State machine problem

rems · Joined: 02 Feb 2012 Posts: 56

I am using a PIC16F887, MPLAB and CCS C.

I am making a combination switch that senses the combo via motion. Due to the nature of certain physical aspects, I need to treat 4 pins as 6 switches. This is why I have the function "scanfaces". It allows me to switch what is i/p and o/p. Basically, I make one pin o/p set low and the other 3 pins are i/p's. Then I make a different pin o/p set low with the rest as i/p's.

I have had luck with the state machine in the bottom of the program when I used 6 different switches. In that case, I used 6 pins, all as i/p's and it worked fine. But now, I'm trying to merge the "scanfaces" function with the state machine and am having a lot of difficulty. I want the led to turn on only when the combo is sensed. As you can see, pins B3 & B4 are shared by different o/p configurations and I think this is part of the problem. I need a way to make them more unique.
Does anyone have any ideas?

Here is my code

Mike Walne · Joined: 19 Feb 2004 Posts: 1785 Location: Boston Spa UK

I don't get what you are trying to do.

Some kind of schematic might help.

Mike

Ttelmah · Joined: 11 Mar 2010 Posts: 19539

Yes, a circuit of the scanning would help.
However just looking at the code, can't see that 'scanfaces' is ever going to work. As it stands, it sets B1, 2, and 3 _all low_, and leaves just B4 as an input.
You need to remember that things are sequential, and happen very fast on a computer. So it starts setting B1 low, then setting B2, B3, and B4 as inputs (actually this can all be done with one instruction - this is exactly what the TRIS statement is for.....). Then only a couple of uSec latter B2 is set as an output, with B3 and B4, once again set as inputs (they already are - things are remembered....), then B3 is set as an output, and B4, once again set as input. At this point, B1, B2, and B3, are _all_ driving low, and only B4 is an input.
Then the routine is existed, and the first test done. At this, B2, is changed to being an input, and immediately read. probably won't work, since it takes _time_ for signals to change.
The similar pattern is then repeated for the other pins....

As a comment, the logic of the code layout also doesn't help understanding. After the 'if' tests, scanfaces is indented, which is normally used as a signe to say 'this is only executed if the test is true', but in fact it is only the single line after the 'if' that if dependant on the test.

It is also 'odd', that you drive the lines 'low', yet seem to test for 'high' as a button made combination.

Post a circuit.

Best Wishes

temtronic · Posted: Wed Mar 28, 2012 5:23 am

You should lookup 'Charlieplexing'....though most use it for outputs, I've used it for input as well.You can get 6 inputs from 3 pins.

John P · Joined: 17 Sep 2003 Posts: 331

Seems pretty trivial. You can get 4 switches where (I'm calling the lines a, b, c and d, nothing to do with port names) a and b cross c and d, and 2 more where a could cross b and c could cross d. Then you can get 4 more where each of a-d can connect to Gnd. If you use pins with internal pullups, there's no need for any external components except the switches. If you can't trust the user to avoid pushing multiple switches, you might need to test for that.

Wouldn't Charlieplexing need diodes? There's no point in using added parts.

rems · Joined: 02 Feb 2012 Posts: 56

Ttelmah,

Thank you, I completely understand your pointers. I was afraid "scanfaces" was happening so fast that it wasn't working. Wow, I didn't realize that the first line after the if is the only one valid for the test condition. That explains a lot. Since my original post, I actually split up "scanfaces" into 3 separate functions.

rems · Joined: 02 Feb 2012 Posts: 56

WOW, my schematic did not look like that when I drew it in the text box before I posted.

++++++++++++++++++++
Schematic fixed by Mod.

How to do:
- Edit text not in forum, but in Windows NotePad (monospaced font).
- Copy & Paste into forum post.
- Enclose in Code Block.
- Preview for correctness.

- Forum Moderator
++++++++++++++++++++

rems · Joined: 02 Feb 2012 Posts: 56

I will look into charlieplexing as well. Thank you for that.

John P · Joined: 17 Sep 2003 Posts: 331

If you want to waste components and your time to wire them up, go ahead.

rems · Joined: 02 Feb 2012 Posts: 56

Here is a link to the schematic.

http://i1249.photobucket.com/albums/hh515/remsphotos/ccs_pic_schematic.jpg

Ttelmah · Joined: 11 Mar 2010 Posts: 19539

This is where {} brackets come in.
Everything inside a pair of these, is (logically) treated as a single statement. So:

John P · Joined: 17 Sep 2003 Posts: 331

Re the schematic, you can eliminate the resistors. Then consider what you can achieve if you don't just use switches to connect the pins to Gnd, but also to each other.

rems · Joined: 02 Feb 2012 Posts: 56

rems · Joined: 02 Feb 2012 Posts: 56

John P · Joined: 17 Sep 2003 Posts: 331

Here's what I think the schematic should be:
http://dl.dropbox.com/u/28291527/sws.jpg

First you would set all 4 outputs low, with the TRIS bits set high (pins hi-Z) and the pullups turned on.

Then set a as the active output, and see if b, c or d go low.
If not, set b as the active output, and see if c or d go low.
If not, set c as the active output, and see if d goes low.
If not, nothing is pressed.

If you had switches to ground as well, you'd insert an initial step where you'd check for anything pulled low while all the TRIS bits were high, before doing anything else.

Actually, if you use connections to Gnd, you could do this job with only 3 pins. 3 switches would connect between the pins (using b, c and d) and 3 would go between each of the pins and Gnd. I don't see why that wouldn't work.