Discussion in 'Headphone Amplifiers and Combo (DAC/Amp) Units' started by k.e., Nov 5, 2016.
Direct link to download: http://www.rme-audio.de/download/fut_madiface_win.zip
I installed the latest firmware update on an ADI-2 DAC, and it solved all my Windows 10 problems. The UAC2 driver now works, and the audio dropouts I had when connected through a hub went away. A very good update indeed.
I have put my thoughts on the SE vs BAL HP output under a quote as after discussion a few posts on and further listening I do not feel they were entirely accurate.
That’s exactly why I opted for pro version - balanced out with HD800 is better to my ears than SE and makes it pretty solid AIO unit without immediate need for external amp.
I’ve put these impressions under a quote as there was a mistake in the main output setting on the RME DAC. As can be seen in my unwitting impressions and the following posts dynamics and micro-dynamics can suffer terribly. I will have to start over.
The NOS filter was selected.
I am not hearing huge differences between my ADI-2 Pro and Gungnir Multibit G5, especially with regard to dynamic range. DAC mode, Auto Ref Level on, +0.0 DBr. PC (max volume) > DAC (USB) > Black Widow 2 > HD6XX. All SE inputs/outputs. Manual (imperfect) level matching.
I will say the dynamic range does noticeably suffer quickly when I start to move away from 0 DBr (with +24 DBu gain set by Auto Ref) and compensate with the volume pot on the BW2. I don't quite understand all the underlying technology, but it feels like the ADI-2's digital attenuation is what is causing all these conflicts between what I read from others and what I hear myself.
Not trying to contradict anyone. Only offering an additional data point from someone with mediocre ears.
Thanks @geniekid, it’s good that you commented because I done fucked up. The unit seems to auto switch what the volume knob is controlling each time you insert or remove various connections. I thought the main output was set to it’s defaults but it was way off (EDIT: I believe volume had gotten set somewhere between -30dbr and -38dbr, with Autoref On). I’ve reset the unit again and the defaults are -15dbr Volume, Auto Ref On, and +13dbu Ref on the Main Output.
I will have to redo the comparison another day as there was enough of a difference to potentially invalidate a lot of what I’d written.
You raise a good point regarding inconsistencies and these settings. Can anyone comment on what the correct settings should be for comparison to other gear like the Gungnir Multibit, are the defaults sufficient?
There's some discussion of this here, specifically with regards to avoiding digital attenuation (read the notes on the "OVR" indications as well). But @geniekid's pointers to set Auto Ref to ON and volume at 0 dB for your chosen output is really what you're after.
Some interesting observations about the RME ADI-2 Pro balanced output using digital attenuation.
These are measurements with ADI-2 Pro set for +19dBu ref:
Digital Volume at 0 dBFS:
Very good performance to begin, though dominated by 3rd harmonic.
Digital Volume at -10 dBFS:
Whoa! Now 2nd harmonic dominated and THD has dropped significantly!
Digital Volume at -20 dBFS:
Only 2nd harmonic remains above noise floor.
Digital Volume at -30 dBFS:
2nd harmonic barely above noise floor.
I'm finding I enjoy having amplifier currently in use set for unity gain and running the ADI-2 Pro between -30 and -10 dBFS with a tendency centered around -20 dBFS, depending on given track RMS compression (mastering squash level.)
Interesting. Maybe your input is getting overloaded, or anolog out gets dirtier on higher levels?
Just from intuition, square waves have odd harmonics and no even harmonics. Maybe there is some slight clipping @ 0 dBFS (+19dBu reference)?
Still, -115dB in THD+N @ -10, -20, and -30 dBFS is pretty awesome. Even -105dB of THD+N (in this case @ 0 dBFS) is very good. As reference, I think an APX515 and an APX525 have a -106 THD+N and -110 THD+N specification respectively.
I believe one could easily use the ADI-2 Pro FS as a measurement device. In some ways it's more attractive to me than the recent RTX6001 which I believe uses similar flagship AKM devices (which I think lead the market in THD+N performance).
@atomicbob, what tool are you using for your measurements? I read "HpW" but I do not recognize it.
The measurements are ADI-2 Pro loopback balanced. It is possible the ADI-2 Pro A/D at +19dBu ref is slightly overloaded compared to the ADI-2 Pro DAC at +19dBu ref. When I have more time this will be something I check against the dScope (been on my round tuit list for almost a year now.)
re: slight clipping, see reply to Hrodulf above.
Tool is HpW-works:
Those of you who bought the RME ADI-2 Pro didn't know you were also acquiring a measurement system, when you add the optional $300 US SW tool from HpW-works. Just be careful about measuring amplifiers with greater than 12.28 Vrms. You will need outboard attenuators to avoid letting out the magic smoke.
I recommend considering +13dBu as the maximum setting with the ADI-2 Pro.
I achieved the best results with the ADI-2 (DAC, not Pro, but I doubt there's a significant audible difference in this case) XLR outputs at between -15 and -5dB, and using the "slow" filter. Control volume with external amplifier potentiometer. To compare fairly to other devices, definitely turn off all DSP (loudness, bass+treble knobs at 0, and PEQ either flat or turned off).
After that, I like to try matching the Harman curve, basing the numbers on Tyll's articles and measurements, as I don't have any fancy measurement equipment.
In light of @atomicbob's measurements, and the fact that I'm as sick as a very sick parrot, that's feeling quite poorly, and would even turn it's beak up at a nice, juicy, cuttlefish, so have little better to do this weekend than evaluate gear ... I spent the day playing with different settings on the RME ADI-2 DAC.
To do this, I pre-programmed the settings I wanted to evaluate/compare into the ADI-2 DAC, mapped them to the four preset buttons on the remote, and had my girlie randomly switch settings at pre-determined intervals (with her recording them as she did so), so all I had was my/her notes to come back to.
Short version/takeaways ...
Dynamics, micro-dynamics in particular, were fucked any time attenuation wasn't at 0 dB.
Listening fatigue/sensitivity (defined as "any time I wanted to REDUCE the volume setting on the amp") occurred faster at 0 dB than at -20 dB. This was regardless of whether I was at +19 dBU (ref) or +7 (+13/+1 if using SE/RCA outs, which I wasn't).
Getting the best out of the RME ADI-2 DAC seems to require, for me at least, an amplifier that has an input stage that's tolerant of rather hot signal levels.
"The Handmaid's Tale" is WAY too slow to watch in real-time*.
I suspect that @atomicbob needs to pair my RME ADI-2 DAC with his Pro, and compare the results. Because, clearly, he doesn't have enough going on ...
*No idea how that slipped in there, and I blame my girlie either way.
So for dynamics 0 dB attenuation is better but in order to avoid listening fatigue 20 dB attenuation is better sounds like a catch 22 to me.
I’d expect seeing lower linearity from an analog system at higher energy levels. In that case I wouldn’t call it an overload. It does fall out of the advertised spec and harmonic content is a tad higher than I’d like to see. Also we are currently looking at the sum of total system harmonic content, right? AD’s usually have a sweetspot to perform at their best, maybe it’s worth adding some extra voltage division to pinpoint what’s causing the extra crud.
Anyways, I got the ADi-2 Pro to run it heavily digitally attenuated, so this makes me sort of happy.
we can make that happen when you are feeling better.
Well, it's not like listening fatigue crops up particularly quickly even at 0 dB ... just faster than at -20 dB. It was still fine for the full run of "Carmen" (3 hours or so).
The real question, for me, is whether that slightly faster onset of fatigue is down to just the increased dynamic resolution (I always find it physically easier to listen to less-dynamic music for longer periods) or is due to the increased harmonics/distortion at 0 dB ... (despite them being very low level).
I don’t think there should be gross dynamic errors even at -20dB, I think RME hinted that their DSP operates at 42bit internal resolution and most high FFT measurements peg the noise floor at what looks like 32bits.
Separate names with a comma.