Cheap MPPTs + Cheap Engine = EMI

[bluewave]

Reporting this in case of interest to others - took me a while to figure out!

Our solar system uses 2 x EPEver ET6415BND 60 amp MPPT controllers in a 24V system.

I have a 6hp Chinese single-cylinder 4 stroke spinning a 24V alternator as backup, using a home-made multi-stage charge controller.

Problem was, when the alternator was started, the LCD screens on the charge controllers became corrupted for a minute or so, then shut down completely (went blank). The solar charging continued apparently normally (and alternator seemed to work fine). MPPT interfaces stayed dead until power cycled. My home-made alternator charge controller and Victron inverter - both of which contain microcontrollers - were fine. Just the EPEver MPPTs were struggling.

Opening up the MPPT charge controllers, they contain a PCB for charging and another one for the interface / screen - so something was mucking up the second PCB.

I assumed was electrical noise from the alternator (brushes etc) - so tried everything, filter capacitors, ferrite beads...despite not being able to find obvious noise with my scope (fairly low-end, so might miss HF stuff). Nothing worked.

Contacted EPEver - to their credit, they got back to me. They helpfully suggested I turn off the chargers when using the generator....

Then I noticed the problem occurred when the field to the alternator was turned off...and in fact when the belts to the alternator were removed! So it is ignition RF EMI! Even though the MPPTs were 3+ feet away from the engine (and the alternator controller and inverter are the same distance or closer, working fine). Investigated further...and discovered the cheap @#$% Chinese engine used a non-resistor spark plug (which are bad for EMI). Stole a resistor plug from a genuine Honda motor I have - and everything instantly came right!

Moral of the story - My Chinese MPPTs are poorly shielded / designed and my Chinese engine saved $0.50 using a cheapie non-resistor spark plug. Each device was hoping all the other devices in the vicinity were not cutting corners like they were...but get two in the same area, and BOOM! 

[eTobey]

This is indeed an interesting case.

Maybe the did this to, because they wanted to make sure, the engine will run (smoothly)?

[David Hess]

I assumed was electrical noise from the alternator (brushes etc) - so tried everything, filter capacitors, ferrite beads...despite not being able to find obvious noise with my scope (fairly low-end, so might miss HF stuff). Nothing worked.

I am surprised none of the filters you tried solved the problem.

My guess is that the circuit design mistakenly includes a floating high impedance node.

[fourtytwo42]

Gone are the days when equipment was tested properly for EMI before being allowed to be sold, it's all self certification without any enforcement these days and as a consequence the consumer environment is littered with EMI sources that nobody does anything about. I lived for many years across the road from someone with a large garage using Chinese electronic ballasts that were so bad they killed my landline DSL but nobody would do anything about it including the so called regulatory authorities. CE means nothing at all!!

[T3sl4co1l]

I am surprised none of the filters you tried solved the problem.

My guess is that the circuit design mistakenly includes a floating high impedance node.

Eh, high impedance maybe; just as likely, induced current in loops between the boards, or voltage drop between them -- for example, two boards joined by a cable have a dumbbell dipole antenna response, with the RF current flowing through only the low-impedance wires in the cable -- VCC/GND usually -- and thus incurring some voltage drop there, and no voltage drop on the signal lines, thus CMRR is poor and noise corrupts the communication interface.

It could still be both, for example if they use I2C between boards.  Almost certainly a terrible idea, but for a short cable to a small board, it can pass commercial levels.  The catch being, I2C uses a couple kohm bus impedance (passive pull-up), so is relatively more sensitive to ambient E-field than most other interfaces (compare CMOS pin drivers in the <100 ohm range).

Ignition basically means steady-state repetitive EFT.

I2C does have the advantage, that it should ignore transients of some ns (officially, up to 50ns); but EFT could still be strong enough, or the subsequent ringing long enough in duration, to stretch beyond that time frame, and corrupt it.

There is still the problem of how it got there; 3ft is a respectable distance, compared to the likely length of spark plug cable, victim cable, etc.  There's a lot of metal on an engine, and it's likely there are grounding metal cowlings around the ignition circuit, alternator, or anything else wired to it. (Why would ignition even be anywhere near the alternator?)  Perhaps there are more plastic housings, or the metal parts are poorly connected (e.g. screws through painted parts, little or no connectivity), no idea.

We are talking pretty ridiculous source levels (10s kV), but depending on the route, we can't throw away too much of that before losing impact.  For example, the cable might have some low 10s of dB CMRR, at relevant frequencies; a transient of 10s of V, maybe 100s even, applied directly to the board -- "direct" in the sense of relative proximity, say we put the two boards in a TEM cell -- might be required to corrupt it.

That leaves the conducted route as a strong contender, but conducted is also relatively easy to filter.  The catch is, if neither unit has a meaningful ground / reference plane with which to filter against, it simply won't do anything; it can't be just shoved inline, it must be wired and placed properly.

Tim

[bluewave]

"I am surprised none of the filters you tried solved the problem.

My guess is that the circuit design mistakenly includes a floating high impedance node."

I'm guessing the EMI was affecting somewhere important deep in the MPPT interface PCB and arriving there via radio waves - so my various filtering efforts at the MPPT DC input terminals (and at the alternator) might not have been in the right place for that.

But I don't really know!

[bluewave]

"Maybe the did this to, because they wanted to make sure, the engine will run (smoothly)?"

Possibly, but seems to be running perfectly with a resistor spark plug.

Hopefully the non-resistor plug isn't compensating for some other corner cutting elsewhere in the ignition! It's a Honda knockoff, and they use resistor plugs.

[bluewave]

I am surprised none of the filters you tried solved the problem.

My guess is that the circuit design mistakenly includes a floating high impedance node.

Eh, high impedance maybe; just as likely, induced current in loops between the boards, or voltage drop between them -- for example, two boards joined by a cable have a dumbbell dipole antenna response, with the RF current flowing through only the low-impedance wires in the cable -- VCC/GND usually -- and thus incurring some voltage drop there, and no voltage drop on the signal lines, thus CMRR is poor and noise corrupts the communication interface.

It could still be both, for example if they use I2C between boards.  Almost certainly a terrible idea, but for a short cable to a small board, it can pass commercial levels.  The catch being, I2C uses a couple kohm bus impedance (passive pull-up), so is relatively more sensitive to ambient E-field than most other interfaces (compare CMOS pin drivers in the <100 ohm range).

Ignition basically means steady-state repetitive EFT.

I2C does have the advantage, that it should ignore transients of some ns (officially, up to 50ns); but EFT could still be strong enough, or the subsequent ringing long enough in duration, to stretch beyond that time frame, and corrupt it.

There is still the problem of how it got there; 3ft is a respectable distance, compared to the likely length of spark plug cable, victim cable, etc.  There's a lot of metal on an engine, and it's likely there are grounding metal cowlings around the ignition circuit, alternator, or anything else wired to it. (Why would ignition even be anywhere near the alternator?)  Perhaps there are more plastic housings, or the metal parts are poorly connected (e.g. screws through painted parts, little or no connectivity), no idea.

We are talking pretty ridiculous source levels (10s kV), but depending on the route, we can't throw away too much of that before losing impact.  For example, the cable might have some low 10s of dB CMRR, at relevant frequencies; a transient of 10s of V, maybe 100s even, applied directly to the board -- "direct" in the sense of relative proximity, say we put the two boards in a TEM cell -- might be required to corrupt it.

That leaves the conducted route as a strong contender, but conducted is also relatively easy to filter.  The catch is, if neither unit has a meaningful ground / reference plane with which to filter against, it simply won't do anything; it can't be just shoved inline, it must be wired and placed properly.

Tim

Getting above my pay grade. I can say the MPPTs are in aluminium housings. The housings and system negative are both tied to a nearby earth peg. I also earthed the engine as a test - didn't help.

[beanflying]

I posted this link in a recent Model Solar Car thread but as you are in this corner of the world you might like to chase them up or look at what they are doing https://www.aerl.com.au/solar-car-controller/

Some reading on the manuals might help if nothing else.

[eTobey]

"Maybe the did this to, because they wanted to make sure, the engine will run (smoothly)?"

Possibly, but seems to be running perfectly with a resistor spark plug.

Have you tried running it at cold temperatures?

[bluewave]

"Maybe the did this to, because they wanted to make sure, the engine will run (smoothly)?"

Possibly, but seems to be running perfectly with a resistor spark plug.

Have you tried running it at cold temperatures?

Good question - no. The location is maritime and temperature range narrow - never see ambient outside of 0 - 30C here.

[BrokenYugo]

AFAIK non resistor plugs are for race cars and airplanes (shielded) where the little bit of extra ignition energy and simplicity is appreciated. I'd guess in your case the engine factory gets them cheaper in bulk. I've always seen and used resistor plugs on all manner of small air cooled engines and they all ran fine, the ignition needs aren't that great and the provided mag or whatever ignition system is generally more than adequate. Fairly low dynamic compression, not a very thick mix to have to fire through, and you're only jumping a .020-.030" gap anyway, not a modern auto engine plug running about double that gap.