in_mad v2 Winamp MAD MP3 input plugin Options

[Emre]

@MoSPDude

did you drop FPM_64BIT compile switch for libmad with new version? which switch did you use? FPM_INTEL or FPM_DEFAULT

[-sanb-]

MoSPDude
BIG THANKS!!!
MY MAN!!!
with respect!
and thank you from the AIMP community

This post has been edited by -sanb-: Mar 4 2008, 00:14

[MoSPDude]

Hi Emre,
I've dropped back to FPM_INTEL (as in my previous in_mad) mainly to see if this helps with peoples problems. As far as I know VS2005 does support 64bit signed int properly however. I kept OPT_ACCURACY on in both cases. I personally didn't notice a difference in speed, though I'd need to test it further. In mad.h, it strongly recommends against FPM_DEFAULT.

If anyone has noticed any major differences as such (eg. improved the stuttering) please feedback.
Cheers

Edit: As this is the first time I've tried FPM_INTEL in Visual Studio, it appears to not take account of OPT_ACCURACY as it did using GCC. I'll have a look into it.

This post has been edited by MoSPDude: Mar 4 2008, 09:23

[Emre]

Hi Emre,
I've dropped back to FPM_INTEL (as in my previous in_mad) mainly to see if this helps with peoples problems. As far as I know VS2005 does support 64bit signed int properly however. I kept OPT_ACCURACY on in both cases. I personally didn't notice a difference in speed, though I'd need to test it further. In mad.h, it strongly recommends against FPM_DEFAULT.

If anyone has noticed any major differences as such (eg. improved the stuttering) please feedback.
Cheers

Edit: As this is the first time I've tried FPM_INTEL in Visual Studio, it appears to not take account of OPT_ACCURACY as it did using GCC. I'll have a look into it.

Hi MoSPDude,

I did some tests with ffdshow's libmad component. i get source code via svn (rev 1882) and i did some modifications in libmad project files. i compiled ff_libmad.dll with vc2005 express and vc2008 express.

for speed test i used foobar2000 with haali's foo_input_ds plugin.
directshow filter chain: source mp3 in matroska container -> haali splitter -> ffdshow audio decoder (libmad)

test results with foobar2000's decoding speed test plugin (cpu core2duo e6550~2.8ghz):

FPM_INTEL VC8 : 115,674x
FPM_INTEL VC9 : 116.268x
FPM_64BIT VC8 : 76.676x
FPM_64BIT VC9 : 77.104x
clsid's build (rev 1882) FPM_INTEL ICL10 (SSE) : 124.217x
xxl's build (rev 1878) FPM_INTEL VC7 : 113.877x

i can't sure about this test accuracy but it seems FPM_64BIT ~%34 slower than FPM_INTEL.

cheers...

[j7n]

Hi, MoSPDude.

The last version (with ini-file) interrupts playback every time when I call the file info box. It walks the MPEG stream and updates average bitrate number. And when it does so, playback stops for half a second. Does not matter if I have opened infobox before on that given mp3.

Also, although I have selected "fast playtime calculation" and display "file average rate", it does not display bitrate from the Xing frame. Instead average bitrate of frames played so far is shown.

[Emre]

...interrupts playback every time when I call the file info box...

unchecking "Link to Winamp EQ" and "Apply ID3v2 EQU information" check boxes will fix this.

[j7n]

I can confirm that disabling EQ solves the problem.

[BSOD2600]

1) I'm still getting occasional short pauses in the playback.

2) After copying/overwriting the existing in_mad* files in my Winamp plugin directory, I launched winamp 5. Went to it's settings, to check things out. Closed the properties. Then played a song. While it was playing, I got the file info (alt+3). Below is what I saw:


3) Closed winamp, reopened. played a track, now I see the old config window:


4) Closed winamp, again. Deleted in_mad_cfg.ini. Reopen winamp. Now I see the newer file properties


5) Another problem with the in_mad_cfg.ini deleted, the in_mad plugin config now disables a bunch of options. Nothing was disabled when I had the in_mad_cfg present. For example:


6) Winamp should write a new in_mad_cfg file, if one wasn't present when it closes, IMO. Currently it does not.

[-sanb-]

BSOD2600
Try to delete this section

HKEY_USERS\s-x-x-xx-...\Software\Winamp\MAD Plug-in

[BSOD2600]

BSOD2600
Try to delete this section

HKEY_USERS\s-x-x-xx-...\Software\Winamp\MAD Plug-in

You mean to use the v2 in_mad plugin, right?

If these keys are no longer going to be used, the plugin should be smart enough NOT to use them at all anymore...

[k.eight.a]

BSOD2600 Try to delete this section

HKEY_USERS\s-x-x-xx-...\Software\Winamp\MAD Plug-in

You mean to use the v2 in_mad plugin, right?

I assume his idea is to delete this entry from the Windows registry. Start > Run > regedit < Then search for this string... ;)

[MoSPDude]

Hi guys,

Sorry for the confusion of it all, I really need to make a bit of a readme.txt up or clarify the settings.

The registry settings still exist by default and are used if the config file doesn't exist - currently a config file isn't made - I plan to add this soon. The reason a lot of settings are disable in the other config tabs is because the option to use Winamp Global configuration is enabled in the General tab and these settings are coming from the Winamp Playback preferences. I also need to add a reminder that when the Unicode option is changed, Winamp needs to be restarted for it to take effect otherwise you get garbled messages and 'Cannot connect to media'. The option to enable the newer unified file properties is also on the General config tab.

I'll also look into the EQ stuttering problem, and I think I will stick with FPM_INTEL even though OPT_ACCURACY isn't used for MSVC.

Cheers guys for the feedback - I have some free time now and will get it sorted.

[BSOD2600]

The reason a lot of settings are disable in the other config tabs is because the option to use Winamp Global configuration is enabled in the General tab and these settings are coming from the Winamp Playback preferences.

Ahh, that bit of info would be very helpful to have documented along with the distribution of the in_mad.dll file.

Do you think it would make sense for the plugin to reflect the global winamp settings which are set, but still leaving it grayed out? Would be easier instead of having to dig through winamps prefs to local all those various settings which the plugin has pulled together into a few tabs.

[_Raziel-BG]

@_Raziel-BG - First make sure your using the latest plug-in currently 0.2.0.14. Are you using basic or advanced title formatting? I need to know the formatting string in either case. As a start, you could try changing the tag lookup order in the config to see if that changes the situation.

Sorry for the late response. Here's the info. Version 0.2.0.20. Basic title formating string is:
[%artist% - ]$if2(%title%,$filepart(%filename%)) (the default one)
Tried messing with the tag preference options and if ID3v1 is before (higher preference) ID3v2, info is displayed fine. However, if ID3v2 is before ID3v1, then tags are messed up. I can't recall if I had tried ID3v1 before ID3v2 in 0.2.0.2 but I do remember that I did SOME form of preference re-ordering and it didn't help.
Just to remind again: the problem with the formating occurs on some songs/albums, not all.

Added: Otachan's input plugin displays everything fine with ID3v2 -> APE -> ID3v1 preference.

Added again: enabling Unicode suppord freezes/crashes Winamp on begin of playback.

This post has been edited by _Raziel-BG: Mar 13 2008, 00:56

[John Clement]

Are there any plans to add the unsynced Lyrics tag? or the cover Artwork tag? At present Winamp supports the artwork, but none of the existing Winamp plugins support the unsynced Lyrics tag. Or if they do, I have not been able to coerce them to work. Your tag plugin certainly finds tags that standard Winamp misses.

[Yashu]

I am somewhat new here, but I was wondering, I have been following the development from the first in_mad mod and now this second version, and my question is, has anyone any link to where testing has been done on the noise shaping and dithering options?

I don't want to ask "what sounds better", I am curious if any serious ABX testing has been done on this plugin, in particular, not against other decoders, but the various options of just this one, perhaps with mpg123, or some other (well documented) one as a control.

I am not really equipped to do the tests myself, but it would be interesting to see the results of noise shaping vs. non, and also the three dither options, and then combinations of them both. I suppose the results could be put on a spreadsheet, and if enough people participated, it would be a really cool experiment.

I don't know if this has been addressed anywhere else, and I am not trying to find out what is "better" per se, but it would be an interesting experiment and I have been wanting to know ever since those options became part of the plugin. Everywhere I have gone to look for information, mostly people are only talking about tagging or replaygain, basically only one mention of some testing done with someone's family, but nothing more than that.

EDIT: Application to mp3 would be interesting, especially if several bitrates are tested. It would be interesting to see if dithering would have an audible effect, and if so, would it be less or more depending on the encoded bitrate. If I were to venture a guess, I would say the noise shaping would not be noticed either way (doesn't LAME employ an LP filter? If the noise shaping algorithm uses freq. above the LP filter, would that make things worse?), but if the dither *was* audible, the preference would be towards triangular weighting. That is my guess. I have toyed with the settings with V0 and 320k LAME, and I can tell you my preference, but that would be against the TOS?

Has this been discussed somewhere else, and if not, what test tones/sounds do you guys use to test?

This post has been edited by Yashu: Apr 1 2008, 01:19

[Sylph]

When can we expect an update?

[Slipstreem]

I'm currently running version 0.15.1b published on the 25th of April (2008) as grabbed from the Winamp website only twenty minutes ago.

Keep up the fine work, MoSPDude.

Cheers, Slipstreem.

[BSOD2600]

I'm currently running version 0.15.1b published on the 25th of April (2008) as grabbed from the Winamp website only twenty minutes ago.

Umm that would be his old/first version. Not the current v2 one this thread is about... I too am using it since v2 is too buggy.

[SkripT2]

Thanks a lot @MoSPDude for this new version. I have an Auzen X-Fi Prelude (Creative X-Fi Pro). I use Mad V2 input plugin with the resolution set to 32 bits and Direct Sound v2.2.6 SSRC output plugin (I like the fadings) with Hardware Acceleration allowed and resampling to 44100 with 24 bits per sample enabled and dithering with no noise shaping and noise distribution in triangular. Could please anyone tell me if that's the best configuration? Should I use the resampling option in the Direct Sound plugin with 24 bits for get a better sound even if Mad is working in 32 bits? Thanks.

This post has been edited by SkripT2: May 7 2008, 00:10

[BSOD2600]

This yet another dead mp3 project?

[Sylph]

I wonder, too...

[MoSPDude]

Hi guys,

I think it's safe to say that for the moment it is a dead project. I've not had any time to make progress on hunting down bugs and issues - been too involved with finishing one large project and starting a new job. If anyone else wishes to have a go at improving the code, then I'm happy to help them understand my messy code.

Other than that, I'm sorry guys! I may pick it up again but it'd have to be once I'm settled back in to a new personal routine

Cheers for your support and bug finding - this is a great forum

[lvqcl]

MoSPDude, in_mad adds tags in ISO 8859 encoding. Non-english characters change to dots. Can this plugin create tags as UTF 16?

added: I downloaded code & replaced 'field->number.value = ID3_FIELD_TEXTENCODING_ISO_8859_1;' by 'field->number.value = ID3_FIELD_TEXTENCODING_UTF_16;' in function id3_field_init(). Don't know is it good, but works.

This post has been edited by lvqcl: Jul 24 2008, 23:51

[Mits2008]

I have been lurking in this forum for years, and I just registered to say a BIG THANKS to MoSPDude for his hard work. I have been using MAD since its inception and I just discovered this new version.

All I can from the first hours of listening is WOW.

My modest 30 year experience in mid-fi tells me that the decoded sound I am experiencing (via SPDIF + external receiver) is crisp, transparent and has definitely an improved sound stage imaging. I am not in the position of making AB tests, but my intuition tells me that this plugin is a real breakthrough and definitely needs both serious testing and ironing of the (not so important for me) details re tagging, streaming etc.

This is NOT another dead mp3 project. As somebody mentioned above, the automatic clipping attenuation alone is worth 50% of mad's value. Since it is virtually impossible for us to replace all those old mp3s with new ones (taking advantage of all the latest Lame features), mad is necessary for the optimal mp3 experience.

Thanks again, and I really pray for the emergence of new developers to pursue this noble cause!