How to software - filter mobil signals from a ADC

[Henrik_A]

Hello,

I have a pressure sensor on the ADC of a 18F6723 giving me a 0 - 5V signal. Everything works fine with the data collection until a mobile phone being around (eg. 15cm) starts to ring. From this point on the ADC goes crazy as the mobil signal is modulized over the pressure signal. I did not find a good electronic solution as the board is already fixed.
Does anyone know what the best software filtering would be to get rid of the mobil modulation?

[FvM]

From an EMC regulation point of view, it's not required for electronic devices to work unaffected with a mobile phone in a 15 cm distance, because the field strength is above required EMC susceptibility limits by several orders of magnitude.

Most likely, the device would fail burst tests, however. So there's probably room for serious improvements and a necessity of a redesign.

Software filtering can be workaround, if the distorted signal is still in it's linear range and the average maintained. Otherwise, something like a interference blanking may work, holding the previous measurement until the interference ends. Depending on the applications signal dynamic, this mode of operation may bring up the danger of supplying wrong measurements for an unknown time interval. It's at least illegal with any safety related measurement.

[SherpaDoug]

If at all possible you need to stop this signal BEFORE it gets to the A/D. Better than filtering it on the PCB is to stop it before it gets to the PCB with a shielded enclosure. If you have a metal box make sure it is grounded to the same ground as the analog circuitry leading to the A/D. If it is a plastic box look hard at a metallic spray paint or an adhesive metal foil to add shielding to the enclosure. If there are wires, or metallic pipes leading into the enclosure they should be grounded or filtered as they pass through the enclosure with maybe a ferrite bead or ballast resistor between where they pass through the shield and where they connect to the PCB.

With strong RF interference it is not likely software can do any good to recover a corrupted signal. Once the A/D saturates all meaningful information is lost. FvM's suggestion to hold the last valid signal is good if the application permits.

If you must do software filtering search for "Olympic scoring" or "Olympic sorting" on this BBS for a simple way to remove sporadic bad readings.
_________________
The search for better is endless. Instead simply find very good and get the job done.

[Ttelmah]

Yes.
_If_ the signal generated what (at your sampling frequency), was effectively 'random' noise, then software filtering, could remove it. However the problem is that the hardware in the chip (particularly the diodes in the input), will tend to first rectify, and then clamp the signal, typically resulting in the noise sending the signal 'off scale', possibly just in one direction, but also possibly to both extremes.
One approach that might work, if your 'normal' signal has a known range, less than the full scale of the ADC, is a combination of a simple average, and an 'olympic' limiting algorithm. So (for example), if you generate a simple average of a number of samples, and at the same time, have a 'flag' cleared at the start of the averaging process. If the incoming value goes outside the 'known acceptable' range, you set the flag, and reject the average, going back to the 'last good' value. The averaging will tend to help for lower level interference, and you effectively use a value coming outside the range, as a indicator that something is wrong.
If the pulses are really short (less than one sample time), you might consider trying a basic 'olympic' average, with something like:

[Guest]

Why is 'Sum' dimensioned as an Int32? The largest value that it can be is 10230 (1023 X10), so the result will fit in an Int16 nicely!

Doug

[Ttelmah]

Because the code was actually based on some being used on a chip with a 14bit ADC...
Yes, you can save time, and space, by switching to int16 for the application involved here.

Best Wishes

[asmboy]

Sherpa Doug is setting you in a good direction: - iI'd like to add - that u did not say what you were using for the ADC reference source.

If it is an external 2 or 3 terminal device - - thing like the LM385 TL431 etetctetc - they can be shockingly GOOD at rectifying the packet pulses emitted by mobile phones and turning it into GLITCHES in the ref output.

When using such things - i use a small series R and capacitor to minimize the issue.

being sure that your DC POWER is well decoupled for RF may be of help also. for any product I design - I assume the end user may be in a 2V/M or stronger , modulated radio field.
( for the NON - RF person - thats about the field strength of a 5KW MW transmitter at a distance of 500 M from the main array - or a cell phone output when dialed up to max TPO at about .25M - kinda your case i think )

I always insist on a ground flood for the PCB and use a simple , "octave above Fmax" PI filter for any inputs and then use a 10uhy multi winding ferrite choke on any power entering.

Serial data flows through 100 ohm ferrite choke bead inductors leaving with a "1/2" pi or ( LC - Lnet ) with the 'c' on the inboard side of the connection.

Be aware that your interference is almost surely happening because your
circuit is conducting radiated energy and rectifying it in an incidental fashion courtesy of all the fine supply breakdown virtual diodes
in the PIC and nearly any CMOS chip ever made.

At cellphone freqs - unintended 1/4 wave resonant structures exist all over the wiring design too - you just don't realize it till it hits the fan.

BTW: my background before semi retirement was in high power RF design
and I've seen good bad and ugly - often enuf to know that there is usually no single answer to RF problems - just more ways to further supress the undesired aspects of ubiquitous RF emitters in our lives.