Just updated Audalytic AH90’s built in Debian streamer to the newest OS “trixie”
And I updated its built in NAA to v6.1.1
Working well.
5 Likes
Hey thanks again and my bad. My brain was associating k13 r2r with this thread since we were talking about it earlier.
People on some other forums are saying its the same as the k11 R2R and Warmer DAC architecture so with a THD+N of 0.0173%, it has a SINAD of 75.2 db.
The math ends up only being 12.2 bits of resolution? Seems disappointing but I doubt its audible?
In the spirit of the thread, supposedly the 1 bit DSD mode is properly working, and the K11 R2R firmware update has it working as well now.
That test is usually done with 24-bit or 32-bit test tone. Result may be different with 16-bit test tone and quality noise shaper
2 Likes
I can probably try K11/K13 given that it is relatively inexpensive. But for DSD it doesn’t seem too promising:
And the D/A stage photos seem to have just PCM R2R stage.
3 Likes
I should be able to test a K13 (which doesn’t need firmware update) next week
But yep doesn’t look good 
Likely be a couple quick measurements and if no good, back in the box 
2 Likes
@dabassgoesboomboom from the manual of the k13 r2r.
"DSD(DSD decoding mode):
The default option is D2P. When playing DSD format tracks, the device internally employs SRC to convert DSD to PCM before decoding via the R2R module.
When the DSD option is set to 1BIT: When playing DSD format tracks, no SRC processing is performed. Instead, DSD is 1bit DSD native decoding. (Note: When 1bit DSD native decoding is selected, the SAM: NOS/OS menu option is disabled.The software background will use NOS mode for decoding both DSD and PCM)."
It is weird none of this is mentioned on the marketing material or product info photos as far as i saw. Also, I know the built in PEQ, etc is bypassed in this mode.
We shall see. The ultimate truth is in the measurements.
I know some people don’t care about measurements but the same people want to know if their DAC is true DSD Direct 
Are you able to get your hands on a k13 r2r? Idk where youre located but im in southern California and could work something out if needed. I would like to know personally and its worth an investment to me. 
I got the k13 today!
When you switch it from DSD 1bit mode to DSD D2P mode, the volume increases by a few dB. Not sure this is a signal that it is handling DSD directly when in 1BIT mode?
It is possible yes, depending on the implementation. But it is not always clear as there may be gain applied to the PCM conversion. If not explicitly applied, DSD should be 6 dB quieter than PCM, also when converted to PCM.
But for example T+A DACs apply 6 dB of analog gain for the DSD conversion, while Holo Audio doesn’t. So on T+A the PCM and DSD have matched levels, while on Holo Audio DSD is 6 dB quieter than PCM.
3 Likes
Okay interesting. Also I discovered right now that the K13 R2R DAC seems to be outputting DSD 256 even though I’m using a Mac Mini M4 over DoP. How is this possible if the K13 is only capable of 384khz?
Actually, I just tested 768khz and it is working? The specs clearly say 384khz is the limit of this DAC?
It plays 768khz in both OS and NOS mode. And also DSD 256 in 1BIT and D2P mode.
How does this work? Is HQplayer somehow bypassing the 384khz limit? I’m assuming the designers made it 384khz for a certain reason even though it can do 768?
I’ve never heard of a DAC playing higher rates than it’s listed to do.
If it can do 705.6/768k, then it can also do DSD256 over DoP. There are some DACs on the market that for this reason support 705.6/768k on the USB interface. For example iFi has such firmware versions for some of their DACs. Going above 384k PCM on those cases typically produces just silence, while DSD256 produces sound.
I have also one Musical Fidelity DAC where maximum PCM input is 192k, but it can also take in and play DSD256.
So not an unusual case…
I see thank you for the insight.
I messaged support to see if there was any technical / quality issues in playing 768khz PCM. The DAC supports DSD256 native according to the manual so I don’t see any issues there.
AI said the following.
" Even in NOS mode, the FiiO K13 R2R hardware is officially rated for a maximum of 384kHz.
- The Resistor Ladder Speed: In an R2R DAC, the resistors have to physically switch states for every single sample. At 768kHz, they are switching 768,000 times per second. Most affordable R2R ladders (including the one in the K13) cannot maintain 24-bit accuracy at that speed.
- The Internal Limit: While the USB receiver can “handshake” at 768kHz, the internal FPGA (the brain that tells the resistors what to do) likely downsamples the signal to 384kHz before it hits the ladder to ensure the resistors can keep up without massive distortion."
Obviously its AI so its probably wrong lol. I know other R2R dacs handle 760khz fine.
There’s no discussion on this since its a newer DAC and I don’t think many are oversampling with it to become aware of this scenario.
I would like to run DSD256 if i can and 768khz when i feel like using PCM, but was wondering if there was any risk like AI is trying to imply.
As always, thank you for the help and wisdom throughout the years!
Yeh I wouldn’t even waste time AI searching for this. This is a niche hobby we are in. AI is giving a lot of wrong answers related to HQPlayer, so not worth the time even typing out questions and reading.
Even if music is playing, seeing the measurements would be interesting… I’ve measured many DACs where music plays at DSD512 but the measurements compared to DSD256 are shocking… Like the DAC looks broken based on the distortion.
Someone just playing music would have no idea if they weren’t aware of the measurements somehow.
1 Like
AI answers are generally rather misleading.
But technically it is true that you cannot keep increasing R2R sampling rate without limits, because of settling time requirements. If you want to maintain the ladder precision, the ladder’s output must settle within ±1/2 LSB voltage in fraction of sample time period.
Less bits you have, higher speed you can run the converter. At the other end you have just one bit and you are switching between two voltage levels. But even then there is optimal rate limit and exceeding that will cause degradation of the output signal. This is why DSD256 is optimal rate in many cases.
The optimal rate depends on the converter hardware. Usually 705.6/768k is not yet a problem though.
1 Like
Fiio K13 R2R. Only brief measurements, too crap to spend too much time on. I wanted to get my AH90 hooked up playing music again ASAP after seeing these measurements 
Even 16-bit PCM output is poor. DSD Direct works but is pretty poor.
Its a shame because this is one of the best DACs I’ve ever seen, visually, with the lighting and glass top.
Oh well. It shows that discrete is not easy - Holo done a great job.
I would EASILY recommend SMSL D200 ROHM over this.
DSD256:
PCM353
PCM705
3 Likes
Now that I have my new Cosmos ADCiso which has practically no more ESS IMD hump (negligible with the APx555) I’m gonna re-measure the SMSL D200 with Cosmos ADCiso + Cosmos Scaler.
This new Cosmos ADCiso also comes with correction/calibration .cal file for REW which designer says makes it flat freq response to 380kHz.
I could now try to make a DAC correction too but I dont have the skill and cant be bothered
Don’t forget T+A! Which is at least as good as Holo Audio.
I have also other pretty well performing discrete implementations like Marantz and TEAC (given HQPlayer’s modulator).
3 Likes
Sorry, but even with that “calibration” file (this is not just about frequency response), ESS based measurement hardware like Cosmos ADC cannot be used for making DAC corrections. They have way too many issues, it is just due to the way the ADC works. Biggest challenge for making DAC corrections is to find suitable calibrated measurement hardware.
And if you use Cosmos Scaler it will add even more errors to the picture. And Cosmos ADC alone has too low input impedance (you need about 100 kOhm).
Thanks for checking in it! Sucks to hear 
Did you get enough info to figure out what the best DAC bits setting would be for it?