Mozg & LensPast
LensPast
Hey, I was just looking at a 1930s Rolleiflex, and I think we could build a hybrid system that combines its bellows with a modern sensor. What do you think?
Mozg
Nice idea—just remember that bellows are a mechanical convolution, not a linear system. You’ll need a micro‑controller to map bellows displacement to sensor gating, maybe a Kalman filter to smooth the analog readout. I once tried a similar hybrid with a Polaroid and the firmware kept rebooting every time the flash fired. Also, the sensor’s dark current might corrupt the 1930s exposure curve; a simple adaptive threshold could help. Sleep? Optional firmware maintenance. What’s the first test you want to run?
LensPast
First thing’s a pure light‑meter comparison: set the bellows to a known focal length, expose a sheet of black‑and‑white paper with the old film’s ISO, then take the same exposure with the sensor and see the differences. That will tell me if the bellows‑to‑sensor mapping is even roughly linear before I bother with any fancy Kalman filters. If that passes, then we can worry about firmware quirks and dark‑current corrections.
Mozg
Sounds solid, but remember bellows stretch isn’t a simple linear function of focal length—it's a logarithmic relationship, so you’ll probably see a 0.1‑unit offset per millimeter. Also, the 1930s film’s reciprocity failure will bite the longer exposures, so your “known ISO” might drift by a factor of two. A quick sensor test at 1/500th, 1/1000th, and 1/2000th will give you a baseline, then you can fit a 2‑parameter curve and feed that into your firmware. If the sensor reads 10% brighter than the film, you’ll have to adjust either the bellows or the exposure meter’s calibration constant. Just be careful to keep the sensor in its linear range—anything beyond 20% of full‑scale and you’ll hit saturation before the bellows even starts to feel like a bellows.
LensPast
Sounds like we’ll need a quick calibration run, then. I’ll set the bellows to a fixed length, fire 1/500, 1/1000, and 1/2000 on the sensor, and then compare the film’s prints. Once we have the curve, we can tweak the bellows mechanics or the exposure meter’s constant. If the sensor is consistently 10% brighter, I’ll just dial the bellows back a bit. Let’s keep it simple and not get lost in code.
Mozg
Nice, just remember the bellows’ stretch curve isn’t perfectly linear—might need a quick fit. If the sensor’s 10% brighter, pull the bellows back a bit, but keep an eye on the exposure meter’s offset. Don’t get caught up in code, but a 2‑point linear fit will do the trick. Once you have that, we can adjust the firmware in a single line. Good luck, and don’t forget to log the raw values for future debugging.
LensPast
Got it, I’ll log every raw readout and keep the math as simple as possible. No over‑complicating it with fancy filters or firmware loops. I’ll pull the bellows back just enough to match the sensor’s 10% brightness, run the 2‑point linear fit, then update the firmware in a single line. Thanks for the heads‑up, and I’ll keep the logs tight for the next tweak.
Mozg
Sounds good, just make sure your log file includes timestamps and the exact bellows gauge reading; that way you can always backtrack if the 10% offset changes when you tweak the firmware. Good luck!