'optocoupler will not solve the problem.'

Hallo Mohan,

that's not true. In most cases optocoupler WILL solve problem. But, it's not the best way, because the reason of common mode noise, means potential difference between ground at input of signal conditioner with AD574 and ground at PC terminal is still alive. Only by making MANY connections between of all such grounds (by even using metalwork like ladders, pipes, a.s.o.) common mode noise can be decreased. Small numbers of ground connections make SMALL numbers of BIG loops, makes high induced voltage inside. But many numbers of connections, forming a ground mesh (finally resulting in idealized ground plane!), make BIG numbers of SMALL ground loops. Resulting areas of these ground loops are so small that induced voltage drastically decreases!
That's the solution when heavy common mode noise has to be struggled with. If you decrease numbers of ground connections, for instance by inserting a ground isolating optocoupler, chance of decreasing common mode noise is wasted.
But this does not mean that optocoupler would not work. As long as isolation barrier is not destroyed (e.g. by induced voltage of it's own loop!) it will work.

Many people actually don't know how a ground loop works. There are mainly two different processes (besides many other disadvantegous effects): 1. Induced voltage, 2. Impoverishing or 'milking' magnetic field.
1. If a wire or metallic connection forms a loop, and ends are close together but not connected to each other, a voltage is induced when a changing magnetic fields crosses the loop. It can be well measured by connecting a voltmeter between both ends.
2. If ends of wire are connected to each other situation drastically differs. Now, a short circuit current flows along wire, driven by the induced voltage. Because now wire milks power from magnetic field, bad effect of ground loop is INCREASED. When this short ciruit current flows along sensitive wires (e.g. at signal conditioning amplifier) voltage drop is amplified as if it were part of signal. But it isn't of course. If now short circuited wire is opened again, example given by inserting a potential barrier (optocoupler) voltage drop of this short ciruit current stops. And signal conditioning amplifier also stops amplifying erroneous signal. Instead of optocoupler some resistance or capacitance will show the same effect.
BUT: Induced voltage between ends of wire is still there. And it's bigger than when having wire short-circuited!
So, with one big loop, you can only choose between two deadly deseases.
Only maximizing numbers of ground loops, forming a ground mesh will solve problems.

How to actually proceed?
Assume you are in an industrial environment and you have some problems with common mode noise. Then, take many big wires and make connections between all metallic parts like ladders, pipes, frames, chassis, enclosures and such stuff. Result should look like an idealized ground plane. Then, have at every equipement green-yellow wire of mains connected to this ground mesh. This minimizes common mode noise so drastically, that most problems are already solved.
When connecting electronic equipement use only symmetrical signal routing. Then, if there's still some common mode noise left, symmetrical input stage substract it from signal, as good as common mode rejection ratio defines.
In such scenario you still can use optocouplers (of course with much success!), but do it never without having this ground mesh: If you connect two points showing much common mode noise there's a big danger that by some spike, burst or ESD event an induced voltage can be built that completely destroys barrier of optocoupler. Quicklier than you can think! Never thought about why optocouplers are becoming destroyed, from time to time???

By the way, obeying above grounding technique is the only remedy against common mode noise which AT THE SAME TIME leads to minimzed RFI (emision and susceptibility) and it's the only solution that does NOT violate the concept of COMPLETELY SURROUNDING FARADAY CAGE. If there were letters bigger than that used, I would take them. Advantage of genius concept of Faraday cage cannot be overemphazised.

Bye,
Kai