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

[Arlo1]

Im just reading the paper Njay linked the SN74LVC2G74DCTR is a flip flop and uses the gate signal and the signal from the comparator. Im trying to understand it as I might have to read it a few times but the paper says

during the current fall time. This is caused by the induced
voltages at the stray inductances in the commutation path. To
avoid a failure command to the driving stage, the flip-flop
disables the fault-signal during the blocking time and turn-off
process of the power MOSFET.

 

[Lebowski]

basically what it says that switching causes a lot of spikes, and to make sure these don't trigger a false FAULT signal the FAULT detection is disabled for a short time

 

[HighHopes]

don't trust what you read. there are IGBTs that are rated for high switching frequency (usually 20khz max) and are in my opinion a much better option than an equivalent power based mosfet module for bus voltgaes >200V because they actually switch slower (nice for noise) are more rugged (short-circuit rated) and fail open not closed .. all good things for a motor drive. best IGBT modules for your application are made by Eupec (now infineon?) and Fuji.

i don't know what document Njay is refering to, but i would be extremely apprehensive about putting a SN74LVC2G74DCTR chip in a gate driver with application of high power high performance. there is so much noise in this type of gate driver that sensitive logic chips like that will struggle to operate. you have to be very careful here. i tried it oncce and it was crap, so i put a 5V CPLD chip to do some logic functions instead and only that was reliable enough and believe me it was a lot harder to do correcly than this simple sentence implies because layout was super special and power supply had a lot of decoupling so it is more than just "use a CPLD". also, in that app-note they use comparator LMV7219 but i think this comparator is not robust enough. logic chips & LMV comparator might work (as it did in their paper) with labratory conditions, but on the road in a high power high performance drive for 20 years, not so much. But feel free to prove me wrong here. i only tried it once, maybe i had a layout mistake or something.. my gut feeling says that is the concept but you need different implementation for robust operation, but you could try putting it exactly as shown in the app-note and prove me wrong. i am happy to be wrong.

arlo, your march 3 schematic, i'm sure its just a draft at this stage, but it is far far away from something you will be happy with. are you extremely sensitive to cost or board space or something?
you have the transformers drawn very strangely, is that your power supply? you should draw them as a DC/DC converter with output of +12V/0/-5V three terminal output and then wire that output into your gate drive circuit. that's just convention, but it helps clarify for the reader. if an intern student showed me a schematic like that i would say "re-draw it" as there is no way for me to easily see that you would actually get +12V & -5V at output of your two transformers because it is totally confusing how they are wired and i don't want to spend the time to follow it along.
try to draw all your schematics so the flow is from left to right, that also makes reading easier. keep the lines as straight as you can and don't be afraid to space it out a bit for clarity. the circuit may look easy to follow to you but that is because you drew it, the rest of us are trying to read it for the first time and the signal flow & connections are not as obvious as it should be.

 

[Arlo1]

don't trust what you read. there are IGBTs that are rated for high switching frequency (usually 20khz max) and are in my opinion a much better option than an equivalent power based mosfet module for bus voltgaes >200V because they actually switch slower (nice for noise) are more rugged (short-circuit rated) and fail open not closed .. all good things for a motor drive. best IGBT modules for your application are made by Eupec (now infineon?) and Fuji.

Yes so to be clear the igbts will be for a separate project
2 projects on the go. 1 is low voltage with paralleled mosfets.
Maybe the same driver will work with some small changes or maybe I will just use your driver.
and by the way the leaf motor has 240 uH inductance so it will not need fast switching.

i don't know what document Njay is refering to, but i would be extremely apprehensive about putting a SN74LVC2G74DCTR chip in a gate driver with application of high power high performance. there is so much noise in this type of gate driver that sensitive logic chips like that will struggle to operate. you have to be very careful here. i tried it oncce and it was crap, so i put a 5V CPLD chip to do some logic functions instead and only that was reliable enough and believe me it was a lot harder to do correcly than this simple sentence implies because layout was super special and power supply had a lot of decoupling so it is more than just "use a CPLD". also, in that app-note they use comparator LMV7219 but i think this comparator is not robust enough. logic chips & LMV comparator might work (as it did in their paper) with labratory conditions, but on the road in a high power high performance drive for 20 years, not so much. But feel free to prove me wrong here. i only tried it once, maybe i had a layout mistake or something.. my gut feeling says that is the concept but you need different implementation for robust operation, but you could try putting it exactly as shown in the app-note and prove me wrong. i am happy to be wrong.

Ok scrap that good to know.

arlo, your march 3 schematic, i'm sure its just a draft at this stage, but it is far far away from something you will be happy with. are you extremely sensitive to cost or board space or something?
you have the transformers drawn very strangely, is that your power supply? you should draw them as a DC/DC converter with output of +12V/0/-5V three terminal output and then wire that output into your gate drive circuit. that's just convention, but it helps clarify for the reader. if an intern student showed me a schematic like that i would say "re-draw it" as there is no way for me to easily see that you would actually get +12V & -5V at output of your two transformers because it is totally confusing how they are wired and i don't want to spend the time to follow it along.
try to draw all your schematics so the flow is from left to right, that also makes reading easier. keep the lines as straight as you can and don't be afraid to space it out a bit for clarity. the circuit may look easy to follow to you but that is because you drew it, the rest of us are trying to read it for the first time and the signal flow & connections are not as obvious as it should be.

Yes its just a start waiting for imput... There is no point designing the whole thing and getting told to redraw it... So. Anyways Im trying to figure out how to add comports to the Kicad Library and up date it.
Thanks for the input.

 

[HighHopes]

i haven't read messages on your thread for couple of weeks so i have a gap.. but here are my thoughts. sounds like you have 2 project like you said, one is low power parallel mosfets with high frequency switching (>20kHz), and the other is high power IGBT module with 20kHz switching.

for high frequency switching i would use IXYS gate drive chip like you show. for high power IGBT module i would use my gate driver over in Ivan's forumn.

for your IXYS gate driver, high frequency lower power MOSFET, focus next on power supply and clean that up. two transformers like that is going to take a lot of board space and add some headache. i would move to pre-made DC/DC converters like zombiess design or see my gate drive.. but i would look at the product brochure to find one that has 3 terminal output (+15V, -5V) and just use that. you should calculate what your power requirements are so you know watt rating of power supply (probably going to be 3W rated), calculate peak current of gate drive so you pick correct IXYS chip (probably 6amp rated). then add power supply to your schematic in legible manor and post for discussion. then move on to the other functions.. design one at a time.

 

[Arlo1]

i haven't read messages on your thread for couple of weeks so i have a gap.. but here are my thoughts. sounds like you have 2 project like you said, one is low power parallel mosfets with high frequency switching (>20kHz), and the other is high power IGBT module with 20kHz switching.

for high frequency switching i would use IXYS gate drive chip like you show. for high power IGBT module i would use my gate driver over in Ivan's forumn.

for your IXYS gate driver, high frequency lower power MOSFET, focus next on power supply and clean that up. two transformers like that is going to take a lot of board space and add some headache. i would move to pre-made DC/DC converters like zombiess design or see my gate drive.. but i would look at the product brochure to find one that has 3 terminal output (+15V, -5V) and just use that. you should calculate what your power requirements are so you know watt rating of power supply (probably going to be 3W rated), calculate peak current of gate drive so you pick correct IXYS chip (probably 6amp rated). then add power supply to your schematic in legible manor and post for discussion. then move on to the other functions.. design one at a time.

1 The transformers represent the DC/DC supplies. I have some coming from aliexpress but I would love to find some +12v and -5v supplies but I can't seem to find them anywhere everything seems to be a = number for both + and -
2 I don't know how to calculate the wattage requirements for the DC/DC
and 3 I think its ok to use the 9 amp peak drivers I have isn't it? I would rather have a bit of head room with something that can switch fast if needed right?

 

[zombiess]

1 The transformers represent the DC/DC supplies. I have some coming from aliexpress but I would love to find some +12v and -5v supplies but I can't seem to find them anywhere everything seems to be a = number for both + and -
2 I don't know how to calculate the wattage requirements for the DC/DC
and 3 I think its ok to use the 9 amp peak drivers I have isn't it? I would rather have a bit of head room with something that can switch fast if needed right?

Here is a quick spreadsheet I put together for estimating power supply requirements. This is what I used to figure out mine. Even 4 parallel IRFP4568's only require about 20mA of supply current @ 21khz which translates into 286mW @ 15v Vgs. The isolated supply charges a cap and the peak currents are pulled from this cap. The average current demand on the isolated DC-DC is usually pretty low, but I am also only driving 4 IRFP4115's which have a low Qg.

View attachment isolated power supply calcs.zip

My own gate isolated supplies have been measured at 20mA draw from the 24V supply when running at 19khz and. That should translate to about 15V @ 30mA into the gate drive capacitor and that was with the boost stage, they are rated for a max of 63mA output @ 15v and are 1W. I have not checked it since I removed it. My 2W 15V supply is left with about 50% overhead. Without the 5V regulator and some of the other things in my driver the current would be quite a bit less. At idle, no PWM it's drawing 15mA from the 24V supply which means the 15V output is about 23mA. With the boost stage I should be hitting peak currents of > 8A, transistors limit it to 15A peak, but with only a single 10uF XR7 cap, I'm not sure how high it can really go on the peaks. That's something I want to check. New design uses 2 10uF XR7 caps in parallel for the boost stage to lower the ESR.

 

[HighHopes]

the gate drive capacitor (or boost capacitors) should be 10x uF rating of the total worse case power switch input capacitance.

here's why:
when the power switch (say 4 parallel mosfets for example) need their effective input capacitance charged up so that they will turn ON where does this charge current come from? the faster the charge current arrives the better it is for the mosfet, so we want a capacitor to deliver this power (fast), we do not NOT want this peak current to come through the DC/DC converter (too slow). so that is the main purpose of the gate drive cap (not the power supply cap, two different caps). since this current has to come fast that means high peak and that means the dielectric must be of the type to easily handle high peak which means should be ceramic, preferably surface mount, gull-wing or C-terminal package (gull is easier to hand solder). now what uF value should this cap be? as it depletes itself to charge up the parallel mosfets, it is bad if the voltage drops too far... so, rule of thumb, the gate drive uF cap value should be 10x greater than the total worse case power switch input capacitance multiplied by 4 for parallel (be careful that sometimes the mosfet's datasheet gives total input cap in units other than micro farad in which case you need to translate accurately into uF unit).

 

[HighHopes]

arlo, define your application first (what are you tryinng to design, what are the key features it needs, what are the critical design objective for that application), then select power bridge operating conditions of inverter (so bus voltage, current, switching frequency, cooling), then pick the power switch (mosfet/IGBT discrete, module) to suit, then pick the gate driver IC, then power supply to that gate driver, then add gate driver functions as necessary. in that order. as you go in order it will limit the choices you have for the next thing to design.

so you might find, for, a full up electric car is application, key features are, reliability & low cost, critical function, 3-phase VSI topology, >20kHz switching frequency, 30kW capable. if this were a product spec you would also have discussion of radiated & conducted emissions. perhaps you have special requirement too, 400A for 10 seconds....

power switch, depends on your battery a lot... lets say you have 400V full charge battery pack. perhaps a 600V in a good layout using a 6-pack module would work. what is the next limiting design requirement? probably 400A for 10s.. so pick a current rated module that can handle that and can also switch >20kHz. does one exist? hope so.

let's say you end up with 600V, 600A, 6-pack module with timing parameters that imply it is suitable for 20kHz switching. what peak current do you need to drive this? probably 16Apk. so your gate drive IC is not good enough, you need a different one. what switching frequency, 20kHz? then that defines how much delay you can have in your circuit, <1us is tolerable. so now you know the timing parameters.

what is the average power to drive this module at 20kHz? probably you will want dual voltage at this power range too, so maybe +15V ON, -10V oFF (for IGBT). now you knkow your power supply.

and so on. as you go through the first pass you will end up with an archetecture at the end. then you go through again but to lower level of detail and start adding things like gate drive cap, turn ON/OFF resistor values, etc.

 

[HighHopes]

for dual voltage power supply, i think you will be far better off using two isolated DC/DC converters tied together. one is +15V the other is +5V connected backwards to get -5V (or if they make also -5V directly that would be good too). it is best if there is single supply with dual output though, so check manufacturer website to see. power supplies can be unregulated if your input to this gate drive DC/DC converter IS regulated (preferred), otherwise they should be regulated type. not all DC/DC converters were made to work well in gate driver application so read the datasheet and look for indication that it will be good. CUI comes to mind, maybe Remco and whoever made zombiess power supply (i forget).

 

[zombiess]

My supplies are cheap but have datsheets. Ones in my current working design are mornsun and the new batch I just ordered are made by Nengda. I think Nengda will custom make dc-dc iso supplies if you order 100 units. I know this isn't realistic for you, I am just tossing it out there that some of these manufacturers will take on small jobs unlike big companies which have MOQs that very few small businesses could ever afford.

I do not know how good the Nengda supplies are, but they appear to have decent specs (same as the mornsun) and are cheap but have a MOQ of 10 per type. With so much manufacturing in China and relabeling / copying going on to different levels of quality its a gamble. My order was delayed a bit since they had to manufacture a new run of one of the 1W types I ordered. I think it will show up today.

I am always a bit leary on ordering stuff so I try to research it as much as possible. I saw many vendors selling ND brand dc-dc converters but no specs so I used some search skills and found the name of the manufacturer, spec sheets and then their online store. My order is < $100 for 20 dc-dc converters so its not that big of a gamble. There were cheaper sellers of the same product, but I figured going with the manufacturer was better. Often time you can get some technical questions answered by an engineer if you know how to dumb down the question properly. It almost always needs to be translated by a non tech to a technical person so it can get tricky. My time spent in hong Kong working with local suppliers taught me a lot.

Most aliexpress vendors and manufactures have been easy to work with in the transactions I have made so far, at least on the electronics side.

It makes me sad to know that at one point there were many small USA based companies who use to do the same but are now pretty much extinct.

 

[Arlo1]

Ok so the application on this thread is the same as always. It will be a powerstage of 4 t0247 or 4 t0 264 package FETs. The FETs I have sitting here are ixfk230n20t and irfp4568 yes they are different but the same driver stage should be able run the both.

 

[HighHopes]

do you plan to make this design schematic & layout open source?

 

[Arlo1]

do you plan to make this design schematic & layout open source?

Yes I don't know if I need to register anything or just post it as I go then when I have a working model start a how to step by step thread.

 

[HighHopes]

next is to pick a minimum bus voltage and calculate what phase current max you think you will get and post it so we can discuss if it is realistic to expect or not. the more detail you can write in how you came up with this phase current value the more we will have to talk about.

 

[HighHopes]

by the way, here is an example of a gate driver power supply that can have dual voltage. this would be more suitable to IGBT.. though could work fine for mosfet. it's $9 and you need 6...

see digitkey part number RKZ-051509D

 

[Arlo1]

next is to pick a minimum bus voltage and calculate what phase current max you think you will get and post it so we can discuss if it is realistic to expect or not. the more detail you can write in how you came up with this phase current value the more we will have to talk about.

The plan will be for 28s lipo but maybe more in the future if things went well. 117v fully charged and 103.6v would be the nominal for 28s which would be the max for the irfp4568 and the plan would be for 350 phase amps for a 10-30 second rating. I have already had success at this level with only 3 fets in parallel !
But the ixfk230n20t could be run at a higher voltage if I wanted and I was originally thinking to go as high as 40s which is 168v fully charged and 148 nominal.
This would likely be able to run a touch more phase amps then the 4568 but not a lot because the package would still limit the amps you can flow though the legs. So maybe 360-370 for a 10-30 second rating.

 

[zombiess]

What is your continuous rating goal?

HighHopes, for the DC to DC could arlo1 get a dual output +/- 12 to 15v 2-3w unit and then use a linear regulator to get the desired -v? Not as ideal as an all in one dc-dc and certainly less efficient, but pretty easy to accomplish and cheap too. Would this create some unknown pitfall? This is how I myself would most likely try to accomplish his goal. It does have the added benefit of being able to change the neg voltage level while testing to get things optimum.

 

[HighHopes]

i know what you mean with the linear regulator.. but i wouldn't do it. you need the +15V (or +12V) and -5V to come direct from DC/DC converter.. not through a linear regulator. i do not have a reference for that statement, it was something that was learned from a guru.

you know with these gate drivers, it performs a critical function and the margin of error gets more narrow as you go up in power .. so you tend to stick with what you know works rather than experiment with new techniques (unless you have specific reason).

 

[zombiess]

That's interesting. I wonder what the reason is to avoid a linear regulator (failure mode?). You mentioned 2 back to back isolated DC-DC converter is OK though, correct?

 

[Arlo1]

That's interesting. I wonder what the reason is to avoid a linear regulator (failure mode?). You mentioned 2 back to back isolated DC-DC converter is OK though, correct?

How does a linear reduce the voltage? Does it not waste some of it into heat... Maybe you loose to much wattage when placing a linear after the DC/DC

 

[zombiess]

That's interesting. I wonder what the reason is to avoid a linear regulator (failure mode?). You mentioned 2 back to back isolated DC-DC converter is OK though, correct?

How does a linear reduce the voltage? Does it not waste some of it into heat... Maybe you loose to much wattage when placing a linear after the DC/DC

It's not too bad, I have six in my controller to get a +5v buss for my optos and the input of the TD350 which only works up to 5V on the input signal, but my controller is only low - medium power. It doesn't even get warm and it's a TO-92 style.

The reasoning must be different under high power, a lot changes and things have to be controlled a lot better.

If you are going to take on the challenge of trying to design this, you are going to need to brush up on you KiCad skills and create custom components and learn to draw better circuit diagrams. Mine use to look sloppy and were hard to read by others, but now they are a bit better but could still use improvement, I tend to jam too much stuff close together since the components are so dang large in KiCad. Making the custom components is easy, I have a large library of stuff I've made for schematic and foot prints. There are additional libraries available online as well, just got one a few weeks ago for a SMD foot print I needed.

 

[HighHopes]

perhaps this gate driver example is exactly what you need, part numbr ACPL-32JT by Avago
low propegation delay for high switching frequency capable
optional boost stage for your leaf inverter (see zombiess gate drive for good boost transistors selection)
completely isolated signal, fault & power supply
fault protection built in
bi-polar power supply built in

what's not to like?

see figure 8-1
http://www.avagotech.com/docs/AV02-4412EN

 

[Arlo1]

perhaps this gate driver example is exactly what you need, part numbr ACPL-32JT by Avago
low propegation delay for high switching frequency capable
optional boost stage for your leaf inverter (see zombiess gate drive for good boost transistors selection)
completely isolated signal, fault & power supply
fault protection built in
bi-polar power supply built in

what's not to like?

see figure 8-1
http://www.avagotech.com/docs/AV02-4412EN

That's awesome IM going to spend what's left of my night studying it.

 

[Arlo1]

First question. I don't see a negative turn off. Is this going to be a problem with the higher voltage if I decide to use this for the leaf inverter.