@jussi_laako Hello, I’d like to suggest/query a feature, don’t know if it makes sense, but I’d like to have a more refined strategy for the choice of filters in settings (for SDM).
For the moment we can choose a filter for “Oversampling 1x” and a second choice for “Oveversampling Nx”.
I remarked that my machine works easily with “gauss-xla” filter up to 96Kbps and fails when 192Kbps stream arrives.
With the current strategy we have to switch to gauss-long for example as as soon as a 96Kbps stream is processed.
Would it be possible to have a different strategy, example Drop list 1 “Low res filter” with an additional toggle button “Up to 96Kps” and the second drop list “High res filter” ?
Just a thoughts, you probably made the current strategy for any valuable reason I didn’t know…
In my opinion it is bad idea to add configuration options just to deal with borderline cases on CPU power. If you want to use gauss-xla for higher rates, better option is to make it work also for all hires content and that is better solved by more powerful CPU or by adding a powerful GPU. No matter what, similar borderline cases will always exist.
Idea behind 1x and Nx settings is that 1x content needs a steep filter because Nyquist frequency is so close to audio band. But “mid-res” and “hi-res” don’t have that issue and can use a bit more relaxed filter slope. So this option is not driven by processing power, but by other reasons.
It was what I thought, I didn’t have the whole picture to understand why things are what they are. Thank you Jussi!
My computer is a Ryzen 3700X with RAM DDR 3000 which which works far better since I replaced the stock CPU FAN cooler by a better one (and more silent at the same time). No overcloking at all.
I’m using the ASDM7ECv2 modulator and upsample in DSD 256 44.1.
Hi @jussi_laako I encountered a strange behavior when the matrix pipeline is enabled with provided PEQ txt files. Could you help on this?
When starting a track, sometimes there is a loud pop sound from either one channel, then that channel becomes no sound. I will need to replay the track, or retry multiple times. It does not always happen on the same track. When this happens, the Limited counter goes by 1.
Some facts and my settings:
HQplayer 4.15.3 on Windows 10
The PEQ filters apply up to 100hz only; no boost, only cut two peaks.
It is PCM playback without any upsampling. No dither as well.
The pop never happened when I set SDM playback.
I do set both min and Max vol to 0. But it works totally fine if I use convolution engine with the wav files generated from the same PEQ filters that I use to generate txt files for matrix pipeline.
Finally had the time to get this dialed in and audition the various filters. I’m saying goodbye to the Audeze Reveal plugin in Roon for my LCD-X 2021’s (even though the Oratory1990 PEQ isn’t an apples to apples comparison to the reveal plugin since Audeze doesn’t track the Harman curve exactly and there may be some other secret sauce going on in their .vst)
I’ve now got an incredibly clean signal path keeping all the DSP in the HQP matrix settings. After a couple days of trying my favorite filters I’ve actually settled on xtr-short-mp using the ASDM7ECv2 modulator. My Mac Mini handles xtr-short-mp at both 1x and Nx just fine since rolling back to 4.15.0 (IIRC it would only run -2s on Nx on subsequent upgrades)
As others have mentioned recently, if you haven’t tried xtr-short-mp give it a go, its the first filter for me that everything falls into place and there are no tradeoffs. Nice warm, organic sound and no fatigue whatsoever. Everything just sounds right to my ears. It should be a crime for my lowly X16 to sound this good!
For Mac Mini users 4.15.0 still seems to be the most optimized if you are running a borderline use case as the CPU load is evenly distributed across the 8 cores.
It does not work for you? It does not for me, have to stop at DSD7 + 256 fs. Or ADSM7. Thing is, sometimes sources at 24/192 cannot play with skipping (rebuffring?) every appr 5 sec. Have set forced multicore to “1”.
@jussi_laako per your request I have a good news for you. Very exciting! Big leap for the mankind. AMSDM7EC 512+fs now can upsampling to DSD1024 (source from 44.1 to 176.4) on my i9 system. i9 12900k w/ DDR5 w/ RTX3070 CUDA offload, installed Ubuntu Server 20.04 w/ hwe-edge and HQPlayer Embedded 4.28.3-107 AMD build. Although the 48 family is still a little bit stuttering (average once per 1min) but I believe it can be optimized.
