Digital Performer is SUPPOSED to be able to record the incoming tempo, but I haven't gotten it to work yet.

Are there any daw that can easily record a variable tempo map and then sequence to that? Usually I'll just pick the matching project tempo and multi channel record that only to find the groove of the track is boring now... Protools let you set the tempo basically per step but this is a painful process at best.

lol math

that explains that!

I absolutely love playing around the groove command. Sometimes I use a basic shuffle (8 and 4) then break out into full compound time by using a groove of 4 and 4 and I just put an H command on The C line of every phrase. It works really well using the two in separate tracks.

Another thing I love to do when I want to go double time or add Demisemiquavers (or 32nd notes for you US people) is just have the groove to 3 and 3 for simple time or 2 and 2 for compound (don't forget the H command at C).

Pretty much gnore what I said in the first section of my post for now. Looking further at the code shows that the algorithm is more sophisticated than I thought. Long term accuracy might very well the stated 0.005%. This could still vary slightly locally. For badly chosen BPMs, the playback speed could vary slightly over the course of a couple of phrases. I'd need to go back and make more calculations to come with a better answer on how big the error would be and over which time.
If you still want to be safe, you should go for 80, 128, 160 BPM. But then again, maybe that's not actually the problem.

The second part i still true.

ah! yeah. LSDJMC2 would probably be a good solution to that. I even know where you can find one if you want.

So does editing the groove setting send the BPM even further off than it was if it's a number not divisible into 4096? (can't remember all my math terms)

No James I'm doing some other stuff that's LSDJ plus MIDI drums.

I was hoping that .005% was wrong, cause I had to edit the shit out of some LSDJ channel recordings to fit in a 180BPM song.

Huh! What are you syncing to that needs that level of precision? Just curious. Is it just to lock to grid in your DAW or instruments from other clocks?

I initially assumed that your post was something about the Zef thing you just did and having to sync 2 gameboy tracks, but as I understand it from what nitro is saying, while LSDJ tempos may not be accurate in an absolutely sense, they should be accurate to other instances of LSDJ (yes? no?)

Anyway, this is probably the most interesting conversation ive seen on cm.o for a while.

Oh good, I'm not an idiot then.  A few of my most recent projects have tempos like 147.752 in order to minimize the amount of editing I have to do in my DAW.  YAY 6 digit BPMs

With the voice of Richard Stallman, "I'd just like to interject for a moment."

LSDj is nowhere near that accurate. It's using a 4096 clock as its tempo base, which is then divided down to the BPM. This means you can set the time between interrupts in increments of about 0.2 milliseconds. The error for the BPM value varies with the BPM. Higher BPM means lower accuracy. However, there is a trick in place to switch between different values on every tick to increase the accuracy. Which increases the accuracy a small bit.

A real world example: 128 BPM is a tempo that divides from the 4096 Hz clock, and is theoretically completely accurate (to the limit of the clock crystal's accuracy.) 129 BPM is a tempo that divides badly from the 4096 Hz clock, and the real world BPM turns out to be 128.805 BPM. This is 0.151% too low. This one of the worst cases. 130 BPM becomes 130.032 BPM, or about 0.0244%.too high. I think Johan may have made a mistake and meant to say that the worst case error is a factor of 0.005, or 0.5% off. That's still fairly good.

Edit: The algorithm seems to be more complicated than I thought, and the accuracy is higher than the calculations above.

Notice how I purposely used the word accuracy instead of precision. Precision is something different than accuracy. LSDj has a relatively bad precision, ie the time between notes may jitter such that the time between each note varies, even if the long term average converges to the 0.5% or smaller error. This is because the CPU may be busy doing other things when a timer interrupt happens. The culprit is typically kits as well high frequency vibrato, and not using these will reduce the jitter. I haven't made any calculations on the actual error from this but might do so in the future.

The only time I've ever found a use for complicated groove tables was the other day when I made 1/32th notes like this:

5 55%
4
4 50%
4
4 50%
4
8 34%
F

So in the end it still evened out to 6 ticks per step.