When news of Terry's sensor appeared on the DPRG mailing list, I asked what made it differnet from the Motorola e-field sensor. Terry posted this response to the ThereminVision List on Yahoo!:

I first started this project using the MC33974 back in mid November. I overnighted the eval kit in from DigiKey thinking I would be all set...

Unfortunately, the MC33974 is very insensitive. It can detect a car seat (it's real use) but it cannot detect to the fine levels needed for robot work. With a 10 bit AtoD (it has analog output), the chip can go to 0.1pF if you are "really" lucky. Most of the details I stuck here:

http://hot-streamer.com/temp/MC33974EvalKitCD/

Although neat for many uses, it was instantly obvious that the chip was not going to work for the range of 0.0001pF which such a system really needs. I gave the kit to a guy at the robot club for his object grabber where it might be more useful (I think I also have a bunch of bare chips in my junk box...). I then started looking into the theremin's circuits. There was a version-I which used all analog circuits that is still described at:

http://thereminvision.com/

However, the new all digital version is vastly better and really made this whole system fly:

http://thereminvision.com/version-2/TV-II-index.html

ThereminVision can detect capacitance levels about to 1/10000th what the Motorola chip can. The Motorola chip does have nine channels and can do shield referenced antennas which is cool, but the sensitivity is just not there.

http://hot-streamer.com/temp/MC33974EvalKitCD/CD33794DWBEVM%20Documentation/MC33794.pdf

The key difference is that the Motorola chip just sets up a 120kHz capacitance controlled oscillator and measures the frequency delta directly. A nice solid thing for critical car safety systems, but too insensitive in our case. However, ThereminVision "heterodynes" the oscillator signal with a reference signal just like the theremin instrument to run the sensitivity way up while still remaining stable. That is the key difference. This is all explained in the manual under "Theory of Operation" on page 5.

http://thereminvision.com/version-2/ThereminVision-II-manual.pdf

What is really sad is that apparently the Motorola chip was originally going to heterodyne the signal against a reference and be very sensitive just like the theremin. I might guess that the analog designers were just not able to get it going or the car application changed there plans with million piece sales prospects. ThereminVision could actually be adjusted to be just as insensitive as the MC33974 if anyone really wanted too. If they had just gone all digital and allowed for the full range of sensitivity adjustment, the Motorola chip might have been perfect for everyone!! But they didn't... Maybe someday they will... The Motorola package is also pretty messy for home soldering and such. They added a bunch of automotive things too like the incandescent dash board light driver circuits (it does have an error signal if it burns out :-p).

If I have not already told you far more than you wanted to know ;-)) Here are two posts that go over this too:

http://groups.yahoo.com/group/FrontRangeRobotics/message/1678

http://groups.yahoo.com/group/FrontRangeRobotics/message/1711

There has been a long running saga about this at:

http://groups.yahoo.com/group/FrontRangeRobotics/messages/1697

All the relevant posts are "MC33947" or "ThereminVision" reports.