GSM interference

[stefandk]

Hi,
I am experimenting with weigh scale - load cell + analog/digital microcontroller MSC1210 (Texas) . When i put my GSM close to cable (from load cell to converter), scale display is changed significantly.
GSM is active - calling to other phone.
With other type ADC (Texas ADS1240) i have not such problem.
How to stop this influence?
regards
stefan

 

[melone]

Make sure your setup can handle airborn frequencies in the 900, 1800, and 1900MHz bands...that means that shielding the cable between the setups (and grounding at ONE END ONLY), and placing the equipment in metal boxes, should eliminate the problem.

 

[stefandk]

Hi Melone,

Metal box is not good for me.
Wires between PCB and load cell are:
to loadcell -> +5V and GND
from loadcell - 2 wires differential signal
shield (grounded only on PCB GND)

When i make short circuit between differential wires - on PCB side, there is no interference (may be ) - at least microcontroller does not die.
regards

 

[Comcokid]

Stefandk
What you have is a typical EMI/EMC type issue. The MSC1210 is an A/D w/8051 core and the ADS1240 is the A/D as a separate peripheral. Obviously, to exercise two different parts, you have two different layouts.

Solving and EMC type issue without using brute force (i.e. shielded box with all shielded I/O) requires careful attention to detail. To stop EMI, you must do one or more of the following:
I) Keep the RF from getting to your circuit - shielded enclosure, shields over sensitive parts of the circuit, multilayer boards with external ground and internal signal routing.
II) shunt the RF being impressed on a possible sensitive trace to ground with a low impedance path for the RF (translation - a small capacitor to ground with minimum connection inductance).
III) Stop an RF signal from traveling along a wire connection or long trace by creating a high impedance for the RF so it doesn't want to travel the wire (translation - ferrite bead, series inductor, or sometime just a series resistance).

What to do:
1) Careful examination of the layout you have A)how long traces are to high impedance points, B) good continuous ground with a minimum slots or notches which provides a low impedance at high frequencies,
2) Proper bypassing at power pins (more than one capacitor with one or two decades of value difference paralleled at VCC with short connections from VCC to ground - i.e. a 0.1uF and a 1nf in parallel. Remember - connection length is equivalent to inductance, and inductance along with capacitance forms a resonant circuit.
3) Proper bypassing at analog signal pins - especially if they are high impedance. Again bypass close to the input pin of the A/D with short connections to ground. Sometimes just 33pf to 120pf is all you need to provide a low impedance path to ground for any RF signal. Be sure to do this at your analog input pins, reference pins, and other pins where the small capacitance will not interfere with the operation of your device.
4) Use one or more ferrite beads over the leads to your load cell at the point where they connect to the board.

 

[melone]

What happens when you put a scope on the differential signals and bring your phone close to the cable?

 

[BrianR]

Your cables are probably acting as antennas carrying signals into the electronics that are driven into overload.

The use of twisted pair wiring will generally give better suppression than shielding the wires although at 900MHz you'll need tight twists.

Use a tightly twisted pair for 5V / GND
Slip some ferrite beads over each end of this pair.

Use twisted pair for the differential signal wires, not just a shielded pair. A shield grounded at both ends would help too.

Grounding the shield at one end is good general practice, but not enough for RF. Try a double shield, one shield grounded at the signal source end only then a second shield insuladed from the first, grounded at both ends. Use ferrite beads at both ends of the cable.

 

[stefandk]

Hi all and Thank You,
I have good enough (not perfect) suppression of EMI induced by GSM with ...
bigger capacitors in power supply(And capacitors are main difference between two boards).For those, who is interesting - on board is linear regulator LM2954IM.Output capacitor was 100microF, for analog part - additional 100microF after 100microH iductance. Capacitors are increased to 2000microF, i hope 1000microF will be enough.

Shielding of signal cable is "standard" , coming with load cell, and helps.

I do not have ferrite beads and this is not tested.

I was trying with LC and RC filters on signal path - without success.

Ferite tube acroos signal cable (close to board) has very good effect.

Best regards and excuse my english.
Stefan

 

[Heydave]

What seems to be happening is a rectification of the RF in the front end of your analog circuit. This does not seem to be balanced between the inputs being measured and thus a DC offset is being created. The amplifiers in which this is happening have junctions held in a linear region making them very good RF rectifiers. Your best bet is probably ferrite beads and small surface mount high frequency caps to short out the RF. The frequency you are dealing with is very high and lead lengths to caps becomes an issue. This inludes the length of leads from the amplifier's ground to the caps. Large value caps often have a high impedance at these frequencies and will probably be near worthless.