McWilliams design was on Regulus at http://www.regulusastro.com/regulus.html , but I understand that John Blackwell has just had a lightning strike. Enlightenment apparently went up in the proverbial flash.
The crux of the problem with the Earth's magnetic field measurements is the very high stability required. We get 40 to 80 nT in 20,000 odd nT North horizontal component and need to at least hold the zero drift to 10%. This implies an overall stability of better than 200 ppm (0.02%) and is somewhat difficult to achieve, considering that the linear expansion of materials alone is 10 to 20 ppm / C Deg. Have a look at the magnetograms on http://samsun.york.ac.uk/samnet/rt_mgram.html The 'problem' with the McWilliams design is that you can't get thin enough steel wire for the suspension, so you have to have a large powerful magnet system and use a long wire to give you a low enough torque / angle. The tempco of Alnico magnets is close to that of steel wire and they match reasonably well. Just about anything else will drift unacceptably. The big pendulum is sensitive to tilt and to vibration. The way the photocells are mounted makes the output sensitive to sideways pendulum motion. The cells need to be both mounted on the same side of the magnet at opposite ends to give high 'common mode' rejection. The CdS cells you buy are not matched, usually have significantly different characteristics and also drift with temperature. They have a S shaped photoconductor and give a non linear 'wobbly' output with position. Large area Si photodiodes with a bridge OPA output work fine. I use current / voltage conversion with 0 V across the photo cell to get insensitivity to temperature changes. The output gain can be altered to give virtually any sensitivity that you are likely to want.
If you go to www.mouser.com, they stock a CdS photo cell, the 621-CL9P5L, which has a STRAIGHT BAR photogap (not a wavy gap), which should give a much more linear trace. Make the movement of the shadow of the vane change the length of the photogap, NOT the width. Do not try to use any of the more expensive CdS cells in TO-5/8/46 metal cans. The cans are ferromagnetic and strongly attract the magnet!
I wrote this to someone who was having problems with the
Your magnetometer magnet naturally wants to point N/S magnetic. If your torsion suspension rotates it through 90 Deg, all of the couple is due to the suspension. The torsional modulus Gm for 1% carbon steel is 8.12 x 10^11 dynes/cm^2 and the temperature coefficient tc is 2.6 x 10^ -4. Gt = Gm(1 - tc(T-15)), where T is the temperature in Deg C.. If the magnet lost field (actually magnetic moment) at the same rate as the wire lost torque, the rotation would be constant, but even with Alnico 5 at -2.3 x 10^ -4, the 'fit' is not perfect, so let's consider the problem a bit further. The normal N/S horizontal field daily variations may be 40 to 70 nT in a field of 0.185 oersted = 18,500 nT. Minor storms may produce 50 to 100 nT, major storms 500 + nT. The difference in the magnet strength and torque coeffs. is 0.3 x 10^ -4 or 0.55 nT / C Deg. indicated, which really isn't too serious. With a Nd-B-Fe magnet at - 0.09% / C Deg., the difference is 6.4 x 10^ -4, so the error is 11.8 nT / C Deg. A 4 C Deg daily temperature change will give about the same effect as your daily field change. Unless you monitor and correct for the temperature changes quite accurately, you won't know the difference. The damping vane MUST be a CROSS and TOTALLY submerged in oil, otherwise you get erratic magnet movement due to changes in the surface tension of the oil. The CdS photocells are non linear, have quite large differences in sensitivity for any two cells and are quite temperature sensitive. Getting a well matched pair may be a problem. Check that the cells you plan to use are NOT MAGNETIC! Some have magnetic wires / housings. CdS cells usually have a S shaped photo gap, so using a knife edge shadow gives a S shaped response, at best! You really need to put a diffusing screen between the shutter edge and the cell. 7 sq mm Si photodiodes used in current mode would be a very much better choice and don't cost much more.
The Magnetics Research fluxgate magnetometer has a drift of 0.35% / C Deg on it's own. 1/2 C Deg overall stability is not very practical. If you use the integrator modification, you can greatly reduce the drift to 0.03% / C Deg. This gives a practical system, but you do need to choose an instrument site which is both free of magnetic interference and has reasonably good temperature control, since a 7 C Deg change can give you a drift = 'normal' daily field variation. I buried my system in the garden, about 2 ft down in a waterproof container. The daily temperature changes at this level are less than 1 C Deg.. I checked and the drifts quoted in the article were actually measured. A factor of 10 improvement would be great. Watch this space... We are working on it!
I attached an IC temperature sensor to the unit and read this out along with the field changes on a computer. Burying instruments seems to be a very practical way of getting low daily temperature drifts. I used a post hole auger and lined the hole with plastic 'soil' pipe. I am now writing up a note on this for MAGNET.