ADC on 16F877A (PCM 3.202)

[Gizz]

Hi all.
Pulling my hair out here.
I have the following:
PCM compiler 3.202
16LF877A running on 3.4V and 9.8MHz
I note from this post
http://www.ccsinfo.com/forum/viewtopic.php?t=19616&highlight=adc+channel
that there is an issue with this version of the compiler and the ADC port setup.
What I am finding is that no matter what I set up in the clock speed or the ports for the ADC, and no matter how much of a delay I give it between channel select and reading, I get the same value on all channels that is present on channel 1.
How can I get hold of an update to the compiler? I bought it in June, it seems a bit carp to have to pay $100 for their error.

My code (snippets):

#include <16F877A.H>
#fuses HS, NOWDT, NOPROTECT, PUT, NOBROWNOUT, NOLVP
#use delay(clock=9831960)
#use rs232(baud=9600,xmit=PIN_C6,rcv=PIN_C7,ERRORS,STREAM=gprs)
#use rs232(baud=9600,xmit=PIN_D5,rcv=PIN_D4,ERRORS,STREAM=diagnostic)
#use i2c(Master, SDA=PIN_C4, SCL=PIN_C3) // Software i2c
#zero_ram

...

enable_interrupts(GLOBAL);
enable_interrupts(INT_RDA);
setup_timer_1(T1_INTERNAL);
enable_interrupts(INT_TIMER1);
setup_adc(ADC_CLOCK_DIV_64); // or internal, 32, etc
setup_adc_ports( AN0_AN1_AN2_AN3_AN4 ); // or any combination

...

if (PortNumber == MAINS_VOLTAGE_PORT)
{
set_adc_channel(1);
delay_ms(100);
read_adc(ADC_START_ONLY);
delay_ms(100);
value = read_adc(ADC_READ_ONLY);
fprintf(diagnostic, "Mains reads %lu\r\n", value);
}
else if (PortNumber == DEPTH_PORT)
{
set_adc_channel(5);
delay_ms(100);
read_adc(ADC_START_ONLY);
delay_ms(100);
value = read_adc(ADC_READ_ONLY);
fprintf(diagnostic, "Depth reads %lu\r\n", value);
}


Depth is always same as mains even though using scope and meter the volts at RA5 is 0.2 and the volts at RA1 goes from .6 to 1.2V (depending on unregulated supply volts).
Chip supply stays at 3.4V.

[PCM programmer]

The first thing I would to find the problem is to simply your
program a lot.

1. Get rid of the "if" statements and just check the Depth channel.
2. Get rid of the fancy stuff with the ADC_START_ONLY, and
just use read_adc();
3. Shut down all interrupts.

In other words, the goal is to prove that you can make this
one thing work. So strip the program down until it only:
1. Setup the adc port.
2. Set the adc channel.
3. Read the adc.
4. Report the value read.

When you post your test program, you need to show your variable
declarations as well. For instance, what is "value" ? Is it a byte or
is it 16-bits ?

The issue may well be caused by the state of the justification bit
in one of the ADCON registers which controls whether it's left or
right justified.

If you make a small but complete test program that only tests
the ADC, as described above, then we can hammer the problem
until we find the answer.

[Mark]

You also have the option of manipulating the registers yourself. So if the error is in their setup functions, it won't affect you.

[Gizz]

Thanks, yes I know I should make it simpler: But I have already tried just using read_adc(), etc.
value is int16, and I use the ADC=10 switch on the #device .....
No matter, I will write a test prog that does just that one simple read.
But a quick question, if you will: If I want to set up the adc ports manually, how do I do this? Do I have to use assembler or is there a C variable declared for the ADC setup ports?
Thanks.

[Ttelmah]

[Gizz]

Thanks everyone, I have this just-about-working now.

It exhibited the same probs with a simple program.
I couldn't get the direct bit programming option to work (too impatient!)
Then I decided to earth out the other 5 unused ADC inputs (0, 3, 4, 6, 7) and lo! The 2 voltages (mains and batt) worked... but no depth.... turns out the depth ADC channel is ch 4 even though its wired to AN5 on the chip (yes I checked, several times!)
LOTS of crosstalk between the channels, and as AN5 was floating it appeared to be linked to AN1/AN2. Of course, when the unit has no internal battery fitted AN1 and AN2 are almost identical.

Ah well! Thanks once again!

[PCM programmer]

This is one of the reasons why I always push for a simple
test program. Once you're convinced that the error couldn't
possibly be in the simple program, then you're forced to take
a close look at everything else. By simplifying things, the
error stands out, and is easy to see.

[Ttelmah]

[Gizz]

You are absolutely correct, Ttelmah, AN4 is on RA5.

Never mind, it all works now! Until the next crisis!