The GM tube is a russian SBT11A (data sheet). It's an alpha tube that looks quite sensitive. It is powered with 400 V. It is actually made by four tubes (my guess is to provide some directionality). In my project they are linked together in parallel.
After looking around a lot I came to the decision to use an off-the-shelf small board: the Theremino. It provides stable high voltage that can be easily set at the level required by the tube. It also has a signal conditioner that can be fed directly into a digital input of the uP.
The controller is an Arduino. Originally the project was for a standalone system with low consumption so I did choose an Arduino Pro Mini 328 3.3V programmed with a FTDI basic breakout board from Sparkfun. Since the project has evolved into a permanent installation the breakout board is always connected to ensure serial communication with the server.
The signal from the GM tube trigger interrupt 0. The interrupt completion routine increases various counters for one second, one minutes, and ten minutes integration.
When requested by the server it sends back the ten minutes integration data via serial communication. It can store up to 48 hours of data, even if normally it is polled every hour (Source).
Major components before installation in enclosure.
Disruption on October 1st
On October 1st I started the assembly into the prototyping box. To my surprise after the assembly I got roughly twice the counts from the same circuit and GM tube. The only explication is that during the assembly I had to plug/unplug the GM tube into the 7pins miniature (valve) socket. This action must have caused a better contact on one of the two sections (see tube datasheet) actually doubling the amount of individual tubes that were active.
Since in the following days I noticed a lot of instability, on October 8th I finally took away the socket soldering directly the HV wire to the GM tube pins.
Mounted in a prototype box.
Originally a Pandaboard running Ubuntu 12.10 was used to retrieve the data from the Arduino controller, store them in a file, and upload the file to the IPS server (GVO). Cron jobs took care of ensuring hourly updates.
This configuration had several drawback: the Pandaboard was configured with fix IP since it was reachable from the Web via DDNS and it was connected to the router by cable (although capable of using WIFI). It is sitting on my desk without any enclosure. It is shared with security cameras that store images via FTP (that's why it has fixed IP).
In order to move the GM detector to a more suitable location I connected it to a Raspberry PI running Raspbian wheezy. It has its own box, a low profile special SD adapter, and a nano WIFI USB adapter. Configuration was straightforward and I have now a detector I can move around wherever there is a plug. When I have to do some maintenance I just login with Putty and Xming via WIFI. The two USB ports in the Raspberry are used for the WIFI adapter and the connection to the Arduino.