isolation of battery to inverter comms

[SeaGal]

I'm trying to get my head around how I've managed to blow a CANBus adapter (and fuse). Does anyone have experience of surge voltages may present on a battery to inverter CANBus link, when a battery is re-connected?

In my case, I have a Solis hybrid (AIO) inverter, connected to my 14.3kWh battery pack. The interface / control system is ESP32 based, connecting to the Solis over CANBus. This control unit is powered by buck converter which is connected to the LFP's.

What happened is that I connected the CANBus cable to the Solis. Then reconnected my batteries via a dual pole fused battery disconnector (so both +ve and -ve at same time). At that point, the CANBus connection failed and, on investigation, the CANBus adapter was destroyed and also a fuse blew on the -ve wire side of the 5V dc supply.

Overall, the setup is working well for over a year, so there doesn't seem to be any inherent problem. Just if I connect the CANBus before the battery, the CANBus adapter seems prone to getting zapped, but the other way round is fine.

The other puzzling thing is that the battery system and hence the ESP/CANBus adapter is fully floating being battery powered - so the battery -ve shouldn't be at different potential to the CANBus connection - or if it was, it would never work.

I'm going to migrate to a galvanically isolated CANBus adapter soon, but in the meantime would love to understand what could cause this issue.

 

[Hedges]

Common mode noise spikes?
Any switch-mode power supplies involved?

We had a Meanwell power supply, AC input, feeding a DUT. USB dongle for I2C between PC and DUT. Another USB or Ethernet to a scope, which was grounded. I2C got blown in 10% of the DUT in our lot. Testing repeated cycles of AC power applied, I captured as high as -8V common-mode spikes on Meanwell DC output.

My solution was an HP linear power supply.

Galvanic isolation could work for you, if that is the problem.

 

[SeaGal]

Any switch-mode power supplies involved?

The ESP32 and CANBus adapter are powered by a step down (60V to 5V) buck.

 

[SeaGal]

Thanks... just researching the causes of common mode spikes, which it seems could be down to different length wires.

I'm interested to understand how a SMPS contributes negatively to common mode spikes compared to a linear PSU, thanks?

 

[Hedges]

That's DC/DC, may not be a problem.
It was phase of connecting AC power to an AC/DC supply that caused our spikes.

If you have a scope, measure common mode of DC to ground (catch with trigger) while switching AC on and off repeatedly.

If one end is galvanically isolated, no current will flow. But if there are two DC paths to earth, or paths with high capacitance, can cause a pulse of current when there is transient voltage.

I'm interested to understand how a SMPS contributes negatively to common mode spikes compared to a linear PSU?

I think it was coupling through from AC to DC.
One supply we had with major EMI issues, in that case a DC/DC converter, the designer added capacitors from input to output. May have attenuated noise, would also couple noise through.

Test on a bench, if you can, to see how they perform.

 

[hwy17]

Basic question here but does Canbus not use a magnetic coupling like an ethernet port does to protect against any DC voltage? Maybe it works entirely differently.

 

[Hedges]

CAN bus appears to be diff pair wires like RS-485, not transformer-isolated like Ethernet.

CAN bus - Wikipedia

en.wikipedia.org

 

[SeaGal]

That's DC/DC, may not be a problem.
It was phase of connecting AC power to an AC/DC supply that caused our spikes.

If you have a scope, measure common mode of DC to ground (catch with trigger) while switching AC on and off repeatedly.

I have, of course, but there's no AC switching. All I was doing was connecting the battery back to the inverter.
Will need to think about how to connect the scope as that will be earthed and the dc battery ports of the inverter are floating and much higher than earth potential, IIRC.

If one end is galvanically isolated, no current will flow. But if there are two DC paths to earth, or paths with high capacitance, can cause a pulse of current when there is transient voltage.

Thanks - that was my thinking too. Not sure why it is not a more general problem though.

 

[SeaGal]

Basic question here but does Canbus not use a magnetic coupling like an ethernet port does to protect against any DC voltage? Maybe it works entirely differently.

I've been using one of these...

 

[AntronX]

Do you have pre-charge resistor between battery and inverter?

 

[Hedges]

You have an inverter - how is it grounded?
Lots of people discovered PV terminals carried common-mode AC, maybe about 60 Vrms. One guy's dog discovered it before we helped him track it down.

Measure AC and DC voltages between everything.

 

[SeaGal]

Do you have pre-charge resistor between battery and inverter?

No precharge.

 

[hwy17]

I've been using one of these...

It looks like there's a different variant of the TJA1050, the TJA1052i that provides galvanic isolation?

To say again, I'm in over my head here so I'm not even exactly sure I'm looking at the right failure point.

How to Implement Power and Signal Isolation for Reliable Operation of CAN Buses

Use isolation to protect CAN buses in challenging automotive and industrial environments, including evaluation boards that can speed the design process.

www.digikey.com

 

[Hedges]

When connecting battery+ to inverter, the massive current of possibly several thousand amps would cause ground voltage offset.
Depending on how other things are connected, could cause current to flow in ground and signals.

For SCC, we've suggested separate wires to battery so it isn't confused by IR drop generated by inverter. But in your case, you want anything coupled by small-signal devices to ride the same (dirty) ground.

 

[SeaGal]

You have an inverter - how is it grounded?

Solis hybrid, grid-tied. Earthing is via ac-grid in connection, as well as earthing point on its heatsink. We have TT earthing arrangement here.

Lots of people discovered PV terminals carried common-mode AC, maybe about 60 Vrms. One guy's dog discovered it before we helped him track it down.

Measure AC and DC voltages between everything.

 

[SeaGal]

When connecting battery+ to inverter, the massive current of possibly several thousand amps would cause ground voltage offset.
Depending on how other things are connected, could cause current to flow in ground and signals.

For SCC, we've suggested separate wires to battery so it isn't confused by IR drop generated by inverter. But in your case, you want anything coupled by small-signal devices to ride the same (dirty) ground.

That makes sense, thanks. As it only happened when transiently connecting the battery sounds like this is the most likely.

 

[Hedges]

It looks like there's a different variant of the TJA1050, the TJA1052i that provides galvanic isolation?

To say again, I'm in over my head here so I'm not even exactly sure I'm looking at the right failure point.

How to Implement Power and Signal Isolation for Reliable Operation of CAN Buses

Use isolation to protect CAN buses in challenging automotive and industrial environments, including evaluation boards that can speed the design process.

www.digikey.com

At this moment I'm watching a Teledyne LeCroy/Eric Bogatin webinar on EMI. More about high frequency impedance control, but ground bounce is one of the ways EMI appears. Ground bounce could be relate to your issue.

Solis hybrid, grid-tied. Earthing is via ac-grid in connection, as well as earthing point on its heatsink. We have TT earthing arrangement here.

Yeah, but I mean multiple devices star grounded (usually what we want), maybe daisy chained off each other, maybe loops.
In your case, connecting AC power to a power supply, or connecting battery to inverter causing inrush, could make ground bounce.

I want star topology "ground" reference for all the analog driver modules of our system, and galvanically isolated signals and power to the modules. But that "ground" reference is the reference plane of circuit boards, distinct from outer shields of cables and boxes. This means nothing should be coax, rather triax or twinax. Any analog (or digital) circuit should have its reference plane bonded to earth ground at exactly one point. (Then to make things more difficult, magnetic fields reach through conductive shields and induce current in any wire loops.)

 

[SeaGal]

How to Implement Power and Signal Isolation for Reliable Operation of CAN Buses

Use isolation to protect CAN buses in challenging automotive and industrial environments, including evaluation boards that can speed the design process.

www.digikey.com

Great article, thanks

 

[SeaGal]

Measure AC and DC voltages between everything.

So that was interesting and not quite what I was expecting. I need to understand this DC-bus concept better.

Battery voltage was 52V. Battery negative to earth -195V dc. Battery positive to earth -143V. AC measurement of each was about 12Vac. This was after dark, so no PV production.

The Solis info screen was showing DC Bus voltage of 391V. It also showed a value that Solis call "ground voltage" of 194V. But from previous discussions on here is expected and AFAIAA is not actually "ground voltage", but the mid-point of the (presumably) floating DC-bus. This has already been discussed on here on a couple of times, including this thread...

Ground voltage

Good afternoon I have recently had a solis 3.6k hybrid inverter installed. This has a pure drive 5kWh battery connected. I have just been checking the running messages and have seen that the ground voltage is 195V. Is this normal or is there a problem with something. Thanks.

diysolarforum.com

It seems this is normal, as @Solar Guppy mentioned in post #29 in that thread.

I guess those can-bus adapters don't like suddenly having their ground voltage changed from being completely isolated to being at -195V wrt ground? Need to think it through more.

 

[Hedges]

Pretty wild having a 48V battery biased to some -150 or -200V with respect to earth.
Is it floating, just charged up? Would a high impedance pull it back to relatively safe voltages?

My system supports either positive or negative ground of battery.
There are 1A PTC fused 48V outputs to put through signaling relays and control power relays, fans, etc.

RS-485 in mine has I think 5V supply, pull up & down resistors and terminators. I'm not sure if its "ground" reference is tied to chassis or not. If so, you'd get current flowing through 1200m cable to remote equipment that lived at an offset. If shielded RJ-45 jacks and cables are used (it uses RJ-45 for both RS-485 and sync cable between inverters), that would connect chassis to chassis.

You checked battery bias vs. ground, did you check logic/comms?

If there is capacitance to a circuit, even though there is galvanic isolation, when common-mode is suddenly applied it forces in current. Which you may be able to measure, or measure voltage rise time and capacitance to determine current.

ICs often spec not to drive inputs more than +/-0.4V relative to rails, but if they can have sufficient series resistance it protects them. Not so practical for fast outputs, but at remote inputs would limit current. Otherwise, when a device is unpowered the data lines can try to bring up its power rails through ESD protection diodes.