Metal P2 (AL substrate) eval test board
Tubular
Posts: 4,701
I recently had some custom Xe-g Led boards made up that used copper pcbs, and it got me thinking about how to lay out a P2 board that could be used for thermal testing and voltage rail manipulation.
The board design below is the result of that thinking. Because vias are difficult, this board relies on SMD male DIL headers, and a second power distribution / fan FR4 PCB that sits above the AL PCB, and does the 3v3 and 1v8 routing.
The Vdd and Vio caps sit under the power headers so are about 3mm further out vs P2eval.
On the AL board, all the Vss ground current arrives at the corner near the TST (grounded) / P0 pin. A 2mm wide channel in the solder mask has been opened up with the intention of beefing up the Vss delivery by soldering a copper wire in parallel, on top of the pcb.
Its been interesting reading comments including by Beau in Ray's thread, it'll be possible to experiment with a couple of distribution methods with this board set. Boards should be back early next week.
Comments
Wow, that's something. Trying to wrap head around that...
Yeah its an odd board (zero vias!!) to support some development/testing we need to do. Its likely that the onboard regs won't even be used and we'll use an external linear lab supply to power it, so we can sweep corner voltages
Having an AL substrate means we should be able to heat/cool the whole thing pretty well, and test with a fan at various airflows
The vesa mounting holes are just because there's room, not really expecting it'll be hanging on back of monitors but who knows stranger things have happened
Certainly very left field! And elaborate in its way. No vias is mind-boggling.
Those narrow pinches for power and ground look scary but I see the high frequency capacitors are all low impedance paths. So at least the Prop2 should be stable at high frequency. That still leaves potential problems with I/O common-mode ground bounce though. It's something I've warned of with some of Rayman's attempts but not sure I've seen a clear example of it occurring as yet. There's always been alternative explanations for limitations so far.
Nice one @Tubular . It'll be great to see how this performs when you get your boards back. It's an intriguing idea you came up with. I like the idea of this fan board doing double duty for power distribution into the P2. Seems to be sufficent pins to feed it power rail current.
I like your trick with placing the bypass caps under the 0.1 inch SMD headers. Surprised they fit there. I guess it's just plastic above them.
What's the component in the middle of this image...two bypass caps and something else?
Yes you're right, though I'm not likely to be running the i/o at high speed at the moment.
I'm curious whether those ADC calibration trails vs temperature/clock freqs (that you plotted) settle down a bit with this board and getting the heat away more efficiently
Yeah in theory the DIL header has 1.9mm between its pins, and those 0402 caps are only 0.5mm wide. Even offset a little it should be fine.
However I'm pretty sure I could get the caps closer to the P2 if I had to - where the 18 33 18 33 labels are (and move the headers out slightly). Basically I realised I could tighten it up after getting everything done, but the reward for a lot of effort might not be that much. We'll see how this first one goes and can go again if needed. I do have a couple of 0805 caps on the 1v8 rail near pins 16 and 48 to compensate for how further away the other bypass caps. I'm just not sure what value makes sense to put there, maybe something like 470NF
Just another 7805 style regulator site, with the usual in and out bulk caps. Could use a TO92, or SOT223 soldered vertically, or even a bigger TO220 and hope the pads don't tear. But as its configured all the 3v3 are connected, so it just needs a single 3v3 reg somewhere. Anything beyond that I need to cut the appropriate ring tracks on the FR4 board that sits above
edit: That particular reg you cropped doesn't have a 5v path into it, like 2 of the others have. I kind of gave up on those inner 4 sites after realising I just needed one single site for 3v3, which can be one of the four outer LDOs
Those outer 5 Volt zero ohm links could be longer. Maybe a custom part to allow wider GND necks ... or rotate them and use a fat link for ground instead.
This gave me an idea: The P2 footprint that I use (I got it from the forum here) has a hole on the middle of the bottom pad. So what if I made the hole bigger and soldered a heatpipe to pad? I think it would be necessary to solder an empty copper tube. Then add water and seal to make it into a heatpipe.
Yeah it was as late addition, 1206 jumpers would be better, and yeah it might be better to rotate.
The 5v is just feeding 3v3 LDOs and i/o, not much current for this test rig so it'll be fine here. But needs further thought for something more universal, you're right
That's a great idea!. I wonder if you could get enough heat into the pipe before damaging the P2's exposed pad though. Perhaps pre-heat it. Watercooled P2, wow.
The big AL substrate should make it easy to try peltier cooling and or heating, and the 50mm fan mounts should help that too (used on the opposite side to the P2)
It'd be nice to get a temp sensor right near that big fat grounding pad to measure temp there. You have a part named THERM (presumed thermistor) nearby but not right at the P2 and somewhat blocked by connectors. It'd measure PCB/ambient temp and rise only once that area heated up. If you respin you might be able to locate something near that fat ground trace which would probably transfer heat out of the P2 faster and be a closer measure to the P2 temp (albeit external not internal which would be far more useful).
Know what you mean. I think we could replace the 10k pullups on P63 and P62 with thermistors too, and those could get closer to the fat ground trace
I'm curious also whether an additional thermistor on P60 (the flash clock / SD chip select) would cause any grief. They're something like 50k nominal, 20~100k ish range from memory
Thermistor pullup on transmit out pin from P2 would be nicer, then you could leave the propplug connected and still measure by floating the TX pin. The latest rev E prop plug seems to show a 470k pull down on this signal but maybe that is quite useful and can help bias the thermistor when not driven by a P2 pin.
Having one thermistor fitted very near the P2 and another away from it could be good for a differential ambient/board vs P2 temp comparison.
I'm not convinced it matters, in that you have an uninterrupted 1.6mm AL substrate immediately below that should spread the heat well in all directions. Also you're still not measuring what you really really want, which is the junction temperature, you're just that little bit closer
Having said that I spent a bit of time mucking about with a potential version 2, and its possible to nest an 0603 thermistor quite close to the Res/P63/P62 pins, but again its not actually contacting the ground, so its still removed from what you want
An alternative better approach may be to connect a bigger thermistor (1206?) across from P62 to the ground, jumping over P63 and Resn, and mostly covering the ground trace to pick up its heat. You'd leave the 10k pullup in, and end up with a divider with the ground leg of the divider variable by temperature, but always above the 1.65v threshold.
Here's an update on first attempt at loading this metal P2 board.
Unfortunately I was unable to clear the shorts between adjacent P2 pins caused by... not ordering a solder stencil. With previous P2 FR4 boards, I've found I can fix shorts using Gel flux and solder wick, but the AL substrate here really resists any effort to heat the pads up, and as a result you can clear the solder higher up the pins, but not at the board level. I guess thats what the AL is there for, to spread the heat effectively.
I'll order a proper stencil but it will be a little while before that comes in. Apart from the shorts everything went smoothly, no tombstoning despite some parts being a size smaller than designed.
Did you try using low melt solder (Indium alloy) on it? That has been like magic for me when needing to remove something from a groundplane. Put the LMS on it with some flux, you can pull it and the other solder off with a desolder wick. I wouldn't have believed how well it worked until I used it.
Great board BTW!
Yeah, looks chilled.
Cool looking board @Tubular , thanks for posting the pic for us to see it. One thing I've found that sometimes helps with fixing fine pitch pad solder bridges is to tip the board sideways while desoldering with wick etc and gravity helps the unwanted molten solder flow in the direction you want. In your case you probably need the solder going towards the chip instead of away and maybe also falling slightly downwards to escape the pins, but it may still help you desolder it. Use plenty of flux. Also hot air may help a bit?
Kind of ironic that the very thing you want to take away from the P2 with that rapid heat transfer is causing these bridges to be a pain to remove.
No, I used regular 60/40 leaded solder, but low temp is a really good idea for the next attempt. I really should try it and get more familiar with it. I do have some somewhere, but its time for a fresh batch I think.
I see also you can get 0.635mm (1/40"?) solder wick, which would be much better than the wider ~3mm stuff i've been using
Thanks evanh. It sure will be chilled (and heated)...
I know what you mean and I'll see if thats possible.
Later on I found my tenma preheater, which might also help a lot. I need to find a way to lock it in place above it, because the clamps that come with the preheater don't go wide enough to accommodate this board. Perhaps I can get there with magnets, or some type of external clamp arrangement (panavise?)
Yeah I feel 'schooled' today after getting about 5 different bits of soldering gear out, several fluxes, magnifiers, brushes, cleaners. It has helped to have a bit of a break from it and read the suggestions here. I'll have another crack at it tomorrow after raiding Altronics for finer solder wick and fresh gel flux.
Unfortunately the 0.5mm pitch is just a bit too fine for making a stencil with our laser, otherwise i'd say it'd be worth going again from scratch.
Bit of an update, I found one of these 100 watt Arcol resistors at home,
https://au.rs-online.com/web/p/chassis-mount-resistors/0188122P?gb=s
I'm thinking of using it as the 'controllable' preheater. It has mounting holes on a 37x35 but I think I can drill new holes on 40x40mm and use the fan mounting holes to couple it tightly to the underside of my metal board
That resistor might also be useful down the track for the temperature testing under P2 control, and it should also fit the proper P2eval board too.
You would only use that Indium alloy solder to remove the chip. You have to remove all of it from the surface to reuse the chip or board, because it melts at like 80 Deg C. That stuff I linked above has saved me a few times when I thought all was lost on a bad solder job.
Ah, got it. I'm going to order some because I think its useful to have in the kit. It does sound pretty amazing.
@Tubular maybe heat the whole board up and they used desoldering gun? No idea if that would actually work...
That's a sticky situation...
Any chance just cutting the bridged pins from the chip would work?
Or, are there some there that you need?
Just had some real progress using the 100 watt resistor to preheat the whole thing to around 150 C, then its easy to work on the top and clear the bridges.
I've cleared the main short that was 1v8 to ground (a true show-stopper), i'm about to go around all the pins and see if any are still shorted to their neighbors, because once I add the inner DIL headers. I think I've damaged a pin or two but it may not matter. Will post photos later
Nice, using the heat sinking to your advantage like that.
That's easier than I would have been thinking using an oven. You've got easy access while the heater maintains temperature.