OK, so I tested most filters and my conclusion is that the MB2 from TNT and bam_comb1 from bambadoo feel closest to the source right now, maybe with the edge at the bam_comb1, it is really nice, keeps the emotional impact and has very little color.
I tested in a loopback configuration (blurring the results due to additional AD and DA conversion of course, but still the only way to do a blind test easily) so I can listen against the original file.
The milder filter slopes of spzzzzkt sound a little bit more coloured than the source, but MinPhaseSlowrev1 is a pleasure to listen to, kind of a bit more exciting than the source file. Very nice!
The winner in the exciting category is NOS, but this one is really coloured, with a very alive bass and prominent highs and a kind of breathing sound, too dirty for me, but I can understand why some people like that configuration.
With the exception of the NOS, all filter configurations sound kind of close to each other (in my setup), and no filter really sucks.
Thanks to all the filter brewers, keep up the good work!
I tested in a loopback configuration (blurring the results due to additional AD and DA conversion of course, but still the only way to do a blind test easily) so I can listen against the original file.
The milder filter slopes of spzzzzkt sound a little bit more coloured than the source, but MinPhaseSlowrev1 is a pleasure to listen to, kind of a bit more exciting than the source file. Very nice!
The winner in the exciting category is NOS, but this one is really coloured, with a very alive bass and prominent highs and a kind of breathing sound, too dirty for me, but I can understand why some people like that configuration.
With the exception of the NOS, all filter configurations sound kind of close to each other (in my setup), and no filter really sucks.
Thanks to all the filter brewers, keep up the good work!
Paul,
The last thing I intend to do is to change the intentions of this thread but thought I would mention what I did last night - finally getting a digital system back together.
Since I have not tried changing anything in the DAC - just wanted to hear if it worked I used dBPoweramp to upsample to 24/192 a few recordings and it is pretty dramatic the difference that makes skipping the first set of filters.
Tonight I will (hopefully) install the proper firmware and your filter files and hear what that does. Seems a worthy comparison even though the thought of a 1 TB disk becoming a 100gB disk is a little daunting.
I hope this SOEKRIS project does not end up requiring THAT though I could live with it.
Willl be interesting to revisit what I listened to last night with your new second stage filter in place.
By the way - does the DAC retain the original filters? Do we need to copy them to retain them? Of course, the question could well be who wants to retain them?!
The last thing I intend to do is to change the intentions of this thread but thought I would mention what I did last night - finally getting a digital system back together.
Since I have not tried changing anything in the DAC - just wanted to hear if it worked I used dBPoweramp to upsample to 24/192 a few recordings and it is pretty dramatic the difference that makes skipping the first set of filters.
Tonight I will (hopefully) install the proper firmware and your filter files and hear what that does. Seems a worthy comparison even though the thought of a 1 TB disk becoming a 100gB disk is a little daunting.
I hope this SOEKRIS project does not end up requiring THAT though I could live with it.
Willl be interesting to revisit what I listened to last night with your new second stage filter in place.
By the way - does the DAC retain the original filters? Do we need to copy them to retain them? Of course, the question could well be who wants to retain them?!
Soekris has posted the original filters in the "original thread"
http://www.diyaudio.com/forums/vend...magnitude-24-bit-384-khz-149.html#post4216998
http://www.diyaudio.com/forums/vend...magnitude-24-bit-384-khz-149.html#post4216998
The materials for the listing test are always 16/44 or you tested also some other Fs ?
Certainly we will see the links to the different filters in the Wiki though there are not so much pages in the original thread and this one about the filters... not the ECDESIGN John' thread
yet !. Ah, I will buy a USB to serie adaptator and read DimDim's blog !
Last edited:
The materials for the listing test are always 16/44 or you tested also some other Fs ?
Certainly we will see the links to the different filters in the Wiki though there are not so much pages in the original thread and this one about the filters... not the ECDESIGN John' thread
Ah, I will buy a USB to serie adaptator and read DimDim's blog !
Just 44/16 and just one track, I was actually only listening for the difference to the original file, not the most emotional sound or so....
If you just listen through that DA without a reference, you might easily prefer some colored conversion like NOS, because it kind of sounds like an enhancer
FYI: There are several software players that will upsample on the fly so that you wouldn't have to save upsampled versions. HQPlayer seems to be the best of the bunch. It will resample all the way up to DSD....
Just so my plans are known:
I'm working on firmware to expand the filters, my plans are:
A) To support 4 FIR1 filters for each sample rate types than can be selected between over the serial port, default will be 2):
1) Linear Phase
2) Medium, optimized mix between Linear Phase and Minimum Phase
3) Soft, mostly Minimum Phase but not that good alias rejection
4) No filter, also called non oversampling, no alias rejection
There should be no reason to upsample elsewhere, the dam1021 have plenty of DSP power to do it in realtime, once the filters are done and tuned.... In principle a PC is way faster, but you can't really load it down to much doing DSP work without making it unusable as a PC....
B) The FIR2 filter will also be modified to be pretty soft....
There will come an official soekris release, all done using rePhase, but it take a little time to make and test all those filters.... Keep working on them here as it give me a lot of input what to do....
I'm working on firmware to expand the filters, my plans are:
A) To support 4 FIR1 filters for each sample rate types than can be selected between over the serial port, default will be 2):
1) Linear Phase
2) Medium, optimized mix between Linear Phase and Minimum Phase
3) Soft, mostly Minimum Phase but not that good alias rejection
4) No filter, also called non oversampling, no alias rejection
There should be no reason to upsample elsewhere, the dam1021 have plenty of DSP power to do it in realtime, once the filters are done and tuned.... In principle a PC is way faster, but you can't really load it down to much doing DSP work without making it unusable as a PC....
B) The FIR2 filter will also be modified to be pretty soft....
There will come an official soekris release, all done using rePhase, but it take a little time to make and test all those filters.... Keep working on them here as it give me a lot of input what to do....
kurlbo,
On the fly would certainly be more efficient but I do not like the idea of the CPU doing that much work while playing music.
Of course, processing a 24/196 file requires more work than a 44, but the other is on a different level entirely.
Hope it all becomes a moot point and feel confident it will.
Take care,
On the fly would certainly be more efficient but I do not like the idea of the CPU doing that much work while playing music.
Of course, processing a 24/196 file requires more work than a 44, but the other is on a different level entirely.
Hope it all becomes a moot point and feel confident it will.
Take care,
I might have missed a previous post this but I'm curious about your setup. Could you expand a little, as I'm not clear what your reference is and how you are making the comparison.
I'm sure it depends on the computing power of your computer but with a relatively new and mid level machine it shouldn't take more than 20% of your cpu capacity....
From computeraudiophile.com:
"Head to Head JRMC19, Foobar+SACD (and HQ Player) doing Redbook to DSD and native DSD
Geoffrey Armstrong of Sound Galleries in Monaco and I did a comparison of players doing Redbook to DSD, Hi Res to DSD and native DSD playback. The results were not what we would have expected a couple of months ago:
Redbook to DSD conversion - voices and acoustic instruments the big winners, more texture and detail
JRMC 19 Redbook > DSD 128 results are very good (CPU load 13%)
Foobar + SACD in Integer (?) mode > DSD 256 is even better, jaw dropping for me as I have never heard Redbook this good. We almost gave up on Foobar until we switched to non 32 bit Floating Point mode (therefore Integer mode ?) in the SACD plug-in, and then the SQ difference was night and day (CPU load 13%)."
So it seems, at least with Jriver and foobar, you might only use 13% of your CPU. That wouldn't seem to be significant....
"Head to Head JRMC19, Foobar+SACD (and HQ Player) doing Redbook to DSD and native DSD
Geoffrey Armstrong of Sound Galleries in Monaco and I did a comparison of players doing Redbook to DSD, Hi Res to DSD and native DSD playback. The results were not what we would have expected a couple of months ago:
Redbook to DSD conversion - voices and acoustic instruments the big winners, more texture and detail
JRMC 19 Redbook > DSD 128 results are very good (CPU load 13%)
Foobar + SACD in Integer (?) mode > DSD 256 is even better, jaw dropping for me as I have never heard Redbook this good. We almost gave up on Foobar until we switched to non 32 bit Floating Point mode (therefore Integer mode ?) in the SACD plug-in, and then the SQ difference was night and day (CPU load 13%)."
So it seems, at least with Jriver and foobar, you might only use 13% of your CPU. That wouldn't seem to be significant....
Thanks Soren. We look forward to. It will facilitate listening tests.
I measured the new filter 1021SA2F2v1.skr. I understand that have changed the filter F2. Keep out measures in multiples of 44 with noise that are not in multiples of 48.
Soren, you may make a difference in the architecture of DAC that may be causing a problem of jitter or calculation in which 352 work?
A greeting.
I measured the new filter 1021SA2F2v1.skr. I understand that have changed the filter F2. Keep out measures in multiples of 44 with noise that are not in multiples of 48.
Soren, you may make a difference in the architecture of DAC that may be causing a problem of jitter or calculation in which 352 work?
A greeting.
I have been listening to Paul's MinPhaseSlowrev1 for the past few days. I found it very pleasant, have very large/deep soundstage and exquisitely detailed. Even Spotify through JRiver sounded really nice if the recording is good. No harshness at all. Can't comment on the other two filters but will try them shortly.
Well, setup is: Dam dac into adc burr brown pcm 4222 evaluation board, which is the master clock also, all connected via spdif with a rme firefac ufx, monitoring dac is a gustard x10. Monitored either through Geithain 904 or Joachim Gerhard Kalasan Speaker. The reference track has just one dac through the gustard x10, the processed tracks have one dac and adc on top of it. I can hear filter differences pretty well, the pcm4222 is not totally transparent, but good enough.
There are still variables, but that is the best testing method for me I can think of, since I want to use the Dam in the studio in precisely that application.
Tobias
HASWELL i3 4140 and a GIGABYTE H81 board. Not an extravagant CPU but worry about those with four cores. CPU busy-ness is a concern. I think the key to good computer audio is to get as little going on as one possibly can.
Working with NTLite and WIN 10 to get as small a c: folder as possible. My goal is to get c: down to 1gB. At 2.5 gB at the moment and stable as can be. NTLite has gone almost as far as it can - from here it will be manual deletions.
Easier to get the OS into memory when it is small and not require huge amounts of memory. Again, too much memory does not sound as good as just enough. I want to stick with 2 gB stick.
The real gains will come with decimating the registry and removing as many items to be polled as possible and still have a stable machine.
Back in the XP days we had a 20mB OS that sounded very fine, infinitely better than an optimized full install. Take that LINUX!
All of those registry entries are constantly being polled. If they are not there there is nothing to poll and a calmer environment ensues. Tiny XP registry was 56 kB, total. No way I will get anywhere close to that with 8/10 but it should be proportional.
Still think 2% usage sounds better! But willing to try ...