CC1120EM-169-RD: TX Simulation and TX Actual S11 VNA Measurement Discrepancy

Tool/software:

I am working on a design for a high volume consumer product.  I am evaluating the CC1120 radio by using the CC1120EM-169, 169MHz evaluation board.  I have two of these boards set up as-is, sending packets back and forth just fine, so I am not having any issues with the software side of things.  I am using 154.4MHz to send the packets back and forth and this works fine even though this is about 15MHz lower than what the eval board is tuned to.  Looking at power, going directly into an R&S FSVA spectrum analyzer, I get the power I expect even though I am a bit below the ideal tuning range of the board.

For my intended design the CC1120 will be used with an external power amplifier, so I would be separating out the transmit and receive sections, and using an external RX/TX switch.  Before I do this I obviously want to be sure I understand the topology of the reference design. 

To do this I have made actual S11 VNA measurements of the eval board, as-is, while it is in receive mode, and while it is in transmit mode where I have set up the CC1120 to just output an un-modulated carrier at about a -10dBm output power.  The measurements were made using an R&S ZNB VNA.

I have also simulated the CC1120 eval board taking into account component parasitics, the SMA connector delay, and the short transmission line traces from the TRX pin, and from the last harmonic filter stage to the SMA connector.  For RX mode I modeled the PA pin as a high resistive impedance as well as the TRX pin, and in TX mode I modeled the TRX pin as being grounded through a low impedance.  I am using the datasheet values at 169MHz for the RX differential impedance, and for the PA output impedance.

In RX mode, the actual VNA measurement agrees well with the RX mode simulation.  

For TX mode, my simulation actually shows a perfect 50+j0 match at 169MHz, but the VNA measurement, while in the ball park of the 50 ohm resistive circle, is rotated upward and then it curves to the right as frequency increases.

I cannot figure out why there is such a discrepancy between the simulation and the real-world VNA measurement.  I am using a -10dBm signal from the VNA for the Hot S11 TX output measurement, direct into the VNA from the eval board's SMA connector, and I'm using 120 averages in order to clean up the star burst display that would result at the frequency my un-modulated carrier tone is at when the VNA sweep hits that frequency.

The un-modulated carrier frequency I am using is 154.4MHz, same frequency as the packet testing with again seems to work great.

I would appreciate any insight as to why there is a disparity between simulation in TX mode, which looks perfect, as if you all at TI really nailed the design, and my actual VNA measurement.

I have attached PDFs showing the simulations, the simulation results.  I will post the VNA measurements, and the raw VNA trace data files separately as it seems I can't add more than 4 files here at one time...

Again, I appreciate any insight as I do need to make sure I understand the CC1120 and its datasheet specifications.

One important question I have:  The datasheet shows different matching impedance requirements for the PA at different frequencies.  At 169MHz, according to the datasheet the PA pin wants to see 80+J0.  Does this change significantly below this, and does having the CC1120 output a 154.4MHz carrier affect a Hot S11 measurement that sweeps from 120MHz to 180MHz because the PA is being affected is some way by this carrier frequency it is amplifying during the VNA measurement?  

Thank you,

DavidTI_CC112x_EvalBoard_RX_VNA_Sim_Unmodified_Diff-Z_RX_SmithPlot.pdfTI_CC112x_EvalBoard_TX_VNA_Sim_Unmodified.pdfTI_CC112x_EvalBoard_TX_VNA_Sim_Unmodified_SmithPlot.pdfTI_CC112x_EvalBoard_RX_VNA_Sim_Unmodified_Diff-Z_RX.pdf