I'm using a AT91SAM7S64 in a project where transmitting data with a modem over an RF channel.
I've found in very rear circumstances my processor is locking up and the only way to reset it is to disconnect the power. The strange thing is that I do have the WDT enabled but it doesn't seem to be functioning when this lockup occurs. I have tested it by butting loops in the program to make it hangup and then the WDT resets the system no problem.
My main problem is that I can't reproduce the situation where the system hangs as I was explaining above. It usually happens during a RF transmition so I suspect it's being caused by the radiation from the antenna but it happens so rarely that I can't reproduce but I would like the WDT to work on these occasions (which it doesn't) because when the processor gets stuck it's in TX mode.
Anyway...my question:

Is there anyway something can cause the WDT oscillator to get stuck and is there any way to work around this ?

hello,
there was an issue with the watchdog timer in revision a of at91sam7s156/128 (see errata of datasheet, search for "The Watchdog Timer May Lock the Device in a Reset State"). but at the at91sam7s64 this problem isn't mentioned.

regards
gerhard

I had a look at it and apparently I'm already doing the workaround.
WDD=0xFFF
WDV=0xFFF

The problem is I am battling to reproduce this lock-up. Murphy says it will only happen every time we are in a presentation of the unit lol (It has happened twice)
They just say in certain rare circumstances and that doesn't help much..

It's ironic that during a project like this you do everything possible to prevent the system from getting stuck but when U want to get it stuck it's impossible

hello,
sounds like a nasty problem.
do you enable user reset?
if xes try to reset the chip in case of hang-up.
if this is working the problem is based on the watchdog.
if not you propably have a emc problem.
are you aware of the schematic checklist (http://www.atmel.com/dyn/resources/prod ... oc6258.pdf)?
there are some hints regarding harsh enviroment.

regards
gerhard

Hi,
So I went through that check-list and tied all the specified pins to ground but it still locks-up.
At least I've been able to reproduce the lock-up by putting my radio antenna and the unit in a specific position.

I enabled the user reset as U recommended and it does reset the controller when I ground the NRST pin during normal operation but when the system locks up it has no effect. The only way to reset the system when it's locked-up is to cut the power (3.3V). This probably means that the WDT could still be functioning but the controller gets stuck.ANother curious thing is when the system is locked-up all the Porifera pins go high.
The problem is that this system is going to be out on the field for hours and it needs to be able to recover from a lock-up like this by itself.

Any suggestions?

hello,
sounds like you really have an emc problem.

maybe the problem is caused by some coupling on the supply rails.
do you have all the capacitors on the supply pins?

do you use the pll? i would assume that this is the most sensitive part of the at91sam7s when it comes to hf imission.

regards
gerhard

As far as I can see I was placing all the caps.
I am using the PLL (need it for USB)

Is there any way to disable the PLL when not using USB (only use it for data download and programming)?
Do you think it will help?

sure, you can use MAINCK as master clock instead of PLLCK. take a look at the datasheet, chapter master clock controller

to be honest i have no idea, but it is very simple to check.

regards
gerhard

So I tried to disable the PLL but it still gets stuck.
I left a post at ATMEL support. We'll see if they respond...

Did anybody come up with a solution for the WDT / CPU hang-up problem? We have EXACTLY the same problem, but using UART transmissions. In our case, this happens only with certain CPU components - we have 10s of identical boards. Only a few of them have this problem; when the CPU is replaced, the problem does not occur.

We have seen this in two different designs; in the first design (a couple of years ago), I simply disabled the watchdog. Since this is now happening in a different design, I can't ignore it, especially since I MUST have the watchdog enabled; we must have something that resets the CPU rather than hangs it up....... We now have customers up in arms...

HELP!

Hi,
Never did get a working solution from Atmel. I eventually went for an external WDT
and I'm now running both. If the one gets stuck the ext WDT will reset.This solution seems to be working for me.
Have a look at the TPS3823 from Texas Ins. It's nice and small

Yup, that's what we did the first time around - we already had an external WD in the system, so disabling the WD in the CPU wasn't an issue.

I'm hoping this is a manufacturing problem with certain series.

Thanks very much for responding so quickly.