This one has taken me a better part of the day to figure out, and hoping=20
someone here can help me.  I'll try to be brief by putting in only the=20
necessary parts, but feel free to ask for any other details needed.

First, I have an app using an 16F872.  It's basically a digital clock wit=
h a=20
5-column 7-seg display.  4 other I/O's -- one pushbutton switch, 2 extern=
al=20
switched signals (connected via transistor switches), and one output to a=
=20
buzzer.  There are a few issues ...

(1) The F872 and the LED display won't fit properly in the given board sp=
ace. =20
The board is non-rectangular, so I have room for an equal area of smaller=
=20
parts (and I state this for a reason).
(2) Poor accuracy with a 5.120 Mhz crystal oscillator.  It gains a few mi=
nutes=20
every 12 hours.
(3) I need to reduce the power consumption.  It needs to be in HS mode fo=
r the=20
5.120Mhz crystal.

Thoughts/ideas I've come up with so far...

(1) Replacing the 16F872 with both a 16F628 and 12F629 will fit on the bo=
ard. =20
The F628 will use the internal oscillator for driving the display and swi=
tch=20
checking, and a secondary oscillator with 32.768khz crystal on Timer1 for=
=20
keeping accurate time.  The ckt can sleep as needed to save power, since =
the=20
display will be blank quite a lot.  The 12F629 will run off it's own inte=
rnal=20
oscillator just drive the LED columns.  It can be configured as a sequenc=
er,=20
shift-register, or (my preference) a 3-line to 5-line decoder.  Also, I c=
an=20
save power on this by letting the display-blanking input signal also cont=
rol=20
the power to this device.Price of both chips happens to be 3 cents *less*=
=20
than the 16F872.

(2) Instead of the 12F629, since I have 3 lines on the F628 to drive the =
5=20
display columns, I thought of a way to multiplex these lines using=20
transistors.  I'd need 5 transistors anyway, but this way it would be=20
configured differently, so no additional transistors needed.  For 2 digit=
s,=20
one output would be used in this manner...

- PIC output 1 to a resistor to Q1 base (NPN).  Emitter to ground, Collec=
tor=20
to common cathode of digit 1.
- PIC output 1 (the same output) to another resistor to Q2 base (PNP this=
=20
time).  Emitter to common cathode of digit 2, collector to ground.

Output-high would make digit 1 active, output-low makes digit 2 active, a=
nd=20
floating (PIC pin switched as an input) would (should?) leave them both=20
inactive.  Another PIC output would handle 2 other digits in a similar=20
fashion, and the third PIC output would directly drive the 5th LED column=
=2E

I've just mocked it up on a breadboard with a 2N4401, 2N4403, 2 LED's and=
 some=20
resistors, but can't get this to work.  Anyone have a clue what I'm missi=
ng=20
here?  Or how I can achieve the same result using mosfets perhaps?

(3) My third thought is to multiplex the input signals using a comparator=
=20
built into the F628.  I can put the transistors in "series" across the po=
wer=20
rails with resistors between the C and E of each.  With careful planning,=
 I=20
should be able to get a reliable set of input voltages that would indicat=
e=20
which switch/signals are on/off.  However, the problem with this (on pape=
r=20
still) is that I've got to control the comparator voltages which means us=
ing=20
the internal Vref.  The only mode that I see from the datasheet that allo=
ws=20
the internal Vref to feed the comparators, also makes 4 ports analog inpu=
ts. =20
If I set only one of these as an input, and the other 3 as outputs, will =
they=20
function as full digital outputs?

There's got to be a way in here somewhere.

Cheers,
-Neil.

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics