As an addendum to the differential A/D and instrumentation amplifiers, and
depending on the precision you require, a standard differential amplifier
using 1,2 or three opamps could be used. As the "common" ground connection
is taken as being a low impedance source, the single op-amp config may well
be adequate. In this configuration run the "common" side to the inverting
input and the "signal" side to the non-inverting input. Common mode error (&
l.f. noise rejection) will  be mainly determined by resistor matching so us
1% or better if you need useful accuracy. This will limit any current loops
to the values determined by your input resistors.
If you need better isolation a double op-amp config will give you high input
Z but somewhat limited common mode range. 3 op-amp will give you best
performance overall - but at the expense of realestate.
Suggested opamp - OP177 for low DC drift & offset (I think a quad pack
version is also available now)


Richard P

> -----Original Message-----
> From: Maverick [mailto:maverick@DRGW.NET]
> Sent: Friday, October 29, 1999 6:01 AM
> To: PICLIST@MITVMA.MIT.EDU
> Subject: Re: [PIC] Measuring Analog Voltages w/o Sharing Grounds
>
>
> >a) Average Voltage to Power circuit.  6V-24V
> >b) Average Current in Power circuit.  Average current during
> the last 100mS.
> >c) Voltage of Control circuit.  Control is a 3-6V source
> >d) Voltage of Monitor circuit.  Monitor is a 4.5-5.5V source
>
> >All data points will be sampled at 10-30 Hz (selected by customer at
> >runtime, with no hardware changes permitted).  8-bit data is
> sufficient at
> >this stage, but they may want 10-12 bit in the future.
>
> >The Monitor circuit is supposed to be small (as small as
> possible) and
> >needs to run at least 1 hour on batteries (NiCd rechargeable
> or Li coin
> >cells).  Of course it's mobile, and will use an RS-232 interface to
> >transfer data to a laptop, either during operation or at the end of a
> >testing period.
>
> >How do I hook these separate analog signals up without
> sharing the ground
> >with the Power circuit?
>
> Depends on what your design constraints are, obviously - how
> much additional
> cost is acceptable, etc...  Do you intend to use the internal PIC A/D
> converters or
> an external part?  If you are using an external A/D, is that
> set in stone
> or flexible?
>
> This post intregued me because I've spent the last four years working
> on EV telemetry and power systems, and we often face similar
> problems in
> terms of noise,
> coupling, floating voltages, and you name it...
>
> Personally, I'd recommend you use true differential delta-sigma A/D
> converters - and no,
> before everyone jumps on me, these aren't necessarily out of the price
> range on any reasonable
> project - Analog's AD7705 is a relatively inexpensive (~$4 in small
> quantity, I think) dual differential 16 bit D-S A/D that has an
> SPI interface and works very nicely under extremely noisy conditions.
> Because of the
> integration-deintegration performed in the D-S process, noise
> and random
> jitter in the data is reduced
> significantly - to the point that one of these was used on
> the output of a
> large (500W) switching power
> supply to measure current through a shunt- and had consistent
> and accurate
> average currents out to 10 to 12 bits
> without a problem.  And with a main ground bus that can have
> ~2-3 volts of
> ripple on it due to currents and inductive
> kickback, this is no small feat!   Power consumption on these is very
> reasonable for battery-based applications - I believe
> they can be standby'd to 10uA or so...
>
> If that isn't an option or you would prefer to use some other A/D
> converter, just put an instrumentation amp on the
> front end - these usually have two drawbacks:  they're
> expensive if you go
> for really high-quality versions, and otherwise
> their common mode rejection isn't the greatest at higher frequencies.
> Since you're only looking for averages, though, you
> could R-C integrate off all the high frequencies and leave
> the in-amp to
> only convert the differential input signal you're
> trying to monitor down to a "ground" referenced voltage for
> your circuit to
> read.  I'm personally a fan of Burr Brown INA2128s, but
> these are costly if you actually have to use them in a
> production circuit.
> Analog also makes some very nice ones at lower cost, I
> believe.  For your application, they also have one other
> drawback I just
> thought of - they're usually current hogs.  On the order of ~20mA.
>
>
> Just my two bits for the day...
>
> ND Holmes
>
>
>
>
> --------------------------------------------------------------
> ------------
>  Nathan D. Holmes   maverick@drgw.net, ndholmes@iastate.edu
>    122 Shepard #3  Box 328  Gilbert, IA 50105  Iowa State
> University - EE
>    http://www.drgw.net/~maverick   PH: 515-663-9368
> --------------------------------------------------------------
> -----------
>