A Late Audio Drift in a Regional Live Programme: Daniel's Case

Daniel produces a weekly regional magazine programme from a small studio outside Cambridge. He rang us in the quieter weeks of mid autumn with a complaint that sounded minor: presenters looked slightly early or late against their own voices, but only some weeks, and never during rehearsal. His crew had already swapped cables, rebooted the vision mixer and blamed the camera feed, and the fault still came back without any pattern they could name.

What caught my attention was the word 'creeping'. A fixed offset is easy to live with, because everyone adjusts and moves on. A drift that grows over a ninety-minute transmission is a different animal, and it usually points to something in the signal chain that is being asked to do two jobs. I asked Daniel to send his routing sketch, however rough, and a note of which sources fed which desk.

He sent a photograph of a whiteboard. It was enough to begin.

The sketch showed three radio microphones and a presenter's lapel feeding a small mixing desk, with the desk output going to an audio embedder and then on to the main switcher alongside the camera signals. A second path took a copy of the same mix through a separate processing unit used for dialogue clean-up, and that copy came back into the chain further downstream.

On paper nothing looked wrong. Each box did what it said. But the clean-up unit added a small amount of delay, and nobody had compensated for it on the return path. When the processed copy was selected for transmission, the voice arrived a few frames behind the picture.

That explained why rehearsals looked fine: the crew rehearsed with the unprocessed mix and only switched to the cleaned feed once the programme was live.

• Two audio paths existed for the same microphones
• Only one path carried processing delay
• Rehearsal used the path without the fault

I visited on a Tuesday and resisted the urge to start turning dials. The first task was to measure. We recorded a clapperboard-style test, a presenter tapping a hard surface in frame, then played the programme output back frame by frame and compared the visible contact with the sound spike on the waveform.

The unprocessed path was within a frame of correct, which is acceptable for most live work. The processed path was three frames late, and the figure was steady rather than creeping. The creeping impression, we discovered, came from the vision mixer occasionally re-timing its inputs after a source change, which made the error appear to wander.

Having a number mattered more than any theory. Daniel's team had spent weeks arguing from impressions; once the offset sat in front of them as three frames, the discussion became practical and calm.

We did not buy anything. The processing unit had a delay setting that nobody had used, and the desk could apply a matching delay to the unprocessed path so that both options arrived together. We set the two paths to line up, then checked again with the same tapping test on both selections and found them within half a frame of each other.

Priya Nair, our audio and sync engineer, reviewed the settings by phone afterwards and suggested one further change: fixing the vision mixer's input timing so that it stopped re-timing on source changes. That removed the wandering impression entirely.

Daniel's crew now switch between the clean and raw mixes freely during transmission, and the picture and voice stay together either way. The whole job took one working day on site and an evening of write-up.

• Matching delay applied to the unprocessed path
• Vision mixer input timing fixed in place
• Both paths tested with the same visible tap

A fix that lives only in a few menu settings is fragile, so before leaving I wrote a one-page check for the crew. It lists the two audio paths, the measured offset for each, the settings that keep them aligned, and a short test to run at the start of each recording day. Tom Ashworth tidied the wording so that a freelancer arriving cold could follow it.

The test is simple: tap a surface on camera, play it back, and look for contact and sound meeting at the same moment. It takes under five minutes and catches most causes of drift before anyone is on air.

Daniel told me the sheet is taped beside the desk, and that a new sound operator used it on her first shift without asking anyone for help.

The case was a reminder that lip-sync faults are rarely mysterious once they are measured. Most come from a path that adds delay quietly, a setting nobody revisited after a change of equipment, or a rehearsal that does not match the live chain.

If your picture and voice drift apart only on certain nights, ask which signal path is different on those nights, and measure each one with the same simple test before replacing anything. The answer is often already sitting in your routing, waiting to be written down.

• Compare rehearsal routing with live routing
• Measure each audio path separately
• Write the offsets down where the crew can see them