The Designer's Guide Community Forum
Design >> RF Design >> LO leakage in direct conversion??????HELP

Message started by Jing on Oct 25th, 2003, 9:28am

Title: LO leakage in direct conversion??????HELP
Post by Jing on Oct 25th, 2003, 9:28am

             I am an university student. I use Cadence to analyze the LO leakage of direct conversion Gilbert cell mixer. It is well known that LO selfmixing is quite serious, sometimes the DC offset can saturate the following circuit. However, my simulations show me that LO feedthrough is not serious at all, when LO = -10dBm, LO feedthrough at RF port = -110dBm. I feel quite strange. Some papers often said their mixers have LO leakage around -40dB or -50dB.
            I have used two methods to obtain the LO leakage. One is PSS+ PXF to get the LO signal gain at the RF port with RF source as DC, the other is only PSS analysis to get the RF port signal at those harmonics of fundermental frequency, therefore the signal power at LO frequency is obtained. Both of the two methods give me very low LO leakage.
            I wonder whether the methods are right or wrong. I am looking for a CMOS Gilbert cell mixer circuit with poor isolation, high LO feedthrough. If somebody has, wound you be so kind to share it to me? Please send it to my Email box
" ", thanks a lot.

Title: Re: LO leakage in direct conversion??????HELP
Post by pancho_hideboo on Oct 25th, 2003, 3:08pm

You don't include parasitic effects due to physical layout.

But without them you can observe large LO->RF leakage.

Your circuit under simulation is perfect ideal and symmetrical ?

Actual gilbert cell is not ideal DBM.
So try to set your circuit in non-ideal situation.
For example,
  * make LO+ and LO- unbalanced
  * make transistor quad or diff-pair unbalanced.
  * make bias resistor unbalanced
  * make loads unbalanced

Don't you observe LO leakage at RF port as differential ?

Title: Re: LO leakage in direct conversion??????HELP
Post by Jing on Oct 26th, 2003, 7:43am

           Thanks for your help.
           Yes, I do not include the parasitic effect due to physical layout. Now my circuit is perfect ideal and symmetrical one. Some books and papers say DC offset in direct conversion is caused by many reasons, such as LO feedthrough, transistor mismatching(unbalance) and so on. LO feedthrough is an inherent problem due to Gilbert cell structure. So what you mean is that unbalance and LO feedthrough are relative each other. If the mixer is perfect ideal and symmetrical, there will be no significant LO feedthrough. Only unbalance can induce the LO feedthrough. Am I right? So the simulation results i got previously and the methods are correct.
          I will try to add in the unbalance for the circuit to see what it looks like. Actually before i tried to separate LO feedthrough from unbalance. I think they are independent. If I add unbalance to the circuit, I will not know what is the contribution of each for the DC offset.

Title: Re: LO leakage in direct conversion??????HELP
Post by rf-design on Oct 30th, 2003, 2:35pm

If you using an I/Q gilbert mixer pair you will see only harmonics of double the LO frequency at the common emitter. If the LO drive is ideal I/Q both common emitters will be 180 degree out of phase. If both common emitters will couple to the I/Q mixer input the double frequency component cancel. What leaves is the four times LO leakage. That is what you get with ideal matched symmetric signals and devices. If you make mismatch analysis you can spec how good your process and your layout should be for some leakage spec.

The Designer's Guide Community Forum » Powered by YaBB 2.2.2!
YaBB © 2000-2008. All Rights Reserved.