RF Calibration & WinCal FAQ

General

What are the VNA calibration coefficients for my impedance standard substrate? (PN XXX-XXX)
A simple answer is in the table below, which can also be found on the bottom of the cover of the blue probe box. A more advanced answer follows. This question may indicate a basic misunderstanding of on-wafer calibration. Let's try to correct it.

In coaxial calibrations the standards are unique devices and since they have a unique connector can have only a single electrical characteristic. In on-wafer probing the electrical behavior of the standards are dependent upon the probe and how it is placed.

The shape and configuration of the standard will also be important. Cascade Microtech makes a variety of impedance standard substrates (ISS's). There are versions for probes with standard or wide pitch with choices for ground-signal or ground-signal-ground tip configurations. In addition there are general purpose, specialty, and custom configurations supporting a wide variety of applications.

What all of this boils down to is that the calibration coefficients are supplied with the probe, not with the impedance standard substrate. For air coplanar (ACP) probes, refer to the lid of the box for a table. When used with the designated ISS the calibration coefficients for a specific probe are found by selecting the tip configuration and pitch entry in the table. For 250 um pitch probes either the standard pitch or the wide-pitch ISS may be used. The calibration coefficients that are normally supplied correspond to the standard pitch ISS. The values will be somewhat different when using the wide pitch ISS.

The only calibration term not supplied with the probe is the Thru standard delay. For standard pitch probes with proper alignment of probe separation using the alignment marks provided on the ISS, the Thru delay will be 1 ps. For wide-pitch ISS's the Thru delay is 4 ps.

C-Open (fF)	GSG L-short (pH)	L-Term (pH)	GSG ISS P/N	Probe Pitch	GS/SG ISS P/N	C-Open (fF)	GS/SG L-short (pH)	L-Term (pH)
-9.3	2.4	-3.5	101-190	100	103-726	-11.0	33.5	36.5
-9.5	3.6	-2.6	101-190	125	103-726	-11.0	41.7	47.2
-9.7	4.8	-1.7	101-190	150	103-726	-11.0	49.8	57.8
-10.1	7.2	0.2	101-190	200	103-726	-11.0	66.2	79.2
-10.5	9.6	2.1	101-190	250	103-726	-11.0	82.5	100.5
-15.7	11.0	-25.0	106-682	250	106-683	-7.0	27.0	0.0
-13.6	15.8	-21.0	106-682	350	106-683	-7.0	28.2	0.0
-12.6	18.2	-19.0	106-682	400	106-683	-7.0	28.8	0.0
-10.5	23.0	-15.0	106-682	500	106-683	-7.0	30.0	0.0
-9.6	28.1	-3.3	106-682	650	106-683	-6.4	42.9	14.1
-9	31.6	4.4	106-682	750	106-683	-6.0	51.6	23.4
-7.5	40.4	23.6	106-682	1000	106-683	-5.0	73.4	46.6
-6	49.1	42.9	106-682	1250	106-683	-4.0	95.1	69.9

Air-Coplanar Probe (ACP) calibration coefficients table. Coefficients for VNA calibration depend on the style and pitch of the probe, as well as the ISS used. The wide-pitch ISS's (106-682 and 106-683) use a 4 ps long Thru line while the standard-pitch ISS's (101-190 and 103-726) have a 1 ps Thru line.

What is this Cal data? Where do I find it? What do I do with it?
Your Cal information consists of three numbers for each probe and a number for the Thru standard. Each Probe: #1 Co Open circuit capacitance, #2 Ls Short circuit inductance, #3 Lterm load inductance. Thru: #1 Delay In most cases the load standards are 50 ohms. Where the 500 ohms that you see in the calkit comes from is in the implementation of the load inductance. The HP 8510 does not directly support entry of load inductance. Instead we use the offset load capability. By choosing a very high impedance (500 ohms) for the offset and the corresponding delay (L/500) we get behavior equivalent to the inductance that we want. This method can also be used for the short circuit inductance.

In all cases the three terms for the probes are supplied with the probe. There is a table of values in the lid of the probe box. Select your configuration and pitch and read the corresponding values. Enter them as described above. The Thru delay is not all that well documented. Use 1 ps for probes with 250 um pitch or less. Larger pitch probes use Thru delay of 4 ps.

What are the electrical lengths of the long transmission lines on my impedance standard substrate?
If TRL calibration is used the delay of the long lines will need to be supplied to WinCal or directly to the VNA. These delays are readily determined from the drawing supplied with the ISS. Simply measure the line using a scale provided on the ISS for this purpose. Remember to allow for the probe contact overlap - about 50 um on each end. Each 130 um (0.130 mm) of length corresponds to approximately 1 ps, so delay may be calculated from T (ps) = Length (mm)/0.13 (best fit).

The 101-190B TRL line lengths as measured after LRRM calibration:

150 um 1 ps
450 um 3.2 ps
900 um 6.6 ps
1800 um 13.5 ps
3500 um 26.6 ps
5250 um 40.4 ps

The 106-682 wide pitch GSG ISS has 4 ps Thru's and lines of 25, 50, and 50 ps. The straight CPW's on the 106-686 GP membrane ISS are 3.8, 5.8, 11.5, 17.1, 22.8, 28.3, 34.2, 39.7, 45.3, 52.7, 60.4, and 67.7 ps.

For more info, refer to "Microwave Wafer Probe Calibration Constants Instruction Manual," part of 101-338 kit
Date: 4-16-98

What is the electrical length for all THRU lines, especially for right angle THRU lines, on the general purpose ISS. (P/N 005-016)? I am going to use these THRU lines for the purpose of SOLR calibration using WinCal.
Here are the delays for the lines in the upper right corner of the 005-016 ISS. They are in order from left to right, then top to bottom. "st" indicates a straight line, "ra" indicates a right angle line.

st 19.3 ps st 8 ps
st 19.3 ps st 2.7 ps st 4.7 ps
st 17 ps st 10.8 ps
st 24.2 ps ra ~5 ps
st 24.2 ps ra ~15 ps
st 41.1 ps

Treat these numbers as estimates - not as precise. The ra numbers are even less precise. The SOLR calibration requires only an estimate of the delay of the line which is used to determine the particular sign of a square root.

After completing the SOLR calibration, look at a polar plot of S21 for the R (reciprocal Thru) standard that was used in the cal. The phase response should be roughly continuous with frequency (for closely spaced measurement frequencies). If there is a step change in phase then the estimate of the Thru delay entry may need to be adjusted and the calibration recomputed.

Once a proper SOLR solution is completed the measurement of the delay of the Thru with the calibration turned on will be accurate and can be used for further work. Note that the best measurement of delay using a VNA is obtained by displaying the transmission phase and adjusting the "electrical delay" (in the 8510 'Response' menu) for constant zero phase over the measurement bandwidth. The 'delay' display format uses a calculation of the difference in phase between successive frequency points and can be very noisy for closely spaced frequencies.

Why do I get a syntax error from the 8510 when WinCal attempts to save a calibration set?
This sometimes occurs when using larger numbers of frequency points. The 8510 does not have memory to support all 8 cal sets filled with maximum size arrays. This error can occur even with several cal sets deleted. Delete a few more to free enough memory.

How do I enter L (inductance) values for parasitics of my standards into my network analyzer cal kit definitions? All I see is the offset delay and offset impedance.
Parasitic inductance of the short and load are required for SOLT calibration, but some network analyzers do not allow one to enter these values directly. Then, the inductance can be modeled as a short piece of transmission line. Assume a maximum impedance that the network analyzer allows (typically 500 Ohms), which will give you offset impedance. Offset delay can then be calculated using equation: td = L / Z_offset. Any other information can be typically left as default. For capacitance of an open, only C_o (0th order model) is used in Cascade Calibration Kit definitions.
For more info, see "Microwave Wafer Probe Calibration Constants Instruction Manual," PN101-338

Why is there a 180 degree step in phase in my S21 phase response when I measure my reciprocal Thru standard after an SOLR calibration?
This kind of step behavior in the transmission phase indicates that the SOLR algorithm has not found the proper root choice in solving for the standard. Adjust your guess for the Thru delay estimate to eliminate the phase step behavior.

Wincal XE

Where can I find FAQ’s for WinCal 2006 questions?
WinCal 2006 is replaced by WinCal XE and the upgrade to XE is free; we encourage you to upgrade to this latest version. Go here for instructions how to get the upgrade.

I’m trying to read 3 and 4-port data using the network driver for my PNA. Why is it not working?
Unfortunately WinCal XE, SP1 has a bug in the network driver for the PNA, which makes it only work for 1 and 2-port data. Note that multiport calibrations only read 1 and 2-port data (but from different port-pairs) so calibrations are not affected The problem applies to measuring more than 2 ports at once in the RF Data Viewer. This is fixed in WinCal XE, SP2.

Why do I get red error events in WinCal when I calibrate the PNA for fewer ports than are displayed on the PNA?
WinCal XE has this limitation, which will be fixed in a future release. The error set calculated must be applicable to the displayed data on the PNA. If, for example, you have a quad display with all four 2-port S-parameters on the PNA and perform a 1-port SOL, you will get an error message in WinCal (red event) when the PNA tries to correct S21 (cannot be done with SOL). WinCal will consider the calibration a failure, but S11 on the PNA is indeed calibrated.

How can I calculate and view derived values from my measured S-parameters?
The “WinCal XE Data Post Processing Tutorial” (new for XE, SP1) provides instruction on using Wincal XE post processing capability. Make sure the main form shows at full size (rightmost toolbar button). Select the Tutorial tab, then press “Other” and the selection of this new tutorial will be available.

Are there any example Reports?
Yes, there are several examples. In the main form, use the menu “Help-Copy Examples to My Documents”. This will copy all example files from the distribution to the “User Home Folder”, typically “My Documents\WinCal XE\Examples” belonging to the currently logged in user. The “User Home Folder” can be changed in Options-Folders tab. Once the examples are there you can open, copy and modify them to your own needs. If you modify them, we recommend that you copy them to a new subfolder to your “User Home Folder” first, so that a new “Help-Copy Examples to My Documents” won’t overwrite your customized version.

How does WinCal XE assist me in validating my calibration?
WinCal XE comes with powerful validation tools. For an overview of these tools go to the main form and use the menu “Help-Documentation Folders”. It will open the Windows Explorer to this folder. Select the subfolder “Application Toolkits” and open the file “Calibration Validation and System Comparison Tools.pdf”

What tools does WinCal XE have to assist me in device characterization?
WinCal XE comes with many tools to assist you with your device characterization tasks. In the main form, use the menu “Help-Documentation Folders”. It will open the Windows Explorer to this folder. Select the subfolder “Application Toolkits” and open the file “Device Characterization Tools.pdf”which describes many of the tools.

Can I create my own Wizards?
Yes. In the main form, use the menu “Wizards-Wizard Script Editor”. A better way may be to start with an existing Wizard and modify it. Select the Wizard tab in the main form. Press “More Wizards” and right click on the Wizard you want to base your custom Wizard on. Right click and select “Create Copy for User” or “Create Copy for Group” depending on if you want to share it or not. The original Wizard gets copied there and you can then right click on this new copy and select “Edit”. After editing you can save it with a new name, but it will not replace the original Wizard even if not renamed. They reside in different folders.

Why do I get a 'Calibration no longer acceptable' result when I monitor my calibration immediately after calibration was completed?
This is an indication of poor repeatability in your measurement system. One way this can occur is when your calkit was radically misdefined or perhaps you measured a short circuit when you should have measured an open. This can create a high sensitivity to measurement system variability.

More likely is that something is wrong in the measurement system. Using your VNA with calibration turned off, compare the magnitude of measured reflection coefficient for a load and a short on each port. Expect more than 10 dB of change in order to get good calibration results.

This experiment can be repeated with coaxial standards at the ends of the cable, and at the VNA front panel to further isolate the problem. If the problem occurs when using the VNA alone there may be a problem with the VNA. Check your VNA setup. Low source power, high port attenuation, or ramp sweep will reduce VNA repeatability. Using inadequate averaging can also reduce dynamic range.

Check all cables and connections. Properly torque all connectors. Make sure that you are using good quality phase stable cables. With a poor cable any strain or inadvertent bump from the user will change the electrical behavior enough to spoil a calibration. Semi-rigid cables may be used but avoid any stresses at the connectors and properly secure the cable to minimize vibration or other motion.

If your system is performing at its best the thresholds for 'good' and 'acceptable' calibrations can be changed to provide guidance for nominal system performance.

For more info, refer to "Calibration and Accuracy Factors Summit High-Frequency Probe Station Reference Manual," PN103-475-A
Date: 4-16-98

Why does the HP8510 Cal Disk (P/N 101-338) specify a -0.5 ps delay for the A. LRM match standard? How do I specify LRM entries when using the HP8510 and WinCal?
This is an interesting pair of questions. The answers are different for direct 8510 entry and for WinCal.

To implement LRM directly on an 8510 we must use the TRL algorithm. This algorithm computes the line versus Thru transmission per unit delay as a function of frequency. In the LRM case this transmission value doesn't really come into play and it is only important that we make sure that the algorithm reaches a solution. We are using a Match termination in place of the second line so our transmission is essentially zero (limited by crosstalk). This is OK since the high attenuation does not impact the TRL solution.

However if we have entered a line (Match) delay of 0 ps we end up with a divide-by-zero problem that causes the algorithm to fail. Since it is not critical, any non-zero number will work for Load delay entry. For example, -0.5 ps, 1 ps, and 5 ps would all yield the same calibration result.

The 101-338 calkit entry for LRM reflect standards (open and short) is technically not exactly correct, but will work just fine in practice. The reference plane is the center of the Thru line so really the reflect standards should be defined with a -0.5 ps offset. In practice this is not necessary since this information is only used to select a sign of a square-root and the phase shift of such a short delay is not enough to cause an improper root selection. For very long Thru's a problem may occur and it will be necessary to properly enter the delays as one-half of the Thru line length.

In all cases the reference plane may be shifted from the center of the Thru line to the probe tips by using the port extensions at each port to specify a delay equal to one-half of the Thru line standard's delay. Since the loss of the Thru line is not compensated by port extensions the loss of longer lines may introduce noticeable error (1 ps lines are very low-loss).

When using WinCal one can use this same procedure to perform an LRM calibration using WinCal's TRL algorithm. In this case all of the above comments apply. This would be unusual. Instead, it is simpler to use the LRM/LRRM algorithms in WinCal. In these cases the actual numbers are entered as described in the help file. If a zero Thru delay is desired it is necessary to enter a tiny number (e.g., 0.01 ps) to work around a minor bug (versions up through 2.23). Also uncheck the "Use computed propagation constant to move reference plane" option.

How come I see a 180 degree step in phase in my S21 phase response when I measure my reciprocal Thru standard after an SOLR calibration?
This kind of step behavior in the transmission phase indicates that the SOLR algorithm has not found the proper root choice in solving for the standard. Adjust your guess for the Thru delay estimate to eliminate the phase step behavior.

How do I use the alignment marks on my calibration standard substrate (ISS)?
The alignment mark was originally designed for use with the WPH probe. The intent was the edge of the ceramic started at the flat left edge when the probe first touches down, and then skates to the inside angle middle point. The correct procedure for ACPs is to touch the ACP down with the end of the tip located half-way between the left edge and the inside angle middle, and then skate until the tip is half-way between the inside angle middle and the right-most points. (Of course the actual tips end up being on either side of the alignment marks).

This positioning also helps to eliminate the problem where Tungsten probes land just off an ISS load (for example) and then scrape off the metal while skating. The positioning that is described should insure that the tips initially land on the metal of the ISS standard.

The ultimate check is to use this alignment and place the probes on one of the standards (load or short). In WinCal with a Cascade semiautomatic probe station, click on one of the calibration buttons (SOLT/LRRM), record the alignment position and then move to the Load or Short (right click on standard in cal form and select move). The chuck will drop and move to the position of the selected standard, while the probes alignment will not change. Make sure that as the probe lands on the standard, it touches down and starts skating at the very beginning of the standard, so as you can see just hair behind the probe as it starts skating, and skates all the way to the end of the standard, leaving just a hair in front so that you can see that the probe did not skate off the standard. Once on the standard, you can again verify this by using the platen lever.

I cannot select any VNA except the Virtual one. What should I look for?
For a Demo installation, where no key is needed, it is actually normal operation. This is a symptom of the hardware key (USB or parallel) either being the wrong one or not attached at all. It could also be a system problem, perhaps with the install of the driver for the key.

Why do I get erratic behavior when accessing a VNA over GPIB on a probe station?
The first thing to check is the obvious physical connections. GPIB contacts can slip out if not screwed in properly. Another common cause is that Nucleus is configured to be talker/listener and is conflicting with WinCal's need to be Controller. Either use a GPIB switch box (and configure WinCal to use it, see below), or configure Nucleus to not be controlled over GPIB at all (Nucleus Hardware Setup - see Nucleus manual). You can try to select the Switchbox in WinCal even if you don't have one. It should still stop Nucleus from using GPIB when WinCal needs it.

WinCal XE / Nucleus COM Setup Alert Box

I'm using Windows Vista, but after a Suspend or Hibernate GPIB doesn't work. Are my settings wrong?
Unfortunately this is a limitation of the National Instrument GPIB driver for Vista. You will see the low level GPIB driver error EPWR. Avoid suspend and hibernate modes while WinCal (or any other GPIB based program) is running. If it happens, restart WinCal to recover. We have a question posted to National Instruments on this issue. On Windows XP, the GPIB driver blocks the system from going into those modes.

My computer is slowing down over time. What could be wrong?
A memory leak was accidentally introduced in SP1. Closing a Report did not return allocated system memory and this would render the system sluggish after opening many large Reports. This is fixed in WinCal XE, SP2.

I save Touchstone 3-port (S3P) files with WinCal, but cannot read them with my other measurement or modeling program. Why?
There was a file syntax flaw regarding the line breaks for a 3-port data set. This is fixed in WinCal XE, SP2.

Can I turn off the prompting for every measurement when I use a manual prober?
In WinCal XE, SP1 we added support for some probe-ISS calibration combinations that required lateral moves (dual probes on a 101-190 ISS for example). Therefore SP1 forced the user to be prompted for all measurements. This requirement has been removed in WinCal XE, SP2. Prompting for all measurements is now an option, and off by default. If you really need to stop and make lateral probe moves between port measurements it has to be checked.

Can I change the limit lines for the WinCal 3 style Validation and Monitor graph?
Yes, that was introduced in WinCal XE, SP2.

I use the ISS 129-239B and WinCal selects User Defined instead of the Loopback Thru. Why?
Some table entries for ISS 129-239B were incorrect (affects B rev only). This is corrected in WinCal XE, SP2.

My 8510 XF VNA has errors when calibrating or reading data. What could that be?
The 8510XF needs a longer timeout setting than the regular 8510C. They share the same WinCal XE driver and the defaults are for the 8510C. It may not be obvious that GPIB timeout is the problem, but just increase GPIB timeouts in WinCal XE System Setup and see if the problem goes away. The error messages in WinCal will be better in a future version. Use these or larger values for an XF model: Cmd Timeout (sec) = 60 and Trace Timeout (sec) = 200.

Wincal 3.x

I’m trying to calibrate my PNA, but get error messages. What could it be?
Wincal 3.x ONLY: Agilent changed the firmware rules with version A06.01. Therefore WinCal 3.x requires a compatibility mode setting on the PNA for proper operation. To set this mode, a registry change must be made on the PNA. On the PNA, go to C:\Program Files\Agilent\Network Analyzer\System and double-click on WinCal32.reg, then click on Yes.
This same information can be confirmed on this Agilent web page: http://na.tm.agilent.com/pna/firmware/rev6summary.html Scroll toward the bottom, to find the instructions under the header “Note for Cascade WinCal 3.2 Users”. Please direct any questions on this patch to Agilent support organization.

I got WinCal 3.2.2 and cannot send the calkit to a PNA. What is wrong?
Agilent changed the firmware rules with version A06.01. To still be able to send a calkit to the VNA you will have to copy a new drvna32.dll file and replace the one installed by WinCal 3.2.2. This only applies to 3.2.2 and should not be done for earlier versions.
Download the file drvna32.dll

Important Note: The shipped WinCal Demo is not affected and should not be changed. It can still send a calkit to the VNA.

How do I configure the TRL calibration for a reference plane at the center of the Thru but using the probes in air as the reflect standard?
In WinCal, deselect checkbox 'Use computed propagation constant to move reference plane' and enter minus one-half of the Thru delay as the reflect offset. For example, for a Thru delay of 10 ps enter -5 ps for the reflect offset.

My ACP probe pitch is in between the ones listed in the table. How do I determine my VNA calibration coefficients?
Interpolation using the closest values will provide good coefficients.

The ultimate check is to use this alignment and place the probes on one of the standards (load or short). In WinCal with a Cascade semiautomatic probe station, click on one of the calibration buttons (SOLT/LRRM), record the alignment position and then click on "Move to Load (Short)." The chuck will drop and move to the position of the selected standard, while the probes alignment will not change. Make sure that as the probe lands on the standard, it touches down and starts skating at the very beginning of the standard, so as you can see just hair behind the probe as it starts skating, and skates all the way to the end of the standard, leaving just a hair in front so that you can see that the probe did not skate off the standard. Once on the standard, you can again verify this by using the platen lever.

Why do I get a 'Calibration no longer acceptable' result when I check stability immediately after calibration?
This is an indication of poor repeatability in your measurement system. One way this can occur is when your calkit was radically misdefined or perhaps you measured a short circuit when you should have measured an open. This can create a high sensitivity to measurement system variability.

If your system is performing at its best the thresholds for 'good' and 'acceptable' calibrations can be changed to provide guidance for nominal system performance.

For more info, refer to "Calibration and Accuracy Factors Summit High-Frequency Probe Station Reference Manual," PN103-475-A
Date: 4-16-98

How come the calibration verification plot for my SOLT or SOLR calibration always shows a flat 0 dB magnitude?
The calibration verification plot is a display of the magnitude of the measurement of the calibrated open. In SOLT and SOLR calibrations this measurement is forced to be exactly what the calibration coefficient for the open specified. Since the open standard is modeled as a small capacitance the load appears purely reflective and shows 0 dB return loss. The verification plot is independent of the "Read stability open from VNA rather than calculate" check box state.

Related Web Pages
Upgrade to WinCal XE from WinCal 2006

Related Files
WinCal XE, SP3 - Readme
WinCal XE, SP3 - Patch