I’m practically only doing DSD output at DSD256 rate using ASDM7EC modulator.
I do PCM output only for testing purposes.
2 Likes
Always round down. Rounding up would pretty much remove the point of entire exercise. So 15 in this case.
Depends on DAC and conditions. So there is no clear answer.
If you use HQPlayer to produce SDM output, DAC Bits setting doesn’t apply at all. Only to PCM output. If you use PCM output to a delta-sigma DAC, you cannot fix it up the same way you do with R2R PCM ladders, you should be using HQPlayer to output SDM instead.
Either choice is fine, up to you which one you prefer.
1 Like
Thank you very much for the brief explanation of my questions, highly appreciated. As for now I am stuck with PCM due to thermal issue of my Mac Mini with higher loads. It works but starts to sound like a helicopter if you know what I mean 
As for the noise shaper I think I prefer the NS5 on my ESS dac upsampled to 384 and the LNS15 on my AKM dac upsampled to 768. Both do a great job imo.
Anyway here is a screenshot of my settings for my AKM dac configured to be used with PCM mainly. Maybe something else catches your eye that could be improved.
I really enjoy HQPlayer, best investment in a long time. I will shill the hell out of it 
5 Likes
poly-sinc-gauss-xla is special !
6 Likes
gauss-xla is about 13% better than gauss-long
But gauss-xla needs about 900% more CPU power !
The price you must pay for 13% better ! So expensive !
3 Likes
Told you gauss-xla is nice, but I use it with adaptive, can’t detect any difference but only about 25% of CPU power needed.
Your 25% total CPU loading number will be different for everyone. Depends on end sample rate, DSD/PCM, CPU/GPU model etc
For me it’s a 10x higher load than gauss-long
This may or may not be a significant issue for someone , depending on those factors mentioned
But it is about 13% sound quality improvement so is worth the price (10x CPU/GPU loading ) in my opinion 
what differences are you guys hearing between gauss long and xla?
How did you come up with 13% better?
Complicated maths. Like the maths of gauss-xla itself !
Are you all doing the gauss xla for 1x as well?
Yes for me. I only change 1x
1x I try more aggressive / faster roll-off filtering
Nx I use slower roll-off filtering. I keep Nx on short-lp and don’t change (that’s the default for every HQP new installation).
1 Like
Is there a simple explanation as to why slower roll off for Nx is a preference over faster roll off?
I just use Jussi’s default for Nx.
For higher sample rate music (which is what Nx is for) you technically CAN use a less aggressive filter since the ‘Nyquist’ frequency for Nx rate music is much further beyond that for 1x rate, outside what many consider to be the audible range.
The trade-off with using slow roll-off filter for 1x is less extended FR within audibility range. But the upside is maybe better time domain performance. No free lunch!
I’ve probably butchered the explanation, so wait for JL to correct it
And of course anyone is free to use whatever their ears prefer.
1 Like
Where does this figure come from? 
I don’t think it is “better”, just different…
1 Like
I tried some different shapers, and somewhat surprisingly I liked one of the simpler ones best: RPDF. Using it with ext3 now.
1 Like
Any post that ends with crazy face emoji 
shouldn’t be taken too seriously
Yes of course ! Goes without saying these days, especially with purely subjective impressions.
Subjectively better can be technically worse in some aspects - e.g. gauss-xla maybe objectively better than gauss-long in frequency domain but maybe technically worse in time domain !
But I have no idea
I hadn’t been doing much with my settings on HQP as I was struggling to run even ASDMEC5 at 128 for some reason so just pottered along with PCM upsampling to 768K with Sinc-M as I liked that. Don’t know why but things like Poly-Sinc-EXT2 just didn’t do it for me.
However just recently I tried it again and really liked it and hadn’t even spotted that EXT3 was put and liked that even more.
But, I really wanted to get SDM working as I felt my PC should be powerful enough - Ryzen 7 3700X with 32gb and a 1080 Ti. Anyway, reading through here I made a couple of changes - switched Multicore DSP to grey box instead of checked, ditto CUDA offload and made sure my HQP profile on my PC had all power saving switched off.
And that’s done it. Now happily running ASDM7EC 256 with Gauss-XLA for 1x and EXT3 for nX. Sounds absolutely immense. I have a new (to me) power amp and its the most ‘real’ I’ve ever heard music from a hifi.
Roon + HQP is a killer combo 
5 Likes
I keep coming back to poly-sinc-gauss-xla ASDM7EC 12288000/1/2
Mostly <50kHz but keep >50 on the same.
CPU i5-10600K with slight over clock to 3.9GHz. 100% stable for more than a month.
1 Like
If you can’t do DSD256-EC with gauss-xla but you can do gauss-long, an interesting comparison:
DSD256-ASDM5EC/7EC + poly-sinc-gauss-long
vs
DSD256-ASDM5/7 (not EC) + poly-sinc-gauss-xla
Obviously if you can do DSD256-EC with gauss-xla then nothing for you to see here