Thanks, Jussi.
I’ve dialed it back to Dsd128 and it’s all good so far.
Between dsd256 and 128 with EC filters… I see that 128 has less cpu load per core on average, but has greater fluctuations of temp.
Just an observation. Happy with 128 and can’t really tell the diff between it and 256.
I have a late 2013 Mac Pro 8 core and can do DSD256 with no EC modulators. However, dropping down to DSD128 with ASDDM7EC modulator is so much more sublime. Currently using sync-M filter. Enjoy the experience.
I had been using the same filters, but I tried the plain old DSD7 modulator and I find it preferable…but thats me.
Thanks, I tried the DSD7 with sync-M Filter at DSD256 does sound nice. I am also trying the DSD7 256 +fs
Out of curiosity, if we are using a non-integer filter, is there any benefit to keeping integer scaling enabled anyway? Would the samples from the base file be “preserved” as such or does it not matter/we should pick the highest sample rate possible?
Only benefit is faster initialization on some filters and somewhat lower CPU load.
That it sort of unrelated to the conversion ratio, having samples “preserved” doesn’t necessarily happen for integer ratios either. You can set “Adaptive output rate” to grayed, in which case highest output rate is used when filter allows it.
Another aspect is that when output rate is static, the DAC doesn’t need to be stopped and clock switched when source sampling rate family changes.
I would say used approach could likely depend on a case. If you upsample only to let’s say 192k, then I would likely keep output sampling rate fixed at 192k instead of switching between 176.4 and 192. When upsampling to DSD, OTOH, in most cases it is necessary to keep output rate multiple of 44.1k in which case output rate needs to be kept for example at 11.2896M.
The cases where I would keep switching output rate are more based on choice of filter where it dictates that the output rate needs to change between the families. If allowed by the DAC (support for 48k-base DSD rates etc).
I think I might be new HQPlayer buyer. Got cuda and everything working I think.
I dont know how things can escalate so fast 
But Im happy that I have something to “play” with.
The EC (Extra Compensation) DSD modulators are intensive. You will need to back off to DSD 256 to run them. They pair well with the ext2 filter imo.
The most intensive filters are the xtr filters. People generally use the -2s variants, which make two passes rather than one. The M tap filters are also intensive.
2 Likes
It looks like I have to go back to dsd128. 
It sounds very good still. Lof of new things to try.
1 Like
Sorry for spamming all the time, but is there problem if Im using convultion filter in roon and passing it after that to hqplayer?
I can’t hear a difference, but Jussi recommends using the convolution engine in HQP. It’s described in the manual.
1 Like
Is anyone using HQP for convolution in multichannel with mono wav filters generated from Audiolense? AL generates 14 filters for my 5.2 speaker setup. Roon’s convolution engine takes in cfgs to route and mix those 14 filters together. In HQP, I see only the ability to have one filter per channel. This means I can use only 7 of the 14 filters. How are you all dealing with this MCH discrepancy? @jussi_laako, maybe this is a good spot in the forum to comment on this point. Thanks. JCR
EDIT: Been playing around with the filters a bit more and reading more. I prefer closed-form-M over closed-form despite the extra delay it adds to playback.
Upsampling to 176/192 and dithering down to 19bits with NS9. All PCM. Running this into a Schiit MB DAC through BNC provided via an ultraDigital. Sounds pretty great!
Hi @Jeffrey_Robbins HQPlayer has a “Matrix Pipeline” functionality that can do the job. The GUI allows to implement whatever is in a Convolver .cfg file. See for example my setup with multiple filters applied to input channels.
If you need help in the implementation just post your .cfg. Keep in mind channels are numbered from 0 in .cfg format and from 1 in HQPlayer, it’s easy to make mistakes!
Very helpful, @alec_eiffel. A few followup questions, then:
-
Why did you pick 0.00 dB vs 1 Lin vs -1 Lin for gain levels and how did you pick that out from your cfgs? (Maybe showing your cfgs used to create the matrix you shared would be helpful.)
-
I note you did not use lines 5, 6, 9 or 10 in the matrix. Was this merely a preference? I thought the line numbering is merely for convenience and has no channel mapping or other effect.
Thanks. JCR
1 Like
Hard to find anything bad in sound with HD660S with convultion filters. I think I have to buy hqplayer.
1 Like
It’s so difficult to find a filter and stick with I find! I just love to fiddle. Anyway of late I have moved from DSD to PCM and am running Poly-Sync-XTRMP and LNS15. It is a little more clinical than EXT2, DSD7EC but I find the soundstage a little more three dimensional as well. Plus I can actually use my PC for other things at the same time as listening 
