Gavin Jackson wrote:

> The reason the circuit is floating, is that it comes from a
> pH probe with a very high output impedance.

  Ah!  It becomes clear now!  That is true, but the trick is that only
*one* electrode (the glass one) has a high impedance, the other is quite
low.

>  I was told that if a supply referenced to ground was used, many
> problems could be introduced.

  Quite so, but not half as many as if you *don't* reference it to
ground.  That suggestion was made in ignorance of the need of telemetry!

  My suggestion:

  1) Ground the solution(!) with a third electrode of whatever material
is practical.  With some luck, it may already be in a metal pipe or
container which can be grounded.  Otherwise *shield* the whole measuring
area with a grounded surrounding shield, either solid metal or mesh.
Reason:  It's far easier to ground out interference (mains) than to
balance it out in the amplifiers or isolate it.

  2) Determine the glass electrode.  You presumably already have a
circuit to measure its "signal" (voltage) at the ultra-high impedance.
Only the buffer part is required.

  3) Provide a medium-high impedance buffer (unity gain) for the
reference electrode.  Frequency compensation must be identical to the
previous buffer.  In fact, the components are not too expensive
nowadays and it will be just as easy to use the same circuit.

  4) Provide a differential amplifier to compare the two with precision,
temperature compensation and the necessary gain.  This should operate
from the standard "instrumentation" amplifier supply of +/- 15V and the
common-mode signal with the test solution grounded/ shielded should fall
well within this range.

  The last three parts should in fact already be part of the design,
being a standard triple-amplifier "instrumentation" design.  Such are
readily available from Analog Devices or the like, already assembled for
the purpose.

  Well, my suggestion anyway.  Any holes to pick?

  Now Russell McMahon <apptech@clear.net.nz> wrote:

> If the floating signal is not "isolated" per se and only floats within
> the power supply range of a standard op-amp (say about +-15 volts )
> then a standard 2 input differential amplifier will give you the
> ground referenced result with as little as 1 op-amp (3 for a full
> instrumentation amplifier).

  Actually, if the floating signal is reasonably low in impedance (say,
less than 1 megohm) and reasonable level (a volt or so), it may still be
measured in this way using resistive dividers up to a few hundred volts.

  If a low-impedance common-mode reference is available (such as the
third electrode/ container previously suggested), a floating buffer may
measure and amplify a higher-impedance signal and relay it to the
"ground" reference via such resistive dividers or a current driver.

> If the floating voltage floats at much higher voltages or true
> isolation is needed then "telemetry" style approaches may be needed.

  I have to agree.

  Cheers,
        Paul B.