First iteration of the 40A variant of my motor driver

[spiral crescent]

I'd put more caps close to the FETs

All that free space: stuff it with capacitance: MLCCs, polymers, electrolytics by order of distance

also your shunts aren't doing anything, you need a kelvin connection to gnd at the shunt

I'm worried about the gate connection inductance with such long tracks

are you doing 2 boards for thermal reasons? I'd put everything one one. Or at least the gate driver on the same board.

40A is very optimistic (unrealistic). That's 20W per FET peak (peak needs to be considered because the silicon thermal mass is very low), what's your thermal solution on this?

[willow igloo]

Yeh, I was worried about that. The idea of two boards as that all the computery bits could be a common board to multiple sizes of FET sub-board, but perhaps the length of the traces makes it unworkable as an idea.

[willow igloo]

The thermal solution is a sheet of aluminium between the two boards, and a heat sync on the back of each board, directly contacting the copper on the back of the FETs and the driver

I mean, I guess I could put my own current sense amplifier on the FET board, and ignore the one in the 8323

[willow igloo]

@spiral crescent given that one leg of the shunt is to ground, how should I go about doing the low side kelvin connection? Just place the current sense contacts close to the resistor and connect it to the ground plane?

[spiral crescent]

I mean I can see it work, but make the connectors and traces nice and fat.

[spiral crescent]

use a net-tie in the schematic, you basically just want a connection that doesn't carry current to measure just the potential difference

with the shunts the issue isn't trace length per se (ignoring interference), you're not pushing any current through them you just want a clean measurement

[willow igloo]

so something like this is acceptable?

[spiral crescent]

yup, absolutely

also just keep in mind that the "current rating" on mosfets is aboslutely unrealistic and assumes heat being removed from the silicon instantly.

I'm planning a 30A ESC and will be looking for 3milliohm rds_on or lower fets

[mild mica]

you could try some GANs

ST has free samples for those

[spiral crescent]

also keep in mind to isolate it electrically, otherwise you short the draits with the heatsink

[willow igloo]

Yeh, I was wondering about how to do that sensibly 😄

Also, pretty sure those FETs aren't going to short the drains, they're QFN chips, but the capacitors need to be avoided - as does everything on the top board

[spiral crescent]

do they now? I only see "preview" status products outside of their 700V range

[mild mica]

i free sampled a few MASTERGAN 2

oh wait no

[spiral crescent]

that's 600v

[mild mica]

the mastergan 1

you need 700V?

[spiral crescent]

no, I need 100V 😂

[mild mica]

then free sample a bunch of mastergans 1

[spiral crescent]

600V @ 3 milliohm would be... quite something

[spiral crescent]

why should I?

[mild mica]

wdym

oh wait, it's 150 mohm

f

[spiral crescent]

yeah. And a 600V 3 mohm fet would probably be a 1k usd behemoth

if those even exist

[mild mica]

well, i found a 71m ohm one

for 4 bucks XD

[spiral crescent]

was just because you were saying you'd be putting a heatsink directly contacting the copper

[willow igloo]

oh, right, brain fart - I for some reason had decided that that was ground, yeh, good call

you get those electrically insulating heat conductive little insert things used in power supplies, don't you?

[spiral crescent]

yup

thermal pads I think they're called

[willow igloo]

yeh, but that can also refer to the thicker, semi-solid thermal paste type things that IIRC are electrically conductive

[spiral crescent]

ah those. but I think they're kinda niche.

pads are also nice because you get good contact even if the board is a bit warped, and it often is

[willow igloo]

so given your concerns about rDS, you think I might be better off with something like CSD18540Q5Bs?

[spiral crescent]

yeah, what's the max voltage you're gonna be running?

[willow igloo]

48V is the target

what I should also do is add another set of mounting holes, so the heat sink mounting isn't the same as the mounting mounting

[spiral crescent]

yup

[spiral crescent]

that's cutting it quite close. Will you be doing regen braking?

[willow igloo]

I hadn't intended on it, no, though now that you mention it I'm wondering about it

<feature creep intensifies>

[spiral crescent]

are you planning on running custom firmware or smth existing?

[willow igloo]

custom - I've got my own firmware on this little baby 2A stick and am planning on trying to make it run some bigger motors too

atm all it does is bog standard trapezoidal drive, so it's pretty naive, but it'll do FOC soon if I have anything to do with it (which I do)

[spiral crescent]

issue with regen breaking is voltage transients. Especially if running on a full battery, it can't absorb a lot of current, so you get voltage spikes which kill the fets. There's softwane side protection (dumping the current into motor windings if voltage gets too high) and hardware side: TVS diodes, tons of bulk capacitance

[willow igloo]

yup, and I certainly need to consider something in that regard, because the intention here is to do some silly things with go-kart like things, so there is going to be the slowing down while battery is full situation to deal with

[spiral crescent]

can be avoided if you have mechanical brakes

[mild mica]

(and you should)

[spiral crescent]

or a brake resistor (but that has to be real phat with a ph)

[willow igloo]

also, yes, I already killed 2 of these DRV8313s by cocking up my commutation timing

@mild mica ppffft, safety nafety

😛

[mild mica]

XD

[willow igloo]

@spiral crescent on the plus side, it could also act as a seat warmer

[mild mica]

regen braking doesn't work with no power, yk

[spiral crescent]

*looks at 40 km/h bobbycar with zero mechanical brakes :S

[mild mica]

...

[willow igloo]

haha

[coarse thunder]

Command Error

The input text has too few parameters.

Usages

.quoteprint text

Run .ve to disable these prompts.

[willow igloo]

but yeh, mechanical brakes will be a thing, but I still need to deal with (for example) it freewheeling on a hill

[spiral crescent]

I mean, I could think of some "safety" system which shorts the motor windings in case of loss of power. But you'd probably go flying from how hard it brakes lol (and fry a motor winding on the way)

[willow igloo]

lol

copper blocks that move in towards a metal disk, and slows it down via eddy currents?

[spiral crescent]

zvs is of course easier with more margin, because you need zero conduction at 48V, then have it start conducting at a couple volts higher, and have full conduction and clamp the voltage below 60v

[mild mica]

nah, a super strong manget will do the job

[spiral crescent]

I mean that's just your mechanical braking system. How you do that is up to you

[willow igloo]

hehe

[spiral crescent]

freewheeling should be fine, unless you end up going faster than top electrical speed

[mild mica]

just use this and literally convert to plasma your wheel

you can't roll if you have no wheels XD

[willow igloo]

then I'd have to solve the "how to close a switch" problem even he could barely solve

also - dude... that guy...

[spiral crescent]

photonic induction showed off some fun switches

[willow igloo]

there's a level of insanity there to aspire to

[mild mica]

can they handle literal megawatts?

[spiral crescent]

smth along that line

[willow igloo]

yeh, the "I built a battery bank in my garden that can briefly emulate the output of a small power plant" is more than a bit nuts

[mild mica]

btw, the batts outputted like 48V

and 9kA+

[mild mica]

you mean create nuclear fusion reactions?

[willow igloo]

yeh, that too

I mean, to be fair, creating fusion reactions isn't that hard - doing it accidentally is still kinda impressive though

doing it in a self sustaining way? Yeh, super hard

[spiral crescent]

youtube.com/watch?v=WAhq_A4EbkE

[mild mica]

he also was the only guy that made a CEO say "f it we ball"

[willow igloo]

I love the fact that the battery company trusted him enough to sponsor him and send him all those batteries

[mild mica]

it was Anydesk

not a battery company

[willow igloo]

oh was it?

still...

[mild mica]

yep

[willow igloo]

the liability

[mild mica]

Anydesk bought the batteries XD

[willow igloo]

I mean - I kinda get it, he seems pretty competent, but also, utterly insane

[mild mica]

and Anydesk was the one that wanted to see a 400 batts vid

lol

[waxen brook]

is this a 2-layer board? Please watch out for GND return current. You have nice planes for supply and each phase, but the same current might flow from sense resistor GND, to the input connector/capacitor.

[willow igloo]

it's 4 layer - or was, it's now 6 layer

[willow igloo]

Still need to figure out what I'm doing with braking

[spiral crescent]

might wanna look into smaller capacitance but higher ripple current rating caps

[willow igloo]

nods

[spiral crescent]

I approve of the MLCCs tho :)

[mild mica]

maybe get some super caps

they can handle very high currents

[spiral crescent]

not very high voltages though...

as in, around 1 or 2 volts

[mild mica]

nah, some handle higher voltages

[willow igloo]

we clearly just need Donut's battery >.>

runs

[spiral crescent]

if you're talking about the supercapacitor banks, I doubt they'll be of much use here

[indigo token]

no

you want solid or hybrid alu caps

much much lower ESR meaning much higher ripple ratings

[mild mica]

oh ok

[indigo token]

ESR bad for ripple rating because that resistance directly correlates to internal heating which determines the ripple rating

Internal heating bad especially for wet electrolytes because it'll make them evaporate/lower lifetime

[willow igloo]

Okay, today's itteration - I'm much happier with this version, still a lot of things to change though.

Solved the ADC proximity issue by sticking a dedicated ADC on the board with the FETs

so now the communication between the boards is the gate drivers, SPI, a "data available" line, and a time synchronisation line

Okay, I couldn't go to bed with those zones looking so crap

[spiral crescent]

can recommend this for bedtime reading: https://www.ti.com/lit/an/slvaf66/slvaf66.pdf

[willow igloo]

I need to figure out how to make that left most zone wider - it's only 8 mm wide

oooh, useful

thank you

[spiral crescent]

:)

[willow igloo]

on a related topic - whoever does documentation at Texas Instruments deserves a raise

their data sheets and guides really are top notch

[spiral crescent]

except the layout examples haha

[indigo token]

honestly just expand the zone straight vertical up through that via grid

the via grid isnt super dense so it should be fine

[willow igloo]

It's beginning to look like a thing

@spiral crescent btw, thanks for the link - much usefulnes was had, I integrated several changes from it

[spiral crescent]

neato :)
Might also be worth putting a pair of fat zvs diodes if you havent done so already

[willow igloo]

Yeh, I was wondering about that. The guide is like “yeh, don’t rely on them”, but… might be a good redundancy

My one slight concern just now is whether my latency on shutting down in the event of fault is low enough, since using that ADC means I need to shut things down in software

Rather than relying on the DRV to shut it down

[spiral crescent]

I prefer to have both softwace and hardware protection:

• dump current into motor windings above threshold voltage (software)
• ZVS (hardware)
• lots of bulk capacitance (hardware)

[willow igloo]

Right - the bit of hardware protection I lose is the V_RS and V_SENSE shutdowns in the case of shoot through. But it’s replaced with software protection. I’ll see if I can figure out how big diodes need to be and throw some on

Defence in depth is definately the right policy for a 2kW device

[spiral crescent]

can also have a look at the moteus firmware for some inspiration, it's extremely well documented in my opinion

[willow igloo]

I think I have a plan for dealing with back emf now - charge the battery to no more than 80%. When driving, if the battery is below 95% SoC charge the battery with it. If above 95% - display a warning and carry on charging. If at 100%, display an error, and set all low side fets to open, and high side fets to closed, and rely on mechanical brakes.

Now just to figure out on which bit of hardware I’m implementing that on - the cart controller or the motor controller