strange results when put on PCB

Chubbs · Joined: 02 May 2005 Posts: 14

Hi all,
I've been using the PIC16F876A on a modified smoke/CO detector. The system worked fine when I connected everything to the PIC on a bread-board. I have now transferred the system to a PCB, changing none of the connections or the code. The problem I now have is that the ADC conversion of the sensor inputs is a little off. For instance the output of the CO sensor is around a constant 0.9v the output from the PIC shown on an LCD jumps around the actual value, you can get 0.94, 0.84, 0.73, 0.97. Even the input from a pot does not give the correct output. The voltages at the pins of the PIC are constant, correct and do not change, however each time a sample is taken a slightly different result is obtained.
If the input from the pot is 2.5v the output may vary from 2.3 to 2.7v. Everything else seems to work ok the program runs as intented except for the conversions. When done on the bread-board the results were perfect, is there any reason anyone may know as to why this is happening?
Any help will be appreciated.
Thanks

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

jbmiller · Guest

Any chance you've changed the power supply feeding the PIC? How about some induced AC voltage ( hummmm ?)? Bad ground wiring or faulty bypass caps?
hope this helps
jay

Chubbs · Joined: 02 May 2005 Posts: 14

Since the system was working fine before the PCB, I've kept everything the same. It has the same crytals, same caps, same everything. I've checked over all the wiring and it seems fine.
Measuring all the voltages at the ADC input pins shows constant correct values , shouldn't this be enough to ensure a correct relult??
What could effect the result after this point?
All the known voltages on the PIC's pins look ok the reference voltages are ok ( 0.01v and 5.02v ).
What should the levesl be for the crystal???

jbmiller · Guest

Hmm..don't get fooled by a digital voltmeter.they only take 3 or 4 readings per second so your voltage my 'look' stable but is actually dynamic. If you have an oscilloscope try looking with it. Is the PCB layout the same as your prototype? Could you have a 'noisey' line close by the adc pins ? Do you have a really good ground bus ? I know, you've probably already thought of these, just trying to help. heck I've got 20 year old wirewrapped boards still running and even a couple of 877s in those white 'experimentor's boards' that run fine. I suspect some kind of induce AC is leaking onto the adc pins. have you gronded the adc pins and read them ? they 'should' be all zeros..... How about the adc reference ? just more ideas....please let us know what you find out, these 'little problems' can be real head scratchers !!

jay

jbmiller · Guest

Hmm..don't get fooled by a digital voltmeter.they only take 3 or 4 readings per second so your voltage my 'look' stable but is actually dynamic. If you have an oscilloscope try looking with it. Is the PCB layout the same as your prototype? Could you have a 'noisey' line close by the adc pins ? Do you have a really good ground bus ? I know, you've probably already thought of these, just trying to help. heck I've got 20 year old wirewrapped boards still running and even a couple of 877s in those white 'experimentor's boards' that run fine. I suspect some kind of induce AC is leaking onto the adc pins. have you gronded the adc pins and read them ? they 'should' be all zeros..... How about the adc reference ? just more ideas....please let us know what you find out, these 'little problems' can be real head scratchers !!

jay

A.Agiannidis · Joined: 04 Jun 2005 Posts: 2

Had a similar situation in the past. Check for some noise to your analog input pin. You will find that on a fast oscilloscope the level on the signal is not steady but in the form of a waveform. Find the period of that waveform and then instead of sampling one value , sample many times for that period. then average the values and you should see a steady, more realistic output. It is however, that you sample as long as the period. Otherwise, the non deterministic start of sampling would incur different results again.

Mark · Joined: 07 Sep 2003 Posts: 2838 Location: Atlanta, GA

Are you reading just one channel or multiple one?
Did you verify that all Vdd's and Vss's are connected?

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

Try the program shown below. This is a minimalist test program
which should work on your board. I tested it with a 16F877A on
my board, with a 1K trimpot connected to pin RA0, so I could vary
the voltage on RA0 between 0 and +5v. It worked fine.
Here is a typical output, with the trimpot left sitting in one position.