Post Go back to editing

adrv9002 s-parameters

Category: Datasheet/Specs
Product Number: ADRV9002

I have quite a number of questions about the s-parameters for the port impedances on the ADRV9002.  First, I downloaded these files from the AD website: ADRV9001_Design_File_Package\ADRV9001_S-parameters_ver.02.  It has both s1p and s2p versions of port impedance files along with an explanatory README file that I believe I understand.  However, there are a number of issues that I have encountered in using these files and comparing them to what is in the adrv9001 user guide Rev. 0.  The first issue that I have is one that I think I know how to deal with but will ask about anyway.  There are two different types of RX port, RXA and RXB and pages 245 and 246 of the user guide show the different impedances that they have.  There are separate s2p files for these as well, and using AWR Microwave Office, I have been able to use these to replicate the Smith Charts for these inputs.  However, there is only one .s1p file and one laminate file for the RX so is the combination supposed to be representative of RXA or RXB?.  I assume that the difference between the two RX port types is mostly in the laminate which connects the die to the package.  So what should the results be using the .s1p?  In the  simulation below, the Blue trace is from the RXA s2p file, the Red trace is from the RXB s2p, and the pink is from the rx s1p used in combination with the rx laminate s4p file. The simulations from s2p files match the user guide exactly, but the s1p based simulation matches neither, although at low frequencies where the laminate details should not have much effect on the impedances.  I have included the schematic I used for the s1p based simulation.  Since I am not using ADS, I used a different method to create load impedances to attach to the laminate 4-port than was described in the user guide and the tutorial.  I used transformers with sqrt(2):1 turns ratios to reduce the s1p impedances in half and used that to load what i think are the correct ports of the laminate.  I also used an ideal 1:1 balun transformer element to convert the other two connections of the laminate to a single ended 100 ohm port. I assume that this is correct.  However, it seems to me that since the s2p file results match the user guide, wouldn't they be better to use in my design?

I have bigger problems with the TX and LO files.  I used the same methodology to simulate the TX port.  Here the fact that there is only one type of s1p file to use with the one laminate file is not an issue, since there is only one kind of TX port.  However, when I tried simulating, neither way exactly reproduced what is in the user guide.  Here I have included both the results and my schematic. Port 1 in Blue is the result for the s2p based simulation and Port 2 in Pink is for the s1p based simulation.  The s1p based trace is fairly close to what is in the user guide on page 245, but is not exactly the same.  Am I doing something incorrectly or is there truly a discrepancy between the user guide and the posted s-parameters?

For the LO port, I was able to use the .s2p file to exactly match the user guide Smith chart on page 261 up until 6 GHz.  However, when I examined the file I found out that it has no data above 6 GHz. In the Smith chart below, it is evident from the markers on the Blue trace, that the data exactly matches the user guide values in the range for which there is valid data, but at frequencies above 6 GHz the simulator extrapolates the data with a straight line.  The Pink trace is what I got using the s1p file (which goes up to 12 GHz, but has very few frequencies above 6 GHz).  It clearly does not match the user guide anywhere but is close at low frequencies.  It looks like if the s2p file had entries above 6 GHz that I would have something useful to work with.

Thread Notes

  • Hi,

    We will get back to you.

    Regards

    Rahul 

  • Hi, 

    I have spent some time to recreate these simulations you have shown and I am getting the same results so you are not doing anything incorrect here. There does seem to be a bit of a mismatch between the S1P and S2P files even when taking the lminate into account so I will investigae that further. In the meantime it might be useful to know that when designing our matching networks, on the evaluation board, I used the S1P files with the laminate and filled in the matching network from there. 

    On the LO, thats a fair point that the simulated S2P should have a bigger range. Again i used the S1P file for this too. I will review the User Guide and queue nessecary changes on this section to the next update when i fully understand the difference. In the meantime if you are to go ahead and use the S1P files you will be able to get a relatively coherent match. Is it matching networks you are looking to design with these files?

    Regards

  • Hi Ruairi,

    Yes, I am trying to design matching networks, but we are trying to make compromises to try and have identical channels that cover the full RX and TX frequency ranges without band specific matching networks while limiting the size, including the use of alternative baluns.  When you say that you used the s1p files for your design, was that to match to the A inputs, the B inputs or both?  I guess that I could look at the eval board schematic to check for myself.

    For the LO I have to figure out the best way to accomplish good performance. The User Guide places some  pretty strict requirements on amplitude and phase balance of the balun that I am not at all sure are met by what is on the eval board, and of course the port impedance varies wildly.  Has the performance of the eval board been characterized with a wide range of frequencies at the LO input?  Do you know what performance degradation occurs when the amplitude and phase balance is not met?

    Since you can use just about any even number divider to set the operating LO from the external input, I am thinking about a fairly high frequency for the input.  Designing for a relatively narrow frequency range at high frequency may be more practical than covering a wide range at lower frequencies.  However, requiring the user to supply an LO near 12 GHz to operate around 2 GHz (for example) may not be palatable.