Relation of P1dB and IP3

I do a amplifier simulation and got the result which IP3 - P1dB is only
6dB. The relation is very different from 9.64dB. What's the reason behind
this? Anybody has the idea?

Thanks,

The rule of thumb [IP3 = P1dB +10] is valid only under a proper bypassing of the two tone products.
Bias bypassing of the F2 ? F1 product does not affect the compression point of the amplifier, but only the IP3 (3rd-order intercept point).

The 9.64 dB difference is the Mathematical / Theoretical relation.
If you are testing a single FET you should get close to this value.
If you are testing a Gain Block you can't assume the same relationship.
One thing you might be doing wrong is the Power setting of your
Two Tones, if the power is too high you will loose IP3. Make sure your
Two Tone Power levels are ~10dB below your -P1 pnt

Sorry, I did not see that this was a simulation. What Software are you
simulating with. Is this a RFIC or Discrete Component/Substrate design?

Any reference for how the 9.64 dB theoretical difference is obtained?

The simulator is spectreRF.
I often measure the two tone when it's power below 10 to 20dB below P1dB.
Because I often get wrong result, the slope is not 3, when the two-tone power
is too small such as 40dB below P1dB. I have no idea about this, anybody know more
about this?
By the way, if I bias the transistor as Class AB mode, the P1db can be as high as
its IP3?

Here's a pseudo-paper I wrote several years back the derives the IP3-P1dB rule, talks about its limitations, and shows how to null out IP3 to get a higher value.

Greg

Look at pg 6 and 7.
http://www.designers-guide.org/Analy...cept-point.pdf

My power amplifier has the same problem as yours. Even the simulated IP3 is less than the P1dB result.

hello everyone,
i designed an LNA but OIP3 values are nearly -6 and -9 dB.. i think they are very small, what should i do to increase them..
with "they" , i mean 2*f1-f2 and 2*f2-f1
shouldn't they be equal, by the way ?
king regards

There are many reasons why they can be different:
1) You're operating at a frequency roll-off point at the fundamental frequency
2) Your 2*RF frequency re-mix product is a significant contributor to the IP3 AND it's at a high frequency roll-off point
3) Your "delta" frequency re-mix product is a significant contributor to the IP3 and it's at a high frequency roll-off.

a) To confirm any of these theories, first bring the two fundamental tones together in frequency, and as you do, you should see the two IM3 tones approach each others.
b) To test if 2RF is the culprit, put a capacitor to ground on the output. You want the collector.drain of the BJT/MOS amplifier to see a low impedance at 2*RF frequencies. This should lower 2*RF contribution and thus the IM3's should get closer together. You also want a capacitor at the base/gate of the BJT/MOS amplifier to shunt any 2*RF components.
c) To test if "delta" frequencies are at fault, you want the output and input to see a low impedance at DC. An inductor to supply that provides a low impedance to the "delta" frequency is desirable at the output. At the input, this is more difficult since your bias network can't be grounded. In fact, it may be the fact that the bias network itself has the "delta" frequency coupled to it and it multiplies it to cause poor IM3 products. If the bias network is a current mirror, you can lower the current gain and increase the reference current in order to lower it's impedance and make it more linear. If this improves IM3, than delta into the BIAS network was your problem.

Hope this helps.

Greg

9.64 dB huh? You guys make me laugh! Why not expect 9.64000 dB? You need to get into the lab more!

rf microelectronics by razavi is a good reference

In theory , IP1dB = IIP3 – 9.6 dB