I work in a metrology lab and I am having the same issues regarding verification/calibration of sliding loads. I’m relatively new to metrology and currently I am trying to master the art of calibrating VNA mechanical calibration kits. I’d like to come up with a method of establishing some Pass/Fail criteria that would be somewhat simpler than the process suggested in the 85052B O&M manual requiring calibrating a system, performing a system verification, and then getting a printout of the residual errors after a calibration has been performed; or, using an airline, precision termination, and time domain gating to measure directivity directly.
I encounter a variety of sliding loads, predominantly HP 911C/D/E and Maury 8035’s in addition to sliding loads from the same manufacturers for N-Type, APC-7, and 2.4mm. I’ll focus on 3.5mm sliding loads as I suspect that whatever solution reveals itself (assuming there is a solution) will be applicable to the other connector style sliding loads. I’ve also noticed that Agilent (HP) used Maury 8035 series sliding loads in the early 85052A calibration kits. I’ve seen one 85052A kit with 911C loads. Most 85052B kits appear to use 911D/E sliding loads, although I have seen a few with Maury 8035 series sliding loads as well. Maury uses 8037 series sliding loads in their kits.
I’ve read Agilent AN 117-1 (great primer for folks like me who are just getting into this) as well as the Sliding Load, Sliding load check, Sliding load calibration, and Sliding Load Verification threads here. The threads point out that the specs for sliding loads are different for each manufacturer, in particular Maury.
911C (Reflection Coefficient)
Moveable Load Element: < 0.1, 2GHz to 10GHz
< 0.035, 10GHz to 26.5GHz
Connector and Transmission Line: < 0.02, 2GHz to 26.5GHz (Male)
< 0.02 + 0.001*F(GHz), 2GHz to 26.5 GHz (Female)
911D/E (Parameter not stated)
Load Stability: <0.004, 3GHz to 26.5GHz (Not sure what this means. If this is reflection coefficient, then return loss would be >48dB, which conflicts with the 85052B kit spec below)
85052A Kit (1250-1891, Return Loss)
Sliding Load: >42dB (Airline portion only) (Similar to Maury specification format)
85052B Kit (911D/E, Return Loss)
Sliding Load: >44dB (residual return loss after calibration)
Maury 8035 series:
Terminating Element (VSWR): 1.090 maximum, 2 — 4 GHz (<1.06 typical)
1.050 maximum, 4 — 34 GHz (<1.03 typical)
Connector (VSWR): <1.02 + 0.002ƒ GHz female
1.01 + 0.001ƒ GHz female (I suspect a typo in here somewhere…)
Airline Accuracy (Return Loss): >44dB
Maury 8037 series:
Terminating Element (VSWR): 1.090, 2 to 4 GHz
1.05, 4 to 34 GHz
Airline Accuracy (Return Loss): 50 dB min return loss (Equivalent return loss of air line impedance)
The formats for the sliding load specifications are different not only between manufacturers, but also within manufacturer models.
Three responses from ‘daveb’ and ‘Dr_joel’ appear to offer a solution:
daveb Re: Sliding Load Posted: Aug 29, 2007 9:36 AM
“…After calibration when measuring the sliding load, you will see the reflection coefficient of the sliding load element. As described previously, the variation in the magnitude of the sliding load element provides an indication of the quality of the calibration.”
Dr_joel Sliding Load Posted: Aug 30, 2007 1:22 PM
Take the difference in linear magnitude (worst case is 0.002), and then take 20*log10(.002)=-54 dB. So, this says that a combination of your sliding load error and your directivity is worst case -54 dB (we can't separate the two). Good cal.
Dr_joel Sliding load calibration Posted: Jan 23, 2009 9:22 AM
“To really measure a sliding load, you would need to meaure the relfection at each slide point, then compute the differences (worst case) to come up with the min-max vswr of the load. The difference is the quality of the load, since it should provide exactly the same gamma at each slide point. The only reason it won't is if the transmission line (airline) is not 50 ohms. The actual value of the return loss does not substantially contribute to the error of a sliding load.”
I calibrated our 8510C using a TRL kit (85052C) and took data for a 911C set of sliding loads, two sets of 911D/E sliding loads, and four sets of 8035 series sliding loads. If I apply ‘Dr_joel’s expression: 20*log10(reflection coefficient) to the reflection coefficient max-min difference across the six slide positions for each measurement frequency, to what do I compare the result to establish proper performance? It appears that I can use the 85052B kit 44dB spec as the minimum performance for 911D/E sliding loads (Only one set of 911D/E’s exceed 44dB across 3GHz to 26.5GHz), but I’m not sure what to use for the other devices. Am I on the right track?
If using the 44dB specification from the 85052B kit is not the correct quantity for evaluating the performance of the 911D/E sliding loads, what should I use? More generally, what would constitute a ‘good cal’ threshold when evaluating any of the sliding loads I listed?
I think you are on the right track, but in making the load comparison, the worst case variation represents the sum of the load error and the directivity error. If the same load is used for calibration as measurement, you can infer that the load is 6 dB better than the measured differences.
I wasn't sure exactly what you were looking for but I attached a spreadsheet with an example of how to electrically verify a sliding load's effective return loss. This calculation is for one test point. The six measurements are from the VNA's Smith Chart format and were taken at the six positions of the 911E sliding load. I hope it helps.
"...in making the load comparison, the worst case variation represents the sum of the load error and the directivity error."
Yes, I got that from your earlier posts on another thread (quoted in my original post). Also, I believe that you indicated in another thread that the actual value - magnitude - of the load reflection was not as important as the accuracy of the circle described on a Smith by the reflections resulting from the five (or six) slide positiions. I assume the accuracy of the circle is a measure of load stability?
"If the same load is used for calibration as measurement, you can infer that the load is 6 dB better than the measured differences."
To expand a bit on my original post. I'm attempting to do calibration of sliding load kits (85050B, 85052B, 85054B, 85056B and the Maury equivalents). My goal is to discover some means to verify proper operation/function of the kit-under-test sliding load without having to perform two calibrations of the network analyzer (saving calibration time for the kit-under-test).
Presently, to verify a sliding load, I have to calibrate the analyzer (I use a TRL kit for this step) and then meaure the kit-under-test shorts and opens to determine the offset phase performance of these components are in spec (this process will be the topic of another thread eventually...), then re-calibrate the analyzer using the kit-under-test and, finally, check the Directivity performance of the analyzer against the analyzer specification to determine if the kit-under-test sliding load is performing properly.
Your process of taking the max-min reflection coefficient across the six slide positions to calculate the return loss at each measurement frequency appears to provide a "figure of merit" that could be used to evaluate the performance of the sliding load. In one of your posts on another thread, some sliding load data were given and the result using your process was 54dB return loss (combined directivity and load error) to which you appended "good cal". That is what got me thinking about using your process to arrive at a figure of merit for comparison to some threshhold pass/fail value. So, I guess the questions at this point are: "What threshhold did you use to make the determination 'good cal'?" and "What value below 54dB would have elicited the response 'bad cal'?"
I posted an expanded description of what I am attempting to do in my response to Dr_Joel's reply. That should flesh out the general idea.
I'm going to have to plead ignorance regarding th spreadsheet. What quantities are being used? (Looks like Xi is return Loss, but for Yi I'm clueless). What math operation are you performing on the data? What do you compare the result to to determine 'pass/fail'?
What I described early would be load stability (using the same sliding load to measure itself. The key aspect of a sliding load is that the center of the circle that is the locus of the slide points is the impedanc of the load.
You can do a TRL calibration and the find the circles and then compute the center and from that compute the impedance of the line portion of the load and that will represent the effective directivity.
The stability usually doesn't have a spec but it must be better than the impedance spec of the load or you cannot get to the impedance spec.
The load's manufacturer should give a spec on the impedanc of the airline, or the residual directivity of the load.
What I described early would be load stability (using the same sliding load to measure itself.
So, that would be calibrating the system with the cal-kit-under-test and then measuring the cal-kit-under-test sliding load return loss at all six positions and comparing the results, expressing the difference in dB? The 911D/E sliding load O&M (00911-90019) specs load stability including connector and airline. When part of the 85052B kit, the sliding load is spec'd as residual directivity of the system. None of the other loads I've come across spec load stability.
The key aspect of a sliding load is that the center of the circle that is the locus of the slide points is the impedanc of the load. You can do a TRL calibration and the find the circles
and then compute the center and from that compute the impedance of the line portion of the load and that will represent the effective directivity.
Oh, looks like I need to do some research to find out how to do this (my vector math is sketchy due to lack of use). Is there an App Note that addresses this?
I've had some time to research this and it looks as though your spreadsheet calculates the circumcenter of the 6-point polygons formed by the points on the Smith Chart representing the six slide positions for each frequency. Is this correct? If so, I would be interested in knowing where you found the equations (or, perhaps, you developed them yourself) so I can get a better idea of exactly what is going on. I'd like to be able to understand exactly what each piece of the overall algorithm is doing.
The spreadsheet appears to work quite well and seems to give the results I am looking for. Thanks.
Thanks Dr_joel. Earlier in this thread 'metatron' supplied a spreadsheet that appears to do the 'circumcenter' calculations for the Smith Chart circles at each frequency. It yields an impedance value at the circumcenter of each circle. Given the spreadsheet calculations and your description of deriving the Effective Directivity, I assume then that all I need to do is calculate the Effective Directivity using the expression: (Calculated Impedance-50)/(Calculated Impedance+50) and convert to dB?