As I've continued to dig into precision electronic levels (those with the ability to measure t*ts in the seconds of arc), I've found an even better mechanical design than that of Timo Kahlmann and colleagues ("Precision Electronic Levels - The Germans Arrive").
I first saw the better mechaical design as Figure 8 (Tiltmeter-horizontal seismograph) in "The design and some applications of sensitive capacitance micrometers", R V Jones and J C S Richards (both of Aberdeen University in Scotland), Journal of Physics E: Scientific Instruments 1973 Volume 6, pages 589-600. (I'm also looking for the article specifically on this tiltmeter.)
The design in Figure 8 is dead simple. The frame is one piece, milled out of a single piece of 0.500" thick metal (Jones used rhodium-plated brass, but for us aluminum will do) that was about 2.25 by 2.25" before cutting. There is a long window cut in from one side, so what remains (ignoring mounting ears) is a blocky letter U.
The pendulum, made of the same sheet, swings in the space between the arms of the U. The hinge element is two pieces of metal foil. The top (stationary) hinge assembly is insulated from but bolted to the tops of the arms of the U, while the bottom (moving) hinge assembly is a part of the pendulum, and the metal foil goes between these two hinge assembiles.
The drive electrodes are two metal crossbars bolted to but insulated from the arms of the U, and not touching the pendulum.
The pendulum swings between the two bars, making a three-plate capacitor, also known as a differential capacitor.
I would follow Kahlmann and add the rare-earth button magnet to Jones' design to provide damping.
The insulation is the key, and was a big issue to Jones. What he ended up using is thin sheet mica, which has a very low and stable temperature coefficient. The bars are held down with screws and very stiff springs, with mica between, to allow thermal cycling without causing screws to loosen or electrodes to creep. Sheet mica is easily obtained, being used for insulating power semiconductors from heatsinks, and being used to make windows in wood stoves. McMaster-Carr carries mica sheets.
The mechanical parts count is small: 1 U-frame, 2 cross-bars, 1 pendulum with hinge recess, 1 top bar with hinge recess, 2 clamp plates (one per side of the hinge), two strips of stainless steel foil, some metal shims to govern the space between bars and pendulum, one button magnet, assorted machine screws, insulated washers, and Belleville or wave washers to maintain steady pressure. Everything is square, so this will be easy to make using a mill. Cutting the U from a single piece of stock greatly improves stability by eliminating some critical joints.
As for electronics, the winning design approach so far is a Blumlein bridge with a center-tapped bifilar-wound pot-core transformer driving the two bars, the center-tap being grounded, and a charge amplifier accepting the unbalance signal from the pendulum. The charge amp feeds a synchronous detector to yield a DC voltage that varies in direct proportion to tilt angle. Use of a charge amp makes the stray capacitance of the pendulum and cable to the amp largely irrelevant, and provides the phase shift needed to condition the unbalance signal for the synchronous detector (also called a lock-in amplifier). The operating frequency will be around 16 KHz, which is the optimum for capacitance transducers of this type. All this circuitry is non-critical and easily built in a home shop.
Joe Gwinn