Hi,
      I have seen in some research papers that IMD3 is measured at 3-dB output power backed off from P1dB....for IMD5 it is 5 dB backoff...
I dont get it...why its not measured at P1 dB or just 1 dB back-off...
why we need to measure it at such low  output power levels..

The main reason for measuring IM products at drive levels significantly below the P1dB compression point is that the individual contributors to compression (IM3, IM5, IM7, etc.) are only clearly identifiable at the lower input amplitudes. That is to say, the classic 3:1 relationship in power for the third-order distortion components and subsequent 5:1 for 5th, 7:1 for 7th and so on are only directly measurable at lower drive levels. By the time compression sets in there are contributions from multiple orders that are significant making it difficult to separate the different order contributors at those power levels.

Also, the concept of intermodulation intercept point is explicitly defined in terms of individual order contributors. There are no composite intercept points which somehow account for all possible nonlinear contributors.

So, the idea is to measure the intermod products where their respective slopes match the order of the intermodulation distortion.

Hope this helps some.

Oh, I didn't understand that you were asking why some high power amplifier vendors specify IM at a few dB backed off from rated power.

I think the vendor is just trying to quantify the distortion performance near the saturated output power where IIP numbers are no longer valid.

But I am not sure how customers use those numbers because the distortion is pretty high (-25dBc), so DPD would be needed anyway to compensate the nonlinearity up to the saturated output power limit.