Terminating 400kHz Digital Signal on 10ft Cable

[Noggin]

I need to terminate an I2C bus on a board in which space is HUGELY constrained. The board of which I have control of will be the master. The slave board is on the other side of a 10ft cable. Data rate is 400 kHz.

I think in general that the clock should be series terminated at the slave's receiver input, but there isn't a series resistor on the board. The data will be bi-directional, so I'm not sure where to terminate that signal. I can probably do parallel or series termination on my board, but probably not both. Both boards have places for pull-ups for both the clock and data signals.

Is there anyway to know which termination option would be best without having the hardware in front of me? On the other hand, 400 KHz really isn't all that fast so maybe it isn't even an issue.

 

[dknguyen]

At the speed of light and fundamental frequency of 400kHz, the wavelength is still 125 times longer longer than the round trip distance of the wire. We want the voltage level in the line to be the same everywhere which is at the very least less than 1/4 the wavelength (power engineers seem to like this number). For digital transmission let's pull a number out of our ass and say 1/10th (others like 1/100th). Using those numbers, you can have frequency harmonics up to 12.5x the fundamental frequency. A square wave has odd harmonics of the fundamental frequency so that means your square wave can be made of the 1st (fundamental), 3rd, 5th, and 9th. That's pretty square for our purposes. My opinion is it's at the moderate edge what the rules of thumb guidelines and that you are probably okay with no termination.

But in case you're curious, Thevnin termination what you seek:

A resistive divider at the each extreme end of the line (past the most extreme node is preferable), with all nodes having a minimally short connection to that line. The resistive divider ratio is sized to match the line impedance and is the most ideal way to terminate a signal (and also the most cumbersome). However, it works with bidirectional lines as it does not matter if a node is receiving or transmitting as the signal does not end there. It continues on it's merry way to the end of the line through the termination impedance which matches the line so absolutely reflections. Whereas single ended terminations are affected by what is the source and what is the sink, and limit the reflections rather than provide a continuous impedance to eliminate them.

https://www.electro-tech-online.com/custompdfs/2011/01/AB023.pdf

And unlike what you said, I do believe the resistor goes at the driver end of the line in series termination.

EDIT: Of course, the pull-up resistor requirements make any kind of resistor-to-supply termination difficult. That's why I'd try to get by without it.

 

[Noggin]

Awesome, thanks. And you're right, the series termination resistor goes at the source not the destination. Just got it mixed up for some reason. I think the PIC's have optional slew rate control built into the I2C outputs, I can turn it on and slow down the edges just a bit. I'll just add series termination to my board on both signals. If nothing else, it should help with EMI (and I can swap them out for ferrite beads if I need to for better EMI reduction).