I'm using a piece of equipment operating on the principle of accelerometry for 2 decades now, and it's amazing how effciently it can pick up noise AKA vibration. Your cell phone has a 3d accelerometer for various tasks, I'm not saying it WILL work but I expect it will. Only question IMO is how much vibration FC adds to the mix.

Come to think of it...I'll run a tracing of my next roast to see if I can distinguish FC from the background noise provided I don't melt the plastic casing around the transducer in the process.

Oren,
FWIW
My first roaster was a Hotop and I had trouble hearing 2nd crack so I bought a stethoscope in hope of hearing cracks through the metal case.
IT DIDN'T WORK. The frequency response of a medical stethoscope is very limited and likely tops out at about 5-khz. Your best bet might be to separate a piezoelectric mike element from the hot surface with an air gap. Step II would be to filter out the background noise with a high pass filter using a graphic EQ or even a capacitor in series if you like experiments.

Have you tried isolating the roast chamber. Another contributor suggested using a simple piece of trimmed cardboard to fill the gap between the lid and lower body of the Gene. I insert this piece of cardboard when I feel FC is approaching and place my ear about 4 inches from the Gene (I too, have my Gene exhaust piped out a window with high temp automotive silicone tubing). I really have had very little difficulty hearing FC with this method. This, I may add, is after years of car stereo, power tool, and firearm abuse to my hearing. I tried stethoscopes and even off-the-shelf hearing aids in the past, but the cardboard method has been my go-to for the last 2 years. Might be a course of action to try before spending considerably more on a microphone?

@Quidgybo - can you post a picture of the cardboard method? It's a little difficult for me to imagine.

Thanks!

With beans for which the first and second cracks are hard to hear, I use a cardboard tube, the sort that's in the centre of paper towels. I use it like those old ear trumpets, with one end near where the chaff collector joins the gene. It's long enough that I don't burn my ear (or hair) - ha ha.

Sometimes I've put the end of the tube in the gap near the rotating drum, however I've found the chaff collector is usually as good a spot as any. I think it helps magnify the sounds.

This is what I use when roasting in my RK drum on my Weber gas grill. I listen at the rear of the grill for early signs of first crack. Once it gets rolling I don't need it any more.

You could try a straight rod, or large screwdriver, that is what mechanics sometimes use to locate funny noises in engines...still think an accelerometer will work, a test should be done today or tomorrow

I can simplify everything for you ... ignore the sound of first crack.

I mark it at 395, regardless of what I hear. Some beans crack early, some later. In my experience, this is less important than the temperatures you are hitting.

I dial each bean in to 1:30-2:00 minutes past 395F to give fully developed roasts, without baking out flavors.

I just gave it a whirl using accelerometry....Proof Of Concept worked, I can clearly differentiate between an empty drum spinning and loaded drum spinning regardless if I put the transducer on the outside casing (using some chipboard to insulate the transducer from the heat) and very similar results attaching the transducer to the BT probe, which theoretically should be the optimal location to pick up vibration.

I have to juggle the data a bit to see if there is a way to differentiate loaded baseline from FC noise, at first glance the difference is small but it may be lurking in the noise.

EDIT; It's not an effect that is readily picked up (if at all), perhaps if the signal analysis whizz kids from Artisan would look at the data they might be able to squeeze something out of it but I only see an average level that is almost the same before (avg 30) and after FC (28)
(taking 60 data samples (1 every sec) well before FC and after rolling FC)

Graph is only shown for reference since the raw data are pretty boring numbers to look at see if you can spot charge