Dear All,

   I am simulating a Gilbert Multiplier. When I start with the noise figure simulation, the results is 140dB. This result is a little carzy. Also I have seen some topics about the simulation of Mixer. But I think mixer is a little different from multiplier.
   For mixer, I think if the switching transistor does not contribute the noise, the analysis is similar with the common-source cascade LNA. But for multiplier, the switching transistor also contributes the noise, but for the input transistor of RF signal, its noise is mainly common-mode noise at the output of multiplier because the input transistor of template signal is always working in saturation region.
   So in this case, how can I simulate the noise figure correctly?

   Some description of my multiplier: double balanced Gilbert multiplier, but there is no current tail at the bottom to increase the voltage headroom. application for 3-5GHz Impulsed UWB receiver. RF input port is 5GHz sinewave port(50ohms), template signal is also 5GHz sinewave port (50ohms). The output port is 50ohms (dc). there is no impedance matching at each port. Because I am not familiar with PSS simulation and QPSS simulation, which one should I use, how can I set up the parameters?

   I am appreciating your help!

Jack

I can't find out any problem in your post. Your post is not design issue but just very simple common usage of simulator.

Jack wrote on Feb 16^th, 2009, 3:30am:
Maybe you are wrong in analysis setting. Do you understand DSB_NF, SSB_NF(USB or LSB) correctly ?
What value is conversion gain ?

Jack wrote on Feb 16^th, 2009, 3:30am:
It depend on your needs.
PSS is a expansion by single frequency basis while QPSS is a expansion by multiple frequency bases which have no common divisor frequency each other.

If your questions are usages of very specific EDA vendor's simulator, post to "The Designer's Guide Community Forum ≫ Simulators ≫ RF Simulators".

Jack wrote on Feb 16^th, 2009, 3:30am:
No, you are wrong.

Jack wrote on Feb 16^th, 2009, 3:30am:
What do you mean by "template signal" ?

Jack wrote on Feb 16^th, 2009, 3:30am:
You mean a difference of Gilbert Cell
  as four quadrant multiplier
      and
  as mixer such as down converter or up converter ?
But there is no difference about treatment of them in simulation, even though they works as multiplier or mixer.

Jack wrote on Feb 16^th, 2009, 3:30am:
What do you confusing ?

Jack wrote on Feb 16^th, 2009, 3:46am:
Not, correct. Rather fixed frequency of LO is rare in mixer application.

Jack wrote on Feb 16^th, 2009, 3:46am:
What  do you mean ?
I think this is zero-IF mixer or matched detection.
Anyway there is no difference about treatment in simulation.

Jack wrote on Feb 16^th, 2009, 6:57am:
This is very common four quadrant multiplier.
There is no special issue about treatment in simulation.
Your diagram is RF matched detection(correlation detection).

Jack wrote on Feb 16^th, 2009, 6:57am:
Every circuits works in the time domain. Nothing special. What are you confusing ?

Until now, I didn't simulate the conversion gain, I just simulate the noise figure,

1. Do I need to set any input matching? Now I am simulate the multiplier seperately. But the multiplier will be connected to LNA at last, 50ohms matching is not fit in this case.

2. I follows the general mixer simulating method. the RF input port, I set the parameters as follows:
50ohms, dc
the Template port: 50ohms, sine, 50mv, 5GHz.
the Output port: 50ohms, dc
PSS setting:

Pnoise setting:

Jack wrote on Feb 16^th, 2009, 7:23am:
Why do you set Reference Sideband as 10 ?
Again do you understand DSB_NF, SSB_NF(USB or LSB) correctly ?

DSB_NF, SSB_NF(USB or LSB)

For DSB_NF and SSB_NF, I think I know the conception, but for USB and LSB, I do not.

Thank you very much!

I change the reference sideband to -1, then the noise figure is decrease much.

Do you think this noise figure is reasonable?

This is my first time to simulate the multiplier, I have no idea of it.