Been running the Smartsuit Pro II for about a year and the magnetometer drift issue is driving me up a wall. Not the normal slow ambient drift. I mean mid-session decay where you're 45 minutes into a capture, your actor's finally hitting their stride on a genuinely great take, and suddenly the hips have rotated 10–15 degrees off and the whole pelvis chain is slowly cooking itself.
Our studio has a steel door frame and an old rack mount server in the corner that I'm fairly convinced the sensors have a personal vendetta against. I've tried repositioning the capture zone but we're working in a constrained space so options are limited.
What I've been doing that sort of helps:
The pelvis/spine drift is usually correctable in post but when it propagates up into the shoulder roots and starts affecting arm rotation it gets messy fast, especially for expressive upper body movement.
Has anyone found a reliable mid-session recalibration method that actually sticks? I've seen mentions of a "mag-clean zone" approach, clear 3m radius of anything ferrous, but that's basically a full studio redesign for us. Also curious whether switching to USB-C tethered vs WiFi has any effect on sensor stability, or if that's a completely separate system from the IMU chain.
And honestly: do Xsens or Perception Neuron suits handle magnetometer interference better in practice, or is this just a hardware reality for all inertial-only setups and I need to accept it?
seconding this. we started treating recal breaks like a camera card swap, just baked into the shoot rhythm, nobody questions it. 45 minutes matches our experience too, though if there's any significant metalwork nearby I'll push it to every 30.
the performers initially pushed back a bit but once they understood it meant fewer retakes from drift killing a good emotional take, they were on board. now they actually remind us when we forget, which is kind of funny.
had nearly the same experience with the Pro II. one thing that helped a bit with the mid-session drift: do a second calibration pose around the 30-minute mark before it gets bad enough to matter, not when you're already seeing it. treat it like scheduled maintenance rather than a fix. doesn't solve the root issue but it cut our data salvage rate way down. still annoying that this is the workflow though ngl.
Something worth checking if you're seeing consistent drift patterns: Rokoko's Contact mode uses body-to-floor contact points as a grounding reference, which helps some people but actively makes things worse in spaces with metal flooring or near heavy electrical equipment. The magnetometer is getting confused by exactly the things Contact mode depends on.
We moved our capture space to a room with wooden floors and away from the electrical panel, and the mid-session drift became significantly more manageable, not eliminated but a different category of problem. Might not be something you can change depending on your facility, but worth auditing the space itself if you haven't already.
the camera card swap framing is genuinely smart for getting director buy-in. nobody argues with "we need to swap the card." borrowed that exact language on our last shoot and got zero pushback, whereas before I'd get "can we just push through" every single time. the one wrinkle is when a performer is finally locked in and stopping feels like you're breaking something real. but if you're 45 minutes in and the data's about to drift, you were going to lose that take anyway. the break is always the right call.
The HVAC thing is no joke. I had a session fall apart because we were shooting near a server room and I didn't even think about the rack infrastructure messing with the magnetometers. Once we moved to a different room on the same floor, drift went from catastrophic to manageable. Completely obvious in hindsight but nobody puts it on the setup checklist. We now literally walk the space with a compass app before any serious shoot.
Something that's easy to miss: are you capturing near any large metal structures or HVAC ductwork? Had a session with consistent one-sided drift that I chased for two shoots before tracing it to a steel support beam directly under the floor on that side of our stage. Moving the volume about two meters over essentially fixed it. Rokoko's documentation is pretty understated about how sensitive inertial suits are to magnetic interference — it's worth doing a quick compass sanity check across your whole volume before a long session, not just at your calibration spot.
one thing that helped us a lot with mid-session drift specifically: we built a quick "drift check" into our shoot prep checklist. before we start rolling on anything important, the performer does a standard T-pose and we eyeball it against a reference. takes 10 seconds. if something's already drifting before we even start, we catch it early instead of 40 minutes in when we've got great performance in the can. doesn't fix the root cause obviously, but it's saved us from re-shooting so many takes.
The Contact mode drift interaction took me an embarrassing amount of time to diagnose. Had a performer who kept developing a slight lateral lean over long takes and I kept blaming the suit or the calibration pose. Turned out Contact mode was grounding against our stage floor, which had subtle unevenness near the edges. Switching to Gyroscope mode for that session fixed it immediately. The Rokoko docs mention the grounding dependency but it's buried, and if you're not already suspicious of Contact mode you'd never go looking for it.
lmao "we need to swap the card" is genuinely the best piece of on-set people management I've picked up in months. no explanation required, no technical credibility needed, universally understood as a Hard Stop. stealing this immediately.
late to this thread but: build a forced recalibration break into long sessions regardless of whether you're seeing drift yet. every 45 minutes, stop, open space, full recal, have the performer do a couple wide-range test movements before rolling again. feels like overhead until you hit minute 90 and discover the last hour has a 12-degree hip bias that crept in so gradually nobody caught it on the monitors. that day cured me of treating recal as a reactive measure.
