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

[Arlo1]

you can also look at zombiess circuit to find out what desat zener he chose and pick from same family.

the 250uA is not so obvious in ACPL-333J datasheet. it is in the blanking charge, typical 0.24mA = 240uA.

Thanks. I'm still digging on this. The data sheets are hard to understand for zener diodes...

On another note I'm sourcing all the parts and I picked X7R rated caps for all the caps on the board. Is there any other pointers you have while I dig up all my parts?

 

[7circle]

So I am not sure what I need for specs... I think a decent amount above the voltage driving the gate which is 15 and -9 which would = 24v so 60v should be lots of head room for spikes.

10amp rated? maybe these? http://www.digikey.ca/product-detail/en/MJD2955TF/MJD2955TFCT-ND/3504460 digikey number MJD2955TFCT-ND
and this for NPN http://www.digikey.ca/product-detail/en/MJD3055TF/MJD3055TFDKR-ND/1473951 digikey number MJD3055TFDKR-ND

Having a look at the specs for the
3055/2955 pairs with
only 2MHz bandwidth
they are much slower, and spec does not include delay/rise/stable/fall times [Edit: they show them in figure 3 and for in their spec .. ]
50MHz for the d44vh11/d45vh11 TO-220
the MJB 44h11/45h11 are 50/40 Mhz

- I am surprised BJT D2PAKs don't go higher in current rating for NPN PNP, but I found Darlington transistors in 15A/D2PAK. Also MOSEFETs are way over 20A in these packages.

It's a challenge trying to find all the Right bits to get it all spot on.

The D44VH11 TO-220 ON-Semi part
15A continous
20A peak
-------------------------------------
Delay Time td − − 50 ns
Rise Time tr − − 250 ns
Storage Time ts − − 700 ns
Fall Time tf − − 90 ns

For (VCC = 20 Vdc, IC = 8.0 Adc, IB1 = IB2 = 0.8 Adc)
---------------------------------------------------------------------
Note with Vcc at 24V and more Base current giving higher Ic times will be longer.

To check how the square wave response is adding up the times
50 + 250 + 700 + 90 = 1090 ns : 1/T = 0.917mHz
This is much lower than 50MHz
as the banwidth test is done with the device not reaching the saturation causing the storage time delay.
So for a device with 2MHz bandwith the its unknown how it will respond as a a square wave, as it may have a short storage time.

Also
Figure 9. Maximum Rated Forward Bias
Safe Operating Area
This is for the Ic RMS which for the driver is less than 100mA .. the average current.
Not the pulse current.
So this figure doesn't help determine if part is okay.

So the D2PAK i linked to the ON-SEMI Datasheet "MJB44H11-D.PDF" {May, 2013 − Rev. 5}
NPN http://au.mouser.com/ProductDetail/...=sGAEpiMZZMutXGli8Ay4kC402l06AIhPRTrtmkHdFBs=
http://www.mouser.com/ds/2/308/MJB44H11-D-273946.pdf

Shows stock at Mouser and digikey for NPN parts but not PNP.

They have similar timing but NPN/PNP are show separately, so they may not be as "Complementary"
MJB44H11, NJVMJB44H11 NPN
-------------------------------------
ON Time td+tr - - 300
Storage Time ts − − 500 ns
Fall Time tf − − 140 ns

(IC = 5 Adc, IB1 = 0.5 Adc)
--------------------------------------

MJB45H11, NJVMJB45H11 PNP
-------------------------------------
ON Time td+tr - - 135
Storage Time ts − − 500 ns
Fall Time tf − − 100 ns

(IC = 5 Adc, IB1 = 0.5 Adc)
--------------------------------------

This close matching of rise/fall times may help with damping the drive circuit to stop over shoot.

Could have been more clearer....


[Edit: they show them in figure 3 and for in their spec .. RE 2955/3055 data sheet
But I find it hard to translate this info to the IGBT driver and effect on "transient dynamics" ... thats IHKFI ]

 

[Arlo1]

So why is it wrong to use mosfets?

 

[Arlo1]

if it were me, i would be WAY more comfortable with a 3W +15 and a 2W -8 connected with common point rather than two 1W bipolar power supplies connected in parallel in an isoalted and unregulated DC/DC converter application.

the parallel option would be acceptable, in my opinion, if the windings shared the same core.. but they do not.

HH I think you need the same wattage rating for both the 15 and -8 or -9 what ever you are using to turn the fets off with.
I found a 9v supply to be easy to find so I chose 15 and 9 which will give me 15 volts on and -9 off. But they both need to be rated the same because the swing is always up to the 15v then down to the -9 it does not go to 0 as a resting point... So its the 15v DC-DC pulling it from -9 to 15 which is up by 24v then its from 15v pulled down to -9v which Is done by the -9 DC-DC so again a 24 volt swing. So they will have an equal load will they not?
Or does the gate constantly consume energy in a igbt unlike a mosfet????
Either way.
I choose 3w for both because it was the only way to get the other numbers I was looking for.

 

[7circle]

Are you refering to ...

so long as you use transistors (not mosfet), have sufficient current/voltage rating and the pair are complementary type, then you're good. actual part number of other reference designs no too important, so feel free to shop around. just meet the above requirements.

The 333J is not compatible to drive a Mosfet booster power power stage (P-FET HI / N-FET LO).
- you would need to invert the Vout before the booster,
and all this will add more parts with rise/fall time delays and cost.
- And 24V on the gate of a N-FET is not good.
To get a booster stage with high side n-fet to pull the igbt gate to 15V it will need say an extra 5v to go bewteen fet gate to source.
Your back to high side driving issues.

But you can build a driver that has an inverted output that can drive a p-fet / n-fet.
But also the uvlo and desat and clamp... need to be added if you ready for the challenge.
So I wonder if a FET booster stage would be faster.



Why are transistors correct?
- A NPN-hi / PNP-lo booster stage output "logic" will follow the input.
So it is simple to implement.

If you design goal is for 2000ns rise and fall times then the BJT stage push/pulling 3ohm gate is looking just in par.

 

[HighHopes]

No mosfet for boost stage. No time to explain today.
Timing of boost transistor only impacts your deadtime. See my posts on how to calculate that. I think you will be fine, especially if u switch <20khz.

 

[Arlo1]

Ok so with lots of Digging I found something complementary. http://www.onsemi.com/pub_link/Collateral/MJD2955-D.PDF
This is about all I can find in the surface mount with 10 amps or more and 60v or more rating. http://www.digikey.ca/scripts/DkSearch/dksus.dll?Detail&itemSeq=159301689&uq=635490596600026525&CSRT=2420770433680999052 Digikey number MJD2955T4GOSCT-ND
and http://www.digikey.ca/scripts/DkSearch/dksus.dll?Detail&itemSeq=159302894&uq=635490596600026525&CSRT=2420770433680999052 Digikey # MJD3055T4GOSCT-ND

 

[HighHopes]

No mosfet for boost stage

7circle basically explained it, the inverting circuit. also, the cost tends to be higher and the losses will be higher in mosfets in boost configuration as comapred to transsistor. there's really no reason why one would pick mosfet boost stage over transistor.

 

[7circle]

For 20kHz PWM and acceptable fast IGBT switching to minimize losses.

What if you want 30kHz or Higher.
How do you decide if the transistors are slowing the driver stage from pushing the gate as fast as needed.

20kHz is Arlin's target so best to stay focused on this.

The 2955/3055 look like there just in spec before they turn to hot jelly. [edit: actually i don't think they will be hot ... so bad metaphore... i prefer a rubber mallet hammering a nail]

 

[Arlo1]

For 20kHz PWM and acceptable fast IGBT switching to minimize losses.

What if you want 30kHz or Higher.
How do you decide if the transistors are slowing the driver stage from pushing the gate as fast as needed.

20kHz is Arlin's target so best to stay focused on this.

The 2955/3055 look like there just in spec before they turn to hot jelly.

I don't need >20 kHz the motor is ~ 830 uH and I will rear measure it.
The only reason for even going close to 20khz will be to make it quiet. Which I'm also not to concerned about.
Remember I come from Nitro powered Top fuel bikes/cars and every gas engine I've had has been loud! A little wine from a electric motor will not be a big deal.
Oh and Lebowski uses Center aligned PWM which feeds the power in on each end of the switching event so its kind like having 2x the PWM frequency.

 

[Arlo1]

Just going over things and putting the order together for all the components and I found the pull down resistors were attached to the common instead of the -9v so I changed that. Now I'm trying to remember what R5 and R6 are for? They attach base to emitter for the clamp transistors. I need to figure out their purpose so I can select a value to add to the order.

 

[7circle]

Oh and Lebowski uses Center aligned PWM which feeds the power in on each end of the switching event so its kind like having 2x the PWM frequency.

Ahh So 20kHz for center aligned setting in lebowski controller acts like 10kHz normal blocked pwm.

      |-------------|              |----------|
------|             |--------------|          |--------
^            ^              ^             ^
<---- n -----><----n+1------><------n+2---><------n+3-->
<<==========blockA=========>><<========blockB=========>>

Looked section 50 of dspic manual - see figure 50-4
iff your real bord.. and want to geek out : http://ww1.microchip.com/downloads/en/DeviceDoc/DS-70579a.pdf

Would be worth understanding what the minimum effective pulse would be.
It might be as long as 5,000nSec but compared to period of 20khz of 50,000nS
Thats a 10% min duty.
But if the effective center aligned PWM is 10Khz then your at 5% minimum block.

But Lebowski may have defined the pwm freq as the block frequency.

The 333J figure 41 has the clamp transistor with R3 to add a stronger base emmiter voltage that will make more current through the collector.
Like R2 in that circuit.

Highhopes was stern on the issue of the gate pull down resistor. "needs a res gate to emitter"

It's there if SHTF. to make sure even if the neg supply is lost the gate will turn off.
If rails and PNP are okay it's not needed. So safer ..to me .. to go directly across gate-emitter. ..
They show the pull down before the Rg. but you have Rg-on and Rg-off.

How did you decide to place it nd where?

 

[Arlo1]

Oh and Lebowski uses Center aligned PWM which feeds the power in on each end of the switching event so its kind like having 2x the PWM frequency.

Ahh So 20kHz for center aligned setting in lebowski controller acts like 10kHz normal blocked pwm.

      |-------------|              |----------|
------|             |--------------|          |--------
^            ^              ^             ^
<---- n -----><----n+1------><------n+2---><------n+3-->
<<==========blockA=========>><<========blockB=========>>

Looked section 50 of dspic manual - see figure 50-4
iff your real bord.. and want to geek out : http://ww1.microchip.com/downloads/en/DeviceDoc/DS-70579a.pdf

Would be worth understanding what the minimum effective pulse would be.
It might be as long as 5,000nSec but compared to period of 20khz of 50,000nS
Thats a 10% min duty.
But if the effective center aligned PWM is 10Khz then your at 5% minimum block.

But Lebowski may have defined the pwm freq as the block frequency

No 20khz will seam like 40khz. The upper and lower both pwm one will be on longer then the other meaning power flows during the start and end of the pwm cycle. I'm comparing this to old controllers that either just pwm the high side or the low side but not both. I'm on my phone but when I get to a computer I will post links.

 

[7circle]

Here's Figure 50-10 _ Dead-Time Insertion in Center-Aligned Mode

The period is doubled.
Causing the freq to be halved.

When you look at the phase current ripple on a scope all will be clear.
That's what makes the high pitch whine. .. mainly

 

[Arlo1]

I don't think you are understanding what I am saying. I will post pictures later. My phone will not open your picture.
But here is a quick explanation. The low side on say phase A turns on then the high side on say phase C turns on then Phase A turns off for a while then back on then phase C turns off then phase A tuns off. That is 1 pwm cycle which happens 20,000 times a second at 20khz but it only feeds power in when both phases are on which happened 2 times in the cycle making it seem like 40khz.

 

[HighHopes]

some quick comments:

<20kHz A little wine

you should open your ipad and download the hearing test app.. then test what a frequency of 15kHz sounds like. turn up the volume. you'll go insane! its not like the turbo bikes you're used to that are loud, loud is one thing.. high frequency whine, even quiet, will drive you crazy. you would rather run a dagger through your eye than sit in a car that has that... i'm just saying.

I found the pull down resistors were attached to the common instead of the -9v so I changed that

it was correct as it was. pull down resistor is there just incase you lose gate drive power (or during startup when gate drive power is not available but DC bus is 400V). so it would be wrong to connect it to -9V.. it must be IGBT gate/emitter.

what the minimum effective pulse would be

for me, the minimum pulse is based on the propegation delays of the gatedriver + IGBT. you want the IGBT to turn FULLY ON. if the PWM pulse is too short to allow this to happen, then that pulse should be ignored. its usually a problem easily solved in software. i don't know if Lewbowski does this or not, but its an easy thing for him to add. and while we're on this subject, the gate drive IC itself should have short pulse rejection feature built in (or you add it with discretes) so that any noise does not cause small 200ns turn ON.

 

[Arlo1]

The motor is sealed with a liquid cooled housing so it will contain a lot of noise. I am running my zero with the motor cut wide open for cooling and I hear some noise but its not likely the pwm and its there but not bad at 400phase amps.
I'm not a apple lemming so no iPad but I'll try what you said with my android powered smart phone.

 

[zombiess]

7circle,

Arlo1 is referring to the unipolar PWM method used in the cheap chinese controllers vs the methods seen in space vector FOC controllers.

The cheap China controllers even switch to full block commutation once they reach full speed and drop all PWM (hence why low inductance motors are known to kill them in short order).

 

[zombiess]

Arlo1, I missed it somewhere in the conversation, but why don't you want to use the TO-220 based D44VH10/ D45VH10 transistors?

Below are two scope shots I took verifying my boost stage on my TD350E gate driver can actually deliver the claimed 15A it was designed for. The purple trace (math function) is the current in amps. You can see it delivered 14.4A in 570nS into a 1.0uF capacitor acting as my MOSFET stand in. Not sure what a 1.0uF translates to in nC. While this is not a test on an actual MOSFET that has Cgd and Cgs creating a Miller effect it does give some idea of the boost stage transistors performance. At lower current values the dI/dT increased and the over all current dropped down a bit, I think into the 12-13A range at 0.5uF hard to recall. The boost stage was fed by two 10uF XR7 caps in parallel. Gate drive power supply was 15V, 1W, so max of 68mA. It wasn't very happy supplying this current level at 19khz so I just did some single pulse tests. Only reason I have these scope shots is I decided to play around on the bench and verify some things + get some quality learning time in. That hour was VERY educational. Driver IC is a TD350E.

purple line is Amps





Here is a picture of one of the gate driver setups that was used in the test. As you can probably guess, 3 parallel IRFP4568 MOSFETs don't need a boost stage to switch at 20khz, but I included them in my design anyways so I can verify the driver design real world.

 

[Arlo1]

Because I'm trying to stick with surface mount the plan was never ever to go even near 20khz but have the system capable of it if needed. The igbts we not selected to go above that. I've run colossus at 14khz and it was not that bad I could just start to hear it. 10amp peak should be OK from the boost Transistors

 

[Arlo1]

some quick comments:

<20kHz A little wine

you should open your ipad and download the hearing test app.. then test what a frequency of 15kHz sounds like. turn up the volume. you'll go insane! its not like the turbo bikes you're used to that are loud, loud is one thing.. high frequency whine, even quiet, will drive you crazy. you would rather run a dagger through your eye than sit in a car that has that... i'm just saying.

LMFAO.

Yes so I played around with the tests both on you tube and my phone I found two tests that show I here ~16-16.5 kHz
Now. What I am pointing out is with center aligned pwm it would sound like 30khz when you have selected 15khz in the PWM menu and as far as the IGBTs are concerned they are switching at 15khz but the power is fed into the motor 2 times a PWM event. I will find the pictures I have.
I can also say I run Colossus which is an open stator out runner and you can hear the windings not the PWM but the electrical RPM and at 14 kHz I heard it a bit but not like the video on you tube showing 14 kHz, it sounded like a lower frequency so it was likely something else I heard. Non the less that was an open stator 2 feet from my ears with no hearing protection on and I did not here any sound from the PWM the leaf motor will be inside the engine bay and the stator is completely sealed and then the housing is water cooled so sound will not exit the motor easily. Let alone make it into the cab.

 

[Arlo1]

The 333J figure 41 has the clamp transistor with R3 to add a stronger base emmiter voltage that will make more current through the collector.
Like R2 in that circuit.

Ok so a value of 47 ohms like R2??

 

[Arlo1]

Here is Center aligned PWM Probe one is phase A - to the A phase leg and Probe 2 is phase B - to B phase leg.

See how the power flows though the motor at the start and the end of the PWM event but not the middle because the phases are both in the Hi state.

 

[Arlo1]

I'm still kind of stuck with the Zener thing. I found this one but its 150mw rated.

Its a bit tricky to wrap your head around the MW rating of the Zener and thinking if I need more then the 250uA or less..... in terms of Zener ratings.
http://www.digikey.ca/scripts/DkSearch/dksus.dll?Detail&itemSeq=159086802&uq=635491906062413063&CSRT=2420770433680999052
Digikey # 641-1021-1-ND

 

[HighHopes]

what is power dissipation of your desat zener, assuming you choise 2.5V zener? P = V * I = 2.5 * 250uA = 625uW. So pick a zener that is rated at least 1.2mW

easy!

only hard part is to find a zener that reaaches the zener voltage with 250uA applied to it. read the datasheet. ignore ones that are ~1mA (common) and look for more hard to find 100 to 300uA ones.