SN65HVD72: Long-term degradation to transceiver pair when motor is active

Overvoltage at any pin can cause damage.

Check all pins with an oscilloscope while the pump is running. Also please show the board layout.

Here are the scope shots of the 115200 baud RS485 pair and the delta with and without the pump motor active for reference. I will check with my team on posting layout shots.

Please note that your oscilloscope has a built-in screenshot function.

I am not able to see if there are any spikes that exceed that absolute maximum ratings. But depending on the board layout, it is also possible for noise to be coupled into the digital signals, whose pins are much less robust.

Here are shots of the board layout of the CCA with the pump driver that is causing the interference/degradation for reference.  U20 is the transceiver.

On the other end of the CCA is the 48V to 24V buck pump driver. J3 is the pump connector:

Here is the stackup as well:

Is there noise on any of the traces near U20?

Please check the signals at all pins of U20 for voltage spikes.

Hi Luke,

So I have a few notes on the issue that you are seeing:

1. Are both sides of the RS-485 bus terminated with 100 Ohms - or is it just one side? I am asking because technically speaking that is "overloading" the RS-485 bus - as 54 ohms is the absolute minimum between A and B - I don't know if that will cause the issue at hand because its barely overloading the bus (I'd expect a little lower output voltage at worst - but I wanted to mention it just in case)

2. Are the devices that fail in the communication - after restarting the system are the parts functional again or are they damaged to a point where you need fresh units installed? 

3. It does look as if the bus is pretty noisy when the pump is on - which isn't not too unexpected - but some potential circuit modifications may prove to be useful in combatting noise errors:

3A) Split termination - i.e. split the termination resistor in two (ideally they are 120 ohms so it would be 2 60 ohms - but I understand you are using 100 ohm which is, I am assuming, why you are using a 100 ohm termination - which then you'd split into 2 50 ohms and add a cap to ground inbetween the resistors - we usually see around 470pF as a common value)

3B) Ferrite beads can also help attenuate some of the noise without too much issue with the needed differential signal.

3C) Common mode chokes work well - but they are larger and based on the layout provided already it may be much harder to modify a board like that without a pretty extensive redesign - but they have shown to help in cases of EMI mitigation. 

3D) I generally don't like suggesting them as they do have a decent amount of negative impact on the bus - but filtering capacitors to ground - so on A and B place 50pF to 100pF capacitors to ground (sometimes up to 470pF ish, but I generally try to avoid anything especially when there is already a diode on the line which does add some capacitance) - will help filter noise. However they could reduce max effective bus length as you are adding additional capacitance to ground on the communication nodes - since this seems long distance it may not be the best way forward - but it is an option. 

3E) Some of the things I have covered above are explained in some more detail in this reference design -  here - it uses a different IC - but the idea remains the same across most RS-485 devices. 

Please let me know about my questions on points 1 and 2 - as the layout looks more or less okay (its fine) and the diode chosen is a standard RS-485 protection diode - so I wouldn't really be worried about noise signals damaging the pins (unless there is a major fault - but I'd imagine there may be a little bit more evidence than just failed communication). 

Best,

Parker Dodson

Thank you for the detailed response. 

1) The 100 ohm termination is only on the controller transceiver. I experimented removing it as well since the band rate is low and it didn't make much of a difference. 

2) The general progression is this. The two board system starts nominal without comms mishap even with the pump active. Eventually, activating the pump prevents communication which automatically causes a reboot, upon which communication returns until the pump turns on again. Switching either board seems to prolong communication for a short while. 

3) I will attempt to white wire in some your suggestions, though as you can see space is minimal. The system is also encapsulated as it functions underwater, so testing can be challenging.  What choke frequency should I aim for? The pump buck switching?

I also am going to try and add a better flyback on the pump along with some small caps to try and mitigate the spikes at the source. 

Thank you again

Hi Luke,

Absolutely no problem!

1. If possible - terminating both sides of the bus would be ideal in clamping down noise - especially when it is being coupled in due to a noisy signal near by (the pump). I understand you are using 100 ohms - so it will be a bit lower than the minimum of 54 ohms - but in terms of application you may slightly undercut the minimum 1.5V between A and B - but the difference of 4 ohms really shouldn't be that concerning to the transceiver in terms of robustness (the device can withstand short circuits - so robustness isn't the concern - but you could have slightly degraded output voltages - however reflections could be reduced when both ends are terminated. I understand it's a lower baud rate - but the energy content in the signal during a signal transition is what reflects first and you will almost have 100% of the incident signal reflected back onto the bus on the unterminated end. It may not be enough - but it should reduce the reflections seen. 

2.  Thanks for the clarification on the issue - I understand what the problem you are seeing is - and based on that making the system more robust to EMI would be the best path forward; which it seems like you are going to try to implement as best as you can with current board space limits. 

3. Completely understand the challenges you face - please see suggestion in point 1 - I'd try that first as it may be more workable without a lot of additional space needed and ultimately that is the first step for clamping down on reflections; typically the other methods are applied when the first is not enough on its own. 

That being said - aiming for the pump frequency or slightly lower would be good for the choke as that does seem to be the main source of the interference. 

Please let me know if you have any other questions or if modification and testing prove fruitful in helping the design. 

Best,

Parker Dodson 

In your opinion, would adding the RC pullup / RC pulldown to the A/B lines be of greater or lessor value than adding the split termination? It may be easier to make that mod.

Hi Luke,

That is a bit of tricky question. 

So overall both of those methods should see improvement in EMI mitigation - you may actually see more reduction in noise with the filters. 

That being said - the split termination will have the least negative impact to the actual data signal, so in general I try to suggest that first; however, it doesn't seem like the differential voltage you are measuring when pump is off is super low  - so the benefit of the filtering may outweigh the data signal attenuation. 

So it would probably with worthwhile to test the filtering method if it is an easier modification - there is some risk with doing this - but since the biggest issue seems to be EMI issues I think the filtering is worth the risk to at least test. 

Please let me know if you have any other questions!

Best,

Parker Dodson