If I have understood you correctly (i.e. you have a IQ modulator direct conversion TX) Sounds as though you have an IQ mismatch problem, most likely due to the LO. You can correct for it somewhat by adjusting the input baseband signal IQ signals relative phase and amplitude.

PCS/DCS is a half duplex system, so the RX should be switched off whilst transmitting.... the fact that switching the RX on causes a change in your SSB "image" rejection maybe down to the fact that the RX causes a change in the LO IQ phase/amplitude matching.

Cheers
aw

Yutao Liu,
Let me restate what you are saying:
Receiver off sideband 21dB
Receiver and LO on 27dB
Un-soldering receiver matching network 30dB.
Questions what state is the receiver and the receiver LO in case III ?
By switching between on and off is there any difference?
if no
Then I would take a look at the isolation between the TX port and the TX mixer, Its possible that when the receiver is on it presents an impedance to the TX port where the TX mixer is better behaved.
There are many ways that imbalances can occur but in my experience the substrate as the cause is not it.  (most of the time when a root cause cannot be determined the substrate is the default culprit, but I have found that metal magnetically/electrically is always the offender, and usually through the supply or ground).  But if the artifact can be observed from external changes then the board chip interface is the likely the smoking gun.  Devise an experiment such as a load pull to see how sideband suppression varies as a function of load impedance.

http://rfcooltools.com

Yutao Liu,

The piece of information you gave where having the rx circuitry on gave better side band suppression.  You should get to bottom of why this is the case?  

As in many isolation situations there can be multiple paths.  Does this new switch achieve better isolation how?  What is the impedance looking into the old switch when the rx is loaded with the matching circuit and with out.  What does the new switch look like?

How does your sideband suppression look when the tx port is unbalanced (if differential) and/or mismatched ?


Power supply coupling can be reduced by onchip bypassing at the circuits you don't wan to talk to each other.  Think about it this way, think of supply coupling reduction from the signal current's perspective.  A current is always looking to find the loop with the lowest impedance.  keep the loops local and of low impedance then they will be less likely to leave through a bond wire.  Does the isolation change when you for example put your finger on the supply?  If it does put a ferrite (choose one that doesn't inhibit normal operation) bead in series with the first on board bypass cap, did the isolation improve ?  


http://rfcooltools.com

This system would normally be run time division multiplexing with the receiver-transmitter never on at the same time.

Suggest you analyze the system that way. Many GSM chip sets will not function with both sides alive at the same time.

hello "http://rfcooltools.com",
Thanks for your suggestion. I am trying the method you supposed in the last post.

Before that, I did an experiment to analyze your problem.  A 1880.32MHz sine wave is input through the RX port to model the transmitted signal when the RX is shut down. And then a spur appears with 640KHz offset the VCO output frequency, which is 3760MHz. The frequency of VCO is 2-divided in the LO path. So I suspect that this spur degrade the sideband suppression of TX.

Furthermore, I do another test to observe the spur. When the VCO output frequency goes far away from 3760MHz, while the input sine wave stays 1880.32MHz, the power of the spur decreased.

Dose the transmitted signal pollute the spectrum of the VCO through the RX port? And this is the culprit of the degradation of the TX sideband suppression? Is there any other ways I can prove my assumption further?

Thanks a lot!
Yutao Liu

Yutao,

The experiment you tried is interesting, but still leads to a few possibilities causing the problem.   Does a simulation yield the similar results?  

When you injected this signal into the rx port, was it still connected to the TX port through the switch?
If yes it is so then it might be the mechanism of coupling you state or at least a similar mechanism.  but since the vco is running at 2x the LO then its probably getting on the divider possibly through a common supply.  This mechanism once found should be an easy fix.

if no,  Does the experiment you tried yield similar results if you remove the rx path matching circuit completely.  If so then I would be more inclined to believe its the transmitters reflected signal causing the imbalance, hopefully this is not the case!

For the yes case
Using a high impedance probe (or a simple stripped coax with a DC block to act as an antenna) and a spectrum analyzer probe the supply for the VCO by injecting the a tone into the disabled rx path.  Also look to see if you get strong relative sidebands around the VCO and 2xVCO this would indicate that the its on the VCO signal.  Record relative levels of all spurs, and try to reproduce them in the simulation by introducing them one at a time until you see in simulation the same as in measured.  

Good luck
http://rfcooltools.com

OK, you are hard biasing the current source off, but the rest of the circuitry is allowed to float.

So, what paths in the system attached to the front end matching network are now powered down but still connected?

When you need good isolation of coupling paths, often the powering down of current sources is not sufficient. Sometimes you need to connect a switch to many of the high impedance paths that are unbiased and inactive, but can serve as zero gain (or lossy) transmission paths.

When in doubt ground it out!