Kelly Controller has no torque, jerks motor

bigchief

100 mW
Joined
Jun 20, 2009
Messages
45
I couldn't figure out how to set the 48v LVC on the controller included with my cheap ebike kit to 36v, so I decided to get a used Kelly KBS48101L-L Controller instead, which I snagged from an eBay listing: https://www.ebay.com/itm/183888487486

I successfully connected this Kelly controller to my PC to program it, but I'm having issues. When I turn the throttle, the motor turns forward, but it does it in an extremely jerky fashion (several loud clicks/clanks per second, almost as if it isn't reading motor hall sensor position correctly), and with almost no torque. If I lift the wheel off the ground while pressing the throttle it will loudly spin up to speed. If I leave it on the ground it has so little torque that I can hold it in place with my hands. I tried changing the "Hall Sensor Type" setting in the Kelly app from 120-degree to 60-degree but that made it even weaker so I changed it back.

1) Why does this Kelly controller have almost no torque and why does it make loud clanky sounds as it spins up? The previous controller works smoothly, it just has the wrong LVC so it cuts off off after 10 seconds.

2) The Kelly programming app gives me an option for both "3-wire pot" throttle and "Hall Effect Sensor". How can I confirm which one I have? Oddly enough, the throttle seems to behave the exact same for either setting.

Fany at Kelly Controllers was graciously helping me figure out the pinouts to this OEM version, but she (understandably) said she can't give further help since it's several years old and not being used in its original application.
 
The red LED does not throw an error code. I'll try soldering the phase wires instead of relying on the JST connectors when I get home.
 
bigchief said:
2) The Kelly programming app gives me an option for both "3-wire pot" throttle and "Hall Effect Sensor". How can I confirm which one I have? Oddly enough, the throttle seems to behave the exact same for either setting.

When throttle actuated closed to WOT...
A potentiometer throttle will output a 0 to 5vdc signal.
A hall sensor throttle will output a ~1 to ~4 vdc signal.

Checking the wiring with the throttles disconnected from the controller...
The potentiometer throttle will vary resistance between wiper (middle lead) and input-output leads, a hall sensor type will not.
 
Gotcha, I dug up a second throttle in my parts bin to compare against (same as this eBay listing https://www.ebay.com/itm/202731390785). I guess they are both Hall Sensors, because if I plug a multimeter to the black and middle pins while the throttles are hooked up to the Kelly controller and powered on, I get 0.82 volts min throttle and 4.18 max throttle, similar deal with the second throttle (0.87 / 4.12).

Still, they both cause the motor to barely budge at all when I max it out. This is with the Kelly Controller programmed to 1v lowest throttle, 4v max throttle.

I also programmed the Kelly controller to expect "3-wire pot" and shorted 2 of the throttle wires on the controller to simulate WOT on a pot throttle, just to rule out these two Hall throttles. It performs the same, where it has nearly 0 torque at a standstill, and a small amount of torque once the bike gets a rolling start. I guess the only other thing to try is to add solder to the crimps of the phase wires going to the motor, and increased the amps on the battery fuse (it seems conservative vs. the wire gauge and battery discharge rating) and see if any of that helps.

One other question, if I am pulling say, 40 amps from the battery, what is the corresponding amps that would be sent along 1 of the 3 phase wires at any given time? I asssume if it draws 40 amps from battery it won't send a complete 40 amps down a single phase wire normally?
 
bigchief said:
One other question, if I am pulling say, 40 amps from the battery, what is the corresponding amps that would be sent along 1 of the 3 phase wires at any given time? I asssume if it draws 40 amps from battery it won't send a complete 40 amps down a single phase wire normally?

I think the general approximation is phase amps is double the battery amps.
 
With regular basic nonprogrammable controllers, when I see this behavior it is a phase/hall combination problem, or a problem with the hall sensors themselves, or a signal / connection problem on either halls or phases.

Even if the Kelly has an autodetect routine, you may still want to try manually experimenting with the phase/hall wiring combination until you get different behavior.


Otherwise...if this is a geared hubmotor, is it rotating it in the correct direction?
 
I went and soldered the motor Hall sensors instead of relying on the JST connection. Then I temporarily increased the fuse rating on my battery so it doesn't blow so quickly, and tried riding around. The motor still feels "gritty" with little torque at low speed; it makes "click" noises many times a second between 1-5mph which increase in frequency and volume until I get past 5mph where it starts to sound like a normal ebike. Oddly loud for a direct drive hub, and it wasn't doing this with the stock controller that was included. If I push off or pedal a bit it gets up to speed faster. Above 10mph it seems to accelerate quietly enough and with some torque but it tops out around 15mph. I know it is a fat-tire bike but that still seems low for 36v at 40a (with 100a peak), and I have the tires pumped up pretty high (30 psi is high for fat tires)

It is about 50 Fahrenheit outside. I rode for 5 minutes. The whole time I was checking the Kelly controller and it remained only warm to the touch. I got home and smelled something possibly burning, so I checked the battery; everything fine there. Then I touched the motor... ouch! It was burning hot even though it was cold outside and I hadn't been riding for 5-10 minutes. The motor is rated for 1000w so going a little above that in the cold shouldn't cause that much heat. Maybe because there's no torque, it is somehow dumping the energy into eddy losses instead or something?
 
Yeah, it's gotta be something with the phase/hall. When I sat down for a while now and thought about it, I recalled that with the original controller (much weaker than the Kelly controller, on paper) I was hitting the throttle while not on the bike, and the torque nearly ripped the bike out of of my hands, and made it pop a wheelie. With the Kelly controller I can max the throttle while I'm not on the bike and still hold the bike in place easily.

The Kelly controller is not throwing any LED codes at the moment, just the single red blink when I first power it up. So it seems to think everything is fine, even when I ride it.

I'll try contacting the hub motor manufacturer to ask them if they know the specs (i.e. 60 vs 120-degree, etc) and triple-check the wiring.
 
it's much simpler than that.

Simply swap the phases around (since you have the halls soldered), just swap any two of them and see if it gets better.

If it doesn't, swap the other two.

If that doesn't, swap the last color pair.

One of those should improve things. If it runs backwards, either use the reverse feature of the controller, or start swapping hall signals until it runs forwards.


During all of this testing, battery current should, wheel unloaded off ground, be an amp or two at most, full throttle. If it is higher than that, it's probably not the right combination.


Almost certanly the halls are spaced at 120 degrees, that's nearly universal for DD hubmotors I've seen.


It's also possible the controller is defective, or has a wiring/etc fault, but it's much more likely to just be the wrong phase/hall combo.
 
I checked again and verified that I have continuity through all the wiring from the controller/wheel/battery, and that all phase/hall wiring is correct. I also did as some Youtube video recommended and used the multimeter to measure 20k ohms from B+ on the controller to the different phase wires (apparently they should all return some value, it reads around 3 for each, which fluctuates).

It definitely isn't 60-degree phase in the Kelly app since that makes the wheel jerk randomly/backwards. But 120 still has no torque. The motor is rated for 45 n.m. This was the original listing: https://www.ebay.com/itm/143339679889
 
amberwolf said:
it's much simpler than that.

Simply swap the phases around (since you have the halls soldered), just swap any two of them and see if it gets better.

If it doesn't, swap the other two.

If that doesn't, swap the last color pair.

One of those should improve things. If it runs backwards, either use the reverse feature of the controller, or start swapping hall signals until it runs forwards.


During all of this testing, battery current should, wheel unloaded off ground, be an amp or two at most, full throttle. If it is higher than that, it's probably not the right combination.


Almost certanly the halls are spaced at 120 degrees, that's nearly universal for DD hubmotors I've seen.


It's also possible the controller is defective, or has a wiring/etc fault, but it's much more likely to just be the wrong phase/hall combo.

You were right, and I was soooo close, but I think I broke something. I'm staring at the weak torque wiring, which was green-green blue-blue yellow-yellow. I'm not colorblind either. But I swapped a bunch of the phase wires around (apparently blue-yellow green-blue yellow-green) and suddenly I had the full, complete torque... only in reverse. So I kept switching them around again and again trying to figure out a combination that could give me forward torque, when suddenly I heard a pop and smelled that familiar blue smoke smell. Now with the phases back the way they were, it doesn't even move forward with weak torque like it used to. It now seems to "limp" like there's only 1 or 2 phases firing. Maybe I blew a mosfet or something else in the controller by drawing too many amps (since the motor is fighting with itself, essentially). Wiring for the "full torque reverse" produces a "limp" too, but a much higher torque limp. I'll take another crack at things tomorrow.

(to clarify, are you saying that once I figured out the "full torque in reverse" combination, then from there it can't be made to go forward from changing the phase wires further and I should have instead started messing with the hall wires?)
 
Your experience clearly proofes that it was the wrong phase/hall combination as amberwolf told you.
Now that you killed the controller you will not make the motor run again properly w/o repairing the controller first. Most probably two phase wires touched, or you touched one of the battery wires with it or something like that.
You will need to open the controller and replace the broken FETs which is not so easy with a Kelly if you haven't done it before and don't have the correct equipment. A hot air station is highly recommended, as a thick powerful soldering iron will fry the PCB quite easily. Best plan is to cut the broken FETs off and desolder the pins, then use a solder pump to suck off the solder from the PCB holes.
 
bigchief said:
You were right, and I was soooo close, but I think I broke something.

I should have instead started messing with the hall wires?

Oh well, probably should have started with the hall combos. :shock: At least you got it for a cheap price. :|
 
izeman said:
Your experience clearly proofes that it was the wrong phase/hall combination as amberwolf told you.
Now that you killed the controller you will not make the motor run again properly w/o repairing the controller first. Most probably two phase wires touched, or you touched one of the battery wires with it or something like that.
You will need to open the controller and replace the broken FETs which is not so easy with a Kelly if you haven't done it before and don't have the correct equipment. A hot air station is highly recommended, as a thick powerful soldering iron will fry the PCB quite easily. Best plan is to cut the broken FETs off and desolder the pins, then use a solder pump to suck off the solder from the PCB holes.

I opened up the controller and the potting material and thermal paste were a mess but the MOSFETs seem easy enough to desolder and replace now that they are exposed.

They say "RJK0882" and then "ON57", I'm not sure if that's letter O or Zero. I don't see a link to buy these MOSFET stateside, but would anyone know if there is a higher quality drop-in replacement I could use instead which would be more readily available, ie Mouser, Amazon etc?

edit: I read the model wrong, it is "RJK0822", datasheet is here: http://datasheetcafe.databank.netdna-cdn.com/wp-content/uploads/2015/09/RJK0822SPN.pdf

I'm asking the company who makes the mosfet, Renesas, if there is newer / better replacements for it.
 
I got home and tested all the Kelly MOSFETS by touching their leads with multimeter continuity tester, using this guide:

https://www.youtube.com/watch?v=RkWy1EirEu8

All four of the circled fets are shorted, across any/all of their 3 leads. The rest seem to be OK. Stupid question, but is this method of testing still valid while the MOSFETS are soldered into the board? Or do I need to remove them first for the short-circuit test to be valid?

P_20191004_020210.jpg
 
No the method is fine. It will detect a DEAD FET. It's no proof that a good one is still good, but if one shorts it's dead for sure.
A correct test would be to test forward voltage in diode testing mode of the desoldered FET. But they will be good for 99.9%.
If you can't find the correct time you may find an alternative, but i don't know how save it is to operate different types in one controller. You still can replace ALL of them if you can't find the correct ones.
Any PLEASE buy from a trusted supplier and not order in China. I ordered A LOT from China, but NEVER when it comes to FETs. Go to mouser or similiar supplier and bite the (price) bullet, you won't regret it.
 
I checked the FET you posted. IMHO one of the best replacements is the well trusted IRFB3077. An excellent FET with almost the same specs (Vmax=75V), but WAY lower RDSon of 3mO compared to 8mO. That means is dissipates about A TENTH of the heat the RJK type does. So your controller will stay cooler and take more stress.
 
izeman said:
I checked the FET you posted. IMHO one of the best replacements is the well trusted IRFB3077. An excellent FET with almost the same specs (Vmax=75V), but WAY lower RDSon of 3mO compared to 8mO. That means is dissipates about A TENTH of the heat the RJK type does. So your controller will stay cooler and take more stress.

Those do look awesome, it looks like they can handle almost 3 times the continuous amps... my battery wires would catch fire long before the mosfets would blow this time around :)

Where would be the best place to buy them from with fast shipping? Arrow seems to give 1-day shipping on $50+ orders and they cost $3.10 each there, so I'd have to buy 18 (only need 12, 14 tops).

edit: I wound up buying them and some extra parts to hit $50. Trump Tariffs and sales tax brought it to $60, yikes.
 
izeman said:
A hot air station is highly recommended, as a thick powerful soldering iron will fry the PCB quite easily. Best plan is to cut the broken FETs off and desolder the pins, then use a solder pump to suck off the solder from the PCB holes.

I cut off the broken FETs, even in this state the pins are very difficult to desolder and remove the excess solder (a lot of the PCB is coating in TIM paste).

Is there any reason I can't just solder the new FET leads directly to the remaining leads from the old FETs? The old FET leads widen at the point I cut them from so there's a large flat surface. The new fets could never get hot enough to melt the butt connections if I solder them, correct? I'm sure it isn't best practice but I don't see any downside.
 
Bumping since I was scrolled off the first page and could still use some help. Is there any reason I can't butt-solder my new mosfets directly to the leads I clipped? Those old leads are in there really good and I don't want to risk burning up the PCB trying to get them out, so can I get away with soldering directly on to them?

mosfet.jpg
 
Back
Top