This article is from the Model Rockets FAQ, by Wolfram von Kiparski with numerous contributions by others.
From George Gassaway:
>>>I'm planning on doing some tests on optimizing streamer
>>>duration. The conventional wisdom favors micafilm, with 1/2" to
>>>3/4" accordian pleats over 3/4 length. I would like to investigate
>>>different materials and pleating techniques. If you have other
>>>materials or techniques you favor, I would appreciate hearing about
>>>them, so that my tests can be as complete as possible.<<<
Well, as for folding techniques, sure. There's a fold I call
"Scorpion's Tail", because the streamer deflects to one side, curled
sort of like an attacking Scorpion. The theory is that by the streamer
being curled like that, the upper part grabs into the air and flaps
around more than an accordion fold does.
The best thing is how relatively easy it is to do compared to making
accordion folds. Roll the streamer up on a 1/4-3/8" diameter dowel,
then squash it flat and iron it to set the folds. Now for really small
diameter models it's not so easy to do that, you'd need less than 1/4"
dowel to do a streamer for an 11mm diameter model, since once it's
flattened and so forth it would be too big to fit inside the tube. But
you get the idea.
And to add another kink, literally, is the idea for double-scorpion
fold. Before rolling the streamer up, first fold the top 6" or 12"
down, and then do the rolling and flattening. The result is an "S"
curve to the streamer.
Another fold, long-forgotten, was the helix fold that Ken Mizoi had a
lot of luck with. He did a normal accordion type of fold, and then
with it folded up he did a diagonal fold from say the upper left to
lower right corner as seen with the whole streamer folded up (NOT
deployed). That made the streamer have a helical twist to it when
I suspect that the optimum fold is out there somewhere that might
involve quite and odd mix of the above. Say a double-scorpion fold at
the top with an accordion helical fold? And whatever is perfect for
one streamer material probably would not work so well for another type
and may tend to be very size and mass-specific. So, be sure to
identify exactly what kind of model would be used, and base the test
streamer sizes and simulated model mass (or actual models)