Crystalyte Controllers - Repair and Modification information

[fechter]

Someone else (I forget now) did some testing on this. It's worth a try.
You'll just hit the current limit faster when you nail the throttle.

[bobmcree]

on the throttle input stage in the crystalyte controllers there is an op amp that subtracts a volt from the throttle signal and then amplifies it so that 1-4v from the throttle is turned into 0-5v before it is fed directly to the soft start cap on the pulse width modulation chip. the output of this chip is fed to a logic gate where it is combined with the motor drive pulses to achieve pwm control of the high side fets. the low side fets are just turned off and on.

the pwm chip is designed to be used a bit differently than the way the crystalyte designer implemented it. there is a feedback amp in the chip that could be used for current feedback to get better motor control, but instead they used that circuit for low voltage cutoff and current limiting. as a consequence, the output of the pwm is directly proportional to the voltage on the soft start cap. the conditioned 0-5v throttle signal is applied directly to this cap, and the ebrake can pull down the cap voltage through a diode. the cycle analyst pulls down the ebrake signal to achieve power limiting, as does the current mode throttle design fechter posted.

reducing the value of the soft start cap which is connected to pin 8 of the ka3525 pwm chip will speed up the ramp up of the pulse widths under all conditions, including the one where you switch the unit on with the throttle stuck on. playing around with the throttle voltage as a function of position can change the throttle response curve a bit, to make it more linear.

when looking at the voltage on this cap be aware that there is an internal circuit on the 3525 that can pull it down when the shutdown signal is asserted on pin 9 by the circuit that tells it the motor is not turning. this can be confusing at first. if you reduce the value of the cap too much you can let the pwm go to full on before the controller "realizes" the motor is stalled and blow a lot of fets as a result. be cautious.

[rf]

Thanks for the heads-up on the throttle delay cap, Bob.

Sounds like fechter's trim pot is the safer fix.


Richard

[bobmcree]

yep, the soft start cap is intended to ramp up the duty cycle on initial start and when releasing the ebrake. if you decrease it a lot you will find the bike does a wheelie when you release the ebrake with the throttle full on. don't tell kids who ride the bike about this, as they tend to blow controllers.

the throttle charges the cap through a fixed resistance, so that controls the rate at which the throttle controls the voltage on the soft start cap. it should be possible to increase the rate of throttle response without sacrificing soft start by decreasing the value of the series resistor where the modified 0-5v feeds in to the soft start cap. on a controller i just looked at it is 5.1k on the opposite side from the throttle connector. you could cut this to 2k and get faster throttle response without sacrificing soft start on ebrake release or startup from a dead stop.

[Jozzer]

Mark @ Team hybrid asked me about increasing the ramp time on the controllers, since he uses them in handcycles. Are there any dangers in increasing the cap size? And by how much?

Ta, Steve

[Doctorbass]

see my tread about the ramp capacitor:

Bob explained the detailed theory... i tried it

http://endless-sphere.com/forums/viewtopic.php?t=1765&highlight=capacitor

2.2uF work great... at least for a couple of month.. and no problem.

Very quick response!

[Jozzer]

I need to go the other way though Doc, for a SlOw response (disabled persons vehicles).

[nemo]

You have 25 A123 cells in series, how many in paralel? 3 or 4 sets?
8.9Ah is strange number considering that A123 should have 2.3Ah. 75cells or 100 cells in total?. Quite some setup! How do you charge the monster ? If you have a link to your page with a detailed description let me know please. I'll be working on a very similar setup for the next year. P.N.

[Jozzer]

4 in parallel. I use a regulated 10A70v PSU for charging, and balance by hand when needed. Pack has a series/parallel relay, so I can charge at half pack voltage if needed, and use an astroflight 112d with a lipodapter (for 3.7v voltage cutoff) as a portable charging solution.
http://endless-sphere.com/forums/viewto ... hlight=kmx
http://endless-sphere.com/forums/viewto ... hlight=kmx

[bobmcree]

Mark @ Team hybrid asked me about increasing the ramp time on the controllers, since he uses them in handcycles. Are there any dangers in increasing the cap size? And by how much?

Ta, Steve

the supply to the cap is current limited so you should be able to increase it almost indefinitely, but the transistor inside the ka3525 pwm chip that discharges the cap when shutdown is asserted will probably blow at some much higher capacitor value from the extra energy it has to discharge. i would try 22, then 33,47,68 uf, all standard values and anything over 10v, if you are careful you can change this cap with the board attached to the metal plate. if you are increasing the value you can leave the existing cap in and just add more in parallel between pin 8 of the 3525 and ground.

you might run into the situation where it starts up so slowly that it always shuts down "thinking" that the motor is not turning.

the soft start cap is really only intended to ramp up the power when you release the brake, or if you power up with the throttle at a high setting. more sophisticated controllers have mechanisms to adjust the throttle response curve, but here it is just a simple RC time constant, so changing the resistor that charges the cap or increasing the cap will have the same effect. you have to remove the board from the plate to get to the resistor so it is a bit more work, but replacing the resistor with a pot would let you adjust the ramp time.

[Jozzer]

Thank you Bob, helpfull as ever

[bobmcree]

i don't want to dissuade anyone who wants to rip out spokes and do wheelies, it is fun. people who don't desire this should read the following:

we should consider that doctorbass has upgraded his controller to fets that can handle 16x the power of the standard ones, so the fact that he has not blown them cannot be taken as an indication that the same mods to a stock controller start up cap would have the same result.

there is a good reason for the soft start cap, and just because the crystalyte designers took shortcuts to save a buck does not mean it can be disabled or treated lightly. there are many good reasons not to instantaneously apply full pwm to the fets when the shutdown signal is de-asserted. if you shorten up the time constant that applies to every startup function, and there may be cases where the ebrake or low batt signal is asserted that you do not want the system to jump back to full power as soon as it is released. i really think it is a better idea to decrease the resistor value that charges the cap from the throttle signal, preserving the soft start function for power-up and ebrake release.

[rf]

Can someone tell me how to locate the current limit adjustment resistor on the 48volt 20amp controller?

Thanks

Richard

[fechter]

From a schematic standpoint, it will be the same as the 35 amp version.
See page 1 of this topic.

I've never actually seen a 20 amp one in person, so I don't know the board layout.

You should be able to trace it out from the KA2535.

You could also just change the shunt if you don't need an adjustable change.

[Jozzer]

http://endless-sphere.com/forums/viewto ... t+limiting
Pic and instructions for 20A version. The other Fechter knows the board layout

[rf]

Can someone tell me how to locate the current limit adjustment resistor on the 48volt 20amp controller?

Okay, I wasn't explicit enough.

I have a CycleAnalyst, so I don't need variable current boost. What's the easiest way to up the max current on a 20amp controller without disturbing the shunt itself?

Thanks

Richard

[bobmcree]

it is much simpler to change the shunt resistance than any other method, either by adding solder to connect two of the shunt conductors together for part of its length or by soldering a piece of wire between battery minus and the minus of the big caps. if you use a piece of #10 wire about 8-10 inches long you can bring it out of the controller and connect the ends together externally for more current, and adjust the wire length to get the desired limit.

because of the way they do the current limiting and low batt by hijacking the error amp in the ka3525 it is difficult to do it repeatably by any other means.

if you have a current limited power supply and a good meter, you can run an amp or two through the shunt on the controller, and you should find it is very close to a milliohm, inferred from the 1 mv/amp voltage drop across it. you can add solder until the voltage goes down to about 80% of the original value, then reprogram the cycle analyst with the new shunt value.

because of nonlinearity in the circuit the change in the shunt will not give you exactly the boost in amps you would expect from the resistance values, more like the square of the change you make is my experience. (if you make the shunt 70% of the original value it will about double the current limit). the adjustable current limiting methods that have been posted rely on first increasing the current by modifying the shunt.

[rf]

it is much simpler to change the shunt resistance than any other method, either by adding solder to connect two of the shunt conductors together for part of its length or by soldering a piece of wire between battery minus and the minus of the big caps. ...

That sounds like the easiest way, except I gather from what others have said here, that will make my CycleAnalyst's current readings inaccurate. Is that not the case?


Thanks

Richard

[rkosiorek]

not really. as i understand the instructions, you have to measure the value of the shunt resistor in your controller and enter the value into the setup regardless. the only time you don't have to do this is if you buy the controller and Cycle analyst at the same time. then they calibrate it for you.

rick

[bobmcree]

rick is correct. the controllers i have measured are within a few percent of the .001 ohm default in the cycle analyst, but the value is totally programmable. you can measure the value before and after the mod by passing a known current between the negative battery lead and the minus end of the big electrolytic caps, and then measuring the voltage across the points where you connected the cycle analyst sense leads to the shunt.

if you have a 1A supply and the voltage across the shunt is .001v that would be typical, and to go from 20a to 30a i expect you will want to reduce the shunt by 20% or so for the first cut, so add solder to half the length of 2 of the conductors and re-measure the voltage across the shunt, then scale the programmed value of the shunt by that factor.

alternately, if you have a good ammeter you want to use for transfer of calibration, compare the cycle analyst readings with those it provides, and scale the default shunt value of .001 ohms by the reciprocal of the ratio you measure to correct the readings.

[rf]

... alternately, if you have a good ammeter you want to use for transfer of calibration, compare the cycle analyst readings with those it provides, and scale the default shunt value of .001 ohms by the reciprocal of the ratio you measure to correct the readings.

Thanks Bob. That sounds good. But won't modifying the shunt with solder or whatever make it nonlinear? Or will calibrating to the middle of the range be good enough?

[rf]

It just occurred to me that modifying the shunt is kind of questionable. If I'm using the CycleAnalyst as my current limiter then messing with it's current sensor is probably a questionable manuever.

Perhaps the resistor network is a better idea. Assuming that won't change CA's readings ...

Richard

[bobmcree]

... alternately, if you have a good ammeter you want to use for transfer of calibration, compare the cycle analyst readings with those it provides, and scale the default shunt value of .001 ohms by the reciprocal of the ratio you measure to correct the readings.

Thanks Bob. That sounds good. But won't modifying the shunt with solder or whatever make it nonlinear? Or will calibrating to the middle of the range be good enough?

it is just a resistor so there is not a problem with linearity. there is the issue of temperature coefficient, but even if you just use copper wire in parallel with the existing shunt the tc of copper wire is about 400 ppm per degree C so if the wire varies in temperature from freezing to boiling water that is still only a few percent error, and it is only part of the shunt so the error due to temperature is not really a problem.

i keep meaining to fill in the couple of gaps in the schematic, and try to figure out a way to use the error amp in the KA3525 to implement a true current feedback throttle. it should be possible with just a few hacks to the pcb and maybe an added op amp.

i am getting continuing requests to bump up the current limit on the controllers i am refitting with the 4110 fets, since they seem to handle 80A or more quite reliably. i will be doing a couple this week set for as close to 100A as i can get, and will post the details. of course anything over 50A I will not warranty except not to be DOA

[bobmcree]

you can always get the external shunt for the cycle analyst or make your own, so that the measurement will be independent of any controller mods. the new ones justin is using have a much better temperature coefficient, but i think they are really only intended to take 50A or so. the stock shunt is made of wire that does not seem to be anything special, and i have seen them wander around by several percent over normal operating temperature. they also have some galvanic issues, voltages created by joining dissimilar metals. adding more solder to the system does not seem to make these problems any more pronounced.

i have replaced the stock shunt resistors in a couple with some kelvin connected resistors that look like little sawhorses. the current goes through 2 legs and the voltage is sensed across the other 2, greatly reducing error.

i know a guy who is routinely increasing the current limit as high as 80A by adding the parallel piece of heavy copper wire and varying the length. this seems to work fine for him.

[fechter]

http://endless-sphere.com/forums/viewtopic.php?t=915&highlight=current+limiting
Pic and instructions for 20A version. The other Fechter knows the board layout

Geez... I guess that brain cell died somewhere along the way.

If you change the 51k resistor in the schematic to something like 25k, you can boost the current limit without changing the shunt value (so the CA will still be accurate). If you go too high, the shunt could overheat.