Expected linearity of diff pair

[exp]

Hi,

Can anyone give me numbers from their experience of which linearity a plain diff pair in a modern process should be able to achieve?

Using a 28nm SOI process I built a simple diff pair with ideal current source loads and tail source (i.e., the output voltage is given by gm/gds). I am using pss and paq to measure IIP3 which is about -20dBm!!! (does not change significantly with different W and currents). This sounds suuuuuper low to me, I was hoping more for +20dBm ...


Thanks!

 

[erikl]

I think the standard definition for IIP3 is

The third-order intercept point relates nonlinear products caused by the third-order nonlinear term to the linearly amplified signal

(see e.g. here), so should be < 0 dB , isn't it?

 

[exp]

I am talking about dBm ... so can be > 0.

For example, for a mixer (only) gives me +30dBm which is very good.

So I am asking mainly about your experiences and if -21dBm can really be true.

If it's easier I would also be happy with HD3 numbers + input amplitude because there is a relationship:

IIP3 = 30 - 10 log6R - 0.5 HD3 + 10 log vi^2

 

[erikl]

I am talking about dBm ... so can be > 0.

Ok, overseen, sorry! So you're talking about the IIP3 point, measured in power units.

For example, for a mixer (only) gives me +30dBm which is very good.

1W ≙ 7V @ 50Ω is rather extreme for a mixer, I'd say.

So I am asking mainly about your experiences and if -21dBm can really be true.

This would mean an intercept point at an output voltage of about 0.5V at an impedance of 25kΩ. Possible ...