Which HQP Filter are you using? [2015-2023]

Ive have room EQ (Convolution) before, it was built into my tag mclaren av32r. What I really meant was I wouldnt know where to start to measuring for HQP.

I have a SPL meter :slight_smile:

Youā€™ll need that too for setting & checking test signal levels!

A good start is with miniDSP UMIK-1 USB mic (make sure to use itā€™s calibration file).

https://www.minidsp.com/products/acoustic-measurement/umik-1

And then follow everything previously mentioned i.e. using a separate program for measuring and generating the convolution files for HQP to use/convolve. Either REW, Audiolense, Acourate.

Detailed questions on using each of those programs are best posted in their specific places, for the best help.

REW: A guide how to do room correction and use it in Roon

Acourate: https://www.audiovero.de/acourateforum/

Audiolense: https://groups.google.com/forum/?hl=en?hl%3Den#!forum/audiolense

Good luck!

Thanks. Yes I have used REW before with my SPL meter. Thanks for your input :+1:

1 Like

So are we saying that for 16 bit there are 2097120 taps and for 24 bit thereā€™s 3145680 taps???

Bits and taps are not relatedā€¦

I used your comment for the calculation. Is that not correct then?
.

705600 / 44100 = 16 which is the rate conversion factor. So you take the base number of taps and multiply by the conversion (upsampling) factor.

1 Like

I seeā€¦everydayā€™s a school day :slight_smile:

So its like a sliding scale then? A basic cd quality file of 44100 upsampled to 352800 has a factor of 8 (352800/44100 = 8) so that makes 1.048,560 taps

If the file was 96khz, it would have a factor of 3.675, which would then have 481,682 taps?

So it would seem the higher the hz the less taps. Im sure theres a reason why the higher the hz the less taps are neededā€¦unless Ive completely misunderstood lol

It is not the frequency, but the ratio, if you convert from 96k to DSD256 at 12.288M, then ratio is 128. Lower ratios need less taps, higher ratios more.

Not sure I still understand lol.

If Iā€™m up sampling to PCM 352.8, are the following calculations correct?

44.1khz = 1,048,560 taps
96khz = 481,682 taps?

Close, but you cannot use sinc-S/M/L to convert between rate families, so 96k needs to convert to 384k.

sinc-S:
44.1k -> 352.8k = 4096 x 8 = 32768 taps
96k -> 384k = 4096 x 4 = 16384 taps

sinc-L:
44.1k -> 352.8k = 131070 x 8 = 1048560 taps
96k -> 384k = 131070 x 4 = 524280 taps

2 Likes

With you now. Thanks for taking your time to explain :+1:

Now I have my new CPU I think I am just about set on filters now with Poly-Sinc-EXT2, ASDMEC7 and DSD256. I have also set up a very light login profile on my W10 workstation. My day to day profile loads a load of guff that isnā€™t needed to run Roon and HQP. It still works but if Iā€™m in for a listening session I use the new profile

That said I did have a play with Sinc-M and Sinc-L and liked those too so need to do some further listening of these once I have my ear tuned to the current settings. Sounds blooming lovely but isnā€™t the point of HQP that we can chase what we as individuals deem perfection? :slight_smile:

At some point too I intend to set up a dedicated HQP server but that is a little way offā€¦

Sounds blooming marvelous

2 Likes

Poly sinc xtr mp2s
Asdm7ec

My ryzen 9 3900x, 32gb ram, 2080 super has a hard time with dsd256. Stuttering intermittently. Wow what are the min requirements to run the above settings?? Dsd128 runs smoothly.

changing the modulator to asdm5ec, dsd256 runs smoothly.

Anyone heard any differences between Asdm5ec and 7ec?

Not me, but Iā€™m the first to admit my ear memory is shocking. I think it could depend on the DAC and system. At one point I thought DSD5 sounded warmer than DSD7, but I havenā€™t been able to distinguish the EC variants. Warmer in this context might just mean slightly less detailed.

My new Ryzen 7 3700X with 32gb runs ASDNMEC7, poly-sinc-ext2, DSD256 just fine so suspect you can get there with your Ryzen 9. Couple of things to check; I have not yet tested it without CUDA offload and am running an nVidia GTX1080 ti. I will test it without CUDA offload at some point however given the heavy lifting CUDA offload can do if you do not have an nVidia card this may impact it.

Also, make sure Multicore DSP is ticked and not blank or greyed out. Without setting this my playback was stuttering too (see screenshot).

Finally, in the early bios releases of the new Ryzen 3xxx series chips the turbo and multicore/multithread was not optimised. Make sure your bios firmware is fully up to date. My older Asus Prime X470 Pro had a new firmware as recently as last November still tweaking the performance of the 3xxx CPUā€™s.

2 Likes

I have just tested the 3700X with CUDA Offload switched off and can confirm it works just fine at ASDM7EX, poly-sinc-ext2, DSD256.

I did try my Ryzen 5 2600 with CUDA offload to a GTX1060 but that didnā€™t have the grunt. It will run ASDM7, DSD512 just fine thoughā€¦

3 Likes

Thanks, @dannybgoode! You were spot on about ticking multicore dsp! Works fine and plays smoothly now.

EC filters seem pretty good. I canā€™t really tell the difference from pcm in terms of dynamics. And the very top end retains the dsd ā€œsmoothnessā€.

1 Like

Good stuff. Yes ASDM7EC sounds just lovely. Still canā€™t quite decide on filters but ext2 does have a lot going for it :slight_smile:

1 Like

Ext2 is a strange beast. Itā€™s at the longer end of the time domain, meaning it has a sharper cut-off and good bandwidth rejection, yet it also has good transient response. Jussi has said he was a bit surprised by the result, which is an interesting observation about human auditory perception itself.

3 Likes