Author Topic: ATmega2560 board overheating any idea's (Read 2962 times)

HendriXML · « **on:** September 27, 2019, 12:01:13 pm »

I recently purchased two ATmega2560 pro boards. One is working fine, one seems to be overheating.

The picture is of the faulty one. The board seems to connect via usb just fine, but after 20 seconds or so. It quits. At that time the lower left part of the board (where the regulators are) is very warm.

With the naked eye nothing special can be seen, but also on the photo I can't see anything unusual. What could be drawing this much power?

I can ask for a refund, but it could be more learning full to fix it myself.

I haven't done any probing yet, but if I had to make a wild guess, I think the 3.3V regulator is misbehaving.

HendriXML · « **Reply #1 on:** September 27, 2019, 12:13:52 pm »

At the usb connector I see something that might be bridging the two bottom pins. Will make a close up shot from that section later. But seems unlikely that that could be the cause.

Psi · « **Reply #2 on:** September 27, 2019, 12:41:23 pm »

Which one of the two regs are getting hot, the 3.3 or 5.0?

Try erase the chip. (to confirm it's not a short happening from code enabling an output which is shorted high/low etc..)

mariush · « **Reply #3 on:** September 27, 2019, 12:54:17 pm »

The usb connector seems a bit misaligned. It almost seems the USB OTG ID pin is shorted to the GND (last two pins).

1117 regulators are known to oscillate in some situations ... most regulators in the 1117 regulators require a capacitor on the output with ESR between 0.1 ohm and 1 ohm ... these boards use ceramic capacitors that have esr lower than 0.1 ohm
It wouldn't hurt to desolder the output capacitors on each regulator (if i were to guess the big ceramic cap below the fuse and one of the two ceramic capacitors below the two resistors marked "220") and replace them with 10..100uF 6.3v or higher electrolytic capacitors.

You could always pour some sanitary alcohol or isopropyl alcohol on the board and turn it on and see where the alcohol evaporates the fastest.. that could be a shorted component. The regulator being hot could be a side effect of another component on the board drawing a lot of power.

magic · « **Reply #4 on:** September 27, 2019, 02:15:08 pm »

Don't try to program anything unless all voltage rails are verified within norm.

And of course start with checking voltage on all rails.

HendriXML · « **Reply #5 on:** September 27, 2019, 02:29:51 pm »

I made a close up of the USB connection. There was something shiny which could be scratched away.

This the most closeup I can reach using my macro objective.

The regulator on the left is BTW the one that heats up painfully hot. Shouldn't that one only do stuff with external power, not on USB?

HendriXML · « **Reply #6 on:** September 27, 2019, 02:53:13 pm »

I've put 5V on the 5 volt line, and then it does not heat up like crazy, the current draw is then 70 mA. I'm not sure whether the board then functions.

Because I cannot use the USB connection yet and I don't now which program it should run. It seems to do some blinky stuff, but not slow.

So the problem seems to be that regulator, and when it is bypassed there isn't a huge current draw.

HendriXML · « **Reply #7 on:** September 27, 2019, 03:58:37 pm »

When I plug the board on a powerbank, there's no excessive heating up and it boots up normal.

When I attach it to a laptop it's 82.4 deg in seconds and still heating up..

Interesting??

An out of control regulator as mariush pointed out could explain the differences between the 2 power sources.

I think I'll try disputing before soldering myself. If it is a design flaw, then a little bit of feedback can't hurt.

HendriXML · « **Reply #8 on:** September 27, 2019, 04:22:21 pm »

Just to be sure, that 5V regulator is not really needed, is it?

I could just remove it, if I use the board with USB or a regulated 5V source.

thinkfat · « **Reply #9 on:** September 27, 2019, 07:33:03 pm »

If you're dead sure there's no short or rather semi-short on the 5V rail I'd say you can remove the 5V regulator. But unless you're sure, something else might be getting hot instead, in your PC or in the power bank. Did you check the 5V rail with e.g. a multimeter? How's the reading in resistance mode and diode mode? Also check the capacitors around the 5V regulator, one might have cracked and is now a "wire".

thinkfat · « **Reply #10 on:** September 27, 2019, 07:43:07 pm »

BTW, the board was on external power while it went hot? How high was the voltage? Don't go too high on these linear regulators. You mentioned there's 70mA going into the 5V rail. That'd be 350mW, which is a bit much for just an ATmega. But anyway, if you tried driving the board e.g. from 12V, the LDO would have to burn up almost 500mW. No wonder it'd get hot. I think the problem is maybe not the 5V regulator but something connected to the 5V rail.

HendriXML · « **Reply #11 on:** September 27, 2019, 07:56:22 pm »

I tested 3 power sources:
1) usb on laptop, resulting in overheating very fast and ongoing
2) psu, 5V at 5V in/output: current draw 70mA
3) powerbank via usb, no overheating

If I remove the regulator I can test wether it even works. I know it had a valid usb connection before, but that got broken each time. I think the usb power was shutdown, but thats just guessing.

What I don't understand is why the 5V regulator only in one situation dissipates that much.

HendriXML · « **Reply #12 on:** September 27, 2019, 08:56:51 pm »

I've taken off the 5 regulator. And now it doesn't become hot any more!!

Kidding.. I now can upload script to it, and have blinky working, yay..

To desolder it I took some thin wire under the 3 pins to add some tension. That way they became loose from the pcb after heating them up with a normal solder tip. Maybe there are better ways but this worked.

The current draw is about 60 mA. I don't think something else is wrong with it, but I may be wrong. The current is a bit high as said, but there's also the usb chip, the 3.3 volt regulator, 2 leds (one blinking). Maybe the script that ran at it was a bit more consuming as well. It certainly wasn't the normal blinky.

Thanks for the suggestions!

SiliconWizard · « **Reply #13 on:** September 27, 2019, 11:14:27 pm »

Don't overclock your Atmega's!!

HendriXML · « **Reply #14 on:** September 28, 2019, 12:03:08 am »

I've read that the 168 and the 328 can run at 20 Mhz, so they are mostly underclocked

I know these aren't powerful processors, but one can still do lot of stuff with them. I really like the smaller "pro" board, but mounting one in a box is not trivial.

There are 2 large holes, not near center of gravity, but at a location that, eh.. had some room left I think.

But why make them so big, how large of a screw does one need to bolt this down. Why not have 4 smaller holes? Will have to think about this.

The idea is to use the photo to get the right dimensions. Have to see wether a reliable workflow can be constructed for that.

Edit - in Fusion 360 one can have 2 points in a canvas image and set the distance in reality between those. After that, it is on scale. That makes it easy to locate the centers of the holes with high accuracy. (I hope) The photo should not have significant perspective errors then.

HendriXML · « **Reply #15 on:** September 28, 2019, 12:53:20 am »

TO 220 washers can probably be used when screwing this board down, without concerns about shorting or hurting the board in other ways.

maginnovision · « **Reply #16 on:** September 28, 2019, 01:30:16 am »

Quote from: SiliconWizard on September 27, 2019, 11:14:27 pm

Don't overclock your Atmega's!!

I overclocked a 2560 to 24MHz with no issues. Batch of 20. However trying beyond 24MHz exactly 0 worked. 50% overclock though not bad. Not even a glitch despite using external SRAM(and on board xmm) and uSD Card with fatFS. It's more about testing than just not doing it.

magic · « **Reply #17 on:** September 28, 2019, 06:54:30 am »

That's low. I have two projects which run AT90USB162 at 36MHz (but only one unit tested). Above that the USART starts to give up

60mA seems quite a lot, measure the other board for comparison.

maginnovision · « **Reply #18 on:** September 28, 2019, 07:24:47 am »

Quote from: magic on September 28, 2019, 06:54:30 am

That's low. I have two projects which run AT90USB162 at 36MHz (but only one unit tested). Above that the USART starts to give up

60mA seems quite a lot, measure the other board for comparison.

That's pretty good! I couldn't get any of the big megas to even startup beyond 24MHz. Even changing the voltage they wouldn't run at 25MHz.

magic · « **Reply #19 on:** September 28, 2019, 08:06:30 am »

Maybe smaller ones are better.

I found this thread while researching OC-bility of AVRs

https://www.avrfreaks.net/comment/2425576#comment-2425576

I could swear I have also seen a video of somebody running a trivial LED blinker clocked from a function gen and pushed it to some 50MHz but I don't remember which chip it was. Of course such code doesn't exercise much of the peripherals or even RAM.
Some people say flash may be the bottleneck.

By the way, IIRC I also got ATmega32U4 to run at 27MHz with USB support, using some weird PLL division factors. Or maybe it was using the internal oscillator to clock USB

USB clock frequency was quite off per the USB spec, but USB2.0 hosts (at least those in computers) are very tolerant of clock skew at USB1.x speeds due to the massive oversampling afforded by a 480Mb/s PHY processing a 12Mb/s signal.

maginnovision · « **Reply #20 on:** September 28, 2019, 08:51:20 am »

Might be the flash, I really couldn't get them to do anything. The project needed most of the IO in the 100 pin package and I went for larger flash models(1280 and 2560) because I needed to store a few 8kB images. Using external SRAM with XMEM interface was 20 IO then I needed another 50 or so IO for interfacing with everything else. 24MHz seemed pretty good. 0 Errors over 192 hour test period with sram, cpld interface, sd card, and display. Now I want to build an overclock test board and order a bunch of different chips to see what I could do.

RoGeorge · « **Reply #21 on:** September 28, 2019, 09:20:05 am »

It's the AMS1117 regulators.

If the PCB is not properly designed, or they use the wrong type or values for the 1117's capacitors, it will self-oscilate and heat itself badly.

HendriXML · « **Reply #22 on:** September 28, 2019, 09:27:50 am »

To my understanding it (AMS1117) didn't get power on the input, so that means it can do crazy stuff when an output from another source is applied?

One thing I asked myself is how they "merge" different sources in this design, when external power would be applied at the input and USB as well. In one of the sources there would actually be a little over voltage situation. (Reversing current)
Or is this a design flaw, of which the heating up is a result?

In that case removing the 5 regulator -if not needed- would be preferable.

maginnovision · « **Reply #23 on:** September 28, 2019, 05:49:36 pm »

Most of my small 5V linear regulators have about 1.2K $\$\Omega\$$ output to ground. One I happen to have nearby burned about .5W with 5V on the output and nothing applied to input.

HendriXML · « **Reply #24 on:** September 29, 2019, 12:59:39 pm »

Quote from: maginnovision on September 28, 2019, 05:49:36 pm

Most of my small 5V linear regulators have about 1.2K $\$\Omega\$$ output to ground. One I happen to have nearby burned about .5W with 5V on the output and nothing applied to input.

So it's safe to say they are not designed to operate in that way. Depending on the power source it can draw a little or a lot (in my case).
If external power is needed, then not using usb and thus programming via ICSP would be an option.


EEVblog Main Site	EEVblog on Youtube	EEVblog on Twitter	EEVblog on Facebook	EEVblog on Odysee

Author Topic: ATmega2560 board overheating any idea's (Read 2962 times)

Share me