1 and -1 (inverting), but sometimes other impedances are given, like 1000 ohms and 5000 ohms. Even dedicated top-quality audio-grade op-amps (dash-dash!) usually aren't performing their best at 600 ohms - distortion starts to rise around 3kHz, and this usually isn't just to a few dB higher by 20kHz - often distortion is up to 20-30dB higher by 20kHz. 2000 ohms is usually a pretty different story. There is still a rise, but it comes later and is smaller. Much later, like at 12-14kHz, and much smaller, like 3-5dB (when the first harmonic is out of hearing range, that's a good thing!) Remember, these op-amps we're talking about are quality, high performance, standalone devices, op-amps which cost as much as the entire PCM5102A. Conveniently, the PCM5102A's datasheet doesn't provide us a THD+n vs. 20Hz - 20kHz chart. It's safe to assume that the included gain stage within the DAC isn't going to be as good as dedicated op-amps which cost an entire 5102A, we should at least aim for twice the resistance with the DAC's output. If the PCM 5102A is loaded with 1200ohms, its distortion rise would occur where the dedicated op-amp's does with a 600 ohm load (it's not guaranteed, just a safer target). Since optimal distortion rise of dedicated op-amps is not until 2000 ohms, the apparent load of the op-amp following the 5102A should be made to be at least 4000 ohms. If one wanted to try to guarantee best channel separation, that's usually around 5000 ohms for standalone op-amps, doubling this to 10,000 ohms could be a help, or maybe not because crosstalk occurs somewhere else within the circuit.
So performance of the 5102A can be enhanced with an added amplification stage. It doesn't need any gain, and you need to ensure that the load that the 5102A sees is very high - as high as possible without meaningfully increasing noise - probably something like 5-10k ohms. Since the internal op-amps included are of lesser quality than those available for purchase

This internal I/V conversion + buffer is a feature of the 5102A. One would choose a 5102A over other DACs with external I/V to save on cost. Its performance is more than good enough for any device being made to a price point where two additional op-amps and a few surface mount capacitors/resistors break the budget, WAY more than good enough. While adding another op-amp to the 5102A can improve its performance, external configuration changes have much more limited potential vs. conventional DACs with external I/V

Anyway, the 5102A surprised/surprises me with the quality of its output. I don't know which op-amp RME chose to amplify it, but whatever it is, it works well with higher impedance headphones like mine (300 ohm)

Does anyone else have experience with the 5102A? If so, what do you think of it?

(sorry if there is a typo, I will proofread and correct tomorrow - seeing your own mistakes is a lot easier after a couple hours, but in a couple hours I'll be sleeping!)

 

My first hifi interface, early 2000s, was an RME card, and I thought it was wonderful! In fact, I preferred its sound to my Cyrus CD player.

After Boaty MacBoatface, though, it's a little hard to take their naming convention seriously