AnsweredAssumed Answered

Data-Based Cal Standards

Question asked by jeff_philips on Jul 31, 2008
Latest reply on Aug 7, 2008 by daveb
Dear Agilent,

I’m using this forum because Stoyan Ganchev will be out tomorrow.  He may have already submitted this issue to the factory.  The case number is 1389162853.

My N5230A network analyzer’s ability to create a custom cal kit with data-based standards is potentially a very useful feature.  It allows me to move the measurement reference planes inside the coaxial connectors down on the microstrip PCB right next to the components I wish to measure.  Thus, de-embedding is eliminated.  Since only one set of standards is needed, it is simpler than a TRL cal.  I’m trying to cover 10 MHz to 10 GHz, which requires multiple sets of TRL standards.

I have built a homemade calibration kit consisting of open, short, and load standards on Rogers 4003.  The reference planes of each standard are located the same distance away from an SMA connector.  50-ohm microstrip lines connect the impedance (open, short, load) at the reference plane to the SMA connector.

I believe I characterized the capacitance at the end of my microstrip open standard and the inductance at the end of my short standard well.  The impedance at the end of my load standard is not a good 50 ohms; in fact, its S11 is only -14 dB at 10 GHz. But I thought the idea behind a data-based load standard is that its impedance need not be a perfect 50 ohms; its impedance merely must be accurately characterized over frequency.  I believe I have accurately characterized my load standard.  I have embedded all this data in a custom cal kit stored on my network analyzer; namely, I entered the capacitance of the open and the inductance of the short directly, and I stored the refection coefficient versus frequency of my load in a data file referenced by the cal kit.

Finally, I performed a 1-port calibration using my homemade cal standards and the custom cal kit.  To verify consistency between the standards, the cal kit, and the network analyzer, I then measured the open, short, and load standards.  I obtained good agreement for the open and short standards.  For example, my cal kit claims that my short standard has 80 pH of inductance.  I measured 77 pH.  Unfortunately, the measured load impedance did not agree well with the impedance implied by the reflection coefficients in the data file referenced by the cal kit.  As I said my load has an S11 (referenced to 50 ohms) of only -14 dB at 10 GHz; the network analyzer reported it at less than -50 dB over the entire 10 MHz to 10 GHz.  Where have I gone wrong?  It appears that the S11 reported by the network analyzer is referenced to the impedances suggested by the data file instead of 50 ohms.

It might help if I describe how I characterized my open, short, and load calibration standards.  The goal is to find the impedance at the reference plane seen looking away from the SMA connector.  To do this, I calibrated my network analyzer at the connectors with an accurate Agilent cal kit.  Next I measured S11 of the open, short, and load standards.  I then modeled these standards inside Genesys.  I modeled the coaxial connector as a length of transmission line.  The length and impedance of the microstrip line between the connector and cal standard is known.  For my open standard, I added a small series capacitance between the end of the microstrip line and ground.  For my short standard, I added a small series inductance between the end of the microstrip line and ground.  I adjusted the length of the transmission line representing the coaxial connector and the capacitance until the measured and modeled data for the open standard agreed reasonably well (to within a few degrees over 10 MHz to 10 GHz).  Retaining the length of the transmission line representing the coaxial connector, I then adjusted the inductance until measured and modeled data for the short standard agreed reasonably well.  It turned out that my load standard resembles 50 ohms in parallel with a small capacitance.  Again I retained the length of the transmission line representing the coaxial connector and adjusted the parallel capacitance until measured and modeled data agreed reasonably well.  Knowing the R-C impedance at the end of the load standard, I calculated the reflection coefficient over frequency and stored it in a data file later referenced by my custom cal kit.

Thank you, Jeff Philips  

Attachments

Outcomes