Gilbert mixer IIP3

The first thing I should point out is that you almost certainly should be using harmonic balance for this simulation, rather than shooting. There's generally few cases where shooting QPSS is advantageous - I normally only use it nowadays for distortion in a switch cap circuit where you have a single clock (possibly with multiple phases) and a large signal input that you want to measure distortion under.

For IP3 of mixers, I'd either recommend hb/hbac (you can also use the older qpss/qpac interface and pick harmonic balance, since you already have it set up), pure hb (three tones) or better still (for efficiency) hb (just LO) plus huac in RapidIP3 mode.

Anyway, that's not the issue here - it's because your frequency selection is wrong. I suspect that the real problem may be that you've got your flo and frf frequencies the wrong way around (i.e. flo should be 2.4G and frf 2.42G), but let's go with what you've picked.

Currently you have:

• flo=2.42G
• frf=2.4G
• fqpac=2.421G

This means that your IM3 is at 2*frf-1*fqpac-flo=-41MHz (so 41MHz). Your 1st order harmonic will be at fqpac-flo=1MHz. It's fairly clear from the frequency multiplier columns in the direct plot form that you have the wrong frequencies - the 3rd order harmonic is dependent on the flo being shifted by 2*flo (which makes no sense for this downconverter mixer), and the 1st order harmonic doesn't involve the LO at all - it's just modulated with the RF input signal. You'd expect to see +/-1 for the flo column, and -2 for the frf column (for the IM3 tone) and +/-1 for the flo column and 0 for the frf column (for the first order tone).

If you're trying to have a second near tone, then if you really meant flo=2.4G and frf=2.42G, then with fqpac=2.421G, the IM3 would be at 19MHz and the IF at 21MHz. Otherwise with your setup the IM3 and IF are 40MHz apart.

Because you have the wrong tones selected in the direct plot form, you're not seeing the 1dB/dB or 3dB/dB slopes for the IF and IM3 curves.

Regards,

Andrew.

What if I do like this:

frf=2.4G , fqpac=2.401G and flo=2.42G  

then I will have 21M and 19M is this correct ?

asrf said:

What if I do like this:

frf=2.4G , fqpac=2.401G and flo=2.42G  

then I will have 21M and 19M is this correct ?

Yes.

still does not make sense

The first plot you have the 1st and 3rd order frequencies the wrong way around (amongst other things). The second plot it looks to me as if you've not set the PAC magnitude on the input source (the "PAC Magnitude (dBm)" field) correctly. Looking at the results, I'd guess you've set it to a fixed value rather than to "prf". You want it swept together with the power of the large signal input. If I set it to a fixed input power, I get a graph very like yours - the giveaway is that the yellow curve is flat (the output power doesn't increase with swept input power) until there's enough power from the moderate tone to saturate the circuit and cause it to start compressing.

Regards,

Andrew.

This is my RF port setting:

I removed the extra pacmag and replaced it with prf. It worked.

Thank you so much for your great help Andrew. 

That works for me. What IC version are you using (Help->About in the CIW will tell you)? Which spectre version? (this will appear at the top of the spectre log file when you run the simulation).

Which version of ADE are you running this in? (ADE L, XL, Explorer or Assembler).

Can you post the input.scs (OK if you want to cut out the circuit - but please leave the parameters at the top, the top level test bench components - i.e the ports, and the options and analysis statements). Note, please post this as text rather than as a screenshot.

Andrew.