Latency of USB I/O on Octatrack
I was asked yesterday about latency figures for USB I/O on the Octatrack. Here are some numbers:
Octatrack USB round-trip latency, measured (image usb-io-tracks-main-cue-abcd from octabam’s GitHub)
I sent a short sweep into each USB input (A, B, C, D, one second apart), recorded all 20 USB outputs, and cross-correlated to get the gain and lag of every path.
Setup: 8 tracks with THRU machines on A, B, C, D, A B, C D, A+B, C+D at AMP VOL +63, with CUE at 127 on every track.
Routing: every path is correct. Unity on single inputs, −6.02 dB per input on the A+B / C+D sums, stereo pairs land on the right sides, and MAIN and CUE both match the predicted sum. All 20 outputs (16 track channels, MAIN, CUE) have the same latency to the sample.
Latency: 11.2 ms round trip with a 32-frame host buffer, no dropouts. Of that, the Octatrack's own share is about 3.8 ms (150–170 samples across runs, depending on where the USB rings settle); the rest is macOS CoreAudio. The unit's share stayed the same while I varied the host buffering, so the remaining headroom is on the Mac side.
In Logic (32-sample I/O buffer): Logic reports 6.6 ms round trip and compensates recordings by that. I played a click track out to input A and recorded it back through a THRU track, and every click landed 168 samples (3.81 ms) later still. That's the Octatrack's internal buffering, which CoreAudio can't see, so Logic doesn't compensate for it. Total round trip in Logic: about 10.4 ms. Setting Logic's Recording Delay to −168 samples lines overdubs up exactly.