Your app note link is fantastic! Thanks for posting. As I was designing the MAX30105 breakout I was excited to get the ability to sense blinking. We used it on a friend's Halloween mask to great effect. We did consider the effects of longterm IR exposure but we didn't have the app note to work with. Here's what I calculate:

For our project we used the Example2_Presence_Sensing sketch. We turned on the red LED to 0x02 (very low) just to let us know the sensor was operating the IR LED. We were able to bring the IR LED power down to 0x0A or approximately 2mA with a sample rate of 400Hz and pulse width of 411us and were still able to detect blinking.

This equates to 1107us channel timing (page 26 of datasheet). I calculate 740uA equivalent current going to the LED. Assuming forward voltage of 1.35V (page 7) power is 1mW. Selecting various configurations on the Intersil XLS I am able to set an effective avg emitted power of 1.62mW (so more than my setup) using the ISL29038 set to 6.25ms sleep and 110mA IRDR current. Assuming 30mm distance from sensor mounted on Halloween mask to the eye, we get corneal exposure of 1.8W/m2 and an exposure limit of 100W/m2. We should be good for any length of time.

That said, less is better. We had very rudimentary filtering and with some better FIRs we could probably bring the sample rate down and overall exposure time quite a bit lower.

My setup example depends heavily on the distance from sensor to eye. The further away the sensor can be the less exposure the eye will have to IR, but you'll need to increase the IR LED power (so six of one, half-dozen of the other). I'd recommend placing the sensor closer to the eye and turning the IR power way down so that you get a cleaner signal, filter more, and see if you can get the sampling rate as low as possible.

Yep. Discussed and fixed on this issue on the repo.

The sensor is advertised by Maxim as being 'capable' of detecting smoke. The issue is that there is not currently an open source algorithm that I know of that has been developed. So the sensor can certainly give you data (I see green, red, and IR LED light be reflected at these levels) but it's a challenge that has not yet been solved turning those ADC values into information (I see 7% smoke in the air!).

What you need is SCL and SDA which are the I2C lines. On the ATMega328 they happen to be on pins A4 and A5. On the Pro Micro they are on D2 and D3. It looks like they are not labeled on the board because of the size, but check out the graphical datasheets on the Pro Micro for more information.

Sounds like you have a baud rate mismatch. Double check to see you have the Serial Monitor set to 115200 baud and see if that clears things up for you.

Frickin awesome question, unfortunately I don't have a definitive answer.

The sensor has a glass lens to protect the inside bits but an additional layer to isolate the SMD components is a great idea.

You should be able to detect a binary 'do I see paint?' easily enough but calibrating that to the amount of paint is tricky. We haven't seen full-blown particle detection projects yet and I'm not sure of the flow rate out of the nozzle and whether the sensor (at 500Hz) would be fast enough to properly capture it.

As long as the transparent particles reflect IR/Red/Green differently (and I bet they do) then the sensor would be able to detect them.

Hmm. Your SAMD21 platform is not pulling in the BUFFER_LENGTH define that is usually in Wire.h. The quickest fix is to throw a

#ifndef BUFFER_LENGTH
#define BUFFER_LENGTH 32
#endif

at the top of MAX30105.h. I'll do some more testing (Wire.h should already have this) and update the repo as needed.