Ableton Forum

i discovered something interesting today... i had loaded a simple drum loop from Reason into Live and zoomed in on the 4/4 kick drum. Should be hitting right on the Warp marker right? (tempo was correct). Well, each kick drum was off a tiny bit, 0.5 - 5 ms on either side of the warp marker for each kick. So i adjusted the warp markers to line up perfectly. Then i recorded the time-corrected beat onto another track, no FX, and looked at the warp markers. They were even more off this time, by up to 10 ms! Is this an issue with Live and Reason, or a general feature of computer-based music, or is it because of my system? i'm running an M-Audio Audiophile 2496 on AMD 2600, 512MB RAM, WinXP.

Was Reason rewired to Live or did you export the drum file from Reason and then import it into Live? The reason I ask is you may be experiencing timing jitter, which is sometimes encountered when rewiring Reason to other applications. If it's the former case, then lowering your sample buffers in Live reduces the problem. If it's the latter case, then it's a problem with Live and/or your setup. Are you on PC or Mac?

There was no ReWiring involved. The loop was exported from Reason at 24/48, then imported to Live, also running at 48kHz. "Beat" warping was used, but this shouldn't matter, as all tempos were the same. Original Loop was at 135 bpm, after time-correcting it was at 135.16 bpm, so i set the master tempo on Live to 135.16 bpm also. I'm running a PC.

How where quantise settings of set/clip?

Global quantization set to "Bar", Clip quantization set to "Global". The clips for sure start at the same time. A 32nd note at 135 bpm is 55 ms, which is way more than the timing issues i'm experiencing. what originally got me interested in this was an interview with bt where he was talking about the inherently "off" timing of MIDI, and how he hates having to time-correct everything. when i saw his Laptop Symphony in Dallas, he had a friend manually adjusting warp markers on Live before his set, using a Powerbook. But if Live introduces its own timing issues, adjusting the warp markers would be pointless busywork. You wouldn't think a millisecond here and there would matter, but it does make a difference, especially when you have many loops going on, all flamming slightly with each other.

okay, i just ran some tests... i loaded a time-corrected bt breakbeat into Live and recorded with no effects into another track at the loop's original tempo, 135 bpm, so no timestretching going on. The waveforms in this loop's original form line up perfectly with the warp-lines. After recording it onto another track, the waveforms are all offset 2.5 ms after the warp-line. So i have determined that i have an overall latency of 2.5 ms. I tried changing this with different Global Latency settings, but no effect.
So i panned the bt breakbeat hard right and loaded a Reason generated beat hard left. The Reason beat was at the same tempo as the bt breakbeat, and had the warp-markers lined up perfectly with the waveforms, thus slightly timestretched, using "Beats". I recorded both loops, at 135 bpm, onto another track, and compared the left- and right-channel waveforms. The were both lined up perfectly together, but offset 2.5 ms after the warp-lines. So time-stretching in "Beats" mode does not appreciably add to latency! I also tried time-stretching both loops at more extreme tempo differences, but the timing was still right on, although with the 2.5 ms latency. Makes sense given the nature of "Beats" mode warping.
So i tried the same comparison tests of a bt breakbeat in "Beats" mode vs. a Warp-Adjusted Reason beat in "Tones" mode. The smallest Grain setting introduced additional latency of up to 10 ms vs. the bt loop. Larger Grain settings increased the latency. The latency stayed the same regardless of whether the beats were stretched to 75 or up to 185 bpm. I also compared "Beats" to "Texture" mode. At Grain Size "2.00" and Flux at "0.00" there was negligible latency added vs. the bt loop. As Grain Size and/or Flux increase, the waveforms drift from sync. Sometimes ahead, sometimes behind the 2.5ms-delayed bt loop, anywhere from 2ms to 30ms depending on Grain Size and Flux.
So apparently the timing is right on with "Beats" mode, but gets a little, or a lot, off with "Tones" and "Texture" modes. As the latency increases with larger Grain Settings, it would seem to pay off to carefully select the smallest Grain Sizes you can get away with. It's still worth it to me to Warp-Adjust my loops whenever possible, just to get the timing as right on as possible. I wonder...is there a way to generate audio that doesn't need time-correcting?

Do you really hear a latency of 2.5 ms ?

Even though this seems like a tiny amount of time I'm still impressed with this guy/girl's findings. That was pretty thorough for a home test. Maybe you should go work a software company as a tester!

Regarding the timing; it's a little disappointing but for my uses I don't think I'd notice, not to say it wouldn't cause a valid pain for others. Makes me think of my old E-mu SP1200. That thing had the most brutal swing and always sounded insanely tight to me. I swear, to this day I think I can spot one being used in a track.

Makes me wonder how Live's timing compares to a modern MIDI sequencer, like an Akai MPC. Anyone got a good idea of a way to compare the two? Maybe just record both simultaneously into Pro Tools or Logic and compare the two, WAY zoomed in on the waves?

i don't think 2.5 ms would be audible as a delay. Delay isn't audible as delay to me until up around 30-50 ms. On the other hand, flangers generally work in the 1 - 10 ms range to generate comb filtering, which i can hear. (Try recording a loop onto another track, then play the original and the copy at the same time...) The good thing about Live's latency is that it is consistent; everything gets offset by 2.5 ms on my system. As long as it's consistent it can be easily controlled, if desired. Either way, there is probably no way to reduce the latency to zero, as software always requires some amount of processing time. I really don't think it's a problem with Live. Even if it was, Live kicks so much arse, i wouldn't really care. I'm more concerned about the inconsistent timing of MIDI- and software-generated music. Mostly i'm just having fun figuring out how to test these things and come to a better understanding of audio and its production.

It also explains the variable phasiness of mixes produced in Live here. I'm generally happy with the sound of Live given the clever stuff it is doing with audio clips but often get that "this just ain't sounding right" feeling with phasing and subsequent volume drop. Through a PA I'm not so concerned but when mixing through dynamics, especially EQ, I have long noticed having to make "unusual" adjustments to compensate when certain clips collide which would not normally be the case in, say, a normal mixing environment. It's similar wih Reason, though a different "problem", ever noticed how a mix springs back to life by routing all tracks through an external host via Rewire and mixing there? It's like night and day. In that case it's more of a "fizziness".

Still, live is such a beautiful work of art I forgive these slight shortcomings.

Your phasey/fizzy correspendent.