EMI Shielding

[kpatz]

My current project is an EMF meter/thermometer combination using a PIC16F818. In order to fit it in the case I want to use, I had to make the circuit as compact as possible (not easy on perfboard!) Anyway, I'm in the testing phase of the circuit and it's picking up EMI/noise from the multiplexed LED display.

Due to space limitations, I put the LEDs on their own board and have it connected to the main board via two pin headers. The EMF amplifier circuit on the main board is sandwiched between these headers, as you can see in the attached sketch.

Last night I made crude EMI shields out of aluminum foil and shipping tape and placed them over the EMF amp circuits and also on the underside of the LED board, and grounded them all to a common point, and this cut the noise roughly in half, but didn't eliminate it completely.

I know the noise is coming from the LED display and drivers, because if I unplug the LED board from the main board, some of the noise goes away. If I set the PIC firmware to not multiplex the LED display at all (leave it off), the noise disappears completely.

Any suggestions? I may try putting a shield on the underside of the main board to see if that helps. I'll also try putting a bigger bypass capacitor on the power going into the EMF circuit as a test (I have a 250µ on there now, anything bigger won't fit).

I suppose if all else fails I could put the EMF sensor/amp on its own board and physically separate it from the main board, but then I'd need to use a different case.

Any suggestions?

 

[crutschow]

Much of the noise is likely coming from the wires going to the display that carry the switching signals from the PIC. You might try using twisted shielded wire for those connections.

Also, do you have a separate ground for the EMF circuits that is only connected at one point to the digital ground? Single-point grounds are important when you are mixing analog and digital circuits.

And you might try placing a small resistor in series with the power to the EMF (before the bypass cap). The resistor should be as large as possible without dropping excessive voltage due to the EMF current draw.

 

[kpatz]

You're probably right about the wires. There are wires going from the PIC to the LED display pin header sockets, and to the 3 column driver circuits. When I made the shields, I put them over the EMF amp components and tucked them in so the EMF bits are inside the shield and the digital circuits and display wires are outside.

With the LED display connected, I get noise that varies depending on which/how many LEDs are lit. If I disconnect the display, then I just get a continuous buzz noise, which is most likely from the column drivers, which are on the main board to the left of the EMF op-amp stage. If I disable the display completely within the firmware (so no LEDs are on and the multiplex pins aren't being strobed) but leave the PIC running, I get no discernible noise.

The EMF circuit's ground is connected at one point IIRC. I'll have to double-check this, but I don't think it's that anyway. When I was experimenting with the bypass caps, I had a scope connected to the EMF power, and was getting voltage fluctuations corresponding to the LED current draw. I added the 250µF and still get fluctuations, but they're smoother, and no change in the noise. When I stuck a 1000µF in parallel, the power fluctuations went away completely on the scope, but the noise didn't go away or change at all. (This was before I added the shielding though).

I'll try the bigger cap again to see if it makes a difference. I thought of the resistor, and instead I put a diode before the cap to see if it would have any effect (I have other reasons for putting a diode in there which I can explain later if you're curious).

FYI, the circuit is powered off a 9V battery. Power is switched on/off through a MOSFET (allowing the power to be cut via the firmware using a button), the EMF circuit gets unregulated power (battery voltage less 1 diode and any drop in the MOSFET), and the digital bits (PIC, LED stuff and temp sensor) are powered through a 78L05.

Maybe I'll make some additional shielding to put on the wires going from the PIC to the LED bits.

 

[crutschow]

If you temporarily power the EMF circuit from a separate battery, it would show whether any of the noise is coming through the supply connection.

As you have determined, it's quite difficult to keep digital noise out of adjacent analog circuits. It often requires a lot of shielding, careful layout, good decoupling, and a separate ground plane for both the analog and digital portion of the circuit.

Shielding both the analog and digital circuits separately, giving a double shield, may help.

You might also try a small ceramic cap (0.1µF) with as short leads as possible, in parallel with the electrolytic to help filter the high frequency noise.

 

[kpatz]

Well, I have good news. The vast majority of the noise was being picked up by the first stage of the amplifier, because I didn't have the EMF pickup coil connected yet. I hooked that up tonight, and made a more precise shield, and now the noise is much, MUCH weaker than it was before.

I may add a "silent" mode to the firmware to switch off the display for sniffing out weak EMF signals though.

Also, none of the noise is coming through the supply rails. Even when I paralleled a 2200µF cap across the 250 as a test, there was no change. Which is good, since there's no room anywhere to put a big 'lytic in there.

 

[Roff]

If you still find you need shielding, Google "Faraday shield" or "Faraday cage".