While looking around for sources of parts for a scintillator/photomultiplier tube cosmic ray telescope I came across a neat project from Berkeley.
Here are the plans for a simple telescope, and here is the operating manual.
Reading these give a good deal of information. I wouldn't follow them directly. You will probably want to deal with data collection differently (this is really primitive for high school classes) and the detector is optimized for portability.
I would recommend placing the detector paddles on a bicycle wheel rim so you can easily aim the device. Scanning through a single angle is fine for this type of work.
Consulting with your local physics department may give good scintillator scraps and access to very expensive optical glue (you won't use very much and there is no sense buying it by the kilo). More importantly it will give you a chance to test it on a working power supply and test the assembly with a radioactive source for calibrating the PMT/power supply and testing the detector for light leaks (you will need thick black plastic and lots of sticky black tape).
I've had very good luck with a 500 nanosecond coincidence window. It allows for considerable sloppiness in design and generally gets you above the PMT and natural radioactive background noise. I do recommend having the ability to adjust the size of this window.
Finding photomultiplier tubes is not easy - especially if you are on a budget. Your local physics department can make good suggestions and may be able to give you some outdated tubes or even give or loan completed detectors. I think you will be surprised at how willing they will be to help an enthusiastic/curious/hard working amateur. Determining detector efficiency and best operating voltages takes a bit of thinking and these guys will have a huge amount of practical experience.
Note that there is a huge amount of flexibility in design. Optimizing the design is probably a secondary consideration on a first telescope. Paddle size, shape, coupler type, PMT type, spacing, etc can be a function of what you can scrounge. If you get to the point where you are thinking of a large array to do calorimetry, design becomes a central consideration (of course let me know! it would be fun to build a meaningful calorimetry telescope - someone else would have to bring the money though).
It is possible to do pulse height analysis from a single detector, but the type of scintillator used (NaI for example) is more expensive and more difficult to work with. The detection electronics are more sophisticated and you loose the ability to adjust your acceptance geometry. You can do it, but I wouldn't go there...
This is well within the capabilities of many of you - particularly if you find a mentor. Playing with nature is much more satisfying than watching tv.
more on other techniques later
Comments