Arlo's power stage Leaf controller runs and drives page 103

[Arlo1]

It would be better to turn on that FET than let any diode catch it, would it not?

I am talking about the PWM i'm looking at it this way.
say I can turn the fet on for 20uS till the max current of 100 amps is achieved and say this is all we can push though the fets then we have to turn it off but because of the inductance of the motor windings the current will continue to flow for about 20uS and it will flow through the diodes. Now the way I would like to get it all working is to have the pwm at a rate where we can turn on the fets for the first time for about 20uS then off for a lesser time like maybe 5uS then on for 5uS and off for 5uS and repeat. This would keep the current bouncing between 100 amps and 75 amps roughly giving us a 87.5 amp average.
We can NOT "just turn the fets on" because they will flow more current then they can handle after a certain period so with fast enough current monitoring we can hopefully run them close to their limit. But doing fast switching speeds and relying on the diodes like this will produce a lot of heat. Hence why I asked about parallel diodes! :wink:

I just used the numbers of 100 amps and 20uS and 5uS as examples BTW.

 

[bearing]

It would be better to turn on that FET than let any diode catch it, would it not?

In the deadtime the body diode will start to conduct. Then when the opposite FET turns on, it will have to discharge the reverse recovery charge of that body diode. The current pulse to discharge the reverse recovery charge is very high if the FET is switched fast. The pulse can be hundreds of amperes when nominal current is 50A, according to my simulations. This will cause high losses during the switch transition, and also cause stray inductances to oscillate with the output capacitance of the FETs. Schottky diodes have a very low reverse recovery charge, which could decrease the switching losses.

 

[texaspyro]

I do not know much about TVS diodes, but they will supress spikes between phases.

I know that they vaporize with a very satisfying KAPOW! when used across the FETs in a CD welder...

 

[Teh Stork]

It would be better to turn on that FET than let any diode catch it, would it not?

I am talking about the PWM i'm looking at it this way.
say I can turn the fet on for 20uS till the max current of 100 amps is achieved and say this is all we can push though the fets then we have to turn it off but because of the inductance of the motor windings the current will continue to flow for about 20uS and it will flow through the diodes. Now the way I would like to get it all working is to have the pwm at a rate where we can turn on the fets for the first time for about 20uS then off for a lesser time like maybe 5uS then on for 5uS and off for 5uS and repeat. This would keep the current bouncing between 100 amps and 75 amps roughly giving us a 87.5 amp average.
We can NOT "just turn the fets on" because they will flow more current then they can handle after a certain period so with fast enough current monitoring we can hopefully run them close to their limit. But doing fast switching speeds and relying on the diodes like this will produce a lot of heat. Hence why I asked about parallel diodes! :wink:

I just used the numbers of 100 amps and 20uS and 5uS as examples BTW.

Hmm, this would only apply at very low rpm's. At 20kHz pwm, one pwm cycle is 50uS. Say that you have 100% duty cycle and a motor time constant of 50uS (L/R). Lets say the motor has a maximum amperage (given by U/R) of 500A. After 5 time constants - the current is 500A. After one, the current is 315A. Given a current limiting of 100A - we would have to clamp the current after some 15 uS! The way most interrupts works is that they shut down the current once 100A is reached - then wait until next pwm cycle starts over just like this.

After the motor starts spinning, the high Z will reduce the current inrush. A higher pwm switch freq at startup is the easiest I can imagine.

Edit: some calc's were abit off :x

 

[Alan B]

If I understand this correctly, when the PWM FET opens up the diode that catches the kickback is the one in the PWM's phase but is the opposite side FET. This opposite FET can be energized to reduce the diode loss significantly. It is like a SPDT switch connecting the motor to power, and alternately to ground, allowing the magnetic stored energy to circulate. The Commutation FET stays closed the whole time on the other end of the motor phase.

Note that this is a separate issue from the proper selection of frequency for the motor's inductance. The switching frequency must be high enough that it can control the rise of current. A low inductance motor requires higher switching frequencies.

 

[amberwolf]

That is, IIRC, what is called synchronous mode. It is harder to do because you have to time it perfectly so you don't wind up with a shoot thru. Most ebike controllers don't bother with it, but would benefit from it.

 

[Alan B]

Yes, that is the term used for this feature on switcher power supplies.

I believe some Kelley controllers do it. The microprocessors and FET drivers have features to support this and handle the timing requirements and make them tunable. Even if the diode clamping FET is late the diode will conduct for a short time and then the FET will clamp it, reducing the overall heat and increasing efficiency. So the timing can be "safe" and the diode will take only a little current.

 

[Arlo1]

Hmmmm I see.

 

[Arlo1]

OK so now if that is to be considered. What about changing the DC/DC supplies to only just a bit above the miller plateau? SO say my irfp4668 have a plateau at 5.5v why not drive them with a 9v supply instead of 12? It will be quicker to discharge and need a smalled resistor so it will be more efficient. But i'm just thinking about quick switching times.

 

[bearing]

Put a diode in parallel with the gate resistor to get a fast turn off time and still have the same turn on time.

 

[zombiess]

Put a diode in parallel with the gate resistor to get a fast turn off time and still have the same turn on time.

Just be careful depending on what kind of driver circuit you are using as too fast of a turn off can cause a Vs overshoot which can cause a temporary high side lockout condition which they say isn't that big of a deal for most scenarios. I know this can be an issue on the IR2110 drivers I'm using. It's one of the reasons I'm using paralleled diodes and resistors to be able to independently tune the turn on/off times of my banks. I believe I read a way to clamp it with a zener diode as well, but to be honest I really need to read up a little more on this and actually get my soldering iron hot and my scope watching what's actually going on. I really need my crazy travel schedule to calm down so I have a clear head with no jet lag and time to make progress. Didn't know I was going to be doing so much travel so soon. Just kills me I have all this stuff I want to do now and so little time to experiment when normally I'm bored off my ass with tons of time for projects. Least I'm making some $.

 

[bearing]

Put a diode in parallel with the gate resistor to get a fast turn off time and still have the same turn on time.

Just be careful depending on what kind of driver circuit you are using as too fast of a turn off can cause a Vs overshoot which can cause a temporary high side lockout condition which they say isn't that big of a deal for most scenarios. I know this can be an issue on the IR2110 drivers I'm using. It's one of the reasons I'm using paralleled diodes and resistors to be able to independently tune the turn on/off times of my banks. I believe I read a way to clamp it with a zener diode as well.

I didn't really understand this. What's the cause of Vs overshoot? what is lookout? what's the IR2110 issue? if using a diode in parallel with the gate resistor is the cause of the problem, how is it solved by "using paralleled diodes and resistors to be able to independently tune the turn on/off times of my banks"?

 

[dnmun]

i thought the gate resistor already has a diode in parallel. there is diode there, where does it connect? brain fade, can't remember how the gate drive works on the high side.

EDIT: ok now i remember, the diode is in series with the gate to hold the charge on the gate while the phase wire, which is ground on the high side, drops and then that causes that pnp transistor to turn on and drain the gate charge to the phase leg.

 

[Arlo1]

i thought the gate resistor already has a diode in parallel. there is diode there, where does it connect? brain fade, can't remember how the gate drive works on the high side.

Gate > source for ON and gate = source for OFF! Diode is inside the fet and some people use other diodes that charge the hi side cap when bootstraping! The source is always the Negative side in comparison to the Drain.

 

[Arlo1]

Here is some scoping of the gates and one of the phases. You can see my PWM low side gate is at the miller plateau at about 500ns I am running 22 ohm gate resistors. I think ~20 per fet is optimal. When I build my multi fet power-stage I will use 10 off the fet driver then plateau match all the fets with the second individual gate resistor. [youtube]n4s1yA3aYgI[/youtube]

 

[Arlo1]

Look what cool things HK has for sale now!!! http://www.hobbyking.com/hobbyking/store/__21192__Blue_Aluminum_Battery_Water_Cooling_Board_2pcs_.html

 

[Harold in CR]

That would go good with their water cooled controllers for boats, no ??

 

[Arlo1]

That would go good with their water cooled controllers for boats, no ??

I am using my own controller to drive this powerstage. But the power stage is all that needs cooling.

 

[Harold in CR]

Gotcha. I was thinking something else.

 

[parabellum]

That would go good with their water cooled controllers for boats, no ??

I am using my own controller to drive this powerstage. But the power stage is all that needs cooling.

Almost ordered them few days ago, but finally did not. IN/Out channels are tinny and will not let significant amount of circulation liquid going trough. To correct this issue will probably not worth the effort.
In your case, few cooper tubes soldered along your cooper bars will be cheaper and far more efficient solution.

 

[Arlo1]

I have a question what do you guys think about parrellel diodes to the diodes in the fets???

If freewheeling mosfet heating is a problem: go for it Will also reduce diode forward drop some, in turn - reducing some of the heating aswell. You should look into TVS diodes aswell, Colton seems to be into them for some of his hi-specific-power builds. I do not know much about TVS diodes, but they will supress spikes between phases. (mounted between A-B, B-C and A-C in the cart build)

I just checked and its not like this. He hooked them up source to drain http://web.mit.edu/first/kart/controller_rev1.pdf

Because the floating supply and optocoupled gate driver are very modular, the MOSFETs used can
be virtually anything. They can be individual small MOSFETs, many small MOSFETs in parallel, or
large MOSFETs with huge current-handling capabilities. They can be in almost any configuration,
including half- and full-bridges. The modular gate driver treats high- and low-side MOSFETs the
same. There are a few general guidelines to follow:
D
S
G
Almost Anything…
1. Use gate resistors. The optocoupled gate drivers can handle a peak output of 2A @ 12V. So the
total effective gate resistance should be no less than 6Ω.
2. Use TVS diodes from source to drain to protect MOSFETs from inductive spikes.
3. For parallel MOSFETs, observe good gate resistor practices to prevent ringing. See next slide.
4. Use a pull-down resistor on all gates to ensure that MOSFETs turn off in the event of a gate
driver failure.
5. Check the gate capacitance specification in the MOSFET data sheet to determine the
approximate turn-on and turn-off time, given your chosen gate resistance. Set the shoot-through
delay at the optocoupler accordingly.

Anyone want to tell me how to spec a diode for this?

 

[Teh Stork]

Anyone want to tell me how to spec a diode for this?

Sorry for the confusion, yes - S-D connected.

I looked a bit into TVS design, my own guidelines were:
1. Clamping voltage at or below mosfet breakdown voltage.
2. Operating voltage at or higher than bus voltage. (This proves to be one of the hard steps, A 100V clamping tvs usually have a operating voltage of 70V(Breakdown is usually 10% over this))
3. Inductive spike energy can be calculated by 1/2*L*I^2. For a 1 uh inductor carrying 200A, this is: 0,5*1*10^-6*200^2 0,02watts in each pulse. You don't see 200A often in normal operation, but lets say you switch at 40kHz - 800W (clearly the calculations are unrealistic).

The reaction time of the TVS is really only limited by the inductance of the connecting leads.

My own conclusion is that unless you want to switch at sub-diode recovery times, a TVS is not needed. The complications of switching at these insanely fast times needs careful design as is.

 

[bearing]

Stop spreading this rubbish/myths!TVS diodes are not needed.

The MOSFETs are more than capabable of dealing with avalanche from stray inductance by themselfes. If you put one or two TVS diodes there, they will fail far earlier than the MOSFETs. If the MOSFETs die from avalanche, then something is wrong with the design to start with, and should not be solved by putting 100grams of TVS diodes there.

 

[Arlo1]

Stop spreading this rubbish/myths!TVS diodes are not needed.

The MOSFETs are more than capabable of dealing with avalanche from stray inductance by themselfes. If you put one or two TVS diodes there, they will fail far earlier than the MOSFETs. If the MOSFETs die from avalanche, then something is wrong with the design to start with, and should not be solved by putting 100grams of TVS diodes there.

Listen bearing. I am looking into reducing switching losses. Shane Coltan is a guru with this stuff so I will consider his design. I already know i need to run a fast PWM frequency so I am looking at what I can do to keep the heat out of the fet for 1 and reduce switching losses for 2.

If you have proof we don't need TVS diodes then get it for me. Show me your hi power controller design and lets see why it works and I will borrow your ideas.

 

[bearing]

You have a point. I have never made anything for these power levels, so I shouldn't make so much noise.

But, in my opinion, the effort should be put at making a design with low stray inductance, rather than trying to kill the ringing caused by that inductance with a TVS or similar. With a good physical design, the heat caused by avalanche will most likely be negligible.