Thank you very much 
Can you may say what dac bits to use based on this measurement of the r26?
Thanks again
16 bits is probably safest (allowing for measurement errors etc)
1 Like
This seems to be a popular DAC at the moment.
Lots of people have been asking Jussi this question on different threads and forum 
3 Likes
@jussi_laako Hello, I have a Rockna Wavedream and I tend to prefer the Hybrid filter (Rockna own filter). The WD accepts accepts as max input frequency up to 384khz, then upsample internally up to 768khz.
I know that itās always better to experiment/test with listening but, technically, which HQP filter would your recommend to use to feed the Rockna used with hybrid filter?
In theory I would say one of the linear HQP filtersā¦then Rockna will do the last mile in Hybrid.
Or is it better (as general rule) to use only filters in HQP and leave the Rockna in NOS?
Thanks in advance
You can choose which combination you prefer. There is so much space between content and Rocknaās filter in this setup, that you can use any of the HQPlayerās filters you like. But in this kind of setup you should always use Rocknaās linear phase filter, this way it doesnāt mess with the HQPlayerās filter.
3 Likes
Thanks @jussi_laako
So I understand that, for the last mile on Rockna, I can use either Linear (if I want to reach the 768Khz) or NOS (if I am ok with playback up to 384Khz).
Hybrid and Minimal filters on Rockna should be avoided because it messes up (I guess āphase wiseā) with HQP filters.
Correct?
Does anyone have good results running the EC modulators for higher bit rates on DSD?
I know they are significantly more demanding but Iām throwing some decent hardware at HQPlayer and the CPU and GPU usage seems to be well within limits but I still get very frequent drop outs!
Specifically: AMD Ryzen 3900X with a GeForce 3090 feeding a Holo May. (Iām omitting my NAA there because Iāve done testing with directly plugging the May into the audio server as well as swapping cables etc!) Running poly-sinc-gauss-long and poly-sinc-gauss-hires-lp with ASDM7ECv2. All 256 bit rates work perfectly. 512 is ok at 44.1 and occasional drop outs at 48 depending on source resolution it seems. Any rate at 1024 gets frequent drop outs.
CUDA offload and Multicore DSP set to grey (so is adaptive). 20 bits set, and various buffers tried. Just to be clear Iām not criticising HQPā¦ loving the results Iāve been getting and I know Iām pushing the limits of what can runā¦ just intrigued if itās doable and if Iām doing anything wrong! Thanks to Jussi for such a cool piece of software.
Also, sorry for noob question: What is the meaning of the bit rate field, and the difference for example between 44.1 x 512 compared to 48 x 512 (other than, higher numbers = more frequency = better)
Try to set Multicore setting to forced (checked) and see if that gives you difference in performance.
Interestingly this really helps for 48 x 512 which is now smooth and CPU resource consumption looks a lot more spread out (thanks!).
Anything @1024 still drops out a lot.
No problem.
I think that 1024 needs insane amounts of CPU power to work smoothly. I wouldnāt be too concerned if it doesnāt work smoothly on your setup.
Yes, exactly, which ever you find better. Probably both provide good results although slightly different.
Yes, this is correct. If you have a non-linear phase filter in chain, it is best placed first in the chain. Otherwise youāll get somewhat strange phase mixture.
1 Like
If you can get 44.1k x512 running OK, you can use that one instead of trying 48k x512. My i5-11600 is similar, 48k x512 is so close to overwhelming the CPU.
x256 vs x512 is much bigger difference than 44.1k x512 vs 48k x512. Mostly the latter one matters if you prefer some filter which cannot do conversions between rate families.
2 Likes
Some people have got EC modulators working at DSD1024, but so far that requires 12th Gen Intel or newer i9 K-series that is seriously overclocked. So not easy thing to get running. With non-EC modulators thereās also plenty of stock hardware that can do it.
RTX3090 will certainly help as offloading filters lifts off quite a bit of load from the CPU allowing higher boost clocks for the fewer high loaded cores.
Once I get my hands on 7000-series Ryzen, I will certainly do some testing. Unfortunately 5x increase in electricity prices and significant inflation in past six months is significantly limiting what I can invest on testing hardware.
Makes sense. I think my goal will be to stick to EC modulators and 512 then play with filters depending on source and genre.
Totally understand the challenges with cost, especially if you have to bump up to an AM5 board. I hope you get there soon!
Try this one working for me on 139k alone. I canāt do any pcm-dsd1024 ECv2 though so youāre not alone.
Add images
Thanks for the suggestion. I get clipping with any of the EC modulators (v2 or not!), but the non EC ones work okā¦ currently running AMSDM7 512+fs at DSD1024 well enough. I guess now I just have to test whether I prefer EC at 512 or not
For reference this is also now with the 3900X set to 4.3GHz with simultaneous multithreading off to try and pull more from single core. Iāll try and OC it a little more later and see if that unlocks the EC modulators at all!
Clipping? Please ensure that your HQPlayer volume is not turned higher than -3 dBFS. If the āLimitedā counter still increments from 0 during playback, turn down the volume further until Limited stays 0 at all times.
Apologies Jussi, I used the wrong language. It is not volume clipping (limited count is 0, volume set at -3 dBFS as you say). I meant dropouts!
Continued here guys
Which HQP Filter are you using? [2023] - Audio Gear Talk / HQ Player - Roon Labs Community
1 Like