I'm using the MK22FN1M0AVLH12, which doesn't have the cystal-less IRC48M option the MK22FN512 does. (NXP folks: I'd have gone with the newer part if I'd known it was the newer part, but the selection guides don't cover that.) I know that the FLL is not capable of meeting the USB certification specs for jitter, but my question is whether it can be used successfully with at least a fair degree of reliability.
The target application sometimes takes a lot of physical abuse and occasionally crystals fail. Occasionally most of the parts can fail, but there's a fair amount of fault-tolerance built in and it'll keep limping along as best it's able. The crystal is an exception, though, and currently a dead crystal means a dead device. I've already written code for a semi-related project on a K02 that detects a failed crystal and switches to a trimmed internal clock, but the complication with this one is USB.
It's questionable whether the frequency accuracy of the trimmed internal reference will be good enough if it works at all, and the clock configuration tools won't even allow you to select the FLL as the USB clock source, but I'm not asking if it's a good idea, just if it's likely to have a 50% chance or better of working long enough to get a replacement in place.
Has anyone tried this? What were your results?