18F4550 Sleep Current Drifting Upward as Temperate Decreases

[foodwatch]

Since these are surface mounted parts it will be difficult to remove the chips without damaging the board (I only have two prototypes built ). I have opened a ticket with Microchip to see if there are any problems related to production. I am ordering new chips and will likely make another board but will wait to see if microchip can add any light to the darkness.

These were not samples, but were purchased from a distributor. I have about 30 in stock with a 2006 datecode, but will see what microchip says (if anything helpful). Thanks

[horkesley]

Hi,

Have you tried cooling the PIC in a fridge/chamber?

I think blasting it with a freezer spray is not cooling the chip equally all over. I Know some chips don't like this.

Regards.
_________________
Horkesley Electronics Limited

[foodwatch]

I have done it in a refrigerated chamber and in a freezer chamber. Same ultimate results, but the refrigeration chamber does not increase the current as much as the freezer. Basically, the colder it gets, the more current it draws until it reaches a limit of about 3ma.

[TIMT]

Not sure if this is of any use but the WDT does seem very sensitive to temperature. It can be used to measure temperature - see AN828

Could this be causing the problem?

http://ww1.microchip.com/downloads/en/AppNotes/00828a.pdf


Tim.
_________________
Tim

[foodwatch]

I looked at the wdt thermometer app you sent. It certainly effects the WDT period, which I did anticipate in my design, but the WDT continues to function at cold temps albeit at a slower rate. I am still waiting for a reply from microchip on the current issue. I have ordered a new set of cpu's and will build another set of boards with the most recent fab and see if things change.

[SherpaDoug]

[foodwatch]

Somewhere I have a copper tool I made to fit my soldering station. It was made to fit the tqfp package... If I can find it I will try... otherwise I can try your suggestion, but with my heavy hand I will likely go through the pin and sever the pad as well.

[treitmey]

Cutting the pins is exactly what I do also.
If you cut right next to the ceramic chip of the tqfp package you won't be
over the pads. Tho you could cut a trace leading to a via under the chip.
But you have to work hard to cut all the way through the trace.

Also look at
http://www.chipquik.com/

[Ttelmah]

The watchdog on the chip concerned, uses a completely different design to the older chips. The application note quoted, does not apply to this chip.
The unit on this chip, runs the watchdog off the internal RC oscillator, which unlike the older units, is calibrated, and quite reasonably accurate. The quoted accuracy, is +/-18% over the entire temperature and voltage range. This is not the problem....

Best Wishes

[foodwatch]

A new CPU arrives tomorrow (latest datecode). I have my fingers crossed. Microchip said they are "studying" the problem and will let me know. Thanks to all for your input...

[foodwatch]

Got new CPU today with 2008 datecode, Microchip support finally contacted me and suspects the high current is due to the wdt oscillator slowing down and drawing more current when cold. They asked me to run tests with wdt turned off in sleep mode. I did this tonight....still drawing much higher current than normal even with nothing at all connected and a simple sleep command. I bounced it back to Microchip. I can't believe no one has ever run an 18f4550 in a cool environment using a battery.

[foodwatch]

I think Microchip is stumped... at least they asked me to do many things and provide data... now that they have the data the correspondence has stopped. I stuffed my pcb one part at a time and adjusted the i/o pins as I added the parts. I now have my sleep current at 6ua and it works beautifully... at a balmy 40-70F. At room temp, all is well, at about 40F the current starts to rise slowly with falling temp... by the time I reach 0F, the cpu is drawing 250-300ua. If I gently heat the case of the cpu with a soldering iron set to 90F for about 3 seconds and we are back to 6ua. All the time the code is functioning correctly... only the current suffers. I have no hair left to pull out....

[Ttelmah]

I would be trying something else at this point.
I wonder, if the problem is actually something like the supply regulator tripping into oscillation. So as you cool the PIC, it's loading characteristic changes a tiny amount, tripping an oscillation in the supply, that is actually causing the massive current. Have you actually tried disconnecting the PIC, from the board supply pins, and feeding it directly?. What is the regulator used?. What components are used to prevent oscillation (capacitor types, sizes, and placement)?. If you are already running with an external supply, and a DVM, in line to measure the current, what smoothing components ar after the DVM shunt, and how are these placed relative to the PIC?.

Best Wishes

[foodwatch]

I have looked at the supply with a scope and 10:1 probe and there is no difference at all in supply ripple and no hint of oscillation. I have tried several different regulators and have settled on Microchip's MCP1703 (a suggestion from an earlier post). This regulator appears to be quite good. I have run the PIC directly from a super well regulated bench supply and a 9v battery all with identical results. I am even using a different board layout than my original...again all the same results. I am stepping up the heat with Microchip today via telephone and see if they at least can simulate and acknowledge the problem... sure seems strange other battery powered PIC users have not noticed this... although not many are in freezers :-) Thanks to all for your suggestions. Maybe I will find away around the lack of eeprom and try the 24j series.

[foodwatch]

After Microchip could not duplicate the problem (they tested with a 40 pin dip instead of tqfp package), I thought I would try the dip and see if I could replicate their successful results... no luck. I then dug into the data sheet and the CCS device file carefully looking for a missed pin to properly terminate. Turns out there is an pin_E3 which I could not find on the pinout of the chip. I set to output low anyway, and the problem went away. A symptom to look for is that the current rises if you touch the top of the chip with a finger (without touching pins of course). I noticed this even at room temperature. Anyway, I guess the lesson is still find and terminate all i/o.... especially analog port pins.

Thanks to all for the suggestions. I am now running with a/d and rs232 at under 8ua.