VESC based +200A powerstage with 200V MOSFETs (videos of it in action)

[squeegee]

Depending on price I'm up for one or two.
BTW what do you figure max battery voltage would be, ie what margin should I plan for overshoot and such.

 

[squeegee]

Oh and what size do you think it will be?

 

[bww129]

I've been thinking about this on and off and after a little digging, found a thermal conductivity figure for Kapton of .12W/m*K.

Apparently not all types of Kapton are created equal when it comes to thermal conductivity.

https://www.dupont.com/products/kapton-mt-plus.html
Dupont's Kapton MT+ specifically made for high thermal conductivity is 0.8 W/mK, which is 6.7 times higher than 0.12 W/mK figure you may have found here http://www.mit.edu/~6.777/matprops/polyimide.htm

https://www.dupont.com/products/kapton-mt.html
Dupont's Kapton MT is 0.46 W/mK.

https://www.dupont.com/content/dam/dupont/amer/us/en/products/ei-transformation/documents/2018-12-11-A218102_Report_Fraunhofer.pdf
Dupont's Kapton FWN targeted at magnet wire insulation is around 0.19 W/mK.

https://www.tapesolutions.saint-gobain.com/sites/imdf.foamsandtapes.com/files/documents/ThermaCool-K271-Tape-TDS-1358.pdf
This interesting silicone-coated Kapton tape (1 mil thick Kapton, 1.5 mils acrylic adhesive, 2 mils silicone rubber) is specified at 0.60 W/mK.

 

[squeegee]

Was buying guitar pots on Mouser and had a look at IRF200P222 outa curiosity.
FWIW, I did see two IXYS parts, one "new" with 220A/6mhoms and another with 300A/4mohms but higher gate charge...oh and 3x $

 

[zombiess]

Was buying guitar pots on Mouser and had a look at IRF200P222 outa curiosity.
FWIW, I did see two IXYS parts, one "new" with 220A/6mhoms and another with 300A/4mohms but higher gate charge...oh and 3x $

This project has been on hold for the last month+ due to another project with the possibility of pay taking precedence, so no updates yet.

The IXYS parts are nice, incredibly expensive and in low supply. The IRF200P222 is nice, lower cost and plenty of supply.

 

[ElectronS]

cool project ... I have 2 questions :

1-what is the ticking sound in the motor testing video ( thermal test ) . is it moving the motor then shutting down then powering again ???

2-When you solve the logic (reset) problem will you make the schematics pdf open source , I am really interesting in this 3 stack with all logic board seems fun

cheeers

 

[zombiess]

cool project ... I have 2 questions :

1-what is the ticking sound in the motor testing video ( thermal test ) . is it moving the motor then shutting down then powering again ???

2-When you solve the logic (reset) problem will you make the schematics pdf open source , I am really interesting in this 3 stack with all logic board seems fun

cheeers

1. It's cogging the motor because I have the opposite motors phases shorted. It takes an very large amount of current to overcome the resistance on a motor this size (~40HP). It also proves that the resolver is tracking the rotor really well as this is sensorless.

2. Not sure where I'm going to take this design yet, I do have the intention of selling power stages /gate drives at some point.

 

[Goonman]

I just stumbled across this from YouTube. Is there any progress on this controller?

 

[zombiess]

I just stumbled across this from YouTube. Is there any progress on this controller?

Yes, I have made some progress on design, but not in production. I have an updated design with manufacturing in mind if I ever get there. The new power stage pulls from what I learned from this one and has 18-24 TO-247's, improves thermal performance and takes up a smaller volume. I still need to prototype it but I need to get parts machined. I have an all new gate drive design as well which reduces cost and drops the required dead time from 850ns to 250ns.

I've cost optimized the VESC design and significantly reduced the size to 80x40mm. The only feature I removed was individual error read outs. It now has one hardware fault indicator in addition to the VESCs software based error detection, however it still retains hardware over current, hardware over voltage detection and op amp based filters in addition to the shoot through protection offered by the gate drives. I added on an isolated DC-DC which runs from 20-30V input. I can shrink the design a bit more if I don't use the 2mm headers, but I want it to be modular during prototyping and testing.





I had to develop a high voltage input SMPS which is compact, efficient and low noise to power everything. Following the modular approach I made this into something which can stand on its own. It works up to 140V DC input with an output of 10-30V, fixed or adjustable at a max of 1.5A (25C room, passive 40x40x11mm heat sink) with 90% efficiency at 24V 1.5A out. It draws < 20uA when powered off which means it won't kill your battery pack like many other supplies can. It has a remote turn on feature which allows it to be enabled with any input voltage from 1.5V-140V so there is no arcing within switch contacts when powering the system up. This supply design was surprisingly tricky but after a several prototypes I have it sorted out.

I also have a version which can do 5A with a small fan and output the same variable range with efficiency of 92% at 24V (non isolated) 5A output with 140V input. Could be useful for powering accessories on PEVs like lighting.



Of course I still have the same problem I usually have, putting everything into an enclosure. Takes me a long time to figure out mechanical designs. I find electrical design easier than mechanical.

 

[mxlemming]

Very cool.

I'm especially interested in your SMPS. The limiting factor for me for a while has been cost effective robust dcdc step downs... The TI lm5xxx ones just blow at the drop of a hat, usually when i do something like short a motor phase (which the FETs aren't affected by).

They also limit the applicable voltage to essentially 20s.

What do you strap this thing to to ride on? I've seen videos of your tiny bike with huge hub motor but guessing you've got something more serious now?

I'm a mechanical engineer by degree and frankly I also find mechanical harder. It looks so easy but the difference between good and bad mechanical design is enormous and immediately visible to everyone, is wide open for opinions and 5x as expensive to prototype.

 

[zombiess]

I'm especially interested in your SMPS. The limiting factor for me for a while has been cost effective robust dcdc step downs... The TI lm5xxx ones just blow at the drop of a hat, usually when i do something like short a motor phase (which the FETs aren't affected by).

They also limit the applicable voltage to essentially 20s.

What do you strap this thing to to ride on? I've seen videos of your tiny bike with huge hub motor but guessing you've got something more serious now?

I'm a mechanical engineer by degree and frankly I also find mechanical harder. It looks so easy but the difference between good and bad mechanical design is enormous and immediately visible to everyone, is wide open for opinions and 5x as expensive to prototype.

I've found SMPS design more challenging than power stage design, power stages are big, dumb and pretty robust. There was a lot of failure involved in developing this SMPS, however I've seen on forums that's not an uncommon experience. Plenty of failed controller chips as they do not tolerate transients beyond spec. That said, I've run power dump tests, load steps, input transients, short circuits, varied input voltage and generally tried to break the design but it is now quite robust. Thermal management was difficult, but this will output 1.5A at 140V input in a 25C room with no airflow and not exceed 75C on the MOSFETs. Just a tiny bit of airflow and it cools off dramatically. The design is dissipating 4.5W of heat at 50W of output. Another issue was getting rid of the switching noise on the output. I have a worst case scenario of 20mV of ripple, but the switching noise was seeing peaks of 500mV with 140V input. Finally got that down to ~50mV. HF Switching noise loves to make it through the inter-winding capacitance of an inductor and ferrite beads didn't seem to do squat to reduce it. Improved layout and bypassing finally did the trick. Making the design cost effective was another challenge, this design runs about $22ea with a manufacturing quantity of 500, that's without an enclosure, but there is no off the shelf solution which meets the requirements.

Were your TI based designs isolated or non-isolated? I wouldn't be surprised if transients killed them from over voltage. I've nuked controller IC's internal gate drives just from the additional load of scope probes.

Here is a short FLIR video comparing one power supply with a heatsink and one without.
https://youtu.be/eLxXqv4K5VI

I don't have a running test vehicle any more, I just use motor dynos since they can provide significantly more load than a vehicle and it's easier to make measurements with them.

I'm an IT guy by trade (enterprise / security architect) with a high school diploma, but I've been doing electronic design for almost 35yrs. Obtaining good information in the pre-internet days was difficult.

 

[mxlemming]

You might consider selling the SMPS without the rest of the kit. You might not approve, but there are quite a lot of OTS controllers with 600V gate drives that could deliver much more power with higher voltage FETsbut the DCDCs would need bypassing... the new flipsky VESC for instance I suspect is like this.

My SMPS on the MESC and my big VESC version of that uses the TI LM5017 reference design buck, non isolated. I have been chasing cheap and small footprint; I like my hobbies to be sufficiently cheap that I don't care if it burns, and so far, only the SMPS has burnt. 22$ is a considerable BoM cost, pretty much the cost of my whole controller... but I need to step this part up, the LM5017s are far and away the weakest link now so more cost must be absorbed.

What's killing them... I'm really unsure. They have a 600 ohm ferrite and 2 ohm resistor and their own ceramic cap in front of the input so the transients would have to be quite long. I'll try adding a TVS but I'm not hopeful. I suspect there's something more subtle than a simple input over voltage, like your suggest the gate drive circuits die it's perhaps serious ground bounce or switch bounce wrecking the bootstrap circuit. Thing is, it's only happened to 3/10 in use so catching what's actually happening is hard.

Guessing that is a part from Analog devices controlling it? I have been thinking through this and come up with nothing remarkable; my thought has been going as far as using an STM32L0xx series chip to drive a 600V gate driver and pair of FETs. I can't see it being too hard to write the firmware for it... But managing the start up where there's no low voltage rail is tricky

set up ~200khz pwm
measure Vin
Measure Vout
Measure current
Set PWM duty cycle as a funciton of Vin Vout and a bit of feedback
Add feed forward term based on current on output...
Set PWM cycle skipping for low current usage...

Done?

 

[zombiess]

You might consider selling the SMPS without the rest of the kit.

22$ is a considerable BoM cost

Guessing that is a part from Analog devices controlling it? I have been thinking through this and come up with nothing remarkable; my thought has been going as far as using an STM32L0xx series chip to drive a 600V gate driver and pair of FETs. I can't see it being too hard to write the firmware for it.

set up ~200khz pwm
measure Vin
Measure Vout
Measure current
Set PWM duty cycle as a funciton of Vin Vout and a bit of feedback
Add feed forward term based on current on output...
Set PWM cycle skipping for low current usage...

Done?

I've considered selling them as a stand alone product. $22ea is the total manufacturing cost for me if I have them assembled,with wires soldered on, but no enclosure. If you need a low cost solution, you can use an isolated AC-DC converter and just feed it DC directly from the pack, many will work from 80V DC to 300VDC input.

I suspect your idea for a DIY approach is a bit more complicated than you might think, I'd advise against it unless it's something you really want to dig into. I don't doubt you can figure it out, but my experience with using off the shelf controllers like this is that there are always things you miss. Writing a custom controller is something I investigated, but I quickly noped out of that idea when I saw what I was up against. Lebowski has done some discontinuous mode SMPS controller design which doesn't look too bad to implement, but I don't know all the downsides.

Here are some issues I hit during design: What losses dominate the choice of a top side switch vs the low side switch? Got enough phase margin? How wide of an input/output range? What current? How do you handle low current operation, pulse skipping, forced continuous mode? Does the control loop handle step loads OK? Will it start up under full load or do you need to have the control loop stable prior to enabling the load? Need to change output caps to one of a different capacity / ESR, time to recalculate the loop again

The list goes on, those are just some of the ones I hit using a commercial LTC part. I have as many hours into this SMPS design as I have into an entire VESC, gate drive and power stage. Your experience may vary from mine, but this one was not easy. The other SMPS designs I've done were relatively easy and worked mostly as expected on the first go, but they were constrained to much smaller voltage in/out ranges where this one is more general purpose. I've done a few designs with LTC parts now and I'm a fan of their design software, so I'll pay the price they charge for the controller IC. I ended up killing about 20 ICs before I sorted out my reliability / assembly issues and from what I've read this is not an unusual experience. I got a lot of practice doing TSSOP package soldering, had to buy a 30x microscope to check my solder joints which looked OK through a magnifying glass but were often poor when viewed under the microscope.

 

[mxlemming]

I've considered selling them as a stand alone product. $22ea is the total manufacturing cost for me if I have them assembled,with wires soldered on, but no enclosure. If you need a low cost solution, you can use an isolated AC-DC converter and just feed it DC directly from the pack, many will work from 80V DC to 300VDC input.

I suspect your idea for a DIY approach is a bit more complicated than you might think, I'd advise against it unless it's something you really want to dig into. I don't doubt you can figure it out, but my experience with using off the shelf controllers like this is that there are always things you miss. Writing a custom controller is something I investigated, but I quickly noped out of that idea when I saw what I was up against. Lebowski has done some discontinuous mode SMPS controller design which doesn't look too bad to implement, but I don't know all the downsides.

Here are some issues I hit during design: What losses dominate the choice of a top side switch vs the low side switch? Got enough phase margin? How wide of an input/output range? What current? How do you handle low current operation, pulse skipping, forced continuous mode? Does the control loop handle step loads OK? Will it start up under full load or do you need to have the control loop stable prior to enabling the load? Need to change output caps to one of a different capacity / ESR, time to recalculate the loop again

The list goes on, those are just some of the ones I hit using a commercial LTC part. I have as many hours into this SMPS design as I have into an entire VESC, gate drive and power stage. Your experience may vary from mine, but this one was not easy. The other SMPS designs I've done were relatively easy and worked mostly as expected on the first go, but they were constrained to much smaller voltage in/out ranges where this one is more general purpose. I've done a few designs with LTC parts now and I'm a fan of their design software, so I'll pay the price they charge for the controller IC. I ended up killing about 20 ICs before I sorted out my reliability / assembly issues and from what I've read this is not an unusual experience. I got a lot of practice doing TSSOP package soldering, had to buy a 30x microscope to check my solder joints which looked OK through a magnifying glass but were often poor when viewed under the microscope.

I have no doubt the SMPS at 5x the frequency of a motor stage and vastly greater requirements for the output stability and noise are harder. Different problem set really; you're unlikely to get the gate drive issues, parasitic turn on etc but the noise... ugh...

This loop stability thing perplexes me. I would have thought these chips would be working as one shot control, cycle by cycle, so the concept of loop stability should be kind of moot.

I wonder if your noise problem would be well tackled like this:

The way I see it, your V+ and GND on the FETs results in about 1/3 of the board being part of the switching loop (green), so there will be inductive pickup on all the lines however well your ferrite works. If you reduce the high frequency switching loop to the absolute bare minimum, those spikes *should* virtually disappear.

You using 2 or 4 layer?

This isn't a project I will take on lightly... I already have too many projects - MESC, big VESC, Ebike, solder station, PT100 reader, battery pack and BMS, now get half the parts to start winding a custom motor... taking on another one is obviously very exciting, but perhaps I'll leave this one to you :lol:

Once I have "finished" the ones above, I want my next project to be a welding project. Going to play with big bits of aluminium.

 

[zombiess]

This loop stability thing perplexes me. I would have thought these chips would be working as one shot control, cycle by cycle, so the concept of loop stability should be kind of moot.

I wonder if your noise problem would be well tackled like this:
Reduce inductance.PNG
The way I see it, your V+ and GND on the FETs results in about 1/3 of the board being part of the switching loop (green), so there will be inductive pickup on all the lines however well your ferrite works. If you reduce the high frequency switching loop to the absolute bare minimum, those spikes *should* virtually disappear.

The waters are often deeper than they first appear. Power electronics work can be a very humbling experience.

My noise issue is not related to layout, but good guess. It's actually related to the inductors capacitance. I learned A LOT on this one project.

My layout decisions are related to mechanical considerations in addition to the electrical ones.

 

[TheRealJohnnyB]

Sent you a message @zombiess

 

[mxlemming]

https://endless-sphere.com/forums/viewtopic.php?f=14&t=114068&p=1747697#p1747697

Hey Zombiess, been a year for me to get round to finishing this one, but I finally got my DCDC with the stm32 and Infineon gate driver up and running! Cheers for the ideas/inspiration, Not going to push this one or attempt to sell it or anything, but there .

Realised 2 days ago I had posted it in the wrong channel, battery stuff not motors but hey.

1.2A continuous is OK up to about 70V, beyond that it starts feeling like it would enjoy a heatsink on the inductor, or perhaps a higher switching frequency.

Still needs some PI loop tuning and fiddling tbh but hey, I am calling it fine It's quite happy enough having the load slammed on and off and can startup into the 1.2A load , so good enough to use. I probably won't though, don't trust my code for this one.

Did not have much trouble with loop stability or zingy stuff, but startup was an absolute pain in the backside. HV regs didn't work and the diode to resistor ended up needed 10x lower resistance than anticipated. Most of my effort involved that challenge.

Note: If anyone thinks it is a good idea to build my one, I advise... Buy one from Zombiess or mouser or... something, he probably gave it 10x the consideration I have and it probably actually works whereas mine is entirely likely to put 90V onto your 12V rail at any moment because (for now at least) it is a hack job...