How safely to send signal between logic and higher current pcb?

Question

I'm designing a PCB with 4 FETs. These are driven in the same board by totem pole drivers, and their load are about 5A total. I am using PWM at 1000Hz.

The signal comes from a logic IC in another board. Is there a safe way to avoid loops and noise without using optocouplers?

The current scheme I am using is:

Logic board to Buffer PCB = 4 signal wires. And each board has its own connection to the PSU. (GND and 12V) The distance is going to be ~20cm max in what I think will be a noisy environment.

Answer 1

I deal with exactly this situation on a regular basis. Start by treating the ground connection at each pair of power boards as a star point for the power supply and controller board. That is: relatively heavy wire between the ground terminals of both power boards. Another wire from that point to the power supply ground. One final wire from that point to the ground on the controller card.

What you want to avoid is having the load current cause a shift in ground potential between the logic board and the power boards. The ground can (will) move around with respect to the power supply ground but so long as the signals between the control and power boards remain constant with respect to each other, you should be good.

Answer 2

Is there a safe way to avoid loops and noise without using optocouplers?

If you can't maintain a common ground between your boards without creating a ground loop, then some kind of isolation may be required. If you don't want to use optocouplers, you could look at transformer-based isolation. There are devices out there that take care of modulating up an essentially dc control signal to pass it through an isolated connection (for example, from Maxim).

If you have a roughly common ground between the boards but don't want to connect ground along the same paths as your control connections, then some kind of differential signalling should work. LVDS seems like overkill for any signal switching at less than 50 Mbps, and current loop seems like overkill for a distance of only 15 cm, but either one ought to be able to withstand a few hundred mV of ground variation between the ends of the connection.

Answer 3

Based on what you've said so far, twisted pair to carry the signals should be adequate. You may well want to use serial termination on the lines. I see no reason to expect major problems with pickup issues, and with a 1 KHz PWM frequency you can accept a certain amount of signal degradation. If you want to add insurance, put a certain amount of low-pass filtering on the signal inputs, followed by a Schmitt trigger.

Also, your load lines should be twisted pair as well, and if you want to really do things right, use shielded twisted pair, with the shield grounded at the driver end, but not the load end.