Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[oldnewbie]

Hi, once again thanks for creating a superb program!

recently upon accidentally running Audio Analyze again, on a file I had already performed this on, I noticed MC came up with different values. I have to say that I had run Audio Analyze on the file probably in 2002 or 2003 (when there was only JRiver Media Jukebox) so my guess is a lot has changed in over a decade and MC now uses a different method for the calculations? So I assume the current method is more accurate?

In any case I wish to run Audio Analyze on a great number of files (thousands never analyzed and some old already analyzed see above). Is it good/safe to just let it run through all or should I use smaller batches (hundreds) at a time?
I don't want incorrect values or much worse somehow corrupt/change the underlying audio of these files… My thought is that even with an i7 processor with hyper threading the files are on a normal hard disk so it will probably take many hours and in that time, there might be some other (system or other) program/service that demands some processing power so my fear is that this could interfere/change the values or worse corrupt the files…
Is is maybe better to not max out the processor and leave some of the cores (or at least virtual cores) for such unexpected events. With this in mind how many files should be simultaneously analyzed?

Lastly (and related to the "how many files should be analyzed in a batch") do I have to sort the files in a special way (probably Album) in order to let the Album Gain value be calculated correctly? My point is: if I list by Artist and a file is from a compilation, the rest of this compilation may be strewn all over the place and I don't know if MC does remember to use all the values from hundreds of files ago in the calculations. Even worse, if there are remasters or reissues (or the occasional single special file) with identical album-titles (let alone same album titles from differing artists) will it now group all those together inside one Album Gain?

Thanks for your answers and your great program!

[DoubtingThomas]

I think the calc for ReplayGain and "volume leveling" changed a bit ago.

I have always believed that the new method of analyzing tracks and determining the correction values does not work as well as the old method.

I'll be curious to hear from you if you feel the same after you re-analyze all your tracks.

[JimH]

R128 was introduced in MC19.

http://wiki.jriver.com/index.php/Volume_Leveling

[Arindelle]

Have to run out, but read the link Jim posted and search on the forums ... there are tons of posts on this. Other than BPM, this allows you to do a lot more than analysis did in the past - also depending on how you set up your DSP studio volume treatment, you no longer have to throw dynamics out the window (unless you prefer to). It will still write "replay gain" values for external devices that may need them.

I do not know why "DoubtingThomas" says it doesn't work as well for him. IMO it works a lot better, and you have a choice if using volume levelling to NOT just have every track at the same peak volume (unless that is important or you are only playing loud music all the time - then you can still configure that). Dynamics won't be quite as good as if you had volume leveling off, but they will at least simulate the differences between quieter passages and loud ones if you want them too.

As for processing on an i7 using local drives?? don't worry about it. If you are doing some heavily intensive stuff like video editing or streaming blu-ray put it at 4 at a time instead of 8, maybe ... I never do on an i5 . For info (again search the forum if you want), v.a. is not as forgiving of corrupt files or files that go past windows rules for long filenames.  If you run it all night, you might end up finding that it hung on a file and that it did not complete.  Might be a good idea to run a tool to see that the files don't have issues.

Oh, to give you an idea : on a fast i5 (8gb ram) 100K flac tracks took a little more than a weekend if memory serves.

[blgentry]

The last time I did a big batch, I used an i7 processor too.  I initially used all 8 cores (8 analyze processes at the same time), but I found that made the whole system pretty slow.  I ended up doing 4 at a time for a while, but found that 2 at a time made the system behave more normally.  I only had a few thousand files to do, so I didn't care if it took a little longer.

I'm going to have to look into more details on volume leveling.  I've been using two options from DSP Studio:  Volume Leveling, and Adaptive Volume, with Adaptive set to Peak Level Normalize.  I thought that preserved EVERY BIT OF DYNAMICS in every file.  If it is impacting dynamics even by a small amount, I want to disable the offending option.  All I'm looking for is a sort of average volume level across my listening session.  I want no alteration of dynamics.

Brian.

[mwillems]

I'm going to have to look into more details on volume leveling.  I've been using two options from DSP Studio:  Volume Leveling, and Adaptive Volume, with Adaptive set to Peak Level Normalize.  I thought that preserved EVERY BIT OF DYNAMICS in every file.  If it is impacting dynamics even by a small amount, I want to disable the offending option.  All I'm looking for is a sort of average volume level across my listening session.  I want no alteration of dynamics.

Brian.

Neither Volume leveling nor peak level normalize perform dynamic range compression.  Some of the other adaptive volume settings do perform dynamic range compression.

However, any reduction in digital volume at all from any source will result in some theoretical loss of dynamic range at the bottom, not through compression, but through that info effectively "falling off the bottom" and getting dithered.  

But given that 16-bit music offers 96dB of effective resolution, and most music has an effective dynamic range well less than 20dB, I wouldn't sweat the small losses from volume attenuation as you're not likely to lose any actual information even with very significant attenuation.  Additionally, if your DAC supports 24-bit output and you're listening to 16-bit files (i.e. CD audio), JRiver automagically pads the bitdepth, so you're not even theoretically losing any information until you've done 48dB of attenuation in that scenario.  

And even assuming there was information down there to lose (which there isn't), unless you're listening at volumes louder than 100dB (unlikely for regular listening in a home setup), you wouldn't be able to hear the info even in laboratory conditions.  Given the fact that the noise floor in a quiet home during the day is typically between 35 and 45dB, and even loud home listening usually doesn't get much above 90dB, it's unlikely you'd even be able to hear any theoretical information loss until you'd attenuated the volume by quite a bit.

So, I wouldn't sweat losing a few dB to volume levelling/peak level normalize.

Is is maybe better to not max out the processor and leave some of the cores (or at least virtual cores) for such unexpected events. With this in mind how many files should be simultaneously analyzed?

Using all the virtual processors will actually make it go slower in my experience.  If you want speed and don't care about using the computer int he meanwhile, my experience has been that analyzing the same number of files as actual cores (4) produces the fastest outcome on my i7's, with doing one more than the number of cores (5) a close second (might be faster depending on the day). Six and up seemed to just slow things down.

[oldnewbie]

Thank you very much for your replies, now I am even more concerned... a program that gets worse on updates? What will come next rising prices? No, all the kidding aside...

I thought I saw values also change on BPM and even intensity... and with me, I mostly thought those were a bit of a gimmick since there were some wich supposedly had high BPMs but curiously did not sound at all like it...

I was thinking of using it on my .flac rips including backups, now I guess I better keep them pristine and maybe just do the lossy formats.  

Jim: In your wiki I read that you keep the older Replay Gain in the tags for other software decoders is there a way then to let MC use the old values in playback? And would that solve your issues DoubtingThomas? Or is even that "old" Replay Gain different to the one in the old MJ?

So the worst thing in doing (too) many simultaneous analysis is that the system slows - no change/wrong values calculated or any more or less disastrous corruption/write errors or audio changes - correct?

Oh and any dithering or unnoticeable loss of range would only occur if the adaptive DSP settings are used, right?

But all this does not answer the questions: it is really, really, safe to Analyze all files at once (both from a correct values and a file/audio corruption/change perspective)?
how many Analysis can/should run at once (without any troubles, see above, occurring)? From your info I would guess is, that with an unused system (only doing Analysis), 4 should be totally fine.
is Album Gain calculated correctly with multiple versions of an album and tracks from an album spread out over the library?
(I am talking hundreds of albums and thousands of songs).

Oh and one more question: how about if you throw out all songs out of the library and re-import the backups (identical) to run the Analyze option on them, will MC keep the memory of the Album Gains etc. and now double the values?

Thanks again!