[bogi]

I have no clue why he would mention maximum digital attenuation. That setting is only relevant if you have a 10v dac with extremely sensitive iems and you need to go even lower than -60db

As maximum attenuation I meant the lowest volume level.

If your range is (-60, 0) as on the picture of the above post then you get 20 individual steps of 3 dB, so very fine adjustment at digital level is not possible. If your area of interest is (-20, 0) or (-30, -10) then you get 20 volume steps of 1dB. I see on the picture @Born2Cuddle has set 0dB or -1 dB level in HQPlayer. Therefore I suggested him to raise the minimum volume level (max. attenuation level) in HQPlayer settings to allow finer digital setting in the volume area of his interest.

 

[Failed Engineer]

I know it's an "old" filter now, but I keep going back to poly-sinc-ext2 after experimenting with newer released filters. Maybe it's just a well worn in shoe but after being smitten by the NA sinc-medium for a bit I'm back to poly-sinc-ext2.

 

[LoryWiv]

I've noticed that certain filters' GPU utilization pattern with CUDA offload activated is fairly steady (EG - with 192 kHz FLAC output to SDM, poly-sinc-gauss-long stays constantly between 20-30%, poly-sinc-gauss-hires-lp stays steady at 30-40%) whereas other filters have much wider swings (EG - Sinc-L may fluctuate over a few seconds time from 10% to 75% GPU utilization). In both scenarios playback is stutter free and GPU temps stay reasonable at 60 - 62%, albeit with GPU fans active more often with Sinc-L. Is this a consequential difference for the GPU in terms of it's performance, health, longevity etc...or just an incidental and unimportant curiosity related to filter design? Running HQPlayer Desktop 5.51. My GPU is EVGA GeForce RTX 3080, 10 GB version. Thanks for any help understanding this CUDA-GPU behavior with different filters and whether it is consequential or not. Always trying to learn!

 

[bogi]

Is this a consequential difference for the GPU in terms of it's performance, health, longevity etc...or just an incidental and unimportant curiosity related to filter design?

It is unimportant, if playback is going well. Similarly with CPU utilization. I am using computers intensively for decades and I didn't observe a relation between CPU/GPU utilization and longetivity.

Since computers are protected against overehating on more levels, you are not able to cause a damage by normal computer operation. The first level of protection is OS level with its thermal throttling. You would observe it as temporary stuttering in playback (maybe 10 - 15 seconds) and then playback continues without stuttering. As result of thermal throttling, no process is interupted, no data is lost, everything continues as run before. Of course, real time applications like HQPlayer, which cannot wait, are affected. That's the whole damage what can happen. If you are getting repeated series of dropouts with very high CPU/GPU utilization, thermal throttling si likely the reason. There are many apps which show CPU/GPU temperatures, it is appropriate to confirm the reason in such a case. If HQPlayer stutters in regular intervals but not in repeating series of dropouts, then it is not about thermal throttling but about insuccicient CPU/GPU clock speed for the requested real time operation.

So you would at first observe OS level thermal throttling as series of dropouts. Another level of thermal protection is part of chip design. If OS level thermal throttling would not be functional (did you hear about such a case?), or if it wouldn't be sufficient (it can happen under some circimstances), then processor level thermal shutdown would come into action. It is not possible to disable CPU/GPU thermal protection. This one of course means terminating everything what was running, without proper shutdown, so data loss could happen in the case of cached and not flushed disk write operations. But it still means no damage for CPU/GPU. Of course in such a case it is appropriate to look for a reason. Some hardware component may fail regardless on CPU/GPU utilization ...

Look for example here:
"If a GPU hits the maximum temperature, the driver will throttle down performance to attempt to bring temperature back underneath the maximum specification. If the GPU temperature continues to increase despite the performance throttling, the GPU will shutdown the system to prevent damage to the graphics card."

 

[LoryWiv]

It is unimportant, if playback is going well. Similarly with CPU utilization. I am using computers intensively for decades and I didn't observe a relation between CPU/GPU utilization and longetivity.

Since computers are protected against overehating on more levels, you are not able to cause a damage by normal computer operation. The first level of protection is OS level with its thermal throttling. You would observe it as temporary stuttering in playback (maybe 10 - 15 seconds) and then playback continues without stuttering. As result of thermal throttling, no process is interupted, no data is lost, everything continues as run before. Of course, real time applications like HQPlayer, which cannot wait, are affected. That's the whole damage what can happen. If you are getting repeated series of dropouts with very high CPU/GPU utilization, thermal throttling si likely the reason. There are many apps which show CPU/GPU temperatures, it is appropriate to confirm the reason in such a case. If HQPlayer stutters in regular intervals but not in repeating series of dropouts, then it is not about thermal throttling but about insuccicient CPU/GPU clock speed for the requested real time operation.

So you would at first observe OS level thermal throttling as series of dropouts. Another level of thermal protection is part of chip design. If OS level thermal throttling would not be functional (did you hear about such a case?), or if it wouldn't be sufficient (it can happen under some circimstances), then processor level thermal shutdown would come into action. It is not possible to disable CPU/GPU thermal protection. This one of course means terminating everything what was running, without proper shutdown, so data loss could happen in the case of cached and not flushed disk write operations. But it still means no damage for CPU/GPU. Of course in such a case it is appropriate to look for a reason. Some hardware component may fail regardless on CPU/GPU utilization ...

Look for example here:
"If a GPU hits the maximum temperature, the driver will throttle down performance to attempt to bring temperature back underneath the maximum specification. If the GPU temperature continues to increase despite the performance throttling, the GPU will shutdown the system to prevent damage to the graphics card."

Thanks @bogi for a clear and helpful response. I appreciate it!

 

[Ghoostknight]

As maximum attenuation I meant the lowest volume level.

If your range is (-60, 0) as on the picture of the above post then you get 20 individual steps of 3 dB, so very fine adjustment at digital level is not possible. If your area of interest is (-20, 0) or (-30, -10) then you get 20 volume steps of 1dB. I see on the picture @Born2Cuddle has set 0dB or -1 dB level in HQPlayer. Therefore I suggested him to raise the minimum volume level (max. attenuation level) in HQPlayer settings to allow finer digital setting in the volume area of his interest.

oh crap, thanks for that... the 3db jumps are kinda huge
EDIT: i changed the setting to -30, -10 and it still jumps in 3 db steps (with the mousewheel) but i just found out that you can use the arrow keys to jump in 1 db steps...


-----


First time trying out HQPlayer since 3 days and i played around a lot with settings... what i still remember what jumps out to me was:

1. closed-form filters have MASSIVE sub bass impact (beside sounding smoother just like the other 1M tap filters)
2. im coming from an FIR EQ in EasyEffects on Linux and the FIR setting in HQPLayer is actually quiet comparable with the FIR setting in the EasyEffects equalizer, i generally preffer FIR EQ over IIR, the difference is subtle but at the same times massive in terms of getting closer to realism (where linear phase is fir and iir is minimal phase)
3. poly- filter sound somewhat more gritty than the "standard" FIR implementation, im not a fan of them and at the same time they are the only real solution if you wanna sample from to anything
4. because the closed-form has massive bass impact which seems somewhat artificial i tend to prefer the sync-m settings in overall performance both are very close in terms of smoothness because of the 1M taps it seems

NOW ... so far sync-m i seem to prefer the most lets see what we have here in terms of additional options:
1. Gaussian - generally speaking (also with the poly filters) i dont seem to like gaussian that much
2. (Gaussian) Apodizing Filters - THIS is the crap! it smoothes edges (like high frequency crap to be honest, you will get what i mean after hearing the difference youself with the right recording...) in recordings i never heared with any system because the errors apodizing is correcting isnt corrected on any other platform i know of beside in hqplayer...
since there are overall only 3 or so apodizing filters i think my choice of choice is pretty clear (atleast for the moment) ... its the sync-mGa filter


the dither options do something but its not as pronounced as the difference between filters imo... its mostly how the background sounds, i went with the recommendation of the manual of NS9 for my 176,4khz upsampling setup and i also slightly preferred it in the comparison


NEXT UP: trying out DSD upsampling, since overall i feel like i tend to like my FIR setup with easyeffects for a particular reason : no long term fatique, i have setup the same EQ on both (tho implemention could atleast slightly vary) but something is still "aggregating" me in the sound compared to my previous setup, does someone share this? it might be the sync-m filter tho there are not much options left for me

 

[Ghoostknight]

If someone is curious what Apodizing errors might sound like, test it with this track its for a dumb reason one of my go to tracks (on speakers ) it reaches a few thousand errors for the whole song

tho it seems like many newer recording have these kind of errors, i feel like specially distorted/clipping songs

the question is... should we take these errors as part of the original recording or see them as "errors", for your ears on higher volumes these definitely feel like "errors", test them with the track above and a bit louder if you dont believe me

the saddest part is: you never heared them before but now that they are gone, you hear them gone! atleast this felt like it for me

 

[bogi]

since there are overall only 3 or so apodizing filters i think my choice of choice is pretty clear (atleast for the moment) ... its the sync-mGa filter

No, there are many apodizing filters, not only those ending with 'a'. Look into the filter table in the manual which is installed together with the application.
But MG/MGa are high quality filters, if you like them, why not to stay with them.

if you like sinc-MGa and sinc-M, you can also try for example sinc-Mx (if your computer can handle it), sinc-Lm, poly-sinc-hb-m, poly-sinc-gauss-halfband.

NEXT UP: trying out DSD upsampling, since overall i feel like i tend to like my FIR setup with easyeffects for a particular reason : no long term fatique, i have setup the same EQ on both (tho implemention could atleast slightly vary) but something is still "aggregating" me in the sound compared to my previous setup, does someone share this? it might be the sync-m filter tho there are not much options left for me

There are more modulators to try and they give different sound. The best ones are *-super, *-light and *ECv3 both of 5th and 7th order. At DSD256 and higher rate you can try also the 512+ ones. I would not generalize so much your initial impreessions until you try more. Tell me what's your target DSD rate and what are you missing in sound, maybe I could then give a suggetioin what to try.

 

[bogi]

the question is... should we take these errors as part of the original recording or see them as "errors", for your ears on higher volumes these definitely feel like "errors", test them with the track above and a bit louder if you dont believe me

Apodizing counter is related to source content regardless on your listening level.

Limited counter is related to volume setting in HQPlayer. Recording mastered up to 0dB level and played at full digital volume (0dB) will most probably cause signal limitation during HQPlayer upsampling and additionally in oversampling delta sigma DACs (not in the case of direct DSD) they likely cause intersample overflows. Therefore it is important to lower digital volume level in HQPlayer by few dB, as I explained few weeks ago in this thread. With for example -6dB you are totally safe and additionally DACs usually measure better (lower distortion) on that level than on 0dB.

 

[Patu]

Sinc-MG/MGa and Sinc-M sound different from each other. I prefer Sinc-M and it has been my choice for DSD256 upsampling (along with ASDM7EC-light) for some time now.

 

[Ghoostknight]

Apodizing counter is related to source content regardless on your listening level.

Limited counter is related to volume setting in HQPlayer. Recording mastered up to 0dB level and played at full digital volume (0dB) will most probably cause signal limitation during HQPlayer upsampling and additionally in oversampling delta sigma DACs (not in the case of direct DSD) they likely cause intersample overflows. Therefore it is important to lower digital volume level in HQPlayer by few dB, as I explained few weeks ago in this thread. With for example -6dB you are totally safe and additionally DACs usually measure better (lower distortion) on that level than on 0dB.

no, im talking about the apodizing errors, not limited errors, for me the apodizing errors just became more "obvious" while listening a bit louder (70 - 80 db is usually "loud" for me)

No, there are many apodizing filters, not only those ending with 'a'. Look into the filter table in the manual which is installed together with the application.
But MG/MGa are high quality filters, if you like them, why not to stay with them.

if you like sinc-MGa and sinc-M, you can also try for example sinc-Mx (if your computer can handle it), sinc-Lm, poly-sinc-hb-m, poly-sinc-gauss-halfband.

oh maybe i have to look again, tho i think the sync-m* is the only 1M taps one

There are more modulators to try and they give different sound. The best ones are *-super, *-light and *ECv3 both of 5th and 7th order. At DSD256 and higher rate you can try also the 512+ ones. I would not generalize so much your initial impreessions until you try more. Tell me what's your target DSD rate and what are you missing in sound, maybe I could then give a suggetioin what to try.

i tried DSD today... oh boy with a sigma delta dac via USB the DSD upsampling makes even more of a difference than PCM upsampling and it somehow makes the music sound less digital/more natural/more analog

i settled with these settings after some "light" testing

tho i still have the worm somewhere... these settings work because poly filters can upsample non integer and i have set it to 48k DSD... is it somehow possible to set it to 48K 256x DSD and check some box so 44,1khz files gets converted to 44,1k DSD and 48Khz Pcm to 48khz DSD ?

with these settings i use it always upsamples to 48khz dsd

 

[suno23]

no, im talking about the apodizing errors, not limited errors, for me the apodizing errors just became more "obvious" while listening a bit louder (70 - 80 db is usually "loud" for me)


oh maybe i have to look again, tho i think the sync-m* is the only 1M taps one


i tried DSD today... oh boy with a sigma delta dac via USB the DSD upsampling makes even more of a difference than PCM upsampling and it somehow makes the music sound less digital/more natural/more analog

i settled with these settings after some "light" testing

tho i still have the worm somewhere... these settings work because poly filters can upsample non integer and i have set it to 48k DSD... is it somehow possible to set it to 48K 256x DSD and check some box so 44,1khz files gets converted to 44,1k DSD and 48Khz Pcm to 48khz DSD ?

with these settings i use it always upsamples to 48khz dsd

In OUTPUT tab theres a box to check for adaptive rate just make sure its checked and not greyed

 

[Ghoostknight]

T

In OUTPUT tab theres a box to check for adaptive rate just make sure its checked and not greyed

AH! thanks alot i knew i missed some setting

 

[LoryWiv]

i tried DSD today... oh boy with a sigma delta dac via USB the DSD upsampling makes even more of a difference than PCM upsampling and it somehow makes the music sound less digital/more natural/more analog

That is my exact experience...once you start converting high quality FLAC to DSD with HQ Player, it is hard to go back! Enjoy.

 

[bogi]

no, im talking about the apodizing errors, not limited errors, for me the apodizing errors just became more "obvious" while listening a bit louder (70 - 80 db is usually "loud" for me)

I was talking about both to show the difference between them, therefore my reaction contained two paragraphs.
Of course, playing louder reveals more low level details, but long term can be inconvenient.

i tried DSD today... oh boy with a sigma delta dac via USB the DSD upsampling makes even more of a difference than PCM upsampling and it somehow makes the music sound less digital/more natural/more analog

Yes, that's a typical experience of people who really tried it.

tho i still have the worm somewhere... these settings work because poly filters can upsample non integer and i have set it to 48k DSD... is it somehow possible to set it to 48K 256x DSD and check some box so 44,1khz files gets converted to 44,1k DSD and 48Khz Pcm to 48khz DSD ?

If your DAC is capable of 48k based DSD rates, check [x] 48k DSD and [x] Adaptive rate - fully ticked, not half ticked (greyed).