Well, a general answer cannot be given since it certainlydepends on the filter order and topology. However, for a second order stage very often the pole frequency wp is identical to the square root taken from the product of the integrators unity gain frequencies.

OK, but what is now your question ? I suppose you have a circuit diagram and corresponding formulas, don´t you ?

At first, I assume you speak about a LOWPASS response, right ?
(because it could be also highpass or bandpass).

For elliptic lowpass filters there are some different topologies available (that means different circuit diagrams) - even when they are derived from a passive prototype.
Each circuit alternative has it own design parametrrs - that means, without a circuit diagram nobody can give you any information.
Sorry, but that´s the problem of filter design: A lot of alternative solutions with different and mostly conflicting pro´s and con´s.
But I suppose you have got some design formulas, don´t you ?

For the active amplifier version of the filter a general rule should be to keep the amplifiers parastic responses well outside the bandwidth of the active filters ole placement.

For example a 500MHz filter should be done using amplifier that have at least 800MHz BW at the worst PVT corner.

That way the response of the amplifier does not interere with the response of the filters defined poles (Bessel, Elliptic,  or Whatever pole constellation)

Generally, that tends to be the approach used, FWIW.

-Jerry

As already mentioned in various replies,

(a) you will need the active element to have a bandwidth a fair bit larger than the highest passband frequency.
(b) the filter topology chosen has a decisive influence on the exact amount.

You could do a sensitivity analysis. This can be done either by solving equations, or else, you just build a simple model of the filter, modelling the nonideal parameters of interest for your active elements (such as bandwidth), and simply do a little bit of parametric simulations to see the effect of the finite bandwidth on overall TF. You know the limits you must fulfil. You can get an idea of the required bandwidth.

However, atleast a very crude analysis on paper is good because then you know what influence you can expect from the finite bandwidth. Also, this allows you a good comparison between different topologies (of course, that can also be simulated, and much faster too).

Regards
Vivek

Thanks for the replies everyone.....

When i say unity gain frequency of the intergrator in Gm-c filters i mean:

ω=Gm/C....(with Gm and C refering to the transconductor and capasitor of the intergrator )