GoldenMotor.com Forum
General Category => Magic Pie & Smart Pie Discussions => Topic started by: Andyl33 on June 12, 2018, 05:18:18 PM
-
I converted my mountain bike with an MP5 and ten 6S 5Ah lipos at the end of 2015 and is my main method of getting around the city and riding trails in the mountains - was overjoyed with it.
I upgraded the battery pack using harvested laptop 18650s with a configuration of 14S16P, and that also worked well, although regen wouldn't work until the charge level dropped a bit.
First controller failure:
Last November, I rode about 20km to return some dead lead acid UPS batteries, and on the way back about 2km from home, I was waiting at some lights. I went to accelerate and nothing happened. Checked all the cables and connections, as it has been known for the throttle connector to undo.
No visible problems, no bluetooth, no life at all... so I took the controller apart as returning it is really not an option.
I checked everything - the main capacitor was a bit bloated, so suspected that - hunted high and low for a similar spec and found nothing. 80V 1000uF in that size is really non-standard. Voltage ratings normally go in steps from 63V to 100V, and any 100V capacitors would be too big to fit in the small space.
So, I thought I'd try a 63V cap, and bought a couple from my local shop with the intention of using a lower voltage battery.
On further inspection, I noticed a really tiny chip on the backside of the top circuit board with a possible sign of failure on it - using a phone camera I managed to identify the marking on it:
SM5B
After some googling it appears that this is an 80V buck converter which seems logical in spite of its tiny size, but the id ambiguously refers to LM34923MM or LM34923MMX.
These chips have very different specifications and are certainly not interchangeable, so I would like to know which one the MP5 uses, and I suspect many other users would like to know!
That was then... I gave up and ordered another controller that appeared on ebay, it arrived a couple of weeks later in February, installed it and it worked... Happy again!
Until last week... :-(
I was coming down a steep hill, using regen and it was working for a while and then it stopped braking. I continued using manual brakes but the throttle didn't work. Nothing. No bluetooth light either. Checked connections etc...
So, had to ride back home 15km with a 40kg bike which wasn't very enjoyable but I survived :-)
So faced with another dead controller, I took it apart, spent an hour delicately cleaning out the silicone potting compound again, and cannot see any possible cause of failure.
What component would likely fail if the motor was generating a higher than expected voltage? The 15810 mosfets look fine, and are apparently rated to 100V.
A circuit diagram would help!
I want to identify and fix the problem, and that SM5B chip seems the likely culprit.
Does anyone know which of the two types it is?
I could buy both kinds and try them and it would either work or not, but surely someone at GM could be able to provide some basic info?
Is it also safe to power up the controller without the motor connected? This would help to check voltages and waveforms.
I think I'll explore using an external controller with it, one that can do 80V or more - I think someone on Endless Sphere did this successfully.
I really can't afford to buy and wait for a new controller every month, but neither can I live without my bike.
(Also on both controllers, the battery indicator light on the throttle failed after a couple of weeks, and I suspect the controllers gave up supplying voltage to it - GM should look at this)
Have to say that I'm pretty disappointed now but hope that the designers of the controller can improve them so that they perform reliably.
-
Hi Andy and(https://i.imgur.com/evDSMvT.png)to the forum.
Unfortunately, I have no idea which of the two types of chip is used. :(
Is it also safe to power up the controller without the motor connected? This would help to check voltages and waveforms.
Over recent years, I have done this countless times using a 0-60V benchtop power supply with different GM controllers and have never experienced any problems.
I also discovered that a mere 12V supply is sufficient to provide adequate power for programming purposes. ;)
(Also on both controllers, the battery indicator light on the throttle failed after a couple of weeks, and I suspect the controllers gave up supplying voltage to it - GM should look at this)
This sounds a bit strange, as the LED battery gauge in the throttle is not dependant upon the controller.
The battery indicator board inside the throttle unit is fed directly with the full battery voltage "en route" to the controller (there's a direct connection to the main battery positive cable inside the molded junction block where the five cables meet near to the controller) therefore it is not being supplied by the controller.
(https://i.imgur.com/deq9ddR.png)
Therefore the battery LEDs should still light up even with a dead controller (unless the failed controller has caused a direct short circuit between battery + and Battery -) resulting in the battery's BMS shutting off the power.
The basic battery gauges on the throttles were originally intended for use with lead acid batteries, therefore the reduced voltage drop during discharge of a Lithium pack, coupled with a higher than 48V nominal voltage (51.8V for a 14S pack or 51.2V for a 16S LiFePO4 pack) makes these gauges virtually useless as far as indicating the battery's actual capacity is concerned:
(https://goldenmotor.com/SMF/index.php?action=dlattach;topic=5324.0;attach=8001;image)
I have just tested a 48V twist throttle and here are my results:
All three LEDs remain brightly lit from 60V down to 48.3V:- The Green LED turned Off abruptly at 48.2V
- The Amber LED turned Off abruptly at 42.7V
- The Red LED faded gradually from ~40V downwards until it was completely Off at 30V
If you put these figures into the above chart, you will see that a 16S LiFePO4 (or 14S lithium battery) would need to be at least 95% discharged before the Green LED would turn Off.
The old style 13S lithium pack would need to be ~93% discharged before the Green LED would turn Off.
With a 48V lead acid battery, the Green light would turn off with 30% capacity still remaining, and the Amber LED would turn Off with ~3% remaining.
The battery gauge on a 24V or 36V throttle will not last long if subjected to a 48V battery, whereas the battery gauge on a 48V throttle should not die, it should simply continue to be totally inaccurate (All 3 LEDs constantly lit up right throughout the majority of the discharge period) with a 14S lithium battery.
However, the accuracy of the 48V gauge can be improved by adding a 500 Ohm variable resistor (https://goldenmotor.com/SMF/index.php?topic=2026.msg9972#msg9972) in series with the battery voltage supply to the gauge's PCB.
(http://www.goldenmotor.com/accessories/images/zoom_tt2.jpg)
(https://i.imgur.com/whIHWOO.png)
If your throttle does not have the voltage printed on the casing like the ones shown above, it is likely to be a 36V throttle, as the 24V and 48V throttles were typically marked, whereas the 36V throttles were not. ::)
The external controller conversion on the motor itself should be relatively easy on the MP5, as it only requires the three phase wires and the five Hall sensor wires to be extended out of the motor to reach the external controller:
(http://goldenmotor.com/SMF/index.php?action=dlattach;topic=4269.0;attach=5632;image)
You will find that the motor will be much noisier if you don't use a sine wave controller, and you will obviously lose the Bluetooth functionality that the MP5 controller provides.
Alan
-
I'd be suspicious of an upper wiring harness/component problem taking out the lower harness main power and eventually the 5 volt dc regulated power which I would guess your mystery chip provides.
Have 5 volts at the board? If not perhaps a temporary external supply will help sort things out. AFTER you check for shorts... ;)
Use these for wiring checks, and voltage check points.
(https://i.imgur.com/TvYOGPa.jpg)
(https://i.imgur.com/PewUb2P.jpg)
Oh, and make sure you have no loose ends! (https://electricbike.com/forum/forum/builds/ebike-building-directions/48457-1st-build-e-bike-rider-magic-pie-v5-rear-bottle-batt-52v-11-5ah-huffy-parkside?p=64253#post64253) :D
Regards,
T.C.
-
Unfortunately, I have no idea which of the two types of chip is used. :(
Thanks for your reply Alan,
I ordered 3 chips from https://www.digikey.at/product-detail/en/texas-instruments/LM34923MM-NOPB/LM34923MM-NOPBCT-ND/3527216 on Wed evening, and they arrived today!
€10 for the chips and €20 for shipping... but much cheaper and quicker than a new controller.
So now I have the delicate task of trying to replace it and hoping it'll work. This thing is *tiny*!
Disconnecting the board and reflowing in the oven would be a hassle, so willl try with a heat gun - unfortunately it doesn't have any thermal control but I could bleed off some of the airflow and direct that more precisely and make a mask with some alu foil to protect other components, or maybe I'm worrying too much!
Image of suspect chip attached, hopefully.
Is it also safe to power up the controller without the motor connected? This would help to check voltages and waveforms.
Over recent years, I have done this countless times using a 0-60V benchtop power supply with different GM controllers and have never experienced any problems.
I also discovered that a mere 12V supply is sufficient to provide adequate power for programming purposes. ;)
Thanks for that info - that's really encouraging.
(Also on both controllers, the battery indicator light on the throttle failed after a couple of weeks, and I suspect the controllers gave up supplying voltage to it - GM should look at this)
This sounds a bit strange, as the LED battery gauge in the throttle is not dependant upon the controller.
The battery indicator board inside the throttle unit is fed directly with the full battery voltage "en route" to the controller (there's a direct connection to the main battery positive cable inside the molded junction block where the five cables meet near to the controller) therefore it is not being supplied by the controller.
If your throttle does not have the voltage printed on the casing like the ones shown above, it is likely to be a 36V throttle, as the 24V and 48V throttles were typically marked, whereas the 36V throttles were not. ::)
I remarked on it because the battery indicator leds started working again when I replaced the controller which surprised me - it's not an issue as I don't particularly need them and usually kept it switched off.
The throttle is a different type than the photo, a two tone gray fascia and black body, but does appear scaled for 48V.
The external controller conversion on the motor itself should be relatively easy on the MP5, as it only requires the three phase wires and the five Hall sensor wires to be extended out of the motor to reach the external controller:
You will find that the motor will be much noisier if you don't use a sine wave controller, and you will obviously lose the Bluetooth functionality that the MP5 controller provides.
If I eventually do go down this route then I would definitely go for a sine wave controller with higher voltage and replace the thin power cable with something much more appropriate. I love the quietness and lack of cogging, and also have 800 18650 cells looking for something to do :-)
I'm sure there are other controllers with bluetooth, but it isn't essential - the BMS I have has bluetooth so it gives me the state of all cells as well as total voltage, current and remaining capacity, a history log for each cell, ability to set thresholds and even a lock to disable output which is useful to save sparking on the main connector.
DIY Tech & Repairs on youtube did a useful review of it, and I bought a 60A version based on that.
https://youtu.be/sGj9UrfEcAw
Bluetooth on my Samsung S3 is a bit temperamental, but it seems to work OK.
-
I'd be suspicious of an upper wiring harness/component problem taking out the lower harness main power and eventually the 5 volt dc regulated power which I would guess your mystery chip provides.
Have 5 volts at the board? If not perhaps a temporary external supply will help sort things out. AFTER you check for shorts... ;)
Use these for wiring checks, and voltage check points.
Oh, and make sure you have no loose ends! (https://electricbike.com/forum/forum/builds/ebike-building-directions/48457-1st-build-e-bike-rider-magic-pie-v5-rear-bottle-batt-52v-11-5ah-huffy-parkside?p=64253#post64253) :D
Regards,
T.C.
Thanks for the tips TC!
My controller has a different layout from the one shown - here are all my layers attached... I did make a list once of what colour wires went to which pins but I've mislaid it and may have to do it again.
I think I need to check the 78M05 too, probably best to desolder it and make up a test rig for it before attempting to change the LM34923 and possibly damaging it too.
If it does all work again, I'll need to find some similar silicone potting compound before it rains!
-
And one more showing the wire colors for reference...
-
Hi Andy,
Thanks a lot for the terrific pictures! Yeah, that pic is from an older Pie 4 I believe. But good to note that the wiring color/terminations are the same. May I use your pictures for some updated illustrations?
Wires to pins? Stock MP5 Pic...
(https://i.imgur.com/fA922YT.png)
So the 78M05 is probably the 5 volt output? Had you lost the 5 volts on both boards?
And who rides in the RAIN anyway? :P
-
Hi Andy,
Thanks a lot for the terrific pictures! Yeah, that pic is from an older Pie 4 I believe. But good to note that the wiring color/terminations are the same. May I use your pictures for some updated illustrations?
Of course!
So the 78M05 is probably the 5 volt output? Had you lost the 5 volts on both boards?
Only tested the one on my 2nd board - I desoldered it, put it across a 3S lipo and got 5V out, so it's good.
I removed the LM34923 chip eventually - it did not come off easily so I used more heat than I think was appropriate and it still didn't budge so I think it was glued as well!
Had to be careful not to nudge any of the other SMD components as they were totally floating... would have been a really really bad time to sneeze!
I eventually used a blade to cut the legs and pick them out with a soldering iron without lifting any tracks.
During this I also partially melted the wire loom, so will replace it with the one from the previous controller.
(This is OK - the cable sheath was damaged anyway when the wheel tore itself out of the drop outs on my previous frame through not using torque arms! Since rectified.
It's really not easy using a Black & Decker heat gun on this without setting fire to and/or melting everything else on the desk, including bits of me that might get in the way.
Now working up the motivation and courage to replace the loom and the chip, check shorts and then power up!
And who rides in the RAIN anyway? :P
I do... (or did!).
In Hungary it isn't just the rain - some roads are so bad that pools of water form in sun-melted grooves which spew tsunamis when a truck goes through them. This is not much fun when it happens, as a solid wall of water exerts quite a force, so it helps to be prepared with full waterproofing, and insulation when it's -25 °C in winter.
As they apparently say in Iceland: There's no such thing as bad weather, just inappropriate clothing!
-
Only tested the one on my 2nd board - I desoldered it, put it across a 3S lipo and got 5V out, so it's good.
I removed the LM34923 chip eventually - it did not come off easily so I used more heat than I think was appropriate and it still didn't budge so I think it was glued as well!
Had to be careful not to nudge any of the other SMD components as they were totally floating... would have been a really really bad time to sneeze!
I eventually used a blade to cut the legs and pick them out with a soldering iron without lifting any tracks.
During this I also partially melted the wire loom, so will replace it with the one from the previous controller.
(This is OK - the cable sheath was damaged anyway when the wheel tore itself out of the drop outs on my previous frame through not using torque arms! Since rectified.
It's really not easy using a Black & Decker heat gun on this without setting fire to and/or melting everything else on the desk, including bits of me that might get in the way.
Now working up the motivation and courage to replace the loom and the chip, check shorts and then power up!
All above my paygrade... :o
Is this the way you see it up close and personal?
(https://i.imgur.com/xgCBFbd.jpg)
Edit: Back from the future. Replaced picture with latest and greatest. ;)
errrrr is speed the same as the pedelec input?
-
errrrr is speed the same as the pedelec input?
No, SP is the throttle signal input, the yellow wire is for the pedelec input.
I think you need to correct two entries as shown here:
Green = THROTTLE SP
Yellow = PAS or PEDELEC TA
(It's actually TA printed on the circuit board not TH, see picture below)
You could even correct the word "Receive" while you are in "edit mode" if you wanted to.
(i before e except after c)
(https://i.imgur.com/5aneEEW.png)
Alan
-
Thank you Alan! Good catches. Will correct. I like the way you can see the labeling in your photo...may I incorporate it? Sorry for the drift Andy. Will start-up a controller input discussion in another thread. ;)
-
The photo was not taken by me, I'm pretty sure it was cropped from a larger photo in a previous post by another member of this forum, but I can't remember who. :-[
I'm sure the original poster wouldn't object to it being used for the benefit of others. ;)
Alan
-
Happy news - finally got motivation together to solder in the new chip and replace the wire loom from the previous controller.
Was not easy aligning by hand something to about 50 micron accuracy while trying to keep it still, dodging the heat gun blast, trying not to overheat it and not disturbing the other components whose solder was also molten.
I had 3 pairs of reading glasses on to be able to resolve the detail, and it looked OK afterwards.
Finally put it all back together again, mounted the wheel, plugged in the control harness and powered up. No bang!
Twitched the throttle and she spun! Yayyyy! So happy!
Now just have to replace a spoke before pumping up and I'll be mobile again and able to give it a more thorough test.
At least if it happens again I have another 2 spare chips, but the thought of going through this again doesn't fill me with joy.
Will report back after a couple of runs.
Cheers,
Andy
-
Outstanding work on the board! ;D My concern would be that you did not find the root cause of the issue, just fixed the results of...did you check things out, harness, wiring, components, regulated voltage amperage draw? After all that hard work I would sure look it over closely. ;)
-
I'm pretty sure the root cause was sustained regen voltage overload by failing while coming down a very steep hill with a 14S battery, 40kg bike and 75kg passenger. It never failed before on similar situations, but it gave up on this one. I had no reason to suspect the cables.
By the look of the chip it had swelled slightly, so it wasn't man enough for the job - hopefully the new chip will fare better.
At least I know the cause, and that the replacement part I shared above works for anyone else that may come across the same problem.
Apparently the controller is notorious for failing, and this is probably the reason why - I hope GM can fix it in a newer version.
So, I got the wheel on, spokes fixed (I really need a complete set of new ones as all the GM spokes failed at the bend, and a rim!)
Pumped it up and the tyre went flat after a couple of hours, so am currently about to fix that...
My woes aren't totally over yet, it never seems to end!
-
My elation was short lived.
I took it out for a run of about 10km, did some shopping rode a bit more then on the way back it went dead again while braking, but this time on the flat. Regen worked many times during the trip but then just gave up. I'm not sure what else the problem could be - the motor won't generate any more voltage and current than it had been doing for the last 2.5 years quite successfully.
Luckily only had about 200m to cycle back, but I'm gutted and faced with either repairing it yet again or getting a new controller which will probably do exactly the same thing.
I want to be able to ride across Europe, but I just cannot trust it.
Perhaps the LM34923MM chip should have been an LM34923MMX?
Is there anyone at all a GM that can answer technical questions?
I've tried their support chat line, but there's never any response.
So, guess I'll have to try again and see how it goes using 12S. :-(
-
Did you end up changing the big 1000uF capacitor or is it the original "bloated" one still in there?
I was just wondering, if the "bloated" capacitor was no longer able to absorb high voltage spikes, could this have resulted in the failure of the new LM34923MM chip (assuming the same chip has failed). :-\
I'd be suspicious of an upper wiring harness/component problem taking out the lower harness main power and eventually the 5 volt dc regulated power which I would guess your mystery chip provides.
Have 5 volts at the board? If not perhaps a temporary external supply will help sort things out. AFTER you check for shorts... ;)
If this chip does provide the +5V supply, it might be worth testing with an external 4.5~5v supply via the pedelec wires to see if the controller still works.
Alan
-
Did you end up changing the big 1000uF capacitor or is it the original "bloated" one still in there?
I was just wondering, if the "bloated" capacitor was no longer able to absorb high voltage spikes, could this have resulted in the failure of the new LM34923MM chip (assuming the same chip has failed). :-\
Good point - the bloated capacitor was in the first controller - the one in the current 2nd looks ok and gives a normal spark on power up. I suppose I could try to charge it through a resistor and observe the charge curve. Desoldering it is a nightmare as the bottom board is a huge heatsink and difficult to get at the legs.
Also, I cannot find any ZhiCheng capacitors on the net, or any of the required 18x25mm in size from anyone!
18mm radius is an immovable limit because of the upper circuit board, but I found some 18x30mm 1000uF 80V and ordered those from China just in case they may be useful one day.
Not impressed at all with this non-standardness!
If this chip does provide the +5V supply, it might be worth testing with an external 4.5~5v supply via the pedelec wires to see if the controller still works.
I haven't taken the controller out yet but I'm pretty sure it's the chip that's gone again - pumping 5V in through the pedelec sounds creative and I can easily do that with a buck board powered by the 58V->12V converter I use for the lights and usb chargers. If that works, then I'd be happy, but it sounds dodgy - the linear regulator would be seeing 5V without having a corresponding input and I don't know what other things might depend on the voltage from the LM34923 chip. It probably puts out more than 5V for the linear regulator to do its thing.
If I try again with another chip, I'll try to glue a heat sink to it but it's supposed to cope with 75V, and 58V is the max I use.
I don't think the motor would put out more voltage than its kv allows, and the high current flows through the fets not the buck chip, so am still a bit mystified why it would fail.
A circuit diagram would help a lot!
-
Hi Andy,
I have just been studying your photos of the controller, and there is a much larger 78M05 on the upper side of the top PCB that I think is far more likely to be the +5V regulator chip, so I'm puzzled as to what the SM5B chip is for.
I was actually looking for a shunt that controls the current (like all of the earlier square wave controllers had) but I could not see one.
I guess it must use a section of copper track on the PCB instead. :-\
As you've already stated, a circuit diagram for this controller would indeed be very useful.
Alan
-
I have just been studying your photos of the controller, and there is a much larger 78M05 on the upper side of the top PCB that I think is far more likely to be the +5V regulator chip, so I'm puzzled as to what the SM5B chip is for.
The SM5B chip is a 75V buck converter, so I think it's taking the high voltage from the battery or the regen and stepping it down to a voltage below 25V for the 78M05 chip to produce the 5V. Seems a bit crazy given its tiny size.
I was actually looking for a shunt that controls the current (like all of the earlier square wave controllers had) but I could not see one.
I guess it must use a section of copper track on the PCB instead. :-\
As you've already stated, a circuit diagram for this controller would indeed be very useful.
High voltages carrying smaller currents can get away with smaller shunts so it may not be obvious - smd components make reverse engineering quite challenging!
I have a hypothesis that the BMS went into charge protection mode and cut the connection to the battery so the motor would be generating with no load to keep the voltage down - however based on what I think I've learned is that the MP5 is supposed to prevent this from happening. Maybe reducing the regen current down from 50A would help, though the BMS never seems to see this amount.
The BMS does provide logs but I'm not convinced of its accuracy. I've attached a couple of edited screenshots showing values for one cell over the duration of the ride, the whole pack, and a zoomed in area at the point of failure.
Resolution is only 1 min. However, if there is a 4V sag at 30A draw, then there's likely to be a 4V surge charging at 30A, and if that happens for longer than the BMS overvoltage duration threshold, then it could cut the BMS and leave the regen without a load, and possibly cause the voltage to exceed 75V.
The failure happened in this last case after slowing down from full speed in one braking action, so it could have caused the BMS to trigger. Pulsed braking or reducing regen current could perhaps avoid this too, but I would really miss it.
It gives me something to experiment with.
I could go back to my 12S lipo pack which doesn't have a BMS as I charge it at 12A with a balancing charger in two sessions.
I'd prefer to use the 14S with the BMS, as it's much higher capacity and I can charge it at 5A in one session and know the cells are protected even if the balancing current has no hope to balance 16 cells in parallel!
So, I think I'll replace the chip again, and use the 14S without the BMS and charge it in stages as I used to do.
If it fails again in a similar situation, then I'll know it wasn't the BMS!