Powering isolated RS485 over CAT6

Question

I'm creating an RS485 network where data and power will be transmitted over a 23AWG CAT6 U/FTP cable. Two twisted pair conductors will be 36V and GND respectively, another twisted pair will carry the data A and B lines. A buck converter will step down the 36V to 5V at each controller. The length of the cable from the master to the first transceiver is roughly 40m, the subsequent length from one transceiver to another is 1m. The total max power on the bus is 36W which means 1A will flow through the power lines or 0.5A per conductor. Based on the datasheet of the cable, the voltage drop will be roughly 2V.

One of the main requirements of the system is that each microcontroller has an isolated power supply from the other microcontroller. This is because the microcontroller will be using its internal ADC to measure the voltage across a LED driver where I do not want the ground to be shared. To achieve this I'm using a DC-DC Isolated Power Module (HLK-1D0505)

The transceiver I'm using at each microcontroller is a SN65176B My question is should I:

1. power my transceiver on the output of the DC-DC Isolated Power Module and connect the data pins directly to the microcontroller or
2. power my transceiver on the input of the DC-DC Isolated Power Module and use some optocouplers to connect the data pins to the microcontroller.

If I go with option 1, will I have any issue with the ground potential difference for the RS485 transceivers?

Answer 1

Isolation is great for line noise and DC level differences , but terrible for CMRR of power supply noise unless you have perfect differential balanced components on each stub with low input capacitance.

This can easily be demonstrated even with USB on a 2m cable from an earthed tower to an isolated Arduino with dedicated floating SMPS.

Choosing 120 term with 12k pullup/down at driver and 120 term. at the end must be matched with selected cable impedance (often 10% tolerance)

2V drop on 36V is only 5%. No Problem. Let's Verify

~ 50m of 2 pairs of AWG23 = 100m loop/2
66.79mohm/m *50m * 1A = 3.35 V drop total < 10% still no problem , unless other loads demand better tolerance.

But distributed SMPS's each isolated injecting CM noise is a big problem perhaps reduced by shield connections only at one end. It depends on several factors. Perhaps high mu Ferrite magnetics.

Answer 2

Option 1 does not isolate so if isolation is what you need or want you can't use it. The SN65176B tolerates only -10 to +15V on the A and B pins, in reference to the GND pin, so there is no isolation.

Option 2 is better as it keeps isolation of bus and MCU. If you calculated 2V as the max ground voltage difference between devices due to current flowing in the ground wire, the RS-485 should handle that just fine.