Cable

• Abusing UTP for I²C

Abusing UTP for I²C

I'd like to abuse category 6 UTP by using it to carry I²C traffic. I used an AVR M-Tester to measure the parasitic capacitance of the different cores in a 30 m length of CAT6 cable.

A	B	Capacitance/pF
White/Green	Green	1703
"	White/Orange	869
"	Blue	770
"	White/Blue	769
"	Orange	869
"	White/Brown	879
"	Brown	872
White/Orange	Orange	1550
Blue	White/Blue	1606
White/Brown	Brown	1565

• Of course it makes sense that the pairs have higher parasitic capacitance since the conductors are closer to each other, but it's useful to know when choosing the wiring scheme. The scheme in use is:

  Position	Use
  1	GND
  2	+5V
  3	SCL
  4	SDA
  5	GND
  6	+12V
  7	GND
  8	+12V

  • It might be useful (or at least not detrimental) to have capacitance across the power supply lines.

  • Would it be better to pair SDA with SCL, or to pair each with either GND or +5V?

• I wonder whether the capacitance of a pair is correlated to its twist rate.

• I²C specifies a limit of 400 pF for bus capacitance.

• The bus capacitance rises as devices are added.

• The parasitic capacitance of the bus lines affects the slew rate of the signals, which limits the maximum frequency at which changes can reliably be discerned.

  • The ATmega328 data sheet recommends pull-ups of no more than 1000 ns ÷ C_b for f_SCL < 100 kHz, where C_b is the capacitance of one bus line in pF. The result is in kΩ.
  • For the worst case (cross-pair) parasitic capacitance of 869 pF, that's 1.25 kΩ (1.36 kΩ using the 0.8473 constant from the TI note).
  • The AVR data sheet specifies that pull-ups should be no stronger than 1.53 kΩ, so there is already a problem even with no devices on the bus (1.36 kΩ < 1.53 kΩ).
  • Luckily, we don't plan to run 100 kHz. On the grounds that t_RISE(max) is 10× at 10 kHz (rather than 100 kHz) then R_PU(max) would be 13.6 kΩ, which is encouraging.
  • With 1k5 Ω pull-ups, the maximum bus loading at 10 kHz should be:
    • 1.5 kΩ = t_RISE(max) / (0.8473 * C_b)
    • t_RISE(max) = 10,000 ns
    • C_b = 7868 pF = 10000 / (1.5 * 0.8473)
    • ..which is well below the parasitic capacitance of even a twisted pair, with plenty of headroom for devices on the bus.