Are these normal phase wire voltages?

[Victor Dupont]

Hi there!

As I am trying to troubleshoot a problem on my electric bike, and got stuck, I tried measuring the voltage on the phase wires, to know whether my controller is faulty.

I would like to know whether that looks normal or abnormal to you, since I have no idea what the correct values are, and I don’t know whether my methodology makes sense.

The way I measured is the following.​

In order to have something that might make sense, I disconnected the phase wires from the motor, so that it doesn’t move. I kept the hall sensors and speedometer wires plugged, so that the controller has the information coming from the motor. I disconnected the brake cutoffs and lamp, to avoid them interfering. I also opened the controller, to be able to measure voltage directly on the circuit board, to rule out potential wire issues.

Basically it’s a pretty normal setup, except for the phase wires that are disconnected and the controller that is open.

I turn the pedal when I want to measure, and then I use a basic multimeter to measure voltage on each phase wire’s solder point.

Again, note that the motor doesn’t move. The pedal movement just ensures that the controller gets the signal to kick off the motor.

In order to change the sensors value, I simply turn the back wheel backwards and measure the hall sensors wire solder points to know where it landed.

Let me know if you think this method of measurement is not usable.

If it's of any use, my controller is a Lishui EPAC Drive System, part number LSW06-90B1CF SB.

Here are my results:​

When idle, meaning pedal is not turning:



When turning pedal:

Some things to explain:​

- I made several experiments for each motor position, and noted the result of each measurement, separated by commas. I made sure to turn the wheel, and thus change the motor position, between each experiment.
- when I include an arrow, such as 0->9, it means that for a few seconds it was at 9, and after maybe 4 or 5 seconds it rose to 9
- twice I had results that were the same as when idle, and controller’s light was flashing differently, not regularly. After I turned the wheel it went back to « normal ». That’s what I indicated as « flashing ».

There are a few things I notice:​

1. when idle, meaning the pedal is not turning, the yellow phase wire has 13 V voltage when measured using DC mode on multimeter. I would have expected 0V.
2. my measurements when pedal is turning seem to hint that in a normal setting, each phase wire would be at either 0 or 35V, which is close to the battery’s voltage, and that the wire with the 35V would be a color that is at 3.3V in the hall sensor cables. This is just my guess, tell me if that’s how it’s supposed to be
3. It seems to me that the most abnormal measurements seem to be on the yellow phase wire, when it’s supposed to be at 35V, and is instead at 6 or 8V

But again, I am just guessing, I have no idea what the correct values are.

Do you know whether these values are normal or not?

I previously tested for shorts between the phase wires, ground and battery positive, to detect a blown transistor, but everything seemed normal.

Other measurement : with motor connected

I also tested voltage on phase wires when they are connected to the motor. That way everything is exactly as in real life, except for brake cutoffs and lamp which are disconnected.

I only tested when it's idle, meaning I am not turning the pedal.

Here are the results:



As you can see, there is voltage even though everything's idle.

If my understanding is correct, phase wires are connected through the windings in the motor. Thus I interpret this as controller sending 13V on yellow phase wire, and because others are connected to it through windings in the motor, they are all at 13V.

Do you know whether these values are normal or not?

For those who are interested in the whole history of this bike’s problem, here it is: Temporary power outages and motor noise

Thanks a lot in advance for the help!

 

[Victor Dupont]

The bootstrap cap charges up very quickly so the average current in the resistor is low. I would just try to duplicate the size of the original.

Awesome! If the average is what matters, then yes definitely. I was worried that the peak current, even if very brief, could fry it.

So to recap, if I understand what you're saying:

• I don't need to worry about wattage, only resistance, because peak current will happen for a very short time only
• 5 ohms is definitely fine, 10 ohms probably ok too because the capacitor will continue charging from where it left off at the next cycle. There won't even be hiccups that the motor could suffer from.

Did I get it right?

 

[Victor Dupont]

Actually, after thinking about it, I feel like there is something I am not understanding.

Relevance or irrelevance of pulse width modulation stage for recharge

The bootstrap capacitor's bottom pin is connected directly to ground only when the bottom FET is turned on. The pulse width modulation stage occurs when it's off, and at that time, it's connected to ground through a 10 kOhm resistor.

Thus it seems to me that during the pulse width modulation stage, the value of the bootstrap resistor doesn't matter at all, since it's completely negligible compared to the 10 kOhm one.

I am even surprised that the capacitor would be able to recharge during that time, with such a low current.

Order of magnitude of time when bottom FET is on

Conversely, if the capacitor is charging when the bottom FET is on, then it's a different cycle. I calculated roughly that if the wheel is spinning at an average speed of 36 km/h, with 15 magnet pairs in the motor, that would mean the bottom FET would stay on for 1 or 2 ms, and off for twice that time.

If the capacitor has a 10 uF capacity, with a 5 ohm resistor that would mean a total time to charge of 250 microseconds.

Thus 1 or 2 ms is plenty to charge it, and conversely 5-10 ohms seems the right order of magnitude to avoid hitting the limit.

It also seems to me that using the 10 ohm resistor would still be safe, it would bring the charging time to half a millisecond, which is still half the available time, given that the bike is unlikely to go much faster than 36 km/h.

Am I understanding things correctly? Is it indeed the case that what matters is the time when the bottom FET is on? Thus not the actual pulse width modulation stage?

 

[Victor Dupont]

Hi guys!

I have happy news:



I still have trouble believing it.

But it's working perfectly now. My partner rode two hours with it to go to work and back.

I'll write up a whole retro on this, including what the problem was, how it generated the symptoms, how I could identify it, and how I fixed it.

But before I do, already there are two things to see with you:

1. thanking you
2. formalizing the methodology I used

and before that, the latest events as number 0.

0. The latest events

Since we last exchanged messages, I picked a capacitor on a scrap motherboard, measured it, it was almost exactly the right capacity.

I tested that its voltage rating would be sufficient, by plugging it 22 V, and checking that it wasn't leaking. I did this by adding a 40 kOhm resistor in series, and checking that it didn't generate any voltage drop across it, thus no current.

I also checked that this capacitor plus the 10 ohm resistor in series were doing well when being plugged 22V.

I then soldered them, and did the same measurements I had done before.

Here were the results:



As you can see:

1. there are still some weird voltages, which could be ghost ones
2. but the yellow phase now reaches 35 V one third of the time too
3. and when the bike is idle, everything is at 0, even without any light bulb

So it seems to me that this method of measuring can spot a difference between a working controller and a faulty one.

Anyway, after that obviously I tested the bike, and all the symptoms are gone. It always starts when supposed to, and doesn't make any noise.

I am very happy, and my partner is super happy.

1. Thanking you

I am deeply grateful for the help you guys have given me. I would obviously not have been able to do this without you.

How can I thank you?

This is a real question.

Please let me know whether I can do something that would be beneficial to you.


2. Formalising the methodology

I am still motivated to contribute to formalising this methodology.

Let me know what you think, but it seems to me that even without adding a light bulb, this kind of measurement could enable someone to know that their controller is faulty without opening anything, with a simple voltmeter:

• from what @fechter said about circuits being identical, I take it that the measurements also should be somewhat symmetric. At least we can say that each phase should have 2 positions when it's reaching the battery's voltage
• also it seems to me that there shouldn't be any voltage when bike is idle, even without any load

I would suggest to propose this methodology without light bulb first, and then only mention that the person can add the bulb in a second time if their first measurements are not conclusive. Two reasons for this:

1. this would enable to spot the abnormal voltage when bike is idle
2. it would make the test simpler, and more straightforward. Also many people who live in a city in Europe don't have neither car nor motorcycle, so no 12V light bulb. Especially the people who have an electric bike to commute.

The simpler the methodology can be, the more useful it will be, because if it only takes 5 minutes, and people don't have to go out to buy something, it's a lot different.

Let me know how to move forward on this.

Again, a huge thank you for all the help you given me along the way. I am deeply grateful.

 

[fechter]

Great to hear it works now. It is a pretty rare kind of failure, but luckily easy enough to fix.

The test methodology is valid other than the "ghost" readings you can get. If no light bulbs around, a resistor across the meter input works probably better. Something between 1k and 10k. If the voltage isn't too high, I've just used my fingers as a resistor to prevent ghost readings.

What you are really looking for is that each phase wire will toggle from zero to battery voltage at some point in the cycle. At any given time, one phase will be floating, and can have random voltages that can be ignored.

A much more common failure is a shorted FET, which is easy to check for with just a meter.

 

[Victor Dupont]

Great to hear it works now. It is a pretty rare kind of failure, but luckily easy enough to fix.

The test methodology is valid other than the "ghost" readings you can get. If no light bulbs around, a resistor across the meter input works probably better. Something between 1k and 10k. If the voltage isn't too high, I've just used my fingers as a resistor to prevent ghost readings.

What you are really looking for is that each phase wire will toggle from zero to battery voltage at some point in the cycle. At any given time, one phase will be floating, and can have random voltages that can be ignored.

A much more common failure is a shorted FET, which is easy to check for with just a meter.

OK.

I think we have enough material to make a methodology. I'll start a Google Document, draft it and invite you guys to comment.

Regarding thanking you, how can I do so?

 

[fechter]

Just a successful repair is thanks enough for me. I enjoy troubleshooting, and when it works out, I'm happy about it.

 

[Victor Dupont]

Got it. Let me know if I can be of any help in the future anyway.

@TommyCat, same question for you, you've been a great help!

 

[TommyCat]

A solved troubleshooting issue is thanks enough, been glad to help.
In fact, I think it should be you who is thanked for the determination, testing, and ultimate successful repair! Job well done!

I do look forward to your summary of results which I'm sure will help others in the future.


Best regards,
T.C.

 

[Victor Dupont]

Thanks for your kind words.

Same, let me know if I can be of any help in the future anyway.

I'm working on a recap and I'll also start a draft of the methodology.

Do you have any thoughts on it, that I should add right away?

Cheers!

 

[TommyCat]

Do you have any thoughts on it, that I should add right away?

No offence, but I recommend you TYPING out the values and information, as opposed to photocopies of handwriting.
But I think you've laid out everything nicely in the past, just condense it.

 

[Victor Dupont]

None taken

Good point, I'll make sure it's all readable and searchable. I'll probably use the pictures only for illustration if needed.

Thanks for the suggestion!