Just remember to keep eye on the Apod-counter on HQPlayer interface. If it increments to figures over 10 during the playback, the source content contains errors that can be only corrected by using an apodizing filter. Otherwise you know you are listening to “corrupted audio”.
3 Likes
Hm, thanks for making me aware of that, I had no idea that HQPlayer had a feature which seems to be similar to clipping detection but for source errors. Interesting.
What would the reason be for those errors contained in the original file? Is it to do with MQA files or lossless FLACs as well? If the latter, then how does the software know what is an error and what is part of the intended original file? Or were you referring to errors not contained in a file itself but created during the process of modulation/oversampling?
Usually it is sourced from the original recording ADC, but sometimes also byproduct of the production process.
It applies to lossless FLACs as it originates from the source, before file has been encoded to something. HQPlayer is running analysis algorithms on the source content to detect this, it runs while playback is ongoing, so you will know the result as the playback proceeds.
MQA also has errors, but in a different way, so these are no indicated by this particular analysis process.
RedBook content is usually worst.
This is from the source content, the analysis runs before any DSP is performed on the content, so you will always see the analysis results regardless what processing you apply. But you know you are safe in such cases if you are using an apodizing filter.
If you use an apodizing filter for content that doesn’t need one, there’s no harm. The only harm is playing content that would need an apodizing filter without such.
sincMX in 705/768 PCM was my favorite on the X26P.
2 Likes
i2S… pins are fine… was working correctly with J River…
Good suggestion… see in now…
@jussi_laako what would be your basic recommendations for the La Voce DAC? (La Voce S3 Discrete R2R ladder DAC). Thanks!
I would run it at 352.8/384k PCM. Use either NS9 or LNS15 noise-shaper. I’d recommend NS9 for this case. And due to lack of measurement data, I would err on safe side and set DAC Bits to 16.
Very helpful, thank you very much!
Thanks for the suggestion, I’ll give it a try. From the technical data in the manual, I can see that it’s a “constant time version of Sinc-M”, “filter length is constant in time”. What does that actually mean? I will of course find out with my own ears what it means for actual playback, but I’m intrigued by the theory behind it and the implications of one filter’s length being constant and the other’s not.
@jussi_laako Tagging you as my hunch is we’ll likely need your expertise. 
Filter length in time is number of taps in filter in relation to sampling rate. In this case in relation to conversion ratio which in turn is related to sampling rate.
1 million tap filter at 768k is 1000000 / 768000 = 1.3 seconds (roughly). Same number of taps at DSD256 rate is 1000000 / 11289600 = 0.089 seconds (roughly), thus 89 milliseconds.
For sinc-Mx reference rate is 16x (705.6/768) at which it is 1 million taps. And then at DSD256 rate that means it is 11289600 / 705600 = 16x longer and thus 16 million taps.
Of course since both the amount of data to process (samples per second) and also the filter length grows as function of sampling rate, the processing power demands grow exponentially.
Most HQPlayer filters are “constant time” vs ratio.
2 Likes
Tell ya, after all that… the PCM 768 really outshines anything I can get on DSD filtering - without a pause / stutter…
Lots of options here…
#jussi or anybody else have recommendations for the Matrix X / I use a USB bridge with II. with a pair of maze elite headphones.
Thank you for your help!
JJ
Can you play native DSD with JRiver or Roon without HQPlayer?
I have not… not sure if that is even an option…
You should try as they both can play native DSD.
I2S pins map can be right for PCM but not for DSD … I suspect this can be the issue
1 Like
I switched my Roon core (and HQPlayer server) from NUC 8i5 (i5 8259U + 16gb RAM + Win11) to Lenovo ThinkCentre m70q Gen 2 (i5 11400T + 16gb RAM + Win11). I use iFi Zen Stream which is limited to 358khz / DSD256. The 8i5 was able to do DSD256 with ASDM7 (poly-sinc-gauss-long) but EC and ECv2 stuttered. Now the m70q has no problem running DSD256 even with the ECv2 (poly-sinc-gauss-long). I haven’t tried the heavier filters yet but it’s good to be able to run the best modulator now. I’ll continue testing, I haven’t bought the license yet.
2 Likes
I’m still amazed at how good Sinc-L sounds with my gear (Focal Utopia, Hifiman HE1000se). I didn’t think that at this stage of the hobby I would come across something that would yet again make me want to re-listen to every single song in my collection. And that’s exactly what’s been happening the last few days.
Thanks @jussi_laako for bringing such sound to the masses. Clarity, separation, DEPTH, and zero harshness/glare… stunning.
7 Likes
Still got some stuttering when upsampling HiRes content (24/96) to DSD256 with ASDM7ECv2 and poly-sinc-gauss-hires-lp. I couldn’t believe it since procexp showed ~15% CPU usage while playing. Well I tried to change the default Multicore DSP setting from gray box to checkmark. Now CPU usage rose to 25-30% but there’s no stuttering anymore? Is HQP possibly now properly utilizing all the cores of the CPU ?
is it with PCM or SDM for you?
@IgorSki PCM. I’ll experiment more with DSD later but so far, on my particular gear, I’ve not liked it enough to justify the extra electricity consumption etc. Sinc-L on PCM sounds like what I imagined Roon’s upsampling on steroids would sound. Which is why I bought a new PC and got HQPlayer in the first place, so mission accomplished. 
1 Like