|Submited on :||Sat, 12th of Jan 2019 - 15:47:23 PM|
|Post ID :||af8i9l|
|Post Name :||t3_af8i9l|
|Post Type :||embed|
|Subreddit Type :||public|
|Subreddit ID :||t5_2qhue|
Awesome! what kind of machine are you going to build with it?
Perfectly legal 25km/h 250W ebike of course. https://www.youtube.com/watch?v=mw0MbFbwiJo
Fried the chinese driver that came in a kit so I decided to build one. It was only a matter of time to be honest. I had 3 chinese motor drivers (two 750W and one 5 kW) and they all died at some point. When I opened the 5 kW driver the design quality was quite poor. Where I used 4€ isolated gate driver, they used two small signal mosfets. Where I used 22 uF film they used 100 nF ceramics. And instead of using one big MOSFET they used four smaller ones ending with 36 MOSFETs total to switch 80A current (I have 12 for 200A).
this is awesome!!
what is that thick bar exposed trace (?) i think below a bank of unpopulated caps
What a nice work you did here ! Are you working in electronics or is it just a hobby ?
Both. Studying electronics and working at the same time. This driver is just a hobby project though.
Nice ! I'm doing the same, keep working as you are it's awesome
I wish I was where you were. It just all fell away when A level Electronics started getting phased out in the UK and now I'm struggling to sustain my workshop on no money at all.
Hey can you answer me this: I don't know much about power electronics, but I thought paralleling fets was an okay move. What are the disadvantages of the paralleled fets, and what does the one honkin big one get you?
Someone can correct me, but when putting FETs in parallel, you really need to watch out for current sharing. Here is a quick link to a research article, some keywords from it might lead you down the rabbit hole to learn more about paralleling FETs: http://www.richardsonrfpd.com/resources/RellDocuments/SYS_31/Dynamic_and_Static_Behavior_SiC_MOSFET.pdf
In short, if you don't parallel FETs correctly, one will heat up faster than the other and eventually fail.
I'll have to study what you've sent here. I thought paralleled fets tended to self balance because heat increased Rds, so they balanced out.
Rds does increase with increasing temperature, but Vgs(th) decreases with increasing temperature.
When used "digitally" in a switching design such as this motor controller, the FET is either on or off; Vgs(th) is irrelevant and the tempco of Rds will usually avoid thermal runaway as you suggested.
When used in a analog amplifier, (i.e. in which the FETs aren't fully on), the effect of Vgs(th) may be greater than the effect due to Rds, leading to poor sharing or thermal runaway. This depends on the operating region, being worse at low Vgs where the change in Vgs(th) has more of an effect on Id.
This is why you see current sharing resistors in series with the source(s) in class AB amplifiers but not in class D ones.
Thanks for this! Would you say this motor driver is a digital application?
I did say that, although I really just meant that the FET doesn't stay very long in the transition between on and off. Digital in the sense of antonym of analog. I'm not trying to imply that we're counting on our fingers or anything like that.
[w]hen used "digitally" in a switching design such as this motor controller ...
The video needs more mpeg by the way.
Looks like Southern Germany or Austria... somewhere in the Alps... ?
Aside from that, can you sort of express the difference of how it feels to ride a that bike compared to a 250W bike?
Close. It's Slovenia.
It feels more like a motorcycle because of its weight. You have to be careful with throttle because torque is always instant and insane. Also you can go 80 km/h uphil. One time I crashed because I was holding the handlebar with only one hand screwing arround and couldn't let go of throttle ... you can imagine how that ended. No broken bones or stitches though. 😁
I'm going to guess a go cart or other small electric vehicle.
Those are some seriously thick traces
nothing against this guys traces though https://www.reddit.com/r/electronics/comments/5i9wxd/i_just_received_my_20_ounce_pcb_soldering_is/
Ha! Dear god the etchant. I hope that was milled
I thought the 1970's HP test equipment boards I got were ridiculous (and they are - double gold plating on the entire board) - but they are nothing like those. Wow.
How is this made??? Can't be etched, can it?
T H I C C
You could probably knock someone out with that thing.
Any chance you're open sourcing this? I would live to build a high power driver but don't have the background d to design it. Completely understandable if don't :)
Looks very impressive!
Software is already open-source. I won't give out gerber or pcb files but I can send you a PCB and required documentation to assemble it.
It's described here: https://vesc-project.com/node/311
Also I'm thinking of finishing this up and start selling assembled units or sell the design to a company or something in between. The demand is high.
How much will that cost?
That would be great, how much would you want for a PCB? ... saw your other comment below, about making and selling the completed driver, there would certainly be high demand for that, at the right price. I would be very interested either way.
Nice one. Did some automotive DCDC work last year for an upcoming hybrid. Was tough as big volumes and extremely low cost, with the usual automotive quality standards...
Lovely clean layout and nice use of busbars. Did you struggle to get the busbars to solder properly?
Any issues with EMC? What about thermals?
What do you mean by the usual quality standard? Are you talking about automotive component quality or quality requirements for the boards? And are you implying that the usual quality standard is high or low?
From reading some data sheets, I though "automotive" parts were supposed to have a longer life and be functional in a wider range of temperatures vs general purpose parts.
Pretty sure he means the automotive design has to be dirt cheap but sort of NASA level reliability. Spec should require putting a square circuit board in a round hole and it needs to be perfectly electrically quiet. I used to do some electronics for electric vehicles and the automotive standards are brutal.
Thank you for your reply. I do some electronic design as a hobby and I am really curious about the requirements/specs of different industries.
I do hobby stuff now as I moved up in management and out of the auto industry. I'll tell you the electronics requirements were really tough. But it is sorry if the reason your car starts everyone for 15 years.
I really enjoyed it and it was so educational seeing everything that we went through to get a car into production. Hot weather, cold weather, emc chamber, bsr, nvh, voltage requirements, etc. Good times but I'm glad to not be in auto any longer.
As a student, after reading the replies I don't see myself working in the automotive sector. I hope I can find a sector with high quality requirements where my anal personality can be an asset without having too much pressure on cutting cost / development time. Maybe I'm still a bit naive.
I'm thinking I could be happy in the medical sector. I'll have to try.
There's an organization called IPC that publishes a book describing all sorts of aspects of electronics assembly, showing what is always required and what may be done for a higher quality.
Thank you for providing some additional info. I appreciate you taking the time.
Sorry, I meant the usual Automotive Quality/Safety standards: ASILB, AECQ components, ambient temp/humidity range, fire protections and strict EMC/EMI specs. Then you've got the over current, short circuit, reverse battery protections on EVERY IO. Turns a simple-ish design into a really complex system.
The verification required is a bastard as well :) Quite astonishing the amount of serious issues under corner cases that show up even when following application notes, etc. "Cold bugs" (-40C operation) were particularly tricky.
Didn't mean to gloat in my original post. Was properly curious about the performance of the board. Very clever layout, that in many ways puts ours to shame.
Thanks for your reply! That is interesting.
Nice design! I'm about to start on my own meager 10kW design soon.
Mind sharing what H-Bridge FETs and gate drivers you're using?
Also, what are the three components on the back? Current-sense?
I'm trying to learn orcad PCB designer right now to make small board alterations to a board we use at work, and I'm completely overwhelmed. This is very cool. Good job.
Orcad is not intuitive, but there's almost always a way to get something accomplished, you just gotta find little workarounds.
Every time I use it, I start out thinking "I'm gonna make my own ORCAD! With blackjack! And hookers!" and by the end of the project, I'm confused as to what my complaints we're in the first place.
What do the tiny vias around the plated slots do? I just designed my first high power board which has slots for .25" QC terms. Wondering if that's something I should have had...
They are there for high current. When the component is soldered those holed are filled with solder making a better connection from internal and external layers to the component.
Generally speaking speaking for two reasons:
Beautiful board man! Just one question, how did you get 20kW out of that usb cable you have plugged in? Very impressive.
It's made from unobtanium. Superconductor at room temperatures.
It took me a while to figure that out. The three posts in the middle of the board are for connecting power and the motor.
why the elecrolytics? doubt they're doing much for you. I bet you'd be better off putting small ceramics right smack next to the FETs
Electrolytic capacitors are there to compensate for inductance of long battery cables. Ceramics are close to useless at high voltages (150V). Instead film capacitors are used. With capacitance of 22 uF +-5% in full voltage and temperature range and only 6 nH of inductance per cap they beat any ceramics except in size.
inductance between the FETs and the caps is quite important. A much smaller value ceramic can be very low ESL and ESR, and also be placed immediately next to the FETs, and therefore have extremely low total inductance. You'll find you need much less bulk dc-link capacitance if you have some smaller capacitance with very low inductance. I know you don't want to believe me, but you should simulate it or rework the board and you'll see.
I ran simulations and did a lot of calculations. Copper connection between mosfets and caps is 25 mm long 150mm wide solid fill on one layer and returning fill on three layers. Total inductance between caps and fets is a bit less than 1 nH.
I get ~ 8V voltage ripple @ 60A load. https://www.youtube.com/watch?v=lDOkBaCXiCE
This is great! thanks for that info - couldn't see the timescale on the scope though. The high frequency ringing is the result of the LC circuit you've created with the copper trace inductance and the capacitor ESL. Electrons are sloshing back and forth between the inductance and the capacitance, causing voltage spike/ringing. This (can) kill FETs and wreak havoc with control circuitry.
You can always increase FET gate resistance, which slows down the switching, but this increases losses/heating in the FETs.
Ringing and the peak value can be brought down two different ways - more bulk capacitance = heavy and ineffective since the the more you add, the further away it gets and thus the more inductance is in its loop. Or you can get the lowest ESL capacitance as close as humanly possible. The lowest mass solution is to get several orders of magnitude less inductance with a much smaller cap. hopefully, inductance goes down less than the capacitance goes down.
The most ideal filter capacitor system (from a mass and size perspective) would be a cascading ladder of small low ESL caps close by, going to progressively larger value caps each placed a little further away. The small caps buffer the high frequency ringing on switch edges. The big film caps should really just be for "bulk" storage, to buffer the fundamental switching frequency voltage drop. You don't need much bulk capacitance to do that. Common error is to try and use bulk capacitors to do both.
Also, electrolytics make very poor bulk capacitors for motor drives. Only use them if cost is the main driver, in that case they can work of course. But since you seem to be able to afford film, you'd be better off just using the same spacial volume filled with film caps.
You can take my advise or leave it, I have built several different high end motor inverters from 1kW on up past 200kW.
This (can) kill FETs and wreak havoc with control circuitry.
Because the voltage can go above Vdsbreakdown for short pulses or is it something like high dv/dt?
I considered mounting ceramic caps as close as possible to mosfets. Something like this. The problem I see is via inductance that can be quite high. Since there is space I'll consider implementing them in the next revision and see how it goes.
only if the peak of exceeds the voltage breakdown of the FET. dv/dt would not damage the switching device. dv/dt might causes issues with digital portion of the inverter, depending on if it could couple back to those circuits - either through parasitic capacitance in digital isolators, transformer/powersupply, or layout proximity of traces/planes.
good work, and good luck. Let us know how it goes!
Great project. This measurement is not valid though, you have made the same mistake everyone makes the first time they try to use a scope to measure a high current switching circuit. What you can see on the scope is not ripple, it is ringing.
The large caps are used to reduce ripple at the fundamental frequency of your PWM (probably 30khz if this is a vesc derivative). What you are trying to measure with the scope on those settings is ringing, you need to increase the time division until you can see two switching edges, then you can see the ripple.
HOWEVER, you can't measure this kind of thing with your scope probes attached the way they are. You are just getting inductive coupling of the switching current magnetic field into the loop created by your scope ground lead. Try to clip your ground lead onto the top of the probe without attaching to the circuit, I bet you get a similar signal. You need to get the loop area of your probe tip->gnd as small as possible.
Ceramics are close to useless at high voltages (150V).
Care to elaborate on that ?
You can get at best 2.2uF 250V ceramic capacitor. It will also cost the same as film cap ten times the capacitance. That's the reason why high voltage inverters use film instead of ceramics. Although it seems like a combination of both is the answer in my case.
I wonder how it compares in terms of ESR/ESL, big chunky film caps will have way way more ESL than small ceramics, also will it be better to have film + electrolytics or ceramics + electrolytics.
I'd say stick some smallish ceramic caps to short out all nasty current loops and that will take care of all the ringing.
What's the stack up (is it the same copper weight throughout?) for this board and how much did it cost you(this looks like it cost a mint)? if you dont mind me asking.
The board is nothing special. 4 layer 1 oz. High current is routed through bus bars. I paid ~100€ for 10 pieces.
Which controller you are using?? Is it stm32 or TI's c2000 range.
Edit: It could also be something else it's just that these two come to mind when I think of EVs.
Those sexy bus bars
How do you even get started on designing circuits like this?
Or do you just need an entire EE degree.
Two months of research and another month of designing. Few years of Altium experience also helps.
It would certainly help. That and lots of practice, and lots of design time
What are the two rounded things next to each of the three BGAs?
Nice job! How did you make the busbars? Can the PCB manufacturer make those as well?
I bought a copper bar cut it to pieces and soldered them to the PCB with reflow.
Male debug header, kinky.
Very cool! Re-purposing the VESC firmware is genius
it's so thicc, I'm so moist if I were to touched it it would short prematurely, without a doubt.