Controller failing after warmup-clues to what may be faulty?

[Magneto]

Hi Guys,

Was hoping to get a couple of clues as to what may be wrong with my controller on my e-bike, it's a 'Greenewheels' very similar to the Wisper bikes. Controller model no is KT36ZWP.

After about 10 mins of riding, under a decent load going up a hill, the power starts backing off then stuttering at a particular frequency, (maybe 5hz?) this is not frequency is not speed related oddly enough. This only lasts for a very short time, it gets worse then shuts down all power to the motor (a brushless geared bafang). After a short wait it will fire up again without re-boot, but has little in the way of power output until the same fail scenario happens all over again. Re-boot will re-start sooner than waiting for it to fire again.

Have checked the halls, all ok even swapped them out to make sure one wasn't failing while warm.

Sounds like overtemp shutdown, although I would expect this to be a smooth tailing off of power rather than being jerky like this.

Have checked FETS, all 6 all look OK, and integrity to heatsink was OK, checked. tightened them and test rode again, no sucess. Re-soldered some of the joints, some pretty bad soldering on this board, so I guess I might try some more of that, but could be a partially dead FET??
Maybe a FET that has lost some of it's shunt capacity through melted solder inside the TO220? Maybe one of those mysterious surface mount controller chips is failing when a bit warm?

Any clues would be very much appreciated, as finding a replacement that has the same connections, features and LVC for a LiManganese 36V may be a challenge, was thinking ecrazyman's 36V 350W unit.

 

[Drunkskunk]

9 out of 10 times, a temp related failure of that type is going to be a capacitor.
But bad solder joints and trouble with anything with a gate (chips, fets, transistors, ect) can fail when warm.

Grab a BBQ thermometer and mark the temp when it starts. It may be nessessary to warm the board on a hot plate up to that temp and then test it componant by componant.

 

[texaspyro]

Sounds like it is hitting LVC on the battery pack. Under load, the voltage drops and controller shuts down. Now, without a load, the voltage rises above LVC, controller starts up again. Lather, rinse, repeat. It starts out with short repeat cycles, that get longer as it takes more and more time for the battery pack to recover.

 

[Magneto]

Many Thanks for the info Tex and Drunkskunk,

Forgot to mention main power to 36V 10w LED headlight and battery status lights remains constant, so BMS cutout not activating.

I did change out a 47uF 50V cap that looked a bit stained on the top, but that did not help, others look ok, but last night, powered her up with PCB removed from the case, so all 6 fets with no heatsink. These did not really warm up much, with a bit of throttle and applied brake load, but the LM317T adjustable voltage Reg was very hot. On my board, this is on it's own, with no heatsink, is that normal for these boards?

Left the power on for a while with no motor activity for 10 minutes, and the LM317T was still very hot, far too hot to hold for more than a second or two without burning my fingers. Even though this thing is probably dropping output to 12v from 36v, I wouldn't have thought the power consumption of the 12v rail (and 5v if there is one? ) would be enough to warm this reg up like this. It also has some minor staining on it where a brown epoxy like substance seems to have bled out. These things have a overtemp cutout AFAIK, so maybe that is the culprit?

 

[texaspyro]

These things have a overtemp cutout AFAIK, so maybe that is the culprit?

Yes, they will shut down on overtemp. Try putting a meter on the output and see what it does when it gets hot. Dropping 24V does not take much current at all for them to overheat, particularly without a heat sink... 40 mA would make it draw 1 watt, and without a heat sink is enough to put it into thermal shutdown.

 

[Magneto]

Thanks for the info Tex,

Yeah maybe it's been running hot for the 4000km it has operated, due to the zero heasink approach by the cowboy engineer who designed this thing, and has finally started to operate less efficiently due to heat damage and is overheating and shutting the whole controller down. Bought a couple more LM317T today to try, (Farnell in Australia, now Element 14 has free same day shipping in Sydney Australia, great stuff. )

 

[Magneto]

Well changed out the LM317T, same story, bike starts to shudder after a couple of decent hill loads..

As Drunkskunk mentioned, maybe there is a dodgy cap in there (9 electros to choose from) that may be way out of spec, causing abnormal 12v rail current and overheating my LM317T, or simply a failure of cap under heated conditions.

Might see if I can wire up the LM317 with probes to test while riding up hills, but wont really help with finding the problem, circuit diagrams impossible to find with this chinese stuff. Might start randomly changing out caps..

have got a 36V350W controller coming from ecrazyman in any case, but worried this will be too long to fit in to the neat under battery case on this bike, the existing controller box is 92mm long, ecrazyman on his specs say 125mm but wondering does that include the mounting tabs, which I could slice off I guess, can anyone confirm the actual enclosure size of these ones from ecrazyman?

I love my bike, but seriously, the reliability of these chinese controllers/motors/BMS's is absolutely crap. Over 4000km I have changed hall sensors, fried a BMS and now the controller is shagged. Soon a cell will probably go bad, as has happened with another guy who I know had this happen very early in the battery's life. If they were more reliable, these bikes would be everywhere by now. The Greenewheels seller on ebay has not shown anything for sale for many months, and did not respond to my e-mail asking if he had a controller to sell me, I suspect he has been overwhelmed by customers experiencing the truly awful quality of these components and the substantial design flaws in them. Maybe the up and coming gear from Shimano/bosch/panasonic may change this situation in coming years.

 

[texaspyro]

I don't think it is a cap problem. In these applications, cap failures tend to be less than subtle. Usually easily diagnosable by the split case and icky goo spewage.

 

[Lessss]

This is where having a second controller to swap out really helps narrow it down to controller or motor.

 

[amberwolf]

One problem with failed caps is that they don't always actively vent goo, or even visibly swell up. Usually they do, but there's no guarantee that you can see the failure.

My first two thoughts given the LM317 temperature are either:

--the input caps (before it in the power supply chain) are no longer filtering the power spikes caused by motor operation, causing the 317 to have surges of voltage higher than normal across it with every PWM pulse and commutation cycle.

--a cap or other component downstream from the 317 has failed or fails after warmup, and is pulling so much current that the 317 can't deal with it.

Another thougth is that the 317 in that design may simply be allowed to always get that hot, which is a poor design choice but not an unheard of one, and that something totally unrelated is actually causing your shutdown.


A test you can do (without an oscilloscope) to see if either of the above might be true:

Turn the bike on and let it sit there for the same amount of time that it takes to overheat in use, and see if the 317 is as hot as before. If it is, then either a downstream component is overloading it or it's a red herring.

To see if it is the spikes causing the overheating, run the bike with the motor wheel off the ground, so it has little load. See if the 317 gets as hot as before in the same amount of time. If not, then the spikes are a likely culprit, and replacing any capacitors on the input side of the system would be a good idea.

 

[Magneto]

Many Thanks for all of the clues amberwolf, lesss and Tex, very much appreciated.

Will look at some of these caps you refer to on the weekend, don't think the power spike filter failure theory will be the issue here however, it definitely does only fail under heavy load, climbing a decent hill for eg, hills that would slow me down to say 10 to 13km/hr with no pedalling, after a warm up and then say a couple of 30 to 60 second periods of this magnitude of load, it starts to stutter, and as mentioned before, regardless of engine RPM, the stuttering happens at a fairly stable frequency of something like 5Hz, then rapidly loses more output power and shuts down completely, battery voltage does not suffer at all so is something in controller shutting down. After a short wait it recovers. I could ride on the flat with no issues it's just when it get a good load.

The 5hz must be the frequency of some cycle like, overtempshutdown, recovery, overtempshutdown, recovery etc until there is no recovery. Or could be discrete component failure, recovery, discrete component failure, recovery.. The stable frequency regardless of wheel rpm may hold a clue I think.

I suppose it could be a motor winding shorting at a particular temp, causing some imbalance in the controller resulting in more cunsumption from the 12V rail, would be pretty rare but is that possible? The 6 fets seem to be running normal temp and all even in temp when loaded.

Could well be the case that it is not the 317 as you say amberwolf, and maybe it is designed to run that hot, I have a high temp probe that I can check idle and loaded operating temperature, from memory these To220's are rated at around 120°C max.

 

[Magneto]

Haven't had the time to check temps or 12v rail in fail mode, but have noticed that when my fail state occurs, the three light battery meter on the throttle is showing the main battery voltage being dragged down to low at the same approx 5hz frequency as the drive failure.

So looks like whatever is failing is capable of sinking a lot of current into heat, pretty sure nothing to do with BMS, so could that be a FET failing after it is warm in that fashion, ie. shorting out somehow when at higher operating temperatures?

 

[Magneto]

Finally received my 6 fet brushless hall effect controller from Keywin, spent 4 hours installing it (trying to fit the extra 10mm of aluminium box in the controller housing took a whole hour and had to cut off the steel mount tabs)

Although the colours matched for phase and hall, ran with a lot of noise, and ended up finding out that two phase wires needed swapping as did two hall signal wires.. Woke up this morning, very keen to see how it went on my commute to work on it for the first time in over a month...

And the problem is still there !

Very frustrating after spending hours changing out parts in the old controller, waiting for the new controller..

In any case the problem I guess is likely a bad cell, gone high internal resistance possibly, or a BMS fault of some type, or even motor winding failure at operating temp. Only happens after a bit of current has been flowing for 10 mins or so.

Don't think a motor winding, as failure mode is still that 5hz pulsing irrespective of speed, which is accompanied by the three LED voltmeter on throttle dipping to low (red) at the same 5hz frequency. If the frequency were related to speed, I would think more likely motor winding?

I guess my next step is to measure individual cell voltage, should this show up a bad cell in this case or would I need to measure under load?

When I have run the battery down to cutout level, the BMS has cut all power from the battery, this particular cutting out is not doing that, it's cutting out and starting up again as I mentioned at 5hz. Could this be another 'brown out' type load reduction by the BMS due to one single cell coming up as low voltage?

The battery is in a specialised made to fit case that slots into the rail, almost vertical, under the seat. 36V 14Ah Lithium ion manganese I believe, 10 cells, balancing BMS. I would just go and get a new Lifepo4 from Ping, but cant get one to match the existing case.

Anyone have any other ideas as to how I can diagnose whether battery cell, BMS or motor winding?

The only other thing it could be is one hall effect chip that I didn't change in the motor (when this problem started I changed the two halls that were original, left one that I had already swapped out for a honeywell due to previous failure) have heard halls can have a failure mode when hot and recover when cold. But motor behavior and noises do not sound like hall failure, not really jerky or noisy, just power coming on and off.

Any suggestions would be much appreciated!

 

[Hillhater]

Before you start pulling anything else apart, the easiest thing to do is to try another battery pack, .
borrow a pack or even 3 SLA's for a quick test.

 

[Magneto]

Good point hillhater, I might be able to get hold of 36v worth of SLA if I try hard enough.

Just had a thought, could be the KSD 9700 5A 65° overtemp sensor strapped to the battery pack. I have a spare one of these so will swap it out, plus can measure temp in battery case with a probe to see if it is getting hot in there. If the temp sensor were to be ok but battery temp going over 65°, is that likely to be a faulty high internal resistance cell generating the heat?

 

[Magneto]

Another chapter to the saga, wasn't the temp sensor, should have remembered these have a cycle frequency of many seconds when triggered, I tested the one inside the battery case to 65° before going any further with that theory. I do have a spare 10Ah battery from the same greenewheels bike, kindly given to me by another owner nearby here in Sydney, when the BMS in mine failed some 8 months ago. (This battery has a bad cell) So I put my old BMS back in (which I had repaired) and went for a test ride.

This time there was no 5Hz failure mode, it just went completely dead when bike was warm and all power from the battery was cut unlike the other BMS, which seems to current limit in that pulsed fashion, and when pushed really hard into that mode, the controller LVC operated, and still had power to meter and headlight. Power was restored after cycling main power switch to controller.

Tested the other way around, put the BMS i have been using into the hand me down battery with the dodgy cell, and sure enough it failed in the 5hz pulsed mode. So even though the BMS's look identical, their LVC behavior is totally different. This is why originally I thought it was not BMS LVC as I thought I knew how this behaved.

So now looks like I will need to test each cell voltage under load (unloaded they all measured between 4.08 and 4.13V) to see which cell has crapped itself.

So much for the $700 this 14Ah battery cost, Joe from Greenewheels claimed 500 full charges and 1000 half charges. I've done 4000km and about 200 charges, some could be described as half charges, and it's already dead. That's about $3.50 per charge which is probably what i would have paid for petrol for the same distance. Anyways, I will try and id the high resistance cells, cut open the heatshrink, extract a good one from the 14ah unit and transplant it into the 10Ah unit, then try and wrap the larger 14Ah cell up with the other 9 10Ah units, probably using glad wrap and gaffer tape, and see if I can flog a little more value out of these very poor Lithium manganese things, then get a LifePo4 and make a custom mount. The LifePo4 will probably be more reliable I hope.

At least now I know I've got two good controllers, two good BMS's, good quality potted halls in my bafang, just need to sort my battery now.