ADC fluctuation PIC16F88

skoberlink · Joined: 27 May 2010 Posts: 52

I'm using a PIC16F88 to read an analog signal and move a motor to a percent of it's range based on that analog signal. I wrote my own function for handling the conversion to avoid using the CCS built-ins. It reads almost properly.

The problem is my motor goes to the right place but then twitches back and forth. If I average several reads and put in a tolerance value for movement (i.e. only move if it has to move at least a certain distance) then the twitching goes away but this significantly reduces my precision. The signal into the PIC is rock solid at exactly the values it's supposed to be.

Any ideas on what might be happening and how to fix it? Here's my read function in case you guys can spot anything:

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

What's your compiler version ?

SherpaDoug · Joined: 07 Sep 2003 Posts: 1640 Location: Cape Cod Mass USA

Look at the numbers coming out of the A/D. How much do they vary? If it is +/- 1 count, that is the best you can do. Is it mostly +/- 1 count with occasional bad values? If so look for something else in the system that happens when the bad value occurs that might be causing it. Also search for "Olympic averaging". It is computationally cheap and good at removing occasional bad values. At the cost of a few more clock cycles and RAM you can also do a "median filter".

On the hardware end make sure your A/D reference is as clean as your signal. One is useless without the other. Double check your grounding.

I will leave your code to others to inspect. It looks OK to me, but I don't use the internal A/Ds much.
_________________
The search for better is endless. Instead simply find very good and get the job done.

skoberlink · Joined: 27 May 2010 Posts: 52

@PCM Programmer
Compiler version is CCS 4.106

@SherpaDoug
I haven't checked the A/D output but it occurred to me while reading your response that this might be the issue. It's a stepper motor and I've been using steps as my tolerance value but the number of steps in one percentage point would change depending on the range of motion given to the motor.

But I'm just thinking "out loud" now. If I'm still having an issue those averages you mentioned looked useful. Also all signals including the reference are clean.

I'm guessing that checking the percent for tolerance instead of the number of steps will fix it. It seems kind of obvious and stupid now...lol

skoberlink · Joined: 27 May 2010 Posts: 52

Well that did not solve it. It was an issue that needed to be fixed but I still get the twitching in the motor. I'll try some of the other average methods mentioned but I'm still open to other suggestions if you guys have any.

EDIT: btw I tried several tolerance values. The ADC reads are +/- at least 10 probably significantly more.

Ttelmah · Joined: 11 Mar 2010 Posts: 19520

It is well worth triple checking how 'rock solid' the input values really are.
A scope for example, will see a signal as 'flat line', that can easily have 10 steps when read with the ADC. Simple DVM's also tend to integrate values, and will not see high frequency noise. If you are using the PIC supply as your 'ref', this _will_ have HF noise on it. The Vref really needs to be from a separate source, and well decoupled, if this is not to be a problem...

Best Wishes

skoberlink · Joined: 27 May 2010 Posts: 52

newguy · Joined: 24 Jun 2004 Posts: 1908

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

SherpaDoug · Joined: 07 Sep 2003 Posts: 1640 Location: Cape Cod Mass USA

It would still be good to know if the A/D values are a wide Gaussian profile (random scatter), or if they are mostly OK with occasional bad readings throw in. If it's the latter, are the bad readings all high or all low? Is there any clear "frequency" to the noise.

Also why did you decide to roll your own A/D code?

As for noise measurement, you may be able to put your scope on AC coupling and then push the gain way up to see small noise signals.
_________________
The search for better is endless. Instead simply find very good and get the job done.

skoberlink · Joined: 27 May 2010 Posts: 52

@newguy
I'm planning to put a .1uF cap on the Vdd line. Do you think that would fix it? I do have a ground plane. I'll look into some of that other stuff.

@PCM programmer
I had missed that line from the data sheet. That could be the problem. Unfortunately I don't have the pins for an external clock. I might be able to turn down the clock speed and see if that helps at all.

@SherpaDoug
I chose to write my own function on the advice of someone more experienced than I. I was having a problem using the built-ins where the voltage coming in to the analog in was at 100% but the motor would only move to 80% of it's range. It had something to do with the math involved in the built-ins. I talked to a friend who has had more experience and he advised me to write my own function. Then I would know exactly what was going on and it was potentially more efficient.

The A/D values seem to be random. I can't see any clear frequency. It all seems completely random.

Thanks for the help so far guys!

skoberlink · Joined: 27 May 2010 Posts: 52

Well slowing it down helped some. I've also discussed it with the guy working on a similar project and we got it down to an acceptable tolerance value. Haven't tried any of the other suggestions yet and probably won't have time to get to it now since it's a little superfluous. Thanks for the help though! All useful info.

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

skoberlink · Joined: 27 May 2010 Posts: 52

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

The 16F88.h file lists these options for the A/D clock, to be used in the
setup_adc() function: