Matt Shackleton

Hi, I'm fairly new to Max. I'm working on a project which needs me to compare a bunch of audio tracks (36 tracks in total, though I might need to use more), figure out which track is loudest at a specific time, and return the track number rather than the content. I can't use [maximum~] since it doesn't say which input track is currently the output.

I have an awful hack at the moment, which involves each track going into a [peakamp~], then the outputs being [pack]ed into a list, then using [maximum] to return the index of the largest value. I'm desperately hoping there's a better way.

Hi, I'm fairly new to Max. I'm working on a project which needs me to compare a bunch of audio tracks (36 tracks in total, though I might ne

finding-loudest-from-multiple-audio-tracks

That sounds completely reasonable. What don't you like about that solution?

peakamp/pack/maximum is not wrong, it is just not complete.

you should store the list for maximum elsewhere, so that you can compare the output of maximum with the list in order to find its position in the list.

peakamp/pack/maximum is not wrong, it is just not complete. 

mzed: The problem with an array of peakamps is it takes ages to set up, since I'll likely need to expand it in the future. Also, I'm concerned about efficiency, since I'd like this to have as little delay as possible.

Roman: the second output of maximum gives the position of the largest value in the list, so my current solution is complete. I'm just asking in the hope that there's some easy solution I've missed.

By "track" do you mean the audio channel of a single multi-channel file? 

If it's the first case you could do this using javascript pretty easily, so that it doesn't matter how many audio channels you are comparing (in a buffer~).

EDIT: second case do it using polybuffer~ with js

By "track" do you mean the audio channel of a single multi-channel file? 
Or are they separate audio files?

I made a javascript and patch for your task: it'll work for a multi-channel audio file. You should be able to adapt it to multiple files and polybuffer without too much trouble.

maxbufsamps.zip

I made a javascript and patch for your task: it'll work for a multi-channel  audio file. You should be able to adapt it to multiple files and polybuffer without too much trouble.
See attached.

Thanks, but that doesn't quite do what I need. The audio channels I am looking at are coming live from a set of microphones, and I need to know which one is currently loudest as they play. I have a metro timer on the array of peakamps, which refreshes them every 200ms or so, but I lose a lot of information if I increase the refresh rate.

you should probably post your patch, or a cut-down version of it illustrating the problem; but it might be the cpu can't cope with 36 channels. What you could do is isolate a single-channel version of the signal processing aspect of it, and place that inside a poly~ with a downsampling argument, and then create 36 or however many instances you need (any number), then tag the outputs (peakamp~ outputs) with the instance number to keep track of the channel, and then do a logical comparison of some sort to find the track with the max amp at that point in time.

you should probably post your patch, or a cut-down version of it illustrating the problem; but it might be the  cpu can't cope with 36 channels. What you could do is isolate a single-channel version of  the signal processing aspect of it, and place that inside a poly~ with a downsampling argument, and then create 36 or however many instances you need (any number), then tag the outputs (peakamp~ outputs) with the instance number to keep track of the channel, and then do a logical comparison of some sort to find the track with the max amp at that point in time.

LoudestTrackExample.maxpat

Here's an example of what I'm currently using. I don't think the CPU is having any trouble; I've had a meter~ on each channel and they were behaving as expected.

not really sure what your problem is, but you need a float argument for maximum if you want it to work with lists of floats ie [maximum 0. ]

don't know what you mean by "lose a lot of information if I increase the refresh rate"

not really sure what your problem is, but you need a float argument for maximum if you want it to work with lists of floats ie [maximum 0. ]
don't know what you mean by "lose a lot of information if I increase the refresh rate"

I mean that if I reduce the refresh rate to 10ms, I lose information about sounds below 100hz because the waveform is smaller than the window.

If you can't think of a better way to do it, that's fine. My current version does what I need it to, and I was mainly asking in the hope that there was an obviously better way that I'd missed.

OK yes that's an example of the heisenberg uncertainty principle, unfortunately you're stuck with that in this particular universe

You could store the levels in a [coll] and query for the max value. This gives you information about the loudest channel and it's level. See the attached example patch which simulates your mic inputs by audio file players. Just set different files to play, loop them, start them, and enable the detection toggle.

The [coll] gets continuously filled with incoming levels, the first column is the channel number and the second column ist the value. The "max" message retrieves the currently highest level and its row number (= channel number).

If this is for some kind of automatic mixing, I think you rather want [average~] and [snapshot~] instead of [peakamp~]. Or a sort of envelope follower to make it even less sensitive (rampsmooth~, slide~ etc). Maybe a combination of all methods.

 The [coll] gets continuously filled with incoming levels, the first column is the channel number and the second column ist the value. The "max" message retrieves the currently highest level and its row number (= channel number).
—
If this is for some kind of automatic mixing, I think you rather want [average~] and [snapshot~] instead of [peakamp~]. Or a sort of envelope follower to make it even less sensitive (rampsmooth~, slide~ etc). Maybe a combination of all methods.

agree w peter re using an averaging or envelope follower, but here's your idea using a poly to make more adaptable/configurable: 

hope it makes sense; you could replace peakamp~ inside the poly~ with any other amplitude following object (like peter suggested) or other more sophisticated process, but it should make extending your patch or ideas easier

agree w peter re using an averaging or envelope follower, but here's your idea using a poly to make more adaptable/configurable: 
save this as "poly_peakamp":


hope it makes sense; you could replace peakamp~ inside the poly~ with any other amplitude following object (like peter suggested) or other more sophisticated process, but it should make extending your patch or ideas easier


Finding loudest from multiple audio tracks