How can I stabilize the ADC reading?

MCUprogrammer · Joined: 08 Sep 2020 Posts: 221

Hello to everyone
I am reading millivolts from adc input. The processor and clock frequency I use have been stated below.

Ttelmah · Joined: 11 Mar 2010 Posts: 19520

It is always best to remove noise earlier. So start in hardware.
The first critical question is what you are using as your Vref?. The reading
has noise from both the signal, and the Vref. Unless the Vref is very stable
and smooth, it is pointless to look further. Genuine Vref sources that
provide a signal to actually offer 12bit resolution are expensive and rare...
Then there is the physical design everywhere round the chip. If there are
ripples on the supplies (and even more importantly on the ground), accuracy
plummets. This is why on some chips the ADC has a separate Vss/Vdd, so
these can be smoothed more to help.
Good analog design is key. Everything introduces noise. Even a resistor
adds noise to a signal. However the levels are small. However heat things up
and noise rises. Make absolutely sure that ground paths do not get crossed
by anything drawing any significant current.
Once you have a stable repeatable analog signal, you can then improve
things by software. It is possible to oversample, and actually extend the
accuracy of an ADC. So with care you can get >13bit results from a 12bit
ADC. The actual filtration method needed depends on the 'nature' of the
remaining noise. If it is normal 'random' noise, simple averaging is good.
However for 'spike' noise, something like an Olympic filter will give better
results.
Start by getting the hardware right. Then look at software.

MCUprogrammer · Joined: 08 Sep 2020 Posts: 221

Mr. Ttelmah, I will take all your words into consideration. I am considering using the LM336 for Vref constant voltage. I reviewed the datasheet. Not very sensitive. Also this information caught my attention:
"± 1% initial tolerance available" I also reviewed the notes for the application. I examined the Absolute Maximum Ratings values. It is not very precise, but I think I can achieve a stable work by creating an offset value with the software.
I also thought of using the constant voltage value inside the processor. for example, VREF_4v096. Do you think it makes sense to use it? The accuracy of tolerance and stable reading is very important. But lm336's low sensitivity will not bother me much. Should I be able to handle the rest with software? But I think PICs are susceptible to interference very quickly. I think I'm facing electronics now. And I have to find ways to offset these damn parasites. The best part is that I think it is a very nice feeling to do research, experiment and come to a solution on these issues. Or because I'm more new, my energy is in full. As time passes, I wonder how I will be : D
_________________
Best Regards...
MCUprogrammer
_______________________________
Work Hard

temtronic · Posted: Wed Dec 16, 2020 6:11 am

From table 27-3 of the datasheet..
specs for the 4.096 VR value

min typ max
3.891 4.096 4.342

Not exactly a precise, accurate voltage reference !!

Though I've never used it, I'd assumed it'd be better.
Obviously the best Vref would be an external device which might cost more than the PIC.

Whatever signal you are reading should be fed into an opamp 'buffer' to reduce any 'bad readings' due to loading the ADC front end . As Mr. T points out, getting the 'hardware' right is the critical first step so the PIC has a known, repeatable, accurate signal to sample. For the 'digital' or 'software' side, there are lots of 'filter' algorithms. Which one you use depends on how much time you have for sampling. For temperature, say using an LM34, I use the 'Olympic averaging'. You take 10 readings, toss out the highest and lowest, then /8 for the average. It's fast (int16 math), gets rid of 'spikes', compact code.

Jay

MCUprogrammer · Joined: 08 Sep 2020 Posts: 221

I'm not a hardware designer. But what can I do about such an issue. For example, what kind of opamp circuit should I set up at the input of adc? Also how can I free the processor from interference. I connected a 100nf capacitor to the VDD legs. What else do you think should I do?
_________________
Best Regards...
MCUprogrammer
_______________________________
Work Hard

temtronic · Posted: Wed Dec 16, 2020 8:20 am

re: opamp.....
1st, select an opamp that is BOTH 'rail-to-rail' on input AND output.
2nd, configure as a 'voltage follower'.
3rd, be sure to add small cap (.1mfd ?) nearr the opamps power pins.
4th, add a 1M ? resistor from the opamp input pin to gnd.

The resistor will reduce 'noise' from entering the opamp/adc circuit
the cap also reduces(filters) 'noise'.
Voltage follower means the output will 'follow' (be the same) as the input.
'Rail to rail' means the input and/or output will 'swing' (go to and from ) VDD to VSS (+5 to 0 volts ). Not ALL opamps are designed as 'rail-to-rail' ! so you have to read the datasheets.....

If you plan on more than 1 analog input, use a 'quad' version opamp. Cost is about the same and usually projects 'grow'.

Jay

Ttelmah · Joined: 11 Mar 2010 Posts: 19520

On the PIC itself that is down to the design of the board, and particularly
it's power supply.
The PIC is not particularly sensitive to noise. In fact it is several orders
of magnitude better than many other parts. All electronics is noise sensitive
and when you start dealing with low level signals the designs become
harder.
If you study a PC motherboard, you will find hundreds of capacitors in the
supply, and then ones adjacent to pretty much every IC. Close to the IC,
with a solid ground plane. Round the analog parts (audio in and microphone),
you will probably find a break in the ground plane often linked by wires
or tracks, to a single point selected for low noise, and possibly inductors
in the supply rails feeding this part and yet more capacitors. All of these
cost money, but they are included because they are needed.
If you want to work with signals at reasonable noise levels, you are going to
have to learn about how to layout and design the board to reduce noise....

MCUprogrammer · Joined: 08 Sep 2020 Posts: 221

Could you please share opamp circuit diagram if Mr Jay exists.
_________________
Best Regards...
MCUprogrammer
_______________________________
Work Hard

temtronic · Posted: Wed Dec 16, 2020 12:25 pm

Use Google 'opamp voltage follower design' or similar keywords, you'll get 1,000s of designs. Keypoints are to use a rail-to-rail, both input and output op amp designed for whatever VDD you're using (+3 or +5), also be sure to use 'bypass caps' close to the power pins.

Jay

bkamen · Posted: Wed Dec 16, 2020 3:13 pm

newguy · Joined: 24 Jun 2004 Posts: 1908

bkamen · Posted: Wed Dec 16, 2020 3:35 pm

newguy · Joined: 24 Jun 2004 Posts: 1908

Ttelmah · Joined: 11 Mar 2010 Posts: 19520

I wonder what percentage of the people here actually suffer from that?.
I suspect it may be quite a few of us. I've been Type 1 for over 55 years.
My grandfather was Type 1, before in-sul-in was available. Only just
survived till it came along. In those days it had a very short shelf life,
so they had to get it from the chemist every day....

MCUprogrammer · Joined: 08 Sep 2020 Posts: 221

I installed the circuit on a perforated copper plate to test it now. What I did in the project:
I reviewed the MCLR link datasheet sheet. I made my setup with the values it specified. I have my OSC connection linked with 16mhz 22pf. I put 100nf capacitor on my VDD pins. I made my ICSP connections. I applied analog filtering to the ADC interference with opamp. I intend to use the opamp as a amplifier if it doesn't provide any benefit. I plan to turn it into a 1/20 gain circuit. Yes, this will be my first project. I have no experience like you. But there is high voltage where I will test the card in the first place. I will give an output according to the voltage information. That's it for now.
_________________
Best Regards...
MCUprogrammer
_______________________________
Work Hard