GSM interference decoupling

[yesway]

We are designing low cost handsfree units for car use and have problems with the infamous gsm interference. The PCB contains simple mic and speaker amps and both pick up quite a good deal of 214Hz hum. Any good general rules would be greatly appreciated. We have tried a lot of basic things like decoupling capacitors, groundplanes, shielded cables, but with very limited success.

 

[cbarn24050]

Hi, what is gsm?

 

[melone]

Obviously, fill all open areas with ground, and also try to place all of your communication circuits in the same area. Then, run a large ground trace around the components. I have also seen a ground plane seperation technique used to help control the amount of RF ground pollution.

Just a thought, good luck and keep us posted.

 

[jasonlitw]

This problem is hard to solve. First , you need to check which part is coupled by GSM TDD noise.
From MIC input or Speaker output ??
Meanwhile , you also need to check your OPamp spec. Does it have high PSRR ??

When RF is transmitting , it will consume more current than receive and cause big voltage drop suddenly.

 

[yesway]

Thanks guys, Melone in special. I tried out a quite different layout with lots and lots of groundplane and vias to tie both layers together, and voila, instant noodles. Well, the result wasnt perfect, but when does that happen? It was a major improvement for sure.

To cbarn24050: gsm in this case is referred to the european digital cell phone system, and its quite nasty interference caused by the pulsing nature of the transmitted rf

jasonlitw: It sounds like you have some experience in this matter?! The mic amp is a single transistor (is it called common collector in english?) with several decoupling caps, and powered from a 5V zener, and it picks up some hum both without the mic connected and even more when the mic with cables comes on. I have tried several layouts, and general layout seems critical. The speaker amp is 2 bridged LM386N-1. This may not be the optimal amp, but is quite low cost though. These amps are powered from a swithing power supply (MC34043) wich is also charging the cell phone battery. The inputs of these amps seem to be hysterically sensitive to unbalanced layout, if I run the unused traces in parallell with the use ones, and then ground them, things improve much.

Well, things are on the right side of acceptable, so basically we are happy. Thanks again, but more help is of course appreciated!