Can I achieve >18 bits of repeatability at a 3KHz bandwidth with the AD598 in a 5 wire LVDT crk?
When using an LVDT you are developing an electro-mechanical system. Many times, the precision and accuracy is limited by the sensor and not the electronics. Typically in an extremely high accuracy system, the mechanical pieces are the limiting factor in performance. When you ask about repeatability at greater than 18-bits, are you actually looking for a ppm level of measurement variance? At an 18-bit level, 1 bit corresponds to approximately 4 ppm.
Repeatability of a measurement (precision of measurement) is determined by noise and short-term drift of the system. Long term drift can sometimes also be considered if precision is required over an extended period of time. The precision of an LVDT system must be inferred from specifications such as the transfer ratio specs, noise, temperature drift and time drift (short and/or long). Precision (repeatability) is meaningful when relative measurements need to be taken. An example would be device calibration.
I think for the AD598, the limiting factor in result repeatability will be the noise and output ripple. In many cases, the output ripple can be 4 mV rms which would correspond to a peak to peak ripple of 5.7 mV. Let's assume you are looking for a 10 volt output range, then the 5.7 mV would represent an uncertainty in your measurement. Unfortunately, this corresponds to a ppm level of uncertainty of 570 ppm. Now this certainly could be reduced by performing an averaging function of the output signal. However, without knowing these exact specifications, along with the type of LVDT and what it is measuring, it would be difficult to say that the repeatability could achieve an 18-bit level.
We have recently released a new LVDT signal conditioner reference design. I would definitely recommend checking it out:
Retrieving data ...