Hi,
I2C is a bi-directional bus, so strictly there should be a buffer at each end of a "long" cable, assuming that the SHT31 and host have similar drive capabilities (I haven't checked the data of either). But for a 3 feet connection a buffer shouldn't be needed at all.
The main requirement is to reduce crosstalk betwen SDA and SCL, which is why they should not be twisted (or ideally even run very close) together. The GND and VCC are assumed to be "clean" (no ac signal) so shouldn't produce any crosstalk; the presence of a dc "bias"on VCC is irrelevant because we only need to consider capacitive coupling through the "insulation". Most of the advice in your link looks very good, particularly using a 12 v pullup (if necessary), provided that the catching/clamp diodes are fitted reliably, preferably at both ends.
But the best aim is to keep the capacitance on SDA and SCL as low as possible, so (for example) using individually-screened coaxial cores would be a bad idea. Unless there is severe "local" electrical interference (e.g. a large Switched-Mode Power Supply), perhaps the simplest interconnection would be a flat "ribbon" cable with the ETH/VCC on the two inside cores and SDA/SCL on the two outside cores.
All the above is based on a single master-slave interface, but if there are multiple slaves then a single "buffer" may help to isolate a "long" cable from a cluster of slaves and the host.
Cheers, Alan.
I2C is a bi-directional bus, so strictly there should be a buffer at each end of a "long" cable, assuming that the SHT31 and host have similar drive capabilities (I haven't checked the data of either). But for a 3 feet connection a buffer shouldn't be needed at all.
The main requirement is to reduce crosstalk betwen SDA and SCL, which is why they should not be twisted (or ideally even run very close) together. The GND and VCC are assumed to be "clean" (no ac signal) so shouldn't produce any crosstalk; the presence of a dc "bias"on VCC is irrelevant because we only need to consider capacitive coupling through the "insulation". Most of the advice in your link looks very good, particularly using a 12 v pullup (if necessary), provided that the catching/clamp diodes are fitted reliably, preferably at both ends.
But the best aim is to keep the capacitance on SDA and SCL as low as possible, so (for example) using individually-screened coaxial cores would be a bad idea. Unless there is severe "local" electrical interference (e.g. a large Switched-Mode Power Supply), perhaps the simplest interconnection would be a flat "ribbon" cable with the ETH/VCC on the two inside cores and SDA/SCL on the two outside cores.
All the above is based on a single master-slave interface, but if there are multiple slaves then a single "buffer" may help to isolate a "long" cable from a cluster of slaves and the host.
Cheers, Alan.