It doesn’t actually work well for stuff longer than a multiple of 64x8, unfortunately.
What I was hoping:
If you start recording with a pickup machine while the sequencer is stopped but STOP recording by tapping the stop key instead of the correct way (rec 2) you can record any arbtrary number of steps in the pickup machine (up to the capacity of the RAM), but you’ve tricked the OT so that it won’t change the tempo to match the legth of the loop. Since PUMs always have timestretch on, I was hoping that even though the audio in the record buffer wouldn’t be exactly the length of the recorded number of steps at the current tempo (because it’s impossible to press stop with sample accutrate timing), it would be timestretched to to actual project tempo. Then you could use it as the master loop length and use a second PUM with the LEN parameter set above 1 to record an even LONGER loop, this time with the sequencer running so that it WAS perfectly quantized to the project tempo.
What actually happens:
When you start the second PUM, its length is set (on the display) to the correct multiple of the master PUM’s length PLUS ONE STEP, and I think it wasn’t actually quantizing the length correctly. I didn’t bother testing any further, but it looks to me like the second PUM quantizes the actual recording time to a multiple of the real record buffer length from the master PUM, and then timestretches it to match the project tempo, the same as if you had loaded it into a flex machine.
The good news is it DOES work well as a method for recording loops up to 4096 steps, because you can set the master PUM’s track to 64 steps with the scale set to 1/8x (for lengths up to 512 steps at x1 scale) and then use the LEN parameter to record loops up to 512x8 (but not every length between 512 and 4096 is possible, since you’re limited to multiples of the master PUM length).
One way you might be able to extend it would be to save the second PUM’s record buffer and then start the process again, but this time use the old “pre-load a file into the PUM’s record buffer” trick to force the master PUM’s length to go above 512 steps with actual sample accurate timing. I haven’t tried it, though, and at that point the RAM limit probably doesn’t make it very useful anyway.