Wednesday, June 8, 2022

Some motor controllers and other electronics WIPs.

Just a bit of what has been going on lately. First of all, the implementation of that fast decay solenoid drive topology in my usual style of blaster manager board:


These are being called E-Core. They have pretty much the same featureset and I/O as S-Core, with the addition of bus voltage sensing vis the ADC and 2 extra (PCINT) tach channels for a total of 4, something that will be standard going forward to provide the best 2 stage cage support with speed monitoring on all wheels.

The powerstage is driven by a LM5109 (or any other basic IR2101 compatible driver). Obviously as from the post on the topology this is not a halfbridge, both switches need to be on at once, and a driver with shoot-through protection won't work. There is an AP3012 boost converter fed from the 5V rail for the gate drive rail in the usual arrangement I use (overkill, I know).

Now for something different (non-blaster):


This is a SimonK controller using singly parallel LFPAK88 FETs in the "sandwich" layout, with generous copper area (to say the least). Board dimensions are 72x100mm.

This is aiming to run a 190kv 6374 motor on a scooter on 6-7S using Nexperia PSMNR55-40SSH (go look that device up, it's insane, by the way; half a milliohm typical at only ~190nC of gate charge). Drivers are Infineon 2EDL23 series. MP2459 (55V rated) for the logic power and again an AP3012 off that for gate drive at ~14V. Since this is going on a vehicle, the power supply overkill with the input sag eliminator and "buck then boost" approach to derive the gate drive rail for maximum voltage headroom are probably justified.

It's experimental I might add, but I'm pretty sure I will have pulled this application off with only six devices; the thermal numbers work out okay-ish even for unipolar PWM (with diode conduction) and some sustained partial throttle that may be seen in use. The low side device that may have diode conduction in the usual arrangement is actually positioned with the drain tab more directly over the double-sided phase node pour where all the vias are, and that will be further heatsunk by a 10AWG phase wire, so... Oh, and I've got a a pair of very heavy solid copper heatsinks made for NVMe drives (surprisingly cheap!) that are planned to be clamped directly onto the top of each rail of FETs with thermal compound. They are not best heatsunk that way but the case is pretty thin compared to usual, and it does help.

And here's another board along the same lines. This one's a bit smaller, designed specifically as a "high" voltage platform for 80-150V class FETs and takes D2PAK-7 devices and 12.5mm caps.

Pretty similar idea; no worrying about the size of the board here, priorities are on robustness/good cooling for SMD FETs as this one is meant mainly to run a brushless string trimmer I found in the trash with a good motor, no battery and a questionable inverter, and other similar sorts of projects involving ~12S. High voltage power tool fixes/conversions, etc. Driver is a FAN7888 as I have a handful of them from a long time ago I never used for blasterscale projects. LTC3638 (140V buck) and a LDO for gate drive and logic power.

 

Back onto blaster stuff, I have been wanting to dump the "ESC" form factor/paradigm since forever, so this is what the ACE-NX is evolving into:

This for reference is only 22mm wide (quite a bit narrower than the NX Inline aka the traditional ESC styled version you might know that the T19 uses and so forth) and about 76 long, so a nice long skinny form factor, just the ticket to either conceal these in apt places within predominantly long, skinny blasters, or to have a nice neat farm of them somewhere when you have a multistage or otherwise a bunch of motors that need driving.

It uses LFPAK56 FETs, the same Infineon 6EDL04 driver in TSSOP28 from the original NX, and due to area/complexity reasons, the same LDO logic power/AP3012 gate drive supply arrangement. DC link caps are 4 10mm units which offers a sound improvement in everything (thermals, ESR, etc.) over one giant capacitor and allows building these with a low total height using i.e. 220 or 330uF parts from Rubycon and others. In addition there are 3 1206 MLCCs. Having a few good snappy ceramics in there is something I am doing in all these new designs.

Oh, by the way, no, the cap footprints do not overlap the FETs in this or any other boards here. Those FET footprints are expanded by a LOT because it makes them easier to solder.

In addition to being a much more electrically optimal layout for the DC bus and so forth these should be thermally quite a bit better than the inline boards. Not that the inline boards have had issues. The bottleneck is probably (as in the inline boards) the phase nodes which have the drain of one of the FETs on them, but then again see the remark about phase wires. Phase wires are not appreciated often as heatsinks but if you work with this stuff or ever see a running controller thermally imaged you quickly realize that as long as they are oversized for ampacity they are a heatsink, not a source, and help get FET heat out of the thing. Hence a move to oversizing the phase pigtails on boards, such as 16AWG.

Not shown (forgot image): a LFPAK88 "Max" version of the same board (i.e. ACE-NX.88m) - which is 34x80mm and has the extra stuff like switchmode logic power, the input filter and so forth and accepts 5 caps. I don't know WHAT would ever reasonably be done in a blaster that this one won't be able to run... And also not shown, the ATmega328-based throttle interface board for the scooter project and the trimmer project and anything similar where you want Hall effect/potentiometer input to PWM throttle, battery low voltage cutoff, battery gauge LEDs, any datalogging or control loops you care for, etc. which is mostly just a MCU and a bunch of passives so little to write home about.

And this is in the ACE-NX Max when I was working things out. Now that's the best you can hope for, for nearly all of a MCU/gate driver lane (what happens when you created the board definition specifically for that driver, lol). Internal pulldowns in the driver.

Actually that's maybe noteworthy. I have really liked Infineon drivers so far. Internal pulldowns on logic inputs, internal pulldown on gates when the chip is not powered (see 2EDL23 series datasheet), internal bootstrap diodes with built-in ballasting resistance. That's quite a few annoying passives and bits deleted from a board. And robust input filtering, with short pulse rejection (read up on that issue with traditional HVIC gate drivers if you design this stuff and want to lose a bit of sleep).