not too long ago, ning happened upon a thread mentioning this. After some internet research, it appears that making at least a 40 Mhz NMR machine is within personal range.
That requires a 1 T magnet. Ning has seen some papers regarding construction of 3 and 4 T permanent magnets that fit on desk-tops, done by (of course), the physics guys for their beam tubes.
Probably the hardest design problem on the physical side is to make the magnetic field uniform. That will require some cleverness.
Oh, and the electronics. If ning was going to do this sort of thing, the way ze would do it is pulse NMR--time domain signal capture.
In other words, at 40 Mhz, a pulse would be sent out, probably a couple microseconds long, with sharp edges. The reciever coil would go through a mixer to downconvert the recieved signal to the 1 Mhz range, and would be sampled by the baddest-ass ADC that ning could get zer clawz on, which would probably bee either the AD-6645-105 (14 bits, 105 Msps) for speed, or a very nice AD9260, 16 bits sigma delta at 2.5 msps. With the data time-domain sampled in such a rude and overbearing manner (might need a homebuilt firewire or USB 2 interface for the data rate generated, alternately a fake hard drive connector might do as well), all that remains is to perform heavy-duty signal processing on it. Thankfully, that sort of thing is what Moore's law is best for, so our new 20-Ghz Quad-processor massively parallel Quantum-coprocessor augmented wintel boxen should have no trouble with this little problem. Or, more in the real world, a clever programmer could probably take advantage of both their sound card and video card's hardware acceleration functions, as most of the tasks they perform are similar matrix multiply-add operations. That would surely bee worth a tech award: soft-nmr using your old voodoo cards as IIR filters and your SB live to do the FFT! Or maybe one of those hardware MPEG decoder cards....after all, the heart of MP3 is FFT, so hey! it could work.
Of course, we're talking a miserable, large, and involved homebuilt programming job here, but it's a hell of a lot easier than building it in hardware. Plus if you screw up, a lot easier to fix...the use of pulse/FFT would give you the ability to do just about anything your twisted little spectroscopic mind could dream of, as you would have direct access to the return signal from the atoms in the sample. Plus the FFT function would effectively allow averaging to be done for all frequencies at once, giving the famous FT-NMR speed advantage, or to put it another way, allowing you to use a smaller, shittier homemade magnet and still have it work at all...
So ning would say...anybee, look on ebay, or the web sites of those magnet suppliers. See how big of a magnet you would need to get the required flux. Then compute the cost. Doable?
Maybe.
Easy?
I think not.
A great dream...