very cool tip!

Cheers Ben!

I was just wondering about a way without getting a commercial product, thamks

Does twisting the pairs deal with capacitive coupling? Like, could one use unterminated ethernet cable?

Dont forget the quality of the wires itself ( real copper or cheaper alluminium )

yes or RS485

@@labiadhchokri2124 that could work, except there is cable count constraint

@@KangJangkrik It's electrically quite similar.

It's a shame you can't convert I2C to canbus or RS485.

You still require transceiver as external chip if you want to do this. Which is no different than this

Differential is mainly to handle external noise that will apply to both lines but leave the differential voltage umchanged. But differerential signals doesn't solve shark tooth or ringing.

The PCA9615 is intended to do differential IIC signaling over long distances

That's to reduce pickup of disturbances. And very best with balanced signals that aren't related to ground but to each other. You can't solve capacitance (creating shark tooths) or inductance (ringing) by twisting. But a proper termination affects reflections at the cable end.

Could you explain why that helps with parasitic capacitance and inductance

@@alexstone691 The parasitics will still be there, in the same way, e.g. a Ethernet wire has them. The trick is to be able to account for their effect: the ringing is due to the wire behaving as transmission line. By matching source and sink impedances to the line you eliminate reflections and ringing. Of course, this only works if the loading is known beforehand.

​@@juanignaciocarrano860 Then you could have any wire length you want right? as long as the circuitry can drive the load of course Also thanks for explaining :D

@@alexstone691 not quite, there is dispersion and wire resistance. The thing is i2c is not really designed for this kind of use. For longer distances some sort of differential signalling is better (eg RS485)

@@alexstone691 You can never have any length wire for two way communication even if assuming zero resistance. The cable still introduces a delay that is proportional to the length. Even if terminating to avoid reflections at the end of the cable, that delay will affect the arrival of the response to a query. So if the master is doing the clocking of the data, the master can only handle a limited amount of response delay before the response no longer matches up with the clocking. Cable delays is also the issue with ethernet cables - the cable has multiple pairs, and the individual pairs will not have identical delay. So going from 1 gbit/s to 10 gbit/s or 100 gbit/s means these delay differences makes bits on different pairs arrive at different times. And higher speed means shorter bit times.