Pre-ringing, minimal phase filters (and others), and apodizing...

[Archimago]

Hey guys, it certainly seems like there has been a move in the hardware world over the last few years towards having multiple filters on DAC's. Looks like most of this is towards having a choice of intermediate and minimal phase filters over the standard linear filter - everything from relatively inexpensive DacMagic to fancy Meridians. Testing methodologies like John Atkinson's measurement reports in Stereophile in recent years seems to be spending more time showing graphs of impulse response and less talk about how many ps of jitter is in the signal.

I have yet to come across any good studies/reports describing audibility of this pre-ringing. Looking at the graphs, it seems insignificant; the amplitude is very low and frequency of the ringing seems to be inaudible to human ears at least. I remember reading about this back in 2006 with the "Ringing False" article in Stereophile and passing it off since even those guys could not agree on the sound (they do have 'golden ears' after all :-)

http://www.stereophile.com/content/r...ng-impressions

Bottom line, has anyone been able to show a significant detectable difference under *controlled* settings for standard linear phase filtering vs. something like minimal phase filters with no pre-ringing say in standard 16/44 music or even test tones (I assume the difference may not be beneficial due to phase shift in the minimal filter)? Is this business with filters yet another solution for a problem that doesn't exist?

BTW Can anyone explain in layman's terms WTF "apodizing" means? I see that term thrown around alot...

Thanks!

[JohnSwenson]

First on the term "apodizing": different people have radically different meanings for the term, so it's very difficult to figure out what any body means.

If I remember correctly the original use for the term in regards to audio digital filtering was about a non-brickwall filter applied to higher sample rate signals. Somewhere along the line it got confused with minimum phase filters (I think one box had both apodizing and minimum phase so the press assumed they were the same thing).

I have done a fairly extensive set of listening tests both by myself and with others concerning digital filters, both hardware implementations and software implementations. I don't have time to go over all the details right now.

The upshot is that the digital filters built into almost all DAC chips are flawed in one way or another. Most of the "digital filters sound bad" sentiment is actually not about pre-ringing etc, the culprit seems to be that the hardware implementations take shortcuts in their implementations in order to both produce good numbers in the spec sheet and meet cost requirements of the chips. Pretty much everybody is assuming these filters are implementing a traditional Sinc function, but in reality they are not. The designers have come up with interesting implementations that that give good numbers but don't take as much chip resources. These shortcuts are what I think are the real issue, not pre-ringing etc.

Some results that tend to back this up are: Take a really good NOS DAC feeding something like a 1704, with async USB front end, very low jitter clocks. Listen to it with a DF1704, NOS, digital filter implemented in an FPGA with just basic Sinc response, and software upsampling to 192 again with just basic Sinc. With the DF1704 it sounds flat and uninvolving. Played NOS it sounds way more alive, interesting and musical, BUT it sounds "dirty", you can hear the aliasing going on. With both the FPGA filter and the software filter it keeps the aliveness etc., BUT the "dirtiness" goes away, it's by far the best sounding. This is NOT some just barely noticeable affect, it's actually quite startling when you first hear it.

Several DAC chips that contain internal digital filters allow you to bypass them and use your own external filter. I have done the same test with several of these with the same result, internal filter sounds flat, NOS sounds much better but dirty, FPGA or software Sinc sounds wonderful.

There is a group of people that are extoling the vitues of software upsampling, but unfortunately most of these people are then feeding this stream into chips that have builtin digital filters implemented with these shortcuts. This severly limits the efectiveness of the software digital filters.

Yes some chips support different varients of filters, but they are all implemented with these shortcuts so it makes the inherant differences very difficult to discern.

There ARE differences between filters, minimum phase, Sinc etc, but these differences are tiny compared to just getting rid of the hardware filters in the first place.

There ARE a few DAC boxes out there that do use FPGA filters instead of the internal filters, but they are few and far between, and usually pretty expensive.

John S.

[Rodney_Gold]

Quite a nice white paper on filters here
http://www.ayre.com/pdf/Ayre_MP_White_Paper.pdf
I have a MDAC (audiolab) with various filters , I can clearly hear differences between them. I haven't bothered to conduct formal tests on an ABX basis , the nearest I got to controlled conditions was having my daughter switch between various filters and me listening...

[JohnB]

Viruskiller (a contributor to these forums) has developed a tweak for the convert.conf/custom-convert.conf that implements something similar to the apodizing filter that is used in Meridian kit. As it uses SOX to resample to 88.2k it can only be used with a Transporter or Touch. He used a combination of looking at impulse wave forms and using his ears to comparing the result with his Meridian apodizing.

[adamdea]

Viruskiller (a contributor to these forums) has developed a tweak for the convert.conf/custom-convert.conf that implements something similar to the apodizing filter that is used in Meridian kit. As it uses SOX to resample to 88.2k it can only be used with a Transporter or Touch. He used a combination of looking at impulse wave forms and using his ears to comparing the result with his Meridian apodizing.

ooh- this is interesting where would i find the constom convert.conf setting. Have they been published?

[Mnyb]

ooh- this is interesting where would i find the constom convert.conf setting. Have they been published?

I think the forum advanced search allows you to filter by user , and I also remember a lot of discussions and treads regarding apodizing a couple of years ago on this forum .
So searching for viruskillers post is a start .

[adamdea]

I think the forum advanced search allows you to filter by user , and I also remember a lot of discussions and treads regarding apodizing a couple of years ago on this forum .
So searching for viruskillers post is a start .

I'm not getting anywhere with user viruskiller and apodizing search term. Perhaps Mr V. Killer will chance upon this thread.

[Wombat]

You can´t simulate the Meridian apodizing filter with sox, only some minimum or immediate phase things that all add smearing to highs that may be audible. The Meridian filter is linear phase up to 18.5kHz afaik.

[JohnB]

ooh- this is interesting where would i find the constom convert.conf setting. Have they been published?

I'm sure that Viruskiller won't mind me posting this (as he posted it to the open access meridianunplugged forum a couple of years ago). As far as I know his final version was:

flc flc * *
# FT:{START=--skip=%t}U:{END=--until=%v}
[flac] -dc --totally-silent $START$ $END$ -- $FILE$ | [sox] -q --replay-gain album -t wav - -t flac -C 0 -b 24 - gain -1 rate -v -M -b 87.5 88200 dither -S

The second and third line are indented but the indentation doesn't work on this forum.

(substitute your touch's MAC for the second asterisk to make it specific to that player)

You can always try it out - if the result is no different or worse you can revert to the original. (If you test it on the convert.conf file I would first make a backup copy.)

[Mnyb]

I'm sure that Viruskiller won't mind me posting this (as he posted it to the open access meridianunplugged forum a couple of years ago). As far as I know his final version was:

flc flc * *
# FT:{START=--skip=%t}U:{END=--until=%v}
[flac] -dc --totally-silent $START$ $END$ -- $FILE$ | [sox] -q --replay-gain album -t wav - -t flac -C 0 -b 24 - gain -1 rate -v -M -b 87.5 88200 dither -S

The second and third line are indented but the indentation doesn't work on this forum.

(substitute your touch's MAC for the second asterisk to make it specific to that player)

You can use code brackets .

flc flc * *
        # FT:{START=--skip=%t}U:{END=--until=%v}
        [flac] -dc --totally-silent $START$ $END$ -- $FILE$ | [sox] -q --replay-gain album -t wav - -t flac -C 0 -b 24 - gain -1 rate -v -M -b 87.5 88200 dither -S