What is the maximum number of musicians who could realistically jam together?

January 7, 2013 9:06 AM

Aside from the basic physics of the speed of sound would there be practical upper limits to the number of musicians, assuming an infinite supply with the necessary talent and disposition, who could jam spontaneously to produce a united song without generating discord?

I've been lucky enough to grow up around the kind of musicians who could together spontaneously pick up a song from one starting a simple beat, and I've started to wonder how crazy that could be taken, any thoughts?
posted by Blasdelb (26 comments total) 2 users marked this as a favorite

the more musicians you have, the more they're going to have to lay back, be careful and leave space for one another, which in a rock context could result in a less driving sound

it seemed to me that when the grateful dead added bruce hornsby as a 2nd keyboard player, the sound was a bit more tentative and less centered at times, which isn't to say they didn't have their moments

so that's 7 musicians there, two of them drummers

if you had people playing single note instruments i'm sure you could bump that up quite a bit with head arrangements

but for spontaneous rock music with no horns, i think two guitars, two keyboards, bass and drums is pretty much it, although three guitars and one keyboard might work
posted by pyramid termite at 3:41 PM on January 7, 2013

Depends what you mean by musicians and jamming. I've been in a room with several hundred people singing in tongues (iin my former born again life) and it sounded wonderful.
posted by unSane at 6:11 PM on January 7, 2013

Nice one, Blasdelb. Cool thought experiment.

I'd say it depends what you mean by "discord", too. There are certainly jazz big bands (14 to 20 musicians) that have used collective improvisation, Mingus & Sun Ra being 2 examples off the top of my head. Not all of the results were "pleasant" in the sort of standard Western music sense. OTOH, bands were often working off an agreed upon set of chord changes, so maybe not quite in line with your parameters.

Aside from the basic physics of the speed of sound

This might actually be pretty relevant though - here's what I'm thinking:

The speed of sound is (ballpark) 1 millisecond per foot, so 10 feet between musicians = 10 ms delay in the sound of ones' instrument reaching the ears of the other. 10 ms may not be enough of a difference to consciously hear, but it's certainly enough for a listener to go, "Hm, something's off here," especially if the 2 instruments are supposed to be playing exactly in time. 20 to 30 ms is definitely into the noticeable range. So, first, at some point you're going to reach a physical limit where the musician over here will be distant enough from the musician over there that they'll be unlikely to play in time.

Then there's the fact that sound waves follow the Inverse Distance Law - for every doubling of distance, there's a reduction in level by half. So there's another physical limit, where any given musician simply won't be able to hear all the others.

(I'm assuming we're gonna leave aside possible technological enhancement, like monitors or a click track.)

Finally, there's the question of how much info our brains can process simultaneously - how many different players can I listen to in any meaningful sense, enough to virtually instantaneously tell that my rhythm isn't meshing with that guy's and this chord isn't working with what that other guy is playing and so on and so forth.

Adding up those three factors (distance, volume, brain power), I'd say that, mmmm . . . . . 20-ish is probably about the limit where things aren't likely to collapse quickly, given your parameter of (if I'm reading you correctly) starting with no predetermined structure.

But I'd also say that imposing even a fairly simple structure or organization could bump the number up quite a bit.
posted by soundguy99 at 11:46 PM on January 7, 2013

In C is more "choose your own adventure" than improvisational, but has been performed with as many as 124 players.
posted by en forme de poire at 8:35 AM on January 8, 2013

To get around the limits of space and time, you could allow musicians to jam "locally" so that each musician only needs to play and update his or her playing based on the 20 or so people around him. Using this scheme the number of players is limited only by the space available.

This is kind of like how the Game of Life can update itself instantaneously without regard to the size of the playing board. The information is local, but there is still a global "sound".
posted by grog at 10:17 AM on January 8, 2013 [1 favorite]

posted by chococat at 2:20 PM on January 9, 2013

after thinking about it for a bit, my guess is that you could get a lot more rhythm instruments together than pitched or tunable ones. If you're talking about a massive group of drummers, grog's point about local-listening giving clues to the overall structure makes the most sense to me.

Just tried to find the largest group of drummers - apparently there was just an attempt at a guinness book record by 15,000 drummers in India. Not sure if it was improvised or a pre-existing structure, but apparently they were playing all together...
posted by dubold at 2:13 AM on January 10, 2013 [1 favorite]

Assume we're talking about a performance which has to be rhythmically coherent globally not just locally, the distance between performers imposes a surprisingly small upper limit. Assume 120 bpm and that all performers must hit the downbeat within 1/16 note of each other and are not using click tracks or a visual cue such as a conductor. That means they must be within 1/32 of a second of each other. Sound travels at 340 m/s so the furthest two performers can be apart and stay grooved with each other is 340/32m - about 10m or 30'.

Assuming each musician needs a square metre to get his/her groove on, then using pi* r^2 we can fit about 75 to 100 people into such a circle.

If you allowed a drummer in the middle and everyone followed them you could double the size of the circle, thus quadrupling the number of performers to around 300-400.

An interesting effect of this is that performers move in and out of phase with each other at a particular tempo at particular distances. In the above scenario the performers will be 1/16 note out at 10m, 1/8 note at 20m and so on. Thus if you placed drummers at phase coherent locations and had musicians around them follow their beat you could maintain rhythmic phase coherence over arbitrarily large distances.

For example, have one master drummer at the centre. Place six drummers around him at a quarter note distance or about 75m. These form a hexagon. Now surround each of those drummers with a circle of performers within a1/16 note or 1/32 note distance

Now repeat this for the six drummers in the hexagon as many times as you like.

Now you have an arbitrarily large collection of musicians all playing on the quarter notes, but as you walk around them you'll feel the '1' shift, always appearing to get later as you walk away from any particular drum circle.
posted by unSane at 4:49 AM on January 10, 2013 [2 favorites]

I suspect in a big group you might find the above structure evolve spontaneously. There's probably a nice way to model this visually using cellular automata.
posted by unSane at 4:51 AM on January 10, 2013

Assuming each musician needs a square metre to get his/her groove on, then using pi* r^2 we can fit about 75 to 100 people into such a circle.

But each performer sits at the center of his or her own circle. You only have to listen to the people near you, and you can still get a globally consistent rhythm.

An interesting effect of this is that performers move in and out of phase with each other at a particular tempo at particular distances.

Yeah, you can't listen to the whole thing at once if the distances get too large. But you can play the whole thing at once, regardless of the distances. That's because playing is always local, but listening can be defined at arbitrary scales.
posted by grog at 8:47 AM on January 10, 2013

You only have to listen to the people near you

There's the rub... it's a quesion of how successfully you can do that. You're probably gonna be hearing a wash of in-time and out-of-time stuff... you know how it sounds if you've ever been in a crowd singing a song.

I really want to model this. I bet there is all sorts of emergent behaviour depending on how much individual cells timing varies, and how sensitive they are to input from far cells and so on.
posted by unSane at 10:08 AM on January 10, 2013

You're probably gonna be hearing a wash of in-time and out-of-time stuff... you know how it sounds if you've ever been in a crowd singing a song.

Hmm, that is a problem. Maybe there's an optimal amplitude of each instrument that maximizes the ratio of perceived loudness between near and far musicians, to make the far off wash as quiet as possible. Or maybe you just play your own instrument at an amplitude that's loud enough to put other players at a certain distance under some prescribed noise floor. That would be interesting to model / optimize.
posted by grog at 12:08 PM on January 10, 2013

For instance, you'd have to put all the loudest instruments as far away from each other as possible, so it starts as an equipotential problem...
posted by grog at 12:09 PM on January 10, 2013

I've played with The Orchestra of Crafty Guitarists several times, on one occasion including over a hundred players. It was certainly an experience.
posted by Grangousier at 3:27 PM on January 10, 2013

I wrote a cellular automata simulation of this today. I did it in Python for speed so it ran pretty slowly but the results were interesting. Each cell (think of them as little drummers) tries to maximize the loudness of the sixteenth note beat, taking into account all of the other drummers, the phase difference introduced by their distance from him, and the fact that distant sounds are quieter. I started with a random arrangement. Then on each tick, each cell thinks about taking a random step in some direction. If that increases the loudness of the beat on the 16th note they move there, otherwise they stay put. They can only get within a certain distance of ech other.

I tried to keep the numbers realistic.

With small numbers of cells, they simply all head towards the middle and sit in a group.

Once you get past 20 or 30 things start to change because the phase difference starts to get too much for the cells on the edges. You start to see arms of little drummers radiating from the main body, spaced at coherent distances, and little scattered knots and tuples. You also start to see that the cells have arranged themselves at multiples of a certain distance. I haven't got any decent screenshots yet but I'll try to do some tomorrow. I might try to translate it into obj c so it runs fast enough to look animated.
posted by unSane at 6:51 PM on January 10, 2013 [2 favorites]

You can grab the code here:


This needs python3.3 but it's a simple tweak to run in 2.7. I also got it running in pythonista on iOS with a few simple changes but it's so slow it's not worth it.
posted by unSane at 7:03 PM on January 10, 2013

Hey unSane, how would you tweak this to run on Python 2.7?
posted by grog at 7:26 PM on January 10, 2013

I think you just have to comment out the print statement with sep='' in it.
posted by unSane at 8:04 PM on January 10, 2013

Your best bet is to download ipython if you have any trouble
posted by unSane at 8:07 PM on January 10, 2013

Here's an example of a shape that 100 little drummers formed themselves into.

For scale, each drummer is 1m across. Speed of sound is 340 m/s, they are playing 120 bpm and trying to maximise the loudness of the 16th notes. At this scale the damping factor of sound in air does not really make a whole load of difference, but I used -0.0003 10dB/m I think, which is pretty standard.

Doing the same thing with 200 drummers now, but it's really slow as on each iteration every drummer has to think about a move and then calculate what the sound would be from every other drummer if they moved there.
posted by unSane at 6:14 AM on January 11, 2013

This is neat! My iPython runs 2.7, but I got the script to work with a "from __future__ import print_function" command. How people ever learned to program before the internet is beyond me.
posted by grog at 7:58 AM on January 11, 2013

unSane & grog, you guys are friggin' awesome! Moar, please.
posted by soundguy99 at 10:33 AM on January 11, 2013

I fiddled with my code some more and straightened some things out and now the little bastards just all head for the center and stay there. I think the problem is that the ones in the middle get trapped and can't move outwards. I also think I'm doing the wrong thing by having them try to find the loudest beat. They should be trying to find the most accurate beat by minimizing the sum of the squares of their phase differences or something, or some combination of the two.
posted by unSane at 10:51 AM on January 11, 2013

OK, now it's kinda working. This one's interesting. If you look at it with your eyes half closed you can sort of see a hexagonal pattern emerging.

This one works as follows: the darkness of the circle (I think of them as little drums) represents how hard the little guy is hitting it. I started with one drummer and then randomly dropped drummers in one at a time, giving them time to find a place (100 iterations) before dropping another drummer in. They're all whacking away on 16th notes at 120 bpm, ie oscillating at 32 Hz and each guy tries to minimize the sum of the squares of his phase difference from all of the others, scaled by both how loud they're hitting the drum and how far they are from him. The sound doesn't fade much from one side of the canvas to the other, maybe 3 or 4 dB I think. Each guy is 1m across. The python code is at the link above, and you can watch the path each guy takes as he's dropped into the drum field, and how the others adjust to his presence.

At 32 Hz the wavelength is about 10m, ie ten circles, which is very roughly the diameter of the hexagons I imagine I am seeing.

I'll do the same thing with 8th notes. If I'm right, and there are hexagons, they should be about twice the size.
posted by unSane at 12:57 PM on January 11, 2013

Ha, my math is totally wrong of course, since the bpm refers to quarter notes not whole notes, but it's how I did it in the code, so it still works.
posted by unSane at 1:07 PM on January 11, 2013

This discussion is great.
posted by box at 7:12 AM on February 17, 2013

« Older softLOUD   |   Who is reading Music? Newer »

You are not logged in, either login or create an account to post comments