lna mixer simulation ads

Im doing a design of front end of RF receiver. I have designed the LNA, mixer and VCO according to my own value by referring to the tutorial from rfic.co.uk. The tutorials are CMOS Gilbert Cell Mixer, CMOS Common Source LNA and LC Oscillator Tutorial. The simulation of the individual design has no problem but problem occurs after combining them.

The output of mixer after a bandpass filter was not the difference between LNA & VCO. The output is still fine with LNA alone connected to the mixer. I was trying to input 2.5GHz for LNA and 2.25GHz for VCO. But the output was 4.097GHz which is the same as measured in LNA but differed from my input of 2.5GHz. I found that the oscillator mode need to be enabled for vco simulation but not for LNA. In fact I dun really understand the settings for simulation.

Im very new in RF and has never done anything similar. I wish to get some advices through this forum...

Thanx

Which tool is this?
May we know the example may be location.
Kindly share your design if possible.

im using ADS for the simulation. The examples i followed were obtained from www.rfic.co.uk. Attached here is my design and problems facing.

It seems we have to work a lot to come out...

First: LNA+Mixer sim. You do not use a local oscillator for the Gilbert cell. This is wrong for the proper mixer operation.
Furthermore the bias of the Gilbert cell is wrong and 18V in the LNA is very high! It is not clear how you biased the Mixer current mirror.

My first suggestion is to switch off input signal and verify bias of all transistors in your circuit.

Mixer is used to translate the signal frequency into another band.

What do you expect from first simulation? Why you did it in this way?

(there's a lot more to say on other schematics, but I would like to proceed step by step)

Mazz

It looks that RF input power is differing in both cases.
May I ask you to share the zap file of this project?

what is the use of local oscillator in the mixer? I attached here my simulation of the mixer alone. Im using 1.80V for LNA, not 18V. I have ran the mixer and LNA simulation individually and I was able to obtained the proper result, so i did the 'first' simulation to see whether the mixer functions properly. The mixer output should be the difference between LNA & VCO am i right? (after the bandpass filter). I'll look into the biasing again...

I set the RF input power to be 30dbm for both cases...

Thanx

OK, we've made the first step.

I don't have ADS intalled, so I cannot look at your schematics, but now the doc is more clear and I have understood how the Mixer is biased and it works properly. 10 dB Gain and 10 dB NF is a common result.

To come back to your question (what is the use of LO in the mixer) it is clearly explained in the introduction of the tutorial you have mentioned, and you are right, in this case, the output mixer freq is the difference between RF and LO. But when you simulated LNA and Mixer you removed the LO, that is wrong.

18V issue: I was referring to LNA input signal peak level. Be aware that you applied +30dBm (that gives you 18V peak level) into LNA that is too large.
When you simulate the mixer you applied -30dBm, that is more correct.

Mazz

I have now changed the RF input power to -30 dbm and ran the simulation together with LO (VCO) and LNA. The result obtained is the same as the previous result. Is my connection between mixer, VCO & LNA correct?

Thanx

Merry Christmas~~~

Merry christmast to you, Dexing.

Now we've come to the last (and major issue) part of the problem.

In VCO stand alone simulations, you let the simulator find the freqs of VCO oscillations.
In Mixer (or in LNA+Mixer) you force the solution freqs.

Putting both (VCO and LNA+Mixer) together, what to do?

I have done it last time some (5?) year ago, but I'll try to remember:
1. First solve the complete circuit with NO RF input (at LNA) as an oscillator. You should find at VCO oscillation nodes (and NOT at mixer output) a very similar signal of the VCO stand alone simulation (It will differ only because now you have a different load on tank. If there is a large difference, you should add a buffer between VCO and Mixer, this is always done in real designs). Print the exact osc freq.

2.In a second simulation (similar to the LNA Mixer simulation) put as 1st tone (that correspond to LO) in the HB the previous saved VCO oscillation freq and activate RF signal at LNA input, but disable the oscillator analysis.
In this way the simulator should find a correct solution for VCO and superimposes on it the RF signal.
I hope I remember well.

A general suggestion is to plot HB solutions in freq domain to see clearly freqs contributions.

Another possible solution is to use transient instead of HB. In transient you do not have predefined freqs, you just need to give a perturbation to VCO to let it run. It will take longer time to reach the steady state (HB gives you directly the steady state solution), but can help to debug the circuit.

Max

Hi dexing,
I too observed the same thing. When we set to osc mode then simulator is not considering the excitation by the RF input power source.
When I toggle this control then only recognizing the input RF power.
I guess that in osc mode it may be checking for internal possible oscillating mode frequencies.

Hi
I was away for few days~~~

Mazz, i attached the result of both the simulations u mentioned. But actually i dun really understand the second simulation. Did u mean that I replace the LO with a signal generator where the signal is the same as the signal I obtained from the stand alone simulation of VCO? n I couldnt find the plot using freq domain :p

kspalla, this means that it is not possible to simulate both of them simultaneously?

dexing

sorry but maybe I wasnt too clear.The schematic should be always the one you have copied in figure 3 of output5.doc. The only modification is to add/remove the RF port.

I see from your doc that, in VCO simulation, the EXACT oscillation freq is 2.25G. This is really strange. The VCO should find its own natural oscillation freq, not the one you have imposed to. check it.

To plot signals in freq domain simply remove the 'ts' function and use 'dBm'.

Mazz

Happy New Year~~~

I have done another simulation, but still with some problem. Regarding the natural oscillation freq of VCO u mentioned, there is a control voltage which determine the oscillation freq, so I just let the voltage floating without putting any value. Is this the way to obtain the natural osc freq? Previously I set the LO to be 2.5G because I wanted the IF to be 250HMz.

Thanx

The answer to you question is:
-as you mention the VCO oscillates at a freq that is controlled by VDD2 (in your schematic).
-what I mean as "natural" freq is the one you obtain without forcing it at 2.5G. If you set LO at 2.5G the solution that the simulator finds is wrong for the VCO.

Did you noticed that:
figure 2. osc freq 2.545
figure 3. osc feq 2.272 - it is changed due to the higher parasitic cap due to Gilbert cell load. This is your REAL oscillation freq. If you want your VCO to run @ 2.5G you need to move control voltage (VDD2) of this simulation in order to obtain 2.5GHz.

figures 5&6. Make no sense for me. If you disable the osc analysis and use only LNA freq (let's call it Frf), the simulator can find only the Frf solution. It seems very strange for me that the simulator converges properly. In any case this is not the real solution.

figure 7. I do not understand.

What you should do now, in my opinion is:
1. find the VDD2 that gives you 2.5G in complete circuit (osc enabled). Take a note of the EXACT osc freq (down to Hz or lower), lets call it Fosc

2.disable the osc and use two tones in HB, Fosc and Frf. The simulator solves the circuit for the two freqs and their linear combination (set by max order, use 4 or 5). In this simulation you will find the correct signa at IF.

Happy new year.

Mazz

Hi Mazz
I juz started my new semester n was bz with other stuffs...

actually i dun quite understand the 'disable the osc' that u mentioned. u mean disabling the osc analysis? if I disable it then the result will be the same as in figure 5&6 I uploaded previously.

Thanx

Yes I mean disable osc analysis.

" if I disable it then the result will be the same as in figure 5&6 I uploaded previously":
This is not correct.
Read carefully my suggestions. You used only one tone in figs 5&6.

Mazz

I have now used two tones for the simulation. But I still couldnt get the proper output IF. Is there still any errors in my simulation?

Thanx

Do you see any convergence issue when disabling osc mode?
I don't remember how the osc port behaves is osc mode is disabled. Try removing it.

Mazz

Hi
Im really sorry for 'disappearing' for such a long time...

Removing the Osc Port was just like disabling the the osc mode, their results are the same. What is a convergence issue?
Is the DC voltage going into the mixer from LNA n VCO affect the result? The DC voltage from VCO is found to be 0.946mV while LNA is 0V.

Thanx

RF input systems.

There is a general problem with simulators to combine the LN, Mixer and Oscillator in most simulators. It is the first an correct way to simulate and optimize the 3 stages alone, and never assume signals with more then -40dBuV to avoid compression and saturation. Also run all S parameters at the same drive level for the LNA and the mixer. What are the DC currents and voltages , and what transistors (foundry ) are you using ?

申明：网友回复良莠不齐，仅供参考。如需专业帮助，请学习易迪拓培训专家讲授的ADS视频培训课程。