transient suppression in regenerative controllers

[Lanchon]

dear forum,

I'm in dire need of info about techniques for transient suppression on the power stage of regen controllers.

knowing what's being successfully used would be great, for example having a description of the power stage of a regen controller including:

-controller ratings
-FETs used
-protection devices used (TVS, etc)
-big capacitors used on the power bus
-decoupling capacitors used on the power bus

even pictures of regen controller boards would help me get an idea of what's working.


the problem:

during regen, current from the motor windings is pulsed into the power bus. during the start of each pulse many factors can contribute to a higher bus voltage including the resistance and inductance of these components: battery, wiring, bus filtering capacitor(s), PCB traces, etc.

the end result is that a voltage peak is seen on the FETs that's higher than the batt voltage. this peak can exceed the FETs max voltage and cause degradation and failure.


why I need this:

I modded a GM 36V non-regen controller (my only option back when I started), I'm using only the power stage (I'm not using the on-board microcontroller). (see the GM 36V here, it's the first one: http://goldenmotor.com/SMF/index.php?topic=341.0 )

I need to decide if it's possible to overvolt this power stage to 48V (16 lifepo4s, 16 * 3.6V = 57.6V) and how to protect it for regen operation. currently it has 12x STP75NF75 FETs (75V, 0.0095 Ohm, 80A) and no protection. I don't want to change the FETs and I have little headroom: 58 to 75V. that'd be enough for non-regen but I don't know if I can operate this in 4 quadrant mode. I've got one 63V 470uF LSHC brand cap on the FET side of the shunt and two similar caps on the battery side. plus 3 bypass caps on the FET side, but no markings on them ). shunt is 10 mOhm. current should not exceed 25~30A.

so here are some questions:

-will the 63V caps fail and take everything down with them, or are they ok?
EDIT: both 36V and 48V GM regens use 2x 1000uF 63V caps on the batt side of the shunt plus a small cap on the other side (can't tell what it is).

-how should I bypass the bus? how many different capacitances should I use? which capacitor types? polyester, multilayer?

-I plan to use 6x TVS'es. both the 36V and 48V GM regens use type 1.5KE62A, so they might be ok (see link above for the 36V regen or this link for the 48V regen: http://endless-sphere.com/forums/viewtopic.php?f=2&t=5142 ). but 1.5KE62A min breakdown is 58.9, dangerously close to batt voltage, and if the motor powers them for more than a few microseconds they'll explode (and the FETs right after). and while sinking 16.7A, these TVS will climb to 85V, 10V over the FET limit. so the question: is it safe to overvolt? what do you think? (note that FETs on the 36V regen are irf2807 (max 75V, just like mine), but I don't know what FETs are used in the 48V regen yet; maybe they handle higher voltages?)

-why are resistors in series with the TVS'es? and what are their value? I found what appears to be an older 36V regen without the resistors (the way I'd put the TVS'es) all blown up. (see here: http://goldenmotor.com/SMF/index.php?topic=228.0 ) I can't understand the function of these resistors or how to dimension them.

-and lastly, in both these regens a small orange diode was added between the gate and source of only the high-side FETs. I guess these are protection TVS'es too, but I don't know. is this a shortcoming of their particular high-side gate driver, or is there a reason to always use these diodes during regen? I'm reluctant to add them unnecessarily; it adds capacitance to the gate.

I need to decide, so opinions and guesstimations are welcome.

anybody has the GM 48V regen or other 48V regen and can take a look at the power stage?

thank you very much for you help!

 

[Lanchon]

some progress:

> both 36V and 48V GM regens use 2x 1000uF 63V caps on the batt side of the shunt plus a small cap on the other side (can't tell what it is).

the small cap is 1uF 160V. (thanks silicium!)

> why are resistors in series with the TVS'es? and what are their value?

two sources now confirm resistors are 0.1 Ohm:
http://endless-sphere.com/forums/viewtopic.php?f=2&t=5142&p=98238#p98018
http://goldenmotor.com/SMF/index.php?topic=350.msg3447#msg3447


so, it appears that a sort of reliable 48V regen can be implemented using 63V bus caps (I'll decouple the bus with smaller caps of course) and 62V nominal breakdown TVS'es in series with 0.1 Ohm resistors.

the only big question remaining is this: does the 48V GM regen come with the same FETs as the 36V unit, the IRF2807 FETs (max 75V)?

if so, then it appears everything would be ok: I could beef up the power stage with TVS'es and do regen. if not, I'm back to zero (or worse, since I would interpret the manufacturer need to use 100V FETs as proof that it can't be done reliably with 75V FETs).

now the problem is that I can't find anyone that has that controller and is willing to open it. if anyone knows which FETs it uses, please tell me. (or maybe someone can point me to another 48V regen controller for which the components in the power stage are more or less know?)


one secondary question remains:

> why are resistors in series with the TVS'es?

during suppression of a 20A transient, voltage across the resistor would be 2V and voltage across the TVS would be around 85V. so basically 97% of the energy will be absorbed by the TVS, which means that the resistor can't be there to help with the transients.

there are a couple of reasons not to add a resistor:

-voltage drop on the resistor increases FET voltage exposure.
-increased inductance in the TVS circuit increases FET voltage exposure.
-resistors have average power limits but also instantaneous power limits which are relatively close (unless you use thick film or better technology). common thin film resistors can't take concentrated energy pulses (thin film means very little active material mass in which to absorb heat), that's why they are ESD-sensitive devices (believe it or not!). so the transient pulses will degrade the resistors, and thus kill the TVS protection.

so why are the resistors there? I think without them the diodes blew when they conducted in the forward (non-zener) direction. they are paralleled to the FET body diodes (which are heat-sinked) and, depending on voltage drops, they might detour a big chunk of the current of the FET diodes through them. not being designed for "continuous" use and without being heat-sinked, it'd be no surprise if they died. (see these pics of an earlier version of the same controller but without the resistors, goes kaboom: http://goldenmotor.com/SMF/index.php?topic=228.0 )

so I think the best solution is to use a bidirectional TVS, such as the 1.5KE62CA (added "C"), without any resistors. it won't conduct forward. what do you think? (why didn't the manufacturer use them? are they more expensive?)


(darn! I need to know what FETs are in those 48V regens!)

 

[Lanchon]

> I've got one 63V 470uF LSHC brand cap on the FET side of the shunt and two similar caps on the battery side. plus 3 bypass caps on the FET side, but no markings on them.

they are 0.1uF 100V, so I'm fine.

I want to add at least one 1uF to 4.7uF low impedance cap on the FET side (but what type? polyester? multilayer? not an aluminum electrolytic), and a 47uF aluminum (or do you recommend something better?).

so I'll have 3x0.1uF, 3.3uF, 47uF and 470uF; that should provide low impedance on a wide frequency range and eat up the transients.

 

[Lanchon]

btw, is there some sort of track record, good or bad, for ST's STP75NF75 FETs? (75V, 0.0095 Ohm, 80A)

 

[fechter]

You would never want your pack to go over ~58v at the max, and most of the time it will be substantially below that, so I wouldn't worry about the 75v rating on the FETs.

Keep in mind that there are several things that can cause your silicon to fail. Voltage is just one of them. Most transistors are rated conservatively, so that your FETs can likely survive well over 75v. Other failure causes are over current, over dissipation, and over temperature.

If your main capacitors are properly rated, I don't think you'll see peaks of more than a few volts over the average. When it comes to caps, more is better. Low ESR is better too, but it is often difficult to get accurate data when shopping. Multilayer ceramic or polywhatever caps usually have a much lower ESR and higher ripple current rating than aluminum electrolytics. It is good to use a combination of these. Many smaller caps will be better than one big one for ESR. The switching frequency will affect the peak to average ratio, so this should be optimized also.

I'm not convinced that TVS's would be effective in this application. I suspect they put resistors in series with them so their current rating is not exceeded. TVS's would also need to dissipate any absorbed energy, which would be wasted. Recycling the energy in the caps is much better.

To prevent overcurrent conditions, you should consider some kind of current measurement and limiting for the regen. This will greatly enhance the reliability and usefulness of the system. If the current limit is set low enough, there will be little chance of blowing anything up. This is also better for the batteries if they don't like high C rate charging.

Pairs of STP75NF75's should be fine for up to 35 amps or so.

 

[Lanchon]

thank you.

> you should consider some kind of current measurement and limiting for the regen

I've got redundant current measures.

> Pairs of STP75NF75's should be fine for up to 35 amps or so.

do you mean 35A in the motor phases or 35A in the battery side? I guess you mean batt side. I've got motor phase current control, so I can set a particular limit without worrying about any "current multiplication" effect brought on by PWM.

 

[fechter]

I was thinking battery side. If you can measure on the motor phases, they can take quite a bit more (like their rated current).

 

[Lanchon]

electrobent confirms that the same FETs are used in the 36V and 48V regens (thank you!). this means that 75V FETs can be used to safely regen on a bus as high as 58V, provided it's properly decoupled and over-voltage protected. this concludes the thread from my side at least. thank you everyone for all the help!

 

[MyPC8MyBrain]

I have one of the 48v golden motor controllers if you want to disect it. It has burned up at least one of the regen diodes and who knows what else. If you are interested let me know and I'll ship it to you. If you can fix it that would be awesome too. PM your address and info to me.

I bought the controller with a 48v 1000w motor kit. The controller fried on day 2 and the motor fried the phase wires last night when i put on a new controller.

And my standard goldenmotor statement:
BE WARY OF ANYTHING FROM GOLDENMOTOR. ONCE YOU SHIP YOUR MONEY YOU WILL NEVER GET ANY RESPONSE IF YOU HAVE AN ISSUE. NONE . TOTAL IGNORE. FU GOLDENMOTOR
So im out 500 bucks for a bunch of burned up e-bike stuff.. lesson learned.

 

[Lanchon]

hi,

thank you very much for your offer, I'll PM you.

sorry to hear both items failed. it's strange about the motor, maybe the new controller delivered unacceptable current levels to the hub. btw, it might be that only the external wires were damaged by the meltdown. there are a couple of posts on GM's forum about replacing the wires; maybe you want to take a look over there, but it seems it's a bit of work...

 

[ZapPat]

I have one of the 48v golden motor controllers if you want to disect it. It has burned up at least one of the regen diodes and who knows what else. If you are interested let me know and I'll ship it to you. If you can fix it that would be awesome too. PM your address and info to me.

I bought the controller with a 48v 1000w motor kit. The controller fried on day 2 and the motor fried the phase wires last night when i put on a new controller.

And my standard goldenmotor statement:
BE WARY OF ANYTHING FROM GOLDENMOTOR. ONCE YOU SHIP YOUR MONEY YOU WILL NEVER GET ANY RESPONSE IF YOU HAVE AN ISSUE. NONE . TOTAL IGNORE. FU GOLDENMOTOR
So im out 500 bucks for a bunch of burned up e-bike stuff.. lesson learned.

Link posted some good info about re-winding a motor that highly resembles a golden motor, maybe check that out if you want to fix it yourself. I have an older (~3 years) golden motor myself, and I have also had many different problems with mine! Very bad QC, and I do agree that they are quite bad at answering emails (even for new sales it seems!).

I had to open up the motor, replace the output phase wires that had melted their plastic sheaths (I replaced them with heavier magnet wires instead of plastic-covered ones), replaced hall sensor wires that had melted onto the motor windings, and will be replacing one hall sensor now too. I even had to add a couple washers on the axle inside the motor since the rotor spacing was narrower than what the side covers were made for. This resulted in the rotor slowly shifting sideways and finally the windings started rubbing on one sideplate, creating the short-circuit that blew my controller and melted the insulation off the phase wires. What an adventure for my first hub motor!!!

Good luck getting things resolved!

PS: If you are not going to fix the motor yourself, maybe you might want to sell it as is...? PM me if so.