# How best to allocate signal wires to 4 twisted pairs that best reduces inteference

I have to transmit digital signals 30meters including power. I am going to use a cat 6 ethernet cable with 4 twisted pairs (8 wires in total). I can already assure that my signal gets to the other end no problem. All that is left is to smartly allocated them to which wire.

I only need 6 wires (5v, GND, LED1, LED2, SDA, SCL). So i have 2 that i can leave it unconnected or can be used as parallel ground or 5v, or which ever.

I would like to allocate the signals to the twisted pairs that best give them immunity to interference, Especially the SDA and SCL line.

With my unexperiend knowledge this is the best allocation i can do

Pair 1

5v and GND

Pair 2

SDA and GND

Pair 3

SCL and GND

Pair 4

LED1 and LED2

• Excellent. Only possible thing you could do differently is swap pair 1 with pair 2. The goal would be to keep power/ground in between the two signals for more shielding. However, I am not sure it matters in a CAT 6 cable. I think all the pairs are equally close to each other in the cable. Any time you run single ended signals through cables, providing a separate ground for each signal will help tremendously for maintaining signal integrity. Commented Oct 7, 2021 at 2:25
• @mkeith assuming there there is 1 more signal wire i need >? which one would be the most optimal to replace? the ground that is paired with 5v? Commented Oct 7, 2021 at 3:52
• Yes, I would say so. The 5V can also act like a ground reference for the AC component of a signal. Especially if you put bypass caps at both ends of the cable from 5V to ground. Commented Oct 7, 2021 at 3:54

The pair allocation is rather optimal, the I2C pins are well shielded for coupling between pairs.

However, CAT6 cable has capacitance of about 50pF per meter, so 30 meters of cable means 1500 pF.

Depending on which I2C speed class you are aiming, that can be a problem, as the specification for Standard 100kHz and Fast 400kHz modes have a 400pF limit. And for Fast mode the speed cannot reach 400kHz with 400pF of capacitance, unless tricks such as active or current source pull-ups are used for acceleration. So just that you know, it can never be within I2C specs, but it should be still possible to lower the speed enough to make it work but slowly.

Which basically is the explanation why I2C is rarely used as an external communication bus when length exceeds a few meters.

• No need to worry about i2c, im using 2 ltc4311 in parallel, even only 1 can do 30meters at 100Khz. I also have tested this myself so im fairly confident. If its reliable though thats another story i will have to see Commented Oct 7, 2021 at 4:58

I have to transmit digital signals 30meters including power.

You don't really have much option other than to drive the SCL and SDA differentially if you want this to be reliable. So, you'd transmit $$\SDA\$$ and $$\\overline{SDA}\$$ on one pair and $$\SCL\$$ and $$\\overline{SCL}\$$ on the other pair. Then use a differential receiver to convert from balanced differential to single ended at the far-end.

I'm assuming that when you said "transmit" that you are just transmitting from master to slave because, if you try and read from distant slaves using a clock generated by the master you'll get into trouble because of the delay time around the loop.

I would like to allocate the signals to the twisted pairs that best give them immunity to interference, Especially the SDA and SCL line.

Twisted pairs utilize a balanced drive signal for them to be effective hence, if you are using twisted pair you need a balanced driver and a differential receiver.