24 MP3's... all big band....and they all sound different over the same speakers..why?

[rasputin1963]

I have been working Sound and Lights for our local community theater on their current theatrical production, a comedy/drama called RED HERRING. I've never controlled room sound in this way before.

 

The show requires 24 snippets of music, played at key moments throughout, all of the same genre: so-called "crime jazz" of the 1950's. (Think: Mancini's PETER GUNN, DRAGNET, and other similar big band pieces).

 

I found 24 pieces in MP3 format, and Normalized all the soundfiles to -1.5 dB

 

But, when played back in context of the show, these different stereo music files all "play" differently over the same set of theater speakers (a pair of 10" YAMAHA speakers mounted high in the overhead ). Though all Normalized to the same loudness, some soundfiles, played-back, sound deafening, others too quiet. At the soundboard I have to keep riding the mixer sliders so I don't deafen our audience (which often include seniors).

 

The take-home lesson is, I see: straight-up Normalization is not the best way to make a bunch of soundfiles "play" roughly the same, volume-wise.

 

Is there a better way to achieve more predictable homogeneity with these 24 music files? I've heard of RMS, but I don't know what it is or how to measure it.

 

ras:rolleyes:

[UstadKhanAli]

Simply normalizing them, as you've found it, does not work. It never has worked.

 

You actually have to approach it as if you are mastering an album in terms of volume. And you do that by listening. Forget your meters. You need to listen.

 

You can also use a compressor. This, of course, only goes so far, but it helps even things out. It's amazing how well it works at radio stations, where there's a wild variety of music and inputs (potentially MP3s, tapes, vinyl, and CDs), but their compressor settings are set on "Kill".

[blue2blue]

Yes, the key to where you need to go is in processing all your files so that the RMS averages (basically average level over a certain amount of time) -- not the maximum peak values -- are about the same.

 

If you have a compressor that will allow you to set a target RMS value, that's pretty ideal.

 

For instance, Sony's Sound Forge (Win only) has an option on their normalize dialog that will allow you to set a target rate. Something like that would be ideal for your circumstances. (Me, I wouldn't really use that particular implementation for high value music work, but for something like this, go for it.)

 

Otherwise, you'll want to go through and use a full-control compressor (two knob 'vintage' jobs are not necessarily going to offer you fine control -- but you might do well enough) and do each file individually. Once you get the hang of it, shouldn't be too bad, particularly if you know your compressor parameters. (A wee birdie whispered in my ear that many contemporary recordists get a bit hazy on the fine points.)

[Philter]

I don't think he even needs to get as far as compressing. Matching levels up just by ear would probably work great. He's just doing that live now and hopefully it's working.

[UstadKhanAli]

Matching levels by ear is by far the best way to go. And as a bonus, it's the best sounding. After that, if you wish, some light compression, and you're ready to go. Do not use your meters or try and use an "average". It does not work. A listener's perception of loudness is very different from a meter reading.

[blue2blue]

I'll agree that doing it carefully by ear can be good -- when one has a good evaluation system in place -- obviously, he must have some sort of calibrated level playback system like Bob Katz' K-System or other in order to do a proper job using one's ears subjectively. And if one only has a handful of tracks, it's not hard at all.

 

However, having done some 30 and 40 track compilations, I know what a pain it can be trying to keep track of relative levels. Having good objective measure is a HUGE help.

 

The human auditory system is a truly amazing thing. It evolved as a highly capable and specialized immediate space mapping tool and threat detection system. And for that, it is truly amazing. Anyone who has seen a blind person negotiate a complex, unfamiliar environment with just echolocation from tapping with his cane or other noises need no further illustration.

 

But the human auditory system is NOT well suited to objective measure. At all.

 

And for that reason, it's important to know when and how to use objective measure in order to assist our aesthetic activities at times when they might otherwise be overwhelmed by the complexity of the task and/or by well-recognized cognitive distortion issues.

 

Use your ears as the final arbiter -- but if you find yourself getting bogged down, the place to look for help is in RMS measurement.

 

BTW, I didn't mean to imply that one necessarily needs to use more compression. By deciding on a target RMS level, getting RMS averages of your files, and then adjusting them up or down in level so their RMS is n the target zone, you can greatly reduce the puzzlement and confusion that can result from playing wack-a-mole with levels in 24 tracks with 24 different timbre and tonal balances from 24 different sources.

 

I stick by my general advice. Sure, use your ears -- but be smart and assist them with good objective measurement.

[blue2blue]

 

Matching levels by ear is by far the best way to go. And as a bonus, it's the best sounding. After that, if you wish, some light compression, and you're ready to go. Do not use your meters or try and use an "average". It does not work. A listener's perception of loudness is very different from a meter reading.

Unless you're very experienced, using live VU or especially peak metering to try to get an idea of average level is probably a fool's errand, to be sure.

 

That is why non-realtime RMS measurement is so important in getting relative levels from one piece of program material to another. It is fundamentally different form realtime, live metering in that you apply the measurement to a representative section of a program stream and the result is an average level that will allow much more on-point direct comparisons with the RMS average of other files.

 

When I was doing a lot of such premaster level setting, I got so, assuming a file was peaking at/just below 0 dB FS, I could typically guess its RMS within a dB or so by sound. It's a pretty good, predictable system for sorting out relative volumes.

[JeffLearman]

I heartily agree with Blue2Blue. Use RMS levels for the quick rough setting for all, at an RMS level that leaves plenty of headroom for adjustment later (e.g., -18dBFS RMS, and quite possibly as low as -24dB).

 

In this pass, you may find you need some compression/limiting to hit the RMS target on some of the tracks. I suggest you use a bit of compression on ALL of the tracks, in order to keep them above the room noise etc ... but use your judgement on this. I don't know whether these pieces are ordinarily compressed in the mastering stage. For earlier eras, they would be compressed simply because the dynamic range of the medium was small. But by the 50's, things were considerably better.

 

After the rough pass, put your tunes in sequence, and find out how each flows to the next and adjust accordingly, pairwise. But keep an eye on that RMS meter, because if you start wandering off the target value in a trend, you'll have a lot more adjustment to do down the chain. If you find yourself in an upward or downword trend

 

This is definitely mastering. You might also find a harmonic balancing plugin helpful, if the tone varies more than you want.

 

Meters are tools. They provide information, but are not substitutes for careful listening and good judgement.

[blue2blue]

I heartily agree with Blue2Blue. Use RMS levels for the quick rough setting for all, at an RMS level that leaves plenty of headroom for adjustment later (e.g., -18dBFS RMS, and quite possibly as low as -24dB). In this pass, you may find you need some compression/limiting to hit the RMS target on some of the tracks. I suggest you use a bit of compression on ALL of the tracks, in order to keep them above the room noise etc ... but use your judgement on this. I don't know whether these pieces are ordinarily compressed in the mastering stage. For earlier eras, they would be compressed simply because the dynamic range of the medium was small. But by the 50's, things were considerably better. After the rough pass, put your tunes in sequence, and find out how each flows to the next and adjust accordingly, pairwise. But keep an eye on that RMS meter, because if you start wandering off the target value in a trend, you'll have a lot more adjustment to do down the chain. If you find yourself in an upward or downword trend This is definitely mastering. You might also find a harmonic balancing plugin helpful, if the tone varies more than you want. Meters are tools. They provide information, but are not substitutes for careful listening and good judgement.

If I could just get you to go around to all the threads I'm tempted to answer, I think I could save myself some work.

 

Good job on further illuminating the process!

[flemtone]

I've got nothing to contribute to this thread, except to say that Crime Jazz completely rules. Nothing better to put one in a 'noir' frame of mind.

 

*puts on fedora, hunches shoulders to adjust trenchcoat, leaves circle of light from streetlamp and saunters off into fog*

[WRGKMC]

The trick to getting songs the same levels is dealing with the percieved loudness, not the RMS, Average or peak levels.

Percieved loudness takes into account how the ears hear the frequency balance of a song.

 

If you match two songs by RMS only as Blue suggests, and one sone has allot of midrange and one has allot of bass,

the song with allot of bass wont be nearly as loud to the ears as the song with allot of midrange. Midrange doesnt

produce nearly as large peaks as bass does and since ears are not linear, the midrange frequencies will be percieved

as being much louder to the ears.

 

If the songs all have the same balanced frequency range, you could use RMS to balance them. In this case where you have a bunch

of different MP3/s recorded by different labels, theres no way they would have simular responces, so the balancing RMS or average

alone wont work well. You would have to use your ears in an AB comparison.

 

Har Bal is the only program I know of that does this pretty much flawlessly. You set up

a reference file, it can be the loudest song in the batch if you choose. That reference file acts as a target

for matching. You simply do a Ctl - M and it matches the percieved loudness of whatever song you have opened.

 

The program is mainly an EQ program for mastering. It scans the frequency responce of every file opened and uses that

static frequency responce to match the volumes. Since percieved loudness changes with frequency responce, the Frequency data is used

while adjusting the levels to match what the ears actually hear.

 

You can also toggle between the target and matching file to compare the matching results.

I been using the program for a good 5 years or more and this volume matching alone is worth the cost of the program.

 

You can do the matching to find the levels need to be limited and then use a different limiter to do the job.

I for example did this with earlier versions of the program because I prefered using Waves L2 limiter.

The newest version of the program has a killer limiter that has no pumping or coloration so I been using that for the most part now.

 

You can also match the frequency responce but that really doesnt work unless the recordings were from the

same recording session, instruments etc. I've used the EQing in balancing up a bunch of MP3s with some mild

EQing so the cuts arent dramatically different. If the percieved loudness is good, and the tracks arent radically different,

the ears will adjust to the EQ responces. Otherwise, matching loudness and EQ is needed. Some tracks have different

frequency rolloffs, especially with MP3's where some of the trebble and bass are missing due to data compression.

[UstadKhanAli]

Unless you're very experienced, using live VU or especially peak metering to try to get an idea of average level is probably a fool's errand, to be sure.

 

Well, you'd be able to do it if the 24 different songs were similar. Perhaps the same band in the same studio all mixed similarly or in some other way similar. But if you've got a variety of mixes from different people, bands, time periods, whatever, it's almost assuredly not going to go well if you go by metering of any kind, particularly the sort you mention. Listening....I hear that works.