Topic: Has the Analyze Audio feature changed the algorithm (Read 19644 times)

Humbledore · « **on:** May 08, 2016, 09:29:20 am »

Hi!
I am noticing that when I reanalyze audio files many of the R128 values are changing. Why is that? Has the algorithm being updated? Or if so, when?
Thanks!

blgentry · « **Reply #1 on:** May 08, 2016, 10:50:36 am »

I'm not sure if this is related, but it's worth sharing:

I rip CDs into a watched folder structure. So MC imports the files as they are ripped. I noticed quite some time ago that the audio analysis values didn't seem to be right for some files. Some played WAY louder or quieter than they should. So I re-analyzed these files and their volume level was now correct on playback.

At some point, I realized that MC is *sometimes* analyzing partial files. These are files that are still being ripped from CD, but MC starts on them before they are complete. Which explains why the audio analysis values were wrong.

Now, I generally analyze audio manually right after I rip a CD. Note that I use an external ripping program and not MC's built in CD ripping function.

FWIW.

Brian.

~OHM~ · « **Reply #2 on:** May 08, 2016, 11:34:44 am »

The windows version of MC there is a option to opt out of analyzing while you rip the cd

Hendrik · « **Reply #3 on:** May 08, 2016, 11:40:34 am »

Quote from: blgentry on May 08, 2016, 10:50:36 am

At some point, I realized that MC is *sometimes* analyzing partial files.

If your ripping application does not lock files while its still writing to them, there is no way to really know its not complete yet.
If this affects your ripping software, I would personally just rip somewhere else thats not a watched folder and import manually and/or move aftwards, or use MC for ripping.

blgentry · « **Reply #4 on:** May 08, 2016, 11:49:08 am »

Quote from: Hendrik on May 08, 2016, 11:40:34 am

If this affects your ripping software, I would personally just rip somewhere else thats not a watched folder and import manually and/or move aftwards, or use MC for ripping.

Yeah, that's all good advice. I should have said originally that I realize that MY workflow is the problem. I was just posting in case the OP has a similar workflow and perhaps a similar effect.

Brian.

Humbledore · « **Reply #5 on:** May 08, 2016, 03:10:02 pm »

Thanks for all answers but the question remains... Why are the values changing? I have always used an explicit analyze using the Analyze Audio tool confirming that the analyze finished correctly.

Vocalpoint · « **Reply #6 on:** May 09, 2016, 10:17:55 am »

Quote from: Humbledore on May 08, 2016, 03:10:02 pm

Thanks for all answers but the question remains... Why are the values changing? I have always used an explicit analyze using the Analyze Audio tool confirming that the analyze finished correctly.

When were your values written/first calculated? Last week? Two versions of MC ago? Might help to determine what time frame we are talking about here?

VP

Humbledore · « **Reply #7 on:** May 09, 2016, 11:03:09 am »

When the new analyze audio tool was introduced in MC a couple of years ago, I reanalyzed all my files. So now I just wonder if something has changed to the algorithm since I have noticed a change in values when I perform a new analyze.
For the actual reanalyze I've started with the oldest songs imported around 2011 and so far what I can see all songs get new values. But it will probably take some time to, I have around 35 000 songs to check/analyze...
Thanks!

JimH · « **Reply #8 on:** May 09, 2016, 11:10:25 am »

You could read about R128 on the forum and the wiki.

https://wiki.jriver.com/index.php/Dynamic_Range

Vocalpoint · « **Reply #9 on:** May 09, 2016, 11:10:38 am »

Quote from: Humbledore on May 09, 2016, 11:03:09 am

When the new analyze audio tool was introduced in MC a couple of years ago, I reanalyzed all my files. So now I just wonder if something has changed to the algorithm since I have noticed a change in values when I perform a new analyze.

I have seen minor changes too in files that were last analyzed several years back. But as long as they played back fine - I never gave too much thought about tweaks to the R128 algorithms - which I one must assume (from an EBU prospective) is logically improved from time to time as the standard evolves.

That said - if the actual specs/code of the EBU 128 algorithm standards have changed - only MC devs could comment on whether or not any of those changes have actually been baked into recent builds of MC...

VP

Humbledore · « **Reply #10 on:** May 09, 2016, 01:07:01 pm »

Another thing I've noticed is that the Peak Level (Sample) value becomes 0 after the reanalyze, see attached image. So it's quite obvious that the algorithms has been tweaked in a way or another. I just wonder why and when...?
Thanks!

Humbledore · « **Reply #11 on:** May 11, 2016, 05:12:16 am »

I think I now have proof that something has changed with the analyze audio algorithm pretty recently. A Marvin Gaye album which I imported March 19, 2016, and for which I manually performed an analyze with the Audio Analyze tool, get new values after a new manually performed reanalyze. See attached image to see for yourself. So I still wonder why this algorithm tweak was introduced and if I need to reanalyze the entire collection again...
Thanks!

Hendrik · « **Reply #12 on:** May 11, 2016, 06:05:18 am »

R128 analysis has not changed for quite a while, certainly not within MC21. Depending on the type of source file, some other things may have changed.
A difference of 0.2 LU in the measured loudness is not something you would hear either way.

JimH · « **Reply #13 on:** May 11, 2016, 07:04:09 am »

Quote from: Humbledore on May 11, 2016, 05:12:16 am

I think I now have proof that something has changed with the analyze audio algorithm pretty recently. A Marvin Gaye album which I imported March 19, 2016, and for which I manually performed an analyze with the Audio Analyze tool, get new values after a new manually performed reanalyze. See attached image to see for yourself. So I still wonder why this algorithm tweak was introduced and if I need to reanalyze the entire collection again...

What version of MC were you using when you first had the file?

Arindelle · « **Reply #14 on:** May 11, 2016, 12:49:57 pm »

just tested this on 5 albums .... I get no changes at all. Curious ? I fin it weird that in the first screenshot, all peak sample levels are at 0 across the board. Mine are not.

All tracks are flac 44.1khz files, both analyzes (what's the plural, duh??) were done manually; none were done during import or rip, for info.

Humbledore · « **Reply #15 on:** May 12, 2016, 12:52:28 pm »

Quote from: JimH on May 11, 2016, 07:04:09 am

What version of MC were you using when you first had the file?

I use the setting 'Latest' so I assume I have the latest possible version when the analyzes are performed.
I attach screenshots, before and after, from another album created September 2015, also analyzed with MC 21. Notice that all the analyze values are changing after the reanalyze. This is the "change pattern" for all the files that are reanalyzed and were created before (around) March 19, 2016.
[img][img]

Vocalpoint · « **Reply #16 on:** May 12, 2016, 01:06:19 pm »

Quote from: Humbledore on May 12, 2016, 12:52:28 pm

I use the setting 'Latest' so I assume I have the latest possible version when the analyzes are performed. I attach screenshots, before and after, from another album created September 2015, also analyzed with MC 21. Notice that all the analyze values are changing after the reanalyze. This is the "change pattern" for all the files that are reanalyzed and were created before (around) March 19, 2016.

I am now seeing weirdness too.

After reading this thread again - I "reanalyzed" David Gilmour's Rattle That Lock - purchased from HD Tracks and imported to MC on 9/22/15.

First analysis was with a very late version of MC 20 as I installed MC 21 on 11/22/15.

The screen caps clearly show that not only has the DR "magically" changed from MC 20 (first pass) to MC21 (second pass) - the Peak Levels in Pass 2 have now all been zeroed?

If nothing has changed within the R128 infrastructure - just what is going on here?

VP

Humbledore · « **Reply #17 on:** May 14, 2016, 04:23:46 am »

Quote from: Vocalpoint on May 12, 2016, 01:06:19 pm

If nothing has changed within the R128 infrastructure - just what is going on here?

I also wondering what is going on? And I would appreciate if someone from the JRiver team could make clear if I should perform a new analyze of my files or if I shall await a correction/mending of the analyze algorithm...
Thanks!

Hendrik · « **Reply #18 on:** May 14, 2016, 04:52:00 am »

Please try to provide more details on what kind of files you are seeing changes with, if possible even providing one file that changed recently. I tried to analyze both audio files and movies that have last been analyzed with MC20 and nothing changes at all.
Like I said, I'm also not aware of any changes in the analysis code that would cause any changes.

Humbledore · « **Reply #19 on:** May 14, 2016, 06:42:51 am »

What more details can I provide I wonder? See previous posts screenshots from me and Vocalpoint...
I just have noticed something happened with the algorithm (probably) in March 2016. No matter file I choose to reanalyze before that time the analyze values are changing...
Thanks!

Vocalpoint · « **Reply #20 on:** May 14, 2016, 06:58:53 am »

Quote from: Humbledore on May 14, 2016, 06:42:51 am

What more details can I provide I wonder? See previous posts screenshots from from me and Vocalpoint... I just have noticed something happened with the algorithm (probably) in March 2016. No matter file I choose to reanalyze before that time the analyze values are changing...
Thanks!

Same here. I can chose almost any FLAC file that I have - that of course - have not been analyzed by the latest version of MC that I am using (21.0.81) - and i am seeing big changes in every file.

I will screencap some more test cases if it will help.

VP

JimH · « **Reply #21 on:** May 14, 2016, 07:06:44 am »

Are they ordinary FLAC files? Or some variant like Lossless FLAC, etc.

Vocalpoint · « **Reply #22 on:** May 14, 2016, 07:24:50 am »

Quote from: JimH on May 14, 2016, 07:06:44 am

Are they ordinary FLAC files? Or some variant like Lossless FLAC, etc.

Not understanding what you mean by "lossless" FLAC.

To my knowledge - every file I have in our master library is a standard FLAC file - and each one should be lossless.

See the attachment for a quick analysis of a single file (in the last five minutes) of a file orginally done a month or two ago in 21.0.42 vs what I get with 21.0.81...

The FLAC panel open (with original analysis) is open on the left - then the fresh analysis is on the right.

Cheers!

VP

Hendrik · « **Reply #23 on:** May 14, 2016, 08:32:22 am »

Quote from: Humbledore on May 14, 2016, 06:42:51 am

What more details can I provide I wonder?

None of the screenshots provide any information about the files themself. No file type, no bit-depth, no sample rate, no nothing.
All the screenshots included are the audio analysis results, which don't tell me anything about the file itself (really don't tell me anything at all). And if I cannot reproduce a problem, I cannot fix a problem.

So considering I cannot reproduce it on any standard files I tried, and I don't know if anything about your files is special, I cannot even start checking anything.
Just to repeat myself, if possible please try to share a file that changed the analysis results, so I can run it with MC20 and MC21 and see if I get different results on that.

Vocalpoint · « **Reply #24 on:** May 14, 2016, 11:34:10 am »

Quote from: Hendrik on May 14, 2016, 08:32:22 am

Just to repeat myself, if possible please try to share a file that changed the analysis results, so I can run it with MC20 and MC21 and see if I get different results on that.

Understood. What is the best way to get some files to you?

And just so we are clear - I am seeing different results here within versions of v21 - nevermind going back to v20. My current example is v21.0.42 to v21.0.81.

VP

Hendrik · « **Reply #25 on:** May 14, 2016, 02:02:22 pm »

If you can put it on any drive service like DropBox, Google Drive, OneDrive, etc and share a link to hendrik at jriver dot com, that would be fine.
If you don't have any of those, and the file isn't way too big (say < 20MB), you could also try mailing it directly, our mail system should handle it.

Vocalpoint · « **Reply #26 on:** May 14, 2016, 02:11:30 pm »

Quote from: Hendrik on May 14, 2016, 02:02:22 pm

If you can put it on any drive service like DropBox, Google Drive, OneDrive, etc and share a link to hendrik at jriver dot com, that would be fine.
If you don't have any of those, and the file isn't way too big (say < 20MB), you could also try mailing it directly, our mail system should handle it.

I will pop a few files up to OneDrive for you. Will notify via PM with details.

Cheers!

VP

Hendrik · « **Reply #27 on:** May 17, 2016, 05:49:35 am »

I analyzed the provided files with the last build of MC20, MC 21.0.42 and 21.0.81, and all produce the exact same values.
For the record, these: http://imgur.com/GCyQhpw

So not sure what to tell you. What kind of CPUs do you have? Some system-specific result would be odd, but maybe not impossible.

Vocalpoint · « **Reply #28 on:** May 17, 2016, 07:29:31 am »

Quote from: Hendrik on May 17, 2016, 05:49:35 am

I analyzed the provided files with the last build of MC20, MC 21.0.42 and 21.0.81, and all produce the exact same values. For the record, these: http://imgur.com/GCyQhpw So not sure what to tell you. What kind of CPUs do you have? Some system-specific result would be odd, but maybe not impossible.

This specific workstation is running an Intel Core i7 @ 3.50GHz

I get different values for every track - until of course it's analyzed by the current version of MC. And then the numbers remain static.

EDIT: FOUND IT!

In my install - last month - I was complaining about the level on files I was converting from FLAC to MP3 for the car. Someone in here suggested that when I highlighted an album and chose Convert Format - within the Convert Format panel - to add Adaptive Volume into the DSP and that should make for files with better volume (And it does).

But I was told that this setting was exclusive to JUST converting files to MP3. vs my normal "listening" DSP settings which only has Volume Leveling engaged at any specific time.

I just tested it - If I turn off Adaptive Volume (but leave Volume Leveling on) under Audio Conversion Options and then go back into MC and choose any file - even one last analyzed with MC 19 - the resulting values remain the same.

But as soon as I return to the Audio Conversion Options dialog and turn Adaptive Volume back on (in addition to Volume Leveling) - and re-analyze the same file - the values change - sometimes dramatically.

So - I have to ask - why are DSP settings (that I thought were specific (and restricted) to just files being converted) - coming into play when doing a simple Analyze Audio from within MC - for files that I thought were being subjected to the DSP settings within the DSP box that appears when I click the "bit perfect" EQ icon in the upper right from within MC?

See the attached files for the dialogs I am visiting in order...

VP

Hendrik · « **Reply #29 on:** May 17, 2016, 07:31:39 am »

That is most definitely a bug, and I'll look into getting this fixed. Good job on finding the interaction!

Vocalpoint · « **Reply #30 on:** May 17, 2016, 07:39:38 am »

Quote from: Hendrik on May 17, 2016, 07:31:39 am

That is most definitely a bug, and I'll look into getting this fixed. Good job on finding the interaction!

Right arm!

VP

Hendrik · « **Reply #31 on:** May 17, 2016, 07:57:19 am »

This will be corrected in the next build.
Fixed: Enabling the "Adaptive Volume" DSP in the "Convert Format" settings could inadvertently affect the results of Audio Analysis.

Vocalpoint · « **Reply #32 on:** May 17, 2016, 07:58:06 am »

Quote from: Hendrik on May 17, 2016, 07:57:19 am

This will be corrected in the next build.
Fixed: Enabling the "Adaptive Volume" DSP in the "Convert Format" settings could inadvertently affect the results of Audio Analysis.

Awesome! Thank you.

VP

Matt · « **Reply #33 on:** May 17, 2016, 08:14:15 am »

Well I was just fixing it as well, but Hendrik beat me to the punch!

Humbledore · « **Reply #34 on:** May 18, 2016, 04:33:14 am »

Quote from: Vocalpoint on May 17, 2016, 07:29:31 am

I just tested it - If I turn off Adaptive Volume (but leave Volume Leveling on) under Audio Conversion Options and then go back into MC and choose any file - even one last analyzed with MC 19 - the resulting values remain the same.

But as soon as I return to the Audio Conversion Options dialog and turn Adaptive Volume back on (in addition to Volume Leveling) - and re-analyze the same file - the values change - sometimes dramatically.

I have also Adaptive Volume enabled in the Convert Format tool (as well as in the DSP Studio settings though). But If I disable the Convert Format Adaptive Volume option the re-analyze values stay the same!
Thank you Vocalpoint, for your detective job!

Vocalpoint · « **Reply #35 on:** May 18, 2016, 06:39:36 am »

Quote from: Humbledore on May 18, 2016, 04:33:14 am

I have also Adaptive Volume enabled in the Convert Format tool (as well as in the DSP Studio settings though). But If I disable the Convert Format Adaptive Volume option the re-analyze values stay the same! Thank you Vocalpoint, for your detective job!

My pleasure guys! Looking forward to the new build.

Cheers!

VP

This2ShallPass · « **Reply #36 on:** June 11, 2016, 04:04:35 pm »

Gosh, I think I'm in way over my head. But my mix of FLAC and MP3 files have been analyzed, and boy howdy, I do not like the algorithm that was used. I am quite certain I can live with compression!

Is there any way out of the mess I happen to be in? I don't understand everything I've read in the wiki about audio analysis, so I think I'd like to un-analyze my files.

Hendrik · « **Reply #37 on:** June 11, 2016, 04:25:27 pm »

Just turn off volume leveling in the DSP Studio, the analysis doesnt change the files.

JimH · « **Reply #38 on:** June 11, 2016, 04:26:20 pm »

What version of MC are you using? Did you try the build from the top of this board?

blgentry · « **Reply #39 on:** June 11, 2016, 05:54:42 pm »

Quote from: This2ShallPass on June 11, 2016, 04:04:35 pm

Gosh, I think I'm in way over my head. But my mix of FLAC and MP3 files have been analyzed, and boy howdy, I do not like the algorithm that was used. I am quite certain I can live with compression!

I don't understand what you are reporting. Volume Leveling and R128 analysis don't relate to compression as far as I know.

Brian.

JimH · « **Reply #40 on:** June 11, 2016, 06:22:24 pm »

You're correct.

RD James · « **Reply #41 on:** June 11, 2016, 08:54:25 pm »

Quote from: This2ShallPass on June 11, 2016, 04:04:35 pm

Gosh, I think I'm in way over my head. But my mix of FLAC and MP3 files have been analyzed, and boy howdy, I do not like the algorithm that was used. I am quite certain I can live with compression!

Is there any way out of the mess I happen to be in? I don't understand everything I've read in the wiki about audio analysis, so I think I'd like to un-analyze my files.

It literally only turns the volume of the track up or down, and is applied as DSP during playback.
It doesn't touch the dynamic range.

The only changes to your audio files are that some new tags are written, the audio itself is untouched.
If you don't want to use leveling, just disable it in DSP Studio.

This2ShallPass · « **Reply #42 on:** June 17, 2016, 09:22:18 pm »

I've done some experimenting, and something's clearly still wrong.

I've unchecked all the boxes in DSP Studio in the left pane, but I still get wild, wild results track-to-track in playback. I have tracks that don't even seem like they're making any sounds at all (but the simple spectrum analyzers bracketing the title area are plenty busy). The next track? Right on time. But all kinds of volume mess.

I went in to DSP studio, tried to load defaults (there really aren't any, are there?), I've got 4 profiles saved, and really... none of them are right.

What on earth did I do?

This2ShallPass · « **Reply #43 on:** June 17, 2016, 09:24:26 pm »

v. 21.0.89, FYI.

blgentry · « **Reply #44 on:** June 18, 2016, 07:51:10 am »

Quote from: This2ShallPass on June 17, 2016, 09:22:18 pm

I've unchecked all the boxes in DSP Studio in the left pane, but I still get wild, wild results track-to-track in playback. I have tracks that don't even seem like they're making any sounds at all (but the simple spectrum analyzers bracketing the title area are plenty busy). The next track? Right on time. But all kinds of volume mess.

That sounds like your playback device might be having a problem. What DAC, or receiver, or other audio device are you using? Have you restarted MC, the device, and your computer?

Brian.

This2ShallPass · « **Reply #45 on:** June 18, 2016, 09:03:48 am »

Yeah, it's just my desktop pc, internal standard audio card, small speakers, plenty of reboots of everything. Been using this setup for years.

I followed the directions for the "analyze audio" feature, applying it to a few files first, but I must not have listened to the results critically enough.

blgentry · « **Reply #46 on:** June 18, 2016, 09:17:22 am »

You're describing zero volume and high volume with the spectrum analyzer showing similar profiles. If that's really what's going on, this has nothing to do with audio analysis or with volume leveling. Something else is going on.

Do your other media players play normally still? Have you checked your cables between sound card and speakers? Speaker power supply cable?

Brian.

This2ShallPass · « **Reply #47 on:** June 24, 2016, 09:48:40 am »

Thanks Brian, everything's plugged in and working fine. This is my main PC, use it everyday, plays all kinds of vid and music, and they play normally.

Seems there's not a "reset to default" or "no processing" option in DSP playback that I can tell. I would like that. Even if I have everything in the left pane unchecked, I seem to have issues.

blgentry · « **Reply #48 on:** June 24, 2016, 09:58:08 am »

Quote

This is my main PC, use it everyday, plays all kinds of vid and music, and they play normally.

I'm assuming you mean from other players? Or do you mean MC plays everything normally except for some files?

Check the Audio Path while you have files playing: Player > Audio Path

That will show you exactly what the DSP studio is doing with the audio. If you have everything unchecked, you shouldn't see anything there.

Brian.

This2ShallPass · « **Reply #49 on:** June 24, 2016, 10:40:38 am »

Listening to Nine Inch Nails' "Ringfinger" and there's definitely something going on. At 1:55 left in the song, the distorted guitars start wailing and they recede TREMENDOUSLY after each of the boardclaps, telling me that some compression is hitting the file REALLY HARD, it's really distracting! Did some A/B listening (using my trusty ATH-ESW9's), streaming the same song to my Android phone has the sound in its more natural state. It does seem to be evened out on the phone, I'm sure. But why would I experience such immense attenuation like that if I'm not processing?

INTERACT FORUM

Author Topic: Has the Analyze Audio feature changed the algorithm (Read 19644 times)