Wet brushless ESC goes up in smoke

Sure, but it would definitely be an advantage if it was in actual contact with all, or at least most, of the MOSFETs.

Yes, the heat sink was just stuck on top of the MOSFETs with some sort of sticky tape. After removal, I could see the impression of the MOSFETs in the tape, but only a few of them had left an impression.

Here's a picture of the underside of the heat sink:

As you can see, only 7 or 8 impressions are visible in the tape. In fact, I think I can see a similar pattern when looking at the surviving MOSFET placements (I have moved one or two of the remaining MOSFETs to get all phases and both sides covered).

The board copper is definitely contributing to the cooling. I had to switch to a thicker soldering tip, because the thin one kept getting stuck as the copper sucked the heat right out of it. I just checked the data sheet: The MOSFETs have *much* lower thermal resistance between junction and foot than they have between junction and ambient.

I specifically asked for decent quality when I bought them, and I paid quite a bit more than I would have paid for the cheapest ones. Now that I have taken them apart, I suspect it might be a ripoff. Just for fun, I am going to order the cheapest ones I can find from China and see how they compare.

WayPoint 25A:

The site is Norwegian, but the product description is in English.

Parts are already ordered. Once repaired, this controller is going to see some harsh water tests :-)

--
RoRo

(...)

RoRo stated that each phase has 3 paralleled P-MOSFETs on each high side, so I don't think a pump cap is involved.

As battery voltage falls, Vgs falls with it, pushing the high side switches toward linearity.

They can then become 'magic smoke emitters'. :)

--Winston

"Robert Roland"

** Solder corrodes very slowly in presence of water - until there is a DC voltage available to cause electrolysis. ** Make sure the PCB is very clean a fee of all solder flux first.

Use a flux cleaning spray or denatured alcohol and warm air to dry.

and

** Either of those last two can result in a repeat event.

..... Phil

Yes, various bench tests will be performed once the controller has been repaired. I am currently ordering parts.

LiPo batteries are not run until "quite low". If their voltage gets too low (about 3V per cell), they could die instantly. As a general rule, I use only 80% of the battery's rated capacity. At this level, each cell has well over 3.5V. The controller is rated for 3 or 4 cells, which essentially means a minimum of 9V. With an adjustable bench power supply, I can push the limits and see what happens.

I will test.

--
RoRo

Sounds like a plan. I recommend slowly heating the board to ~150 F or so, soaking at temperature for 1/2 hour, then dipping it in varnish until it cools (in the varnish) then letting it drip and dry.

Oil based varnish.

Heating drives off any moisture and cooling in varnish will allow any contracting air pockets to suck in varnish, and the coating will be thicker after it cools.

Switches, sockets etc.. will not work with this method.

For your idea of a bag, (opening it to allow water vapor to escape when you're not using it). I recommend adding a bit of dry cloth, sponge, or desiccant bag. In the event that some water does get in, you don't want it rolling around in there, but give it someplace to go.

I have a waterproofed electronic camera that I mount to the bow of my kayak. SOP is to keep a #10 metal can with about an inch of super dry desiccant in it, and pack the whole camera and case in desiccant over night. Then I assemble it, when I'm ready to use it. The water proofing is good (will stay submerged for hours without any ingress - gasketed case), but just a slight amount of water vapor will condense on the glass port, so pictures look out of focus.

One of these days I hope to find a camera light enough to waterproof and attach to the kite that pulls the kayak. The "chewing gum - spy" cameras work, but the photo quality is abysmal.

Can't go wrong with Castle Creations.

--
http://www.wescottdesign.com

I think the way to go would be to bring the heat sink down to the copper, rather than the FET case. I've seen quite a few designs where the board is thermally connected to a rather substantial case for thermal transfer, but I'm not sure what modern practice is when the board is little and the 'case' is heat-shrink tubing.

Go to the local hobby shop and peer at a name-brand ESC and see what they do, perhaps.

--
http://www.wescottdesign.com

If overheating was the actual problem, I'd agree. In normal operation, however, the controller does not get much more than hand warm, so cooling is probably sufficient. And once the commutation jams or the MOSFETs shoot through, no amount of cooling is going to save the day.

On designs that produce considerable amounts of heat, I'd expect they'd use transistors that have a proper metal tang to bolt securely to a heatsink?

--
RoRo

Yes and yes. First, the ESC can't count on good airflow, and second you don't want to burn up power in the ESC -- you want it going to the motor.

And if you have the Gabor sisters advising you on your circuit design, you know that nothing's going to help you if you have serious "shoosht-through".

The dividing line just keeps moving out, though -- more and more, thermal management is being done by dumping heat into the PC board, then (if necessary) removing heat from same.

If you waltz through the available selection of power parts, you'll find _lots_ of surface mount chips that feature large tabs that are designed to solder onto the board for heat dissipation. Some (I can never remember the package name) are essentially TO-220 parts with the tab trimmed and tinned, and the leads formed to contact the board.

TO-220 (D-PAK^2??) seems to be the current dividing line between "reasonable to move to through-hole" and "let's keep this all surface mount".

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" was written for you.
See details at http://www.wescottdesign.com/actfes/actfes.html

Excellent! Please keep us posted.

--Winston

I just poked around with the scope a bit. It does definitely NOT work.

So, one high-side MOSFET never turns on, which means that two of the six phase combinations are lost. The remaining four are apparently enough to get the hard disk motor going.

Have to wait for parts now, before I can do any more testing.

--
RoRo

(...)

Can you 'reverse engineer' the connections from the Atmel chip to your three bipolar transistors? It'd be reassuring to see that all three Atmel 'high side' outputs are operating properly.

--Winston

I have had one half-hearted go at it, but got nothing sensible. I will take another look.

--
RoRo

OK, I figured out the gate drive. The signal looks normal.

--
RoRo

Here's the microcontroller spec. That should be helpful.

Looks like the PORT B pins are a popular source of your PWM signals, particularly 13, 14 and 15:

--Winston

Excellent news!

--Winston

Is it possible that your processor is acting strange when insufficient voltage is present at intermitting moments? Maybe the drive circuit from the uC does not have proper protection in case a long STh/Olap takes place during a uC reset or reboot caused by unexpected dips in power? I can only assume you're using a 3 pole full bridge with a 3 wire motor? I kind of suspect some problems in phase firing if so. Maybe you should be using fuse Links.

Jamie