Those cute little Pololu stepper driver boards using the Allegro A4988 chip have one conspicuous problem: there’s no good way to heatsink the chip. The doc recommends heatsinking for currents around 1 A and some informal testing shows it will trip out on thermal protect around 800 mA, so heatsinking really isn’t optional.
A thermal pad from the chip bonds to vias that conduct heat through the PCB to the bottom surface copper layer: putting a heatsink on the top doesn’t help as much as one on the bottom. What I’m doing here is a first pass at a bulk heatsink that would work with several of the driver chips lined up in a row; this one is ugly and doesn’t work well, but it should let me do some further electrical tests.
The general idea is to clamp the heatsink around the board, with the chip as the top-side pressure point. The catch: no room for an actual heatsink underneath, because that’s where the connector pins live. You could mount the board upside-down, but then there’s no good way to tweak the stepper current trimpot. That may not be a problem after you get things set up, although I’d hate to unplug and replug the board for each adjustment.
So I think a reasonable solution involves a metal strip to conduct the heat out the ends and up to the heatsink. What I’ve done here does not accomplish that; I’m just feeling around the parameter space.
You can’t get too enthusiastic with the clamping force, lest you crush the chip, so moderate pressure is the rule of the day. However, the chip sits low on the board, surrounded by taller components, so I put a drop of epoxy on top and flipped it over to produce a short thermally conductive column that’s higher than everything else:
The blue sheet comes from a trimmed-down TO-220 transistor heatsink pad; it’s thermally conductive silicone, provides a bit of compliance against the PCB, and insulates the REF trimpot test point from the heatsink.
The result looks OK, but it would be better to embed a small metal block between thinner epoxy layers to get better thermal conductivity:
Although most of the heat goes out the bottom, you still need something on the top to take the spring pressure. I trimmed down the TO-220 heatsink that came with that silicone pad; it must mount off-center to permit access to the trimpot but, alas, blocks the voltage monitoring pad and both sense resistors. A length of 45-mil music wire bent into a flat M provides the spring:
The side view show how the kludge fits together:
The final result is truly ugly. The epoxy column didn’t turn out nearly as parallel to the PCB as I’d like, so some filing and finishing will be in order.
Now, to find out if it’ll allow the chip to run above 1 A for at least a while.
The Smell of Molten Projects in the Morning 
I just put four plastic spacers (snapped from the same type of pin strip) on the corner pins to raise the board up and added a 25mm fan and cowl running the length of three boards. The chips then run cold with currents > 1A.
Nice-
In a grisly sort of way… [grin]
So there’s enough conduction off the bottom of the PCB to completely eliminate the need for a heatsink: I like it!
This one went onto a solderless breadboard amid a tangle of other junk, where there wasn’t much room for active cooling, and runs somewhat more than warm to the touch at 1 A. Blowing air across it would certainly improve the situation, but eliminating the heatsink entirely would be even better.
I’ve been doodling a four-axis breakout board with more attention to cooling than what I’ve seen on Thingiverse & elsewhere; mounting a tiny fan in the middle wouldn’t do it a bit of harm.
Thanks for doing the experiment!
Any chance you could put solder on the thermal vias, and then solder a copper strip to that? Essentially do the same thing you did with a copper strip rather than aluminum, and use the solder for attachment rather than a clamping mechanism?
Alternatively, mill a chip-sized island onto the bottom of the block that goes on top, to avoid the issues with the epoxy not being flat and maybe getting a bit more heat out that way?
They’re neatly tucked away under the solder mask, so you’ve start by chewing that off. Then you’d mask the exposed trimpot test point. Then you’d have to heat the whole thing enough to melt the solder, without also melting the solder under the chip… and that’s the part that scares me.
Better to do something non-destructive to the board. That way you don’t form a deep emotional attachment to it: when it blows up, pop another one and move on…
I actually thought about that, but it seemed like a lot of work compared to a blob of epoxy…
Given I didn’t put a small block of metal atop the chip, I’m obviously not thinking nearly hard enough. [sigh]
Masked vias do complicate matters. When I’m soldering boards at work — because more than half our LED driver chips have thermal pads — I leave the vias unmasked (inasmuch as I design the board), solder the chip’s legs down, then turn the board over and look (via scope) through one via while I run solder down another, and when I see the solder run across the DAP and up the adjacent vias I know it’s attached well. Which is to say: if you did what I was suggesting, I’d hope the solder under the chip *would* melt.
With that said, I really like some of the other suggestions in this thread.
Which works for sticking the chip to the vias, but the next step of cooking the whole affair enough to solder a copper heatsink to the bottom without dislodging any of the SMD parts on the top seems too dicey for my taste. Basically, I’d be reworking the entire board to add the heatsink… which would give me entirely too much of an emotional attachment to the result.
But now I know how to treat those vias and the chip’s thermal pad when I get around to hand-soldering such a gadget myself: thanks!
http://www.newegg.com/Product/Product.aspx?Item=N82E16835708012&cm_re=copper-_-35-708-012-_-Product
With the stock tape or remove that and use a super thin layer of thermal epoxy:
http://www.google.com/products/catalog?q=thermal+epoxy&hl=en&safe=off&prmd=ivns&biw=1445&bih=1162&um=1&ie=UTF-8&tbm=shop&cid=16705661584947719806&sa=X&ei=ZHf7Tam0IIeisAObocXdBQ&ved=0CGkQ8wIwAA
A music wire spring might fit right down the middle of one of those!
Ed – Setup a mini shroud around a fan that you have and attach these. Then direct specifically where you want. You could even make it multi-tenacled – Kinda steam punk sci fi ubber cool….
http://www.thingiverse.com/thing:9006
The unholy offspring of an air compressor and a squid…
Since the DAP thread is as deep as wordpress allows it to be, I thought I’d mumble some other things about soldering chips with exposed pads on the bottom. Hotplates rock: they’re definitely the best way. Put solder on the pad on the board, heat it on a hotplate, load the chip, then solder all the pins down by hand. You can even do this with a double-sided board if you’re bored: take the silkscreen layer and cut it oversize in 0.25″ aluminium so you have a heat conductor with cutouts for all the parts. It’s been my experience that if you’re careful even a D2PAK will hang on the bottom of the board.
If you’re designing the board, bring the pad out from under the chip and do a topside soldermask cutout so you can put the soldering iron there. Add solder all over, put the iron on the cutout, then slide the chip onto the molten solder.
For removal, heatgun or hotplate are about the only options, and neither is particularly attractive, but hotplate is much less likely to pull a pad.
Reflowing solder through the vias works quite well, but without a microscope and clear vias, it’s not so useful.
I like that, although it sounds like the sort of thing one would do early on a low-caffeine morning…
That’s dynamic tension between you & dbliss, with me trying to keep the deeper comments from
crushing
down
like
this
even
on
big
screens.
Dunno what the right nesting depth is, but WP’s idea of nesting (at least in this theme) isn’t helping.
Go with David’s recommendations, not mine, any time there’s any doubt whatsoever.
Ed – I just spun a pololu board derivative for one of my cupcakes (http://www.thingiverse.com/thing:8586) and honestly using the makerbot steppers they are really only warm to the touch. Now when I get around to building a reprap/mbot motherboard Im thinking tight arrangement of the 4983s with a vga style cooling heatsink and fan. But for the mbot steppers they don’t seem to heat things up that much. YMMV of course…
The data sheet says the A4988 has about 0.8 ohm total driver resistance at 1.5 A. Could be just over 600 mohm, could be just under 900 mohm.
If you’re using the original 14 V steppers at 300 mA the driver dissipates under 100 mW, which wouldn’t budge the PCB much off ambient.
If they’re the “new” steppers at 800 mA, then it’s up to half a watt: warm to the touch.
I’m running an unloaded 2.3 ohm stepper at 1 A peak, so the driver dissipates 800 mW and pushes that kludged heatsink to the hot side of warm-to-the-touch. The datasheet indicates it’d run perfectly well with no heatsink, but I’m that type of guy…
More current than that definitely requires active cooling!