IIP3s are different on left and right side of LO frequency

We have tested out chip, but found out that IIP3 value measured on + side is much lower than - side.

I couldn't figure out what cause this. From mathematical view, they should be same. Can anyone explain this phenomenon to me?

Thanks,

If an amplifier gets imbalance between IMD products, the amplifier exhibits so named ?memory effect?, which is cause by a lag between AM-AM and AM-PM characteristics.
Generally this is due to a poor bias decoupling at low frequencies causing a distortion of the envelope currents which result in IMD asymmetry.

Thanks, vfone,

Can you explain it a little bit more, or do you have some paper regarding this issue. I am so glad that you know the answer of this.

Thanks a lot.

http://ee.eng.usf.edu/WAMI/resources...urce_study.pdf

http://herkules.oulu.fi/isbn9514265149/html/x1900.html

http://pasymposium.ucsd.edu/papers20...d_PosterA4.pdf

Just wondering if you tested this at many frequencies across the AMP's Bandwidth.
You might be on the edge of a filter for example...

Yes, I tested it across the whole band - actually, the receiver is designed for wide band - it exhibits the same behavior, but the differences between left side and right side are different across the band.

ths a lot for vfone

hi, now I encounter the same problem. Anyone can tell me how to get the proper IIP3 values according to the measured results with the imbalance between IMD products? Thanks a lot!

Select the worse IM3

Thanks for tony_lth! But is there any reason to select the worse IM3? Say, Why not select the average value of the two IM3?

Such as for PA IM3, the spectrum from PA must meet the requirement of MASK. The worse IM3 maybe break the MASK, but the better one maybe not. At this situation, the left spectrum and the right one can't be averaged. So the IM3 should select the worse one.

OK, I catch you. Thank you, tony!