Author Topic: ATmega TQFP package overheating  (Read 21104 times)

0 Members and 1 Guest are viewing this topic.

Offline richardbb

  • Newbie
  • Posts: 8
  • Country: gb
Re: ATmega TQFP package overheating
« Reply #50 on: September 07, 2013, 11:11:13 pm »
The datasheet says AVCC must be connected always - it powers some of port C as well. Maybe there is some internal path causing the overload. If you don't care about ADC noise you can just wire it directly to VCC
 

Offline Psi

  • Super Contributor
  • ***
  • Posts: 9889
  • Country: nz
Re: ATmega TQFP package overheating
« Reply #51 on: September 07, 2013, 11:34:40 pm »
Based on your circuit i cant see any obvious reason for the excess voltage. (Assuming your regulator is ok)

However, here's a few notes/questions that may help us narrow down where the problem is.

1) Connect AVCC to VCC,  this shouldn't be the problem but your supposed to connect it.

2) Disconnect "Temp in" and "PWM out" from other things until you get the mcu working and 5V showing on VCC.

3) Did you get excess voltage with a new ATMega containing no code?
Yes = proves the problem is not to do with the mcu switching the fet and causing some sort of inductive spikes
No = Remove the fet and confirm 5V is present and the mcu is working (flash an led or something)

4) Check if you're getting this overvoltage issue with the ISP cable unplugged.
I very much doubt this is the issue but some programmers support 12V high voltage programming. It shouldn't be doing that but might pay to check its not trying to feed in 12V.

5) Do you by and chance have a large length of wire connected to this circuit? like 5+ meters
(eg, you might be building something that runs over a length of cable.)  Any time you have a long length of wire and something that switching you can produce nasty higher voltages.  I was once switching 24V through a fet from a micro and driving the signal into a 100 foot length of cat5. I was getting 160V ringing on the cable!

5) Can we have a photo of your circuit to check. (in case there's something wired up in a different way to your diagram /pcb  and you haven't noticed it)

6) Put a 10k pulldown on the gate of that fet.

7) Double check that the VCC feed to your 2.8k and 10k resistors is actually from the 5V side of the reg and not the input side.

8 ) The copper tracks on your pcb, is that on the top or bottom of the pcb.  I'm just wondering if there is a mirroring problem and the reg is connected up inverted.
(eg, reg pads are numbered for chip placement on bottom of pcb but the pads are on the top layer, or vise versa)
« Last Edit: September 08, 2013, 12:14:33 am by Psi »
Greek letter 'Psi' (not Pounds per Square Inch)
 

Online Andy Watson

  • Super Contributor
  • ***
  • Posts: 2074
Re: ATmega TQFP package overheating
« Reply #52 on: September 07, 2013, 11:35:18 pm »
Well 5.14V on the zener, 4.67 after the reg. Looks pretty civil, will it all suddenly turn into a mysterious voltage multiplier when i put the atmega back in ?
The regulator has a 1.7V drop out voltage - you're not giving it much chance to regulate.  I would guess that the regulator is unstable - what is R1 doing? I would also bypass C1 with a lower value cap - just to ensure there's no possibility of inductance creeping in.
« Last Edit: September 07, 2013, 11:45:20 pm by Andy Watson »
 

Online tom66

  • Super Contributor
  • ***
  • Posts: 6678
  • Country: gb
  • Electronics Hobbyist & FPGA/Embedded Systems EE
Re: ATmega TQFP package overheating
« Reply #53 on: September 07, 2013, 11:39:44 pm »
Have you got the minimum input and output capacitance near the regulator, tied your regulator's power ground to one point (input and output capacitor grounds tied together at the regulator ground), and ensured you have decoupling caps generously placed around the ATmega?
 

Offline David_AVD

  • Super Contributor
  • ***
  • Posts: 2797
  • Country: au
Re: ATmega TQFP package overheating
« Reply #54 on: September 07, 2013, 11:48:22 pm »
Sorry, but that layout is pretty ordinary.  The 0V line needs to be thought out better.

The 100n cap should be nice and close to the micro; right on the 0V and Vcc pins, not lurking off to the side on long traces.

As far as I know Avcc should not be left floating, even if not using the A2D.  I assume you're using the A2D with that "temp in" anyway?

I also notice you have pins 1 and 8 4 and 5 on the 78L05 connected to 0V.  The spec sheet says "NC" and i always take that to mean "don't connect to anything", not "can connect to anything".
« Last Edit: September 08, 2013, 12:20:56 am by David_AVD »
 

Offline Psi

  • Super Contributor
  • ***
  • Posts: 9889
  • Country: nz
Re: ATmega TQFP package overheating
« Reply #55 on: September 07, 2013, 11:58:55 pm »
Most of the time "NC" is physically not connected but yeah, always a bit risks using them.
It would be a good idea to check with the DMM if he has a spare reg IC handy.

They tend to say DNC for pins which shouldn't be connected to anything.
Greek letter 'Psi' (not Pounds per Square Inch)
 

Online Andy Watson

  • Super Contributor
  • ***
  • Posts: 2074
Re: ATmega TQFP package overheating
« Reply #56 on: September 08, 2013, 12:04:39 am »
You might also consider what happens with the unused pins on the processor. I'm not too familiar with the ATMEL parts, but in general, allowing unused inputs (especially digital inputs) to float can lead to excessive current consumption and/or oscillation.
 

Offline Psi

  • Super Contributor
  • ***
  • Posts: 9889
  • Country: nz
Re: ATmega TQFP package overheating
« Reply #57 on: September 08, 2013, 12:06:39 am »
Floating inputs on a ATMega is fine.  It's not good practice but wont cause this sort of catastrophic failure.

I'm fairly sure its a wiring/connection error somewhere, regulator input voltage is getting onto the mcu somehow.

Having the regulator soldered around the wrong way would do it and would be very easy to accidentally do, since the pcb was designed with the reg rotated 180deg.
« Last Edit: September 08, 2013, 12:13:47 am by Psi »
Greek letter 'Psi' (not Pounds per Square Inch)
 

Offline David_AVD

  • Super Contributor
  • ***
  • Posts: 2797
  • Country: au
Re: ATmega TQFP package overheating
« Reply #58 on: September 08, 2013, 12:19:28 am »
The 0V routing is pretty horrific.  People often underestimate the importance of good power and ground routing.  The having the common 0V connection (wire) as the star point would be much better.
 

Offline Psi

  • Super Contributor
  • ***
  • Posts: 9889
  • Country: nz
Re: ATmega TQFP package overheating
« Reply #59 on: September 08, 2013, 12:23:51 am »
Agreed, but i doubt that's his issue.

I've never heard of bad grounding/power causing more volts to appear :)  its usually the other way around.
Might cause oscillation that the DMM could misread as DC i guess.

Best to use the scope instead for debugging weird issues like this.
« Last Edit: September 08, 2013, 12:26:27 am by Psi »
Greek letter 'Psi' (not Pounds per Square Inch)
 

Online SimonTopic starter

  • Global Moderator
  • *****
  • Posts: 17728
  • Country: gb
  • Did that just blow up? No? might work after all !!
    • Simon's Electronics
Re: ATmega TQFP package overheating
« Reply #60 on: September 08, 2013, 07:08:07 am »
Thanks for all your comments guys. please note as per further up I have removed and bridged the regulator and swapped the 18V zener for a 5.1V one so no more power problems.

Yes the MCU overheats before being programmed

I have run this circuit on the DIP version of the MCU no problems.

With 5V through out the supply and sizable resistors from it to any I/O I fail to see where it is drawing so much power from.
 

Offline jahonen

  • Super Contributor
  • ***
  • Posts: 1052
  • Country: fi
Re: ATmega TQFP package overheating
« Reply #61 on: September 08, 2013, 08:36:41 am »
How much load there is to the regulator, I think those 78xx-regulators need few milliamperes of output load to work properly.

Regards,
Janne
 

Offline Psi

  • Super Contributor
  • ***
  • Posts: 9889
  • Country: nz
Re: ATmega TQFP package overheating
« Reply #62 on: September 08, 2013, 08:46:35 am »
Can we get a photo of the actual pcb?  top and bottom?
Greek letter 'Psi' (not Pounds per Square Inch)
 

Online SimonTopic starter

  • Global Moderator
  • *****
  • Posts: 17728
  • Country: gb
  • Did that just blow up? No? might work after all !!
    • Simon's Electronics
Re: ATmega TQFP package overheating
« Reply #63 on: September 08, 2013, 11:08:53 am »
How much load there is to the regulator, I think those 78xx-regulators need few milliamperes of output load to work properly.

Regards,
Janne

I am using a zener now, still same result
 

Online Andy Watson

  • Super Contributor
  • ***
  • Posts: 2074
Re: ATmega TQFP package overheating
« Reply #64 on: September 08, 2013, 11:21:08 am »
Try 10-100nF on the pins of the processor supply - otherwise there is a relatively large loop back to C3.
 

Online SimonTopic starter

  • Global Moderator
  • *****
  • Posts: 17728
  • Country: gb
  • Did that just blow up? No? might work after all !!
    • Simon's Electronics
Re: ATmega TQFP package overheating
« Reply #65 on: September 08, 2013, 11:25:05 am »
photos attached
 

Offline David_AVD

  • Super Contributor
  • ***
  • Posts: 2797
  • Country: au
Re: ATmega TQFP package overheating
« Reply #66 on: September 08, 2013, 12:11:07 pm »
Where does that TEMP wire come from?  Is it connected when the fault appears?
 

Online SimonTopic starter

  • Global Moderator
  • *****
  • Posts: 17728
  • Country: gb
  • Did that just blow up? No? might work after all !!
    • Simon's Electronics
Re: ATmega TQFP package overheating
« Reply #67 on: September 08, 2013, 12:12:11 pm »
it's a resistor to ground
 

Offline mmagin

  • Frequent Contributor
  • **
  • Posts: 610
  • Country: us
Re: ATmega TQFP package overheating
« Reply #68 on: September 08, 2013, 05:57:37 pm »
6) Put a 10k pulldown on the gate of that fet.

This was bugging me as well.
I don't think it's necessarily related to this problem, but it's a darned good idea anyway.  When that IO pin is Hi-Z, the gate is effectively floating.
 

Online SimonTopic starter

  • Global Moderator
  • *****
  • Posts: 17728
  • Country: gb
  • Did that just blow up? No? might work after all !!
    • Simon's Electronics
Re: ATmega TQFP package overheating
« Reply #69 on: September 08, 2013, 06:09:07 pm »
well it will always be working so long as the micro is on and no not to do with this problem. I'm redesigning based on a tiny85
 

Offline madworm

  • Frequent Contributor
  • **
  • Posts: 371
  • Country: de
Re: ATmega TQFP package overheating
« Reply #70 on: September 08, 2013, 06:20:48 pm »
1) The board is gunked up pretty good. May not be an issue, but... it lacks style.

2) The alignment of the TQFP package is quite a bit off. Who knows what's going on underneath.

3) If the chip gets hot instantly and can't be programmed anymore, it's probably dead. I've killed one or two myself by accidentally applying over-voltage (about 12V).

4) If you redesign it anyway, add the FET pull-down resistor. Otherwise it might switch on willy-nilly.
 

Offline mmagin

  • Frequent Contributor
  • **
  • Posts: 610
  • Country: us
Re: ATmega TQFP package overheating
« Reply #71 on: September 08, 2013, 06:43:14 pm »
well it will always be working so long as the micro is on and no not to do with this problem. I'm redesigning based on a tiny85

I was thinking of the fact that the IO lines not involved in programming go Hi-Z while programming. (Unless I'm misremembering.)
 

Online SimonTopic starter

  • Global Moderator
  • *****
  • Posts: 17728
  • Country: gb
  • Did that just blow up? No? might work after all !!
    • Simon's Electronics
Re: ATmega TQFP package overheating
« Reply #72 on: September 08, 2013, 06:50:30 pm »
well the pwm will be working all of the time, it is the sole purpose of the circuit. the PWM output will not be in use while programming and infact on the tiny85 version the pwm line will be a programming line too so i prefer to not put a load on it.
 

Offline miceuz

  • Frequent Contributor
  • **
  • Posts: 387
  • Country: lt
    • chirp - a soil moisture meter / plant watering alarm
Re: ATmega TQFP package overheating
« Reply #73 on: September 08, 2013, 06:58:02 pm »
As you will be making a new board for a new design, leave copper on the bottom side and attach it to ground. I usually leave copper pour on the top side as well and try to route traces so that I don't split it too much. But don't leave unconnected copper pours - if some area becomes an "island" - either remove it (there are automatic options in Eagle and Kicad) or just add via to a bottom ground plane.

Flux is conductive. I once had a problem when my AVR board was not working till I cleaned the flux from it.

Regarding your voltage regulator problem -- I would guess something was very very wrong with the pinout of IC. Can you measure how much current is being drawn?

I'm not sure if unconnected AVCC might cause any problems, but datasheet explicitely tells you to connect it.

Read http://www.atmel.com/images/atmel-2521-avr-hardware-design-considerations_application-note_avr042.pdf on how to do everything "right" according to Atmel.

Offline miceuz

  • Frequent Contributor
  • **
  • Posts: 387
  • Country: lt
    • chirp - a soil moisture meter / plant watering alarm
Re: ATmega TQFP package overheating
« Reply #74 on: September 08, 2013, 07:02:37 pm »
well the pwm will be working all of the time, it is the sole purpose of the circuit. the PWM output will not be in use while programming and infact on the tiny85 version the pwm line will be a programming line too so i prefer to not put a load on it.

Once I had a problem in exact situation - just before etching a board I've substituted BJT's for mosfets and forgot to add pulldown resistors. No wonder I couldn't program the board when mosfets were randomly switching 4 high power IR LEDs 100mA each when powered from USB only - PC was resetting the USB port because of high current draw.


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf