supply voltage problem

championx · Joined: 28 Feb 2006 Posts: 151

Hi! here I am with another strange thing.

I have a board that measures some variables and send it over UART to a TTL device.

My board operates with 12vcc provided by a wall power supply.

The other device, the one that receive the data operates with an internal battery.

Everything works fine... BUT... when i disconnect the wall power supply (my board lost power but the other device don't) wait a few seconds and then reconnect the power supply, then my board doesn't work.

What i think is happening is that the pic get some current over the TTL lines and it doesn't restart and go to a strange state... then.. never get back to earth... Confused

These are my fuses:

Ttelmah · Joined: 11 Mar 2010 Posts: 19481

One possibility is it is nothing to do with voltages.
What is your UART setup line on the 'slave' device?.
Unless it has 'ERRORS', the UART could be hung by a continuous zero. Once hung, the code will not recover, unless you either do it yourself, or use 'ERRORS' which adds code to do this.

championx · Joined: 28 Feb 2006 Posts: 151

Hi Ttelmah, thanks for your answer, yes, it have the ERRORS added on the UART line.

I will add a led to check if its a UART hang or a pic hang...

Any other suggestion?

thanks!

Ttelmah · Joined: 11 Mar 2010 Posts: 19481

What pins are you using for the 'TTL' device?.
Do you have any signal protection on the lines?.
If you are unplugging the master, then the slave becomes ungrounded. What does it actually connect to?.

championx · Joined: 28 Feb 2006 Posts: 151

I'm using the uart pins to connect to the TTL device.

No protection on these lines, just a 1k resistor in series.

Both grounds are the same and always connected.

If I just cut the board positive, wait a few seconds and then reconnect it then the board doesn't work anymore.

newguy · Joined: 24 Jun 2004 Posts: 1907

Grab a multimeter, cut the main board's power again, and then measure main's Vcc. Measure any digital lines that run back to the main processor, namely the (still connected) UART's TX and RX lines. You might be getting some partial feed through the logic lines and that's causing the processor to go into a weird lockup state. Does the oscillator still run while it's connected like this (can you measure a clock signal)? Are you using the watchdog on main?

championx · Joined: 28 Feb 2006 Posts: 151

Hi newguy, thanks for your help, i will make these measurements and post back.

I'm using the watchdog timer too.

temtronic · Posted: Wed Jan 11, 2017 2:05 pm

comment.

Get rid of the watchdog timer ! There is zero need for it during testing, up until the product is ready to be shipped to clients. THEN and only then, enable and test for 1-2 weeks.
While it _might_ not be causing a problem, trouble shooting is a process of 'eliminations'. You have to get rid of 99% of the code to SEE the 1% that might be the cause.

While it sounds like backfeed of power might be the issue, I'd like to know how you've connected the 2 devices as you say ground is always connected.

Jay

championx · Joined: 28 Feb 2006 Posts: 151

Thanks temtronic, yes, the ground between the device and my board is always connected. And the ground from the power supply to the board is always connected too. The only wire that is disconnected is the positive coming from the power supply.

Ttelmah · Joined: 11 Mar 2010 Posts: 19481

What actual _pins_. Point is this affects what can be affected. Some pins on the PIC do not have the protection others have. I'd never rely on the internal protection diodes for something going 'off board'. It is possible if you are getting a nasty spike, for the FET's to actually become latched. If that happens, a soft reset won't restart the chip (had to be hardware).

gjs_rsdi · Joined: 06 Feb 2006 Posts: 468 Location: Bali

Ttelmah · Joined: 11 Mar 2010 Posts: 19481

Also, another very important thing is how the grounds actually _route_. One (almost classic) would be that when the 'main' supply goes off the ground potential moves relative to the ground of the second system. There used to be a quite famous telescope sold commercially, which if you connected it to a computer would give problems when the computer switched off. The reason was that though it had a ground, there was in some cases a small potential difference between this and the computer. This connected through the serial port (a few years ago now...), and the spike on switch on/off could actually damage the input transistors...
The solution was to opto-isolate the signals, so that the grounds were not actually connected. Really there should only ever be one actual ground connection if the systems are permanently linked. Two grounds is a formula for an earth (or ground) loop (look it up...).

jeremiah · Joined: 20 Jul 2010 Posts: 1343