Those things seem impossibly tiny for that.

http://www.intersil.com/data/an/an1325.pdf

See pages 5 & 6, discusses also the effects of surface mount package sizes on ESL.

I had a GSM RF scare years ago. My expensive tube computer monitor (yes many many years ago) was out of its warranty period and the screen was flickering every now and then. I was thinking that I would soon have to replace it. What it turned out to be was a poor selection of a cell phone charging dock placement.

Two things I wanted to comment on, though I am not sure if anyone will notice new comments on an old entry.

I, like Martin, am not convinced of using multiple bypass capacitors by default – without analyzing it for the particular case in hand. Take a look at this short paper: “Effectiveness of Multiple Decoupling Capacitors”, Clayton R. Paul, IEEE Transactions on Electromagnetic Compatibility, Vol 34, No. 2, May 1992. I am sorry I cannot post it.

I noticed you used solder wick to connect the bottom layer to the top. I used to do the same thing. (I’m a beginner, so ‘used to’ here refers to just two or three boards.) Solder wick is flat, wide, etc.; better than using a single wire I used to think. And I used to cut traces on the top board like you did. Lately I am realizing that I don’t really know if that serves as a proper ground plane by default. Especially after having read these articles recently:
http://www.cvel.clemson.edu/emc/index.html

Thank you for the blog.

He is a really good writer, and I always enjoy his posts.

Looks nice !

But I see now how it may be beneficial to use several caps in order to lower the total parasitic inductance to ground and thus get better attenuation at frequencies where the parasitic inductance is dominant.

However, I did a simulation for this particular circuit using typical parasitics for 0402 caps and it seems like using 33//10 instead of 43 puts an extra pole very near 1800MHz. Leading to a ~20dB decrease in attenuation at that frequency. That seems bad since this is one of the bands that the filter should block effectively.

The beneficial effects of distributing the capacitance doesn’t seem to kick in until above 2GHz.

Maybe I have it backwards, you and other people here are certainly more knowledgeable than me. The simulation schematics and results are here:

http://imgur.com/3L1wK.gif

It is common to use multiple caps in parallel as each cap will have a reactance at different frequencies, it is not uncommon in Digital circuits to see a 100nf, 10nf and 1nf in parallel on each power pin of a VLSI chip.

Kat.

I’d also like to share a link to an article with some related information. It also has some of the diagrams you drew:

http://www.infoelectronica.ro/date_upload/lucrari/498768(perturbatii%20in%20telefonie%20GSM).pdf?PHPSESSID=4f3ea4a78f00c79ea31cf1e8fdaaf54e

Another good blog post, keep it up!

Every part of the design was by a separate person, one designed the printer controller, one the comms manager, another the blue tooth etc. I designed the battery charger (based on an Atmel AVR Tiny 26) it worked great on my work bench,(apart from a few exploding batteries, but that was a problem with the cheap China batteries we tried but in the final product it failed as did some of the other areas, after a few weeks, we traced it to the POS engine, it was a GSM Tx Rx module, with all the smarts, card reader, display and keypad.

We tracked it down to the POS engine having a higher than normal Tx power level, it was swamping everything, to the point of coming in through the silicon of the Uc’s.

In the end we had to put RF traps on every conductor that left the printer and a silver coating on the inside of the case.

Regards,

Kat.