ReClock 1.8.5.8

[James]

Watching PAL DVDs?
Here's why you need ReClock: http://www.schmidt-web.info/malte/english.html
(unfortunately the sound examples were removed from the site... )

http://sandbox.slysoft.com/beta/SetupReClock1858.exe

1.8.5.8 - 10/01/2010
* New: "24 bit integer padded to 32 bit" in PCM output formats is back - and it works with the original ATI HDMI driver, and probably other HDMI interfaces. Hurrah!
* New: Added option to expand mono to both front speakers
* New: Added "NoVista" registry value where ReClock can be forced to behave as it would under XP with Vista & Windows 7
* Fix: Memory leak in resampler plugin
* Some fixes and improvements

Source code for GPL'ed code is available here:
http://oss.slysoft.com/ReClock/

 

[leeperry]

oh noes, a new version again

lotsa promising changes, will look into it right away..maybe I'll prefer 24 padded to 32 over plain 32 on my new ESI EWDM drivers. much thanks!

PS: indeed I do! on the acapella intro of "Jazzanova feat. Thief - Lie.flac" the vocals sound more artifacty in plain 32..man, does it sound good like this

many audio players do pad 24 to 32 anyway, might explain...it's drivers dependent of course.

 

[dbezerra]

With so many new Reclock updates I must confess that I am a bit lost... What are the recommended settings to achieve WASAPI bit exact audio out of TMT 2 (v.131) and W7MC, using an ATI 4670 and Realtek HDMI driver?

Thanks
Daniel.

 

[efjay]

Can someone advise how to get Reclock to load? I have installed the last two versions but both of them do not load when i open either wmp or media center. I am using Windows 7 32-bit.

 

[Ingram]

I believe I have found the source of my frame drops when watching Rome on Blu-Ray in TMT3.

I was using Reclock's AC3 encoder to convert the DTS-HD Master Track (after TMT has downsampled it) to a 5.1 DD track. I was getting dropped frames randomly. Yet! if I watch it with PDVD8 I get butter smooth results.

I've now turned off the AC3 encoder and switched to DTS Connect on my soundcard, and now Rome is playing just as smoothly.

Thought I should bring this up, since I least suspected this to be the cause for my problems.

Is there something that can be done? Log file?

 

[el Filou]

James you're a god, 24-bit output finally works on my ATI under XP!
(and the mono->stereo workaround is really nice too!)

However now something confuses me: I think someone here has written in another thread that ffdshow prefers to output, in order: 32 int, 16 int, 24 int, 32 float (I think, based on my tests... can't find the original post again, so please correct if that"s wrong).
I also understand that outputting 32 int is useless if you're not doing any processing.
How can I get the best quality for 24-bit *and* 16-bit sources at the same time then, by setting ffdshow to only 24 int or to only 32 float?
Previously James said making ffdshow output 32 float directly would be exactly the same as outputting 16/24 int and letting ReClock do the conversion to float, but with the recent changes to the int=>float conversions, does that position still stands?
But I'm also afraid that forcing ffdshow to convert everything to 24 int wouldn't be optimal either because ReClock treats int input as full range and ffdshow would pad 16 int output to 24 int and then the conversion to float wouldn't be optimal, am I right? Or am I wrong and ffdshow converts 16-bit sources to full-range 24 int (which would be worse still, surely)?
I can't understand the ffdshow developers' reasoning behind giving 16-bit output precedence over 24, even for 24-bit lossless sources...

Thanks if someone can clear my mind on this, this is really confusing! (and annoying because ffdshow is really convenient with all the decoders in one filter)

Also, does anyone (leeperry? ) know any DirectShow decoders for the common lossy codecs that decode and output in 32 float natively?
I'm mainly interested in MP2, AC3 and DTS as these are the lossy codecs I use most.

 

[leeperry]

Previously James said making ffdshow output 32 float directly would be exactly the same as outputting 16/24 int and letting ReClock do the conversion to float, but with the recent changes to the int=>float conversions, does that position still stands?
[..]
Also, does anyone (leeperry? ) know any DirectShow decoders for the common lossy codecs that decode and output in 32 float natively?
I'm mainly interested in MP2, AC3 and DTS as these are the lossy codecs I use most.

1) it does yes, 32float from ffdshow and you're done.
2) Gabest's Audio Decoder does it all in 32fp(when set properly in its config)...liba52/libdts in ffdshow work in 32fp, libavcodec is 16int only...but I still think that it sounds far better for DTS than libdts, so 16int it is then for me.

DTS has been reverse engineered(specs are not open like AC3), and libdts really sounds ugly to me(I dunno what DTS decoder Gabest's DTS decoder uses, but ffdshow's "number of channels" condition doesn't work w/ Gabest's Audio Decoder or AC3Filter, so I've got no use for either of these two).

I put a bug report on their tracker like 1 year ago about this, it's a good guess noone'll ever fix it. ffdshow has only one active coder atm and he cares about new features(HDMI audio, DXVA), so the bug you're mentioning w/ outputting 16int on 24int sources will prolly never be fixed either.

 

[yesgrey3]

How can I get the best quality for 24-bit *and* 16-bit sources at the same time then, by setting ffdshow to only 24 int or to only 32 float?

You should check 16 bit int, 24 bit int, 32 bit int and 32 bit floating point.
This way, ffdshow will always output the format nearest to the source sound, and you will let reclock perform the final conversion.

 

[el Filou]

1) it does yes, 32float from ffdshow and you're done.

Thanks!

2) Gabest's Audio Decoder does it all in 32fp(when set properly in its config)...liba52/libdts in ffdshow work in 32fp, libavcodec is 16int only...but I still think that it sounds far better for DTS than libdts, so 16int it is then for me.

DTS has been reverse engineers(specs are not open like AC3), and libdts really sounds ugly to me(I dunno what DTS decoder Gabest's DTS decoder uses, but ffdshow's "number of channels" condition doesn't work w/ Gabest's Audio Decoder or AC3Filter, so I've got no use for either of these two).

I put a bug report on their tracker like 1 year ago about this, it's a good guess noone'll ever fix it. ffdshow has only one active coder atm and he cares about new features(HDMI audio, DXVA), so the bug you're mentioning w/ outputting 16int on 24int sources will prolly never be fixed either.

In MPC(-HC) when you keep the mouse cursor over an internal filter name, it will show various info. For the Gabest audio decoders for AC3/DTS/MPEG it says "based on liba52/libts/libmad". Don't know if that info is up-to-date, but I would tend to think so as these haven't seen any big development for a while.
At least with DTS Gabest decoder sounds as bad as ffdshow's one in libdts mode, I'm with you on that one: it's really obvious (I've got a reference scene in one particular movie for that in which with libdts the heavy rain sounds reeeeally metallic).
With AC3 I didn't try to listen for differences between the two yet, but from what I understand ffdshow needs to decode AC3 for it to be able to decode TrueHD anyway (or does it? ), so to hell with that, I'll keep ffdshow in libavcodec... fiddling with media types to try to use Gabest only for MPEG and basic AC3 would drive me crazy anyway.

You should check 16 bit int, 24 bit int, 32 bit int and 32 bit floating point.
This way, ffdshow will always output the format nearest to the source sound, and you will let reclock perform the final conversion.

Well, it certainly doesn't behave like that on my system!
If I check everything it will always output 32 int. If I check everything except 32 int, then it will always output 16 int, no matter what the source is (even when fed 24-bit LPCM for example ! :bang.
And I don't do any processing of any kind with it, just decode and output.

I'll stick with only 32 float then.
ReClock 1.8.5.8 is the holy grail for me. 8)

 

[Ingram]

I'm curious what kind of sound setup you people have going on? Are you using analogue out, SPDIF, or HDMI?

In my circumstances, all of this 24bit conversion and what have you is useless to me because I can only either output DD5.1 via Reclocks AC3 encoder or use DTS Connect on my soundcard.

All of my rips and Blu-Rays are either always 16bit natively or are downsampled to such, so I guess until I get a better receiver and 1080P/24 TV there isn't much I can do...

 

[yesgrey3]

Well, it certainly doesn't behave like that on my system!
If I check everything it will always output 32 int.

If ffdshow outputs 32 int is because it's the closest one, trust it.

And I don't do any processing of any kind with it, just decode and output.

If you don't do any processing, it's natural that ffdshow does not output 32 floating point.

I'll stick with only 32 float then.

Don't. Stick with what I suggested, it would be preferable... though, you should not hear any difference.

I'm curious what kind of sound setup you people have going on? Are you using analogue out, SPDIF, or HDMI?

I'm using analog out with my RME FF400.

 

[leeperry]

You should check 16 bit int, 24 bit int, 32 bit int and 32 bit floating point.
This way, ffdshow will always output the format nearest to the source sound

good point! that's the answer to my problem, I've learned to never trust "auto" stuff(whatever in overclocking or projectors) but it does work on 16/44.1 24/96 FLAC fed from madflac, MP3(decoded to 32fp w/ Gabest's Decoder, I love how it sounds), and for movies and 5.1 stuff I always post-process like hell anyway to get a nice stereo binaural mixdown.

With AC3 I didn't try to listen for differences between the two yet, but from what I understand ffdshow needs to decode AC3 for it to be able to decode TrueHD anyway (or does it? ), so to hell with that, I'll keep ffdshow in libavcodec... fiddling with media types to try to use Gabest only for MPEG and basic AC3 would drive me crazy anyway.

well the AC3 specs are opened, and liba52 sounds great to me...plus it's 32fp and I do a lot of PP, so the better!

Well, it certainly doesn't behave like that on my system!
If I check everything it will always output 32 int. If I check everything except 32 int, then it will always output 16 int, no matter what the source is (even when fed 24-bit LPCM for example ! :bang.
And I don't do any processing of any kind with it, just decode and output.

I got PP set to 32fp, and I use madflac..everything's cool here.

I'll stick with only 32 float then.

well, yesgrey3 said that the int>32fp code in ffdshow was ever so slighty inaccurate...it doesn't use the "2^15 trick" that allows to transport lossless integer over 32fp, so it's slightly killing the actual number of bits if I got it right

it's not even clear if you guys are going 2^15 when feeding the resampler btw

ReClock 1.8.5.8 is the holy grail

damn right!

 

[efjay]

So nice to get so much help, thanks a lot.

 

[leeperry]

I'm curious what kind of sound setup you people have going on? Are you using analogue out, SPDIF, or HDMI?

In my circumstances, all of this 24bit conversion and what have you is useless to me because I can only either output DD5.1 via Reclocks AC3 encoder or use DTS Connect on my soundcard.

All of my rips and Blu-Rays are either always 16bit natively or are downsampled to such, so I guess until I get a better receiver and 1080P/24 TV there isn't much I can do...

I use Sony MDR-CD3000 headphones connected to a pimped Prodigy AudioTrak Prodigy HD2 Advance DE, very cheap and awesome sounding w/ some good opamps

well, if any of your movies has lossy audio, you'd be better off decoding them to 32fp, output from Reclock in 24bit, and oversample to 96/192kHz(but HDMI 1.2 is limited to 5.1 16/48 anyway I think?)

So nice to get so much help, thanks a lot.

prolly noone uses WMP here? especially on W7, they went MediaFoundation, that makes things far more complicated I think.

 

[James]

but HDMI 1.2 is limited to 5.1 16/48 anyway I think?

You think wrong.
7.1 24/192 shouldn't be a problem, 24/96 certainly isn't.

 

[Ingram]

I'm using SPDIF not the HDMI.

I've tried all sorts of combinations, but the only ones that I can *SEE* from my Receiver menu is 44.1 and 48Khz. Anything else doesn't work.

I can get reclock going with 192Khz resampling, but DTS Connect still outputs it as 48Khz, which I imagine is probably a really bad thing to be doing.

Even so, the audio kept dropped out when I tried that.

 

[leeperry]

You think wrong.
7.1 24/192 shouldn't be a problem, 24/96 certainly isn't.

apparently the "maximum supported audio stream bandwidth is 6.144 Mbps" on the 4800 :
http://ati.amd.com/products/radeonhd4800/specs.html

16-bit/48 LPCM 5.1 is 4608 Kbps
24-bit/48 LPCM 5.1 is 6912 Kbps
16-bit/48 LPCM 7.1 is 6144 Kbps
24-bit/48 LPCM 7.1 is 9216 Kbps
24-bit/96 LPCM 5.1 is 13824 Kbps

do you circumvent the DRM crapola through Reclock? I thought you needed PAP to do that?

and sorry for asking, but is there a way you could let yesgrey3 have a look at the int>32fp Reclock code used to feed the resampler please? maybe there's room for improvement as well(2^15 yada yada)

 

[leeperry]

I can get reclock going with 192Khz resampling, but DTS Connect still outputs it as 48Khz

you can output stereo 24/192 over S/PDIF w/ some high end interfaces like the M2tech Hiface...but for multichannel 192 shouldn't possible, you need to find the bottleneck...either DTS Connect, your audio drivers, the amp, the S/PDIF specs of your interface, etc..

PS: Kazuya got 16/96 to work I think.

 

[James]

apparently the "maximum supported audio stream bandwidth is 6.144 Mbps" on the 4800 :
http://ati.amd.com/products/radeonhd4800/specs.html

Must be wrong. My 4850 outputs 7.1 192/24 no problem.

 

[James]

do you circumvent the DRM crapola through Reclock? I thought you needed PAP to do that?

No. You don't need PAP to play sound.