Document glitch filter values

[mbenkmann]

I'm reading through the code to get an in-depth understanding and I'm hanging on this part

// Set filter
if(busFreq >= 48000000)
    *(i2c->FLT) = 4;
else
    *(i2c->FLT) = busFreq/12000000;

Could you document this better. How did you arrive at these values? Why have the glitch factor depend on the bus frequency for slower buses and have a fixed value of 4 for faster busses. Why have it depend on the bus frequency at all? Shouldn't the glitch filter depend on the I2C frequency or more precisely be based on the number of bus clocks per i2c clock pulse width?
And why such a low value? If I compute this correctly, the above code would set the glitch filter to just about 40ns for a Teensy 3.2 with F_BUS 36000000. I haven't done any experiments but my gut feeling is that just by touching a jump wire on my breadboard I could get glitches longer than that.

Wouldn't it make sense to use the highest value for the glitch filter that does not interfere with the data rate?

[nox771]

Actually the glitch filter settings were from the original T3 Wire code. I don't think I modified them because the existing settings always seemed to work (it probably goes back to version1, I don't recall).

I don't know the internal workings of the I2C peripheral, but I assume it is a synchronous system. It is likely sampling the I2C data at the bus rate, and this filter is more of a digital filter to guard against high-speed double pulsing at the crossover threshold. It is not setup in an analog filtering context. This type of filter will manifest as a latency to the I2C state machine, so I would be wary of setting values too high (it may skew timing or mask events). This is probably why the filter is decreased at low busFreq.