Contact an Expert

Thread: N5247A 4-port on-wafer multi-line TRL calibration routines


Permlink Replies: 16 - Pages: 2 [ 1 2 | Next ] - Last Post: Jul 17, 2012 12:17 PM Last Post By: bhokkan Threads: [ Previous | Next ]
KevinK

Posts: 9
Registered: 07/05/12
N5247A 4-port on-wafer multi-line TRL calibration routines
Posted: Jul 5, 2012 7:51 AM
Click to report abuse...   Click to reply to this thread Reply
Hello,

I am looking to do a 4-port on-wafer multi-line TRL calibration routine, and I am wondering exactly what the best approach is. I have read through the network analyzer help, but it gives a detail of the 4-port TRL cal using a PNA with an external test set using the valid delta match cal. This PNA has 4-ports with reference and test receivers at each port, so I am not sure if the routine is still applicable.

I have a custom calibration board that has multiple lines and appropriate line lengths for a standard 2-port TRL. I also have an on-wafer 90 deg bend for transmission to orthogonal ports. Would I also need 90 deg bends at all TRL line lengths?

I have in the past used software for a 4-port on wafer SOLT cal, but I would much prefer a front panel multi-line cal with capacitance correction. Does the PNA-X support this routine?

Thanks,
Kevin
Dr_joel


Posts: 2,687
Registered: 12/01/05
Re: N5247A 4-port on-wafer multi-line TRL calibration routines
Posted: Jul 5, 2012 9:22 PM   in response to: KevinK in response to: KevinK
Click to report abuse...   Click to reply to this thread Reply
You might post the model and FW revision to help us.

PNA doesn't support overdetermined "Multi-line TRL" ala the NIST formulation; but if you have multiple lines to define a broad bandwidth, then you can use multi-step TRL that applies normal TRL to each of the freqeuncy segments for which the lines are determined.

If you prober is East-West on one port pair, and North-south on the other, then cleverly selecting port thru methods can allow you to use unknown thru for the 90 degree bends.

Select the thru paths for N/S to have TRL; then EW to have TRL, then NE or SW to have unknown thru. You only need three thru measurments for full 4 port calibration.

If you want to do source and receiver power calibrations, those can be easily added by detacing the prober from port 1 and using a power meter and a 1 port calibration on port 1 (part of the guided power cal) and this power is automatically moved to the probe-tip, and the S-parameter thru portion also automatically computes the loss to each other receiver and each source so you have a full power and S-parameter cal.
KevinK

Posts: 9
Registered: 07/05/12
Re: N5247A 4-port on-wafer multi-line TRL calibration routines
Posted: Jul 6, 2012 7:28 AM   in response to: Dr_joel in response to: Dr_joel
Click to report abuse...   Click to reply to this thread Reply
Hi Dr. Joel,

The model N5247A is firmware A.09.50.13.

Is overdetermined TRL having lines with overlapping bands? I am very familiar with the NIST TRL and less familiar with multi-step TRL.

It sounds as if I need to list the bandwidth of each line and let these bandwidths overlap (they do). Next, the algorithm will select which line is fitting most appropriately to the band when there is an overlap?

EDIT: I am reading that the order in which a line is listed determines its priority for overlapping bands

Also, I am uncertain about if a capacitance (constant C) correction is available to correct for imperfect Z0 conditions of the line.

Thanks,
Kevin

Edited by: KevinK on Jul 6, 2012 8:37 AM

Dr_joel


Posts: 2,687
Registered: 12/01/05
Re: N5247A 4-port on-wafer multi-line TRL calibration routines
Posted: Jul 6, 2012 1:29 PM   in response to: KevinK in response to: KevinK
Click to report abuse...   Click to reply to this thread Reply
Your reading is correct that you must choose the order to use the lines and set the frequencies appropriately. For the least chance of any issue, set the frequencies for each line to not overlape: 1.000-3.000GHz, 3.000-9.000GHz. I think you have to have no regions not covered, so be sure the frequency for the lower and upper line use at least the same number.

If the line's Zo is not 50 ohms, I think you must find the Zo value and enter it for that line. I'm not familiar with in capacitance correction that you mention, but you can define the Z0 of the lines to be different from the system Zo.

KevinK

Posts: 9
Registered: 07/05/12
Re: N5247A 4-port on-wafer multi-line TRL calibration routines
Posted: Jul 6, 2012 3:12 PM   in response to: Dr_joel in response to: Dr_joel
Click to report abuse...   Click to reply to this thread Reply
Dr. Joel,

I will experiment with ensuring that the lines do not overlap.

I have worked through several 2-port multiline TRL calls on the front panel, but I get the same problem each time.

I can't seem to get correct phase data when I measure and one port standard (open short load for example) Please see the image to understand. What could be causing this?

All the lines display the correct phase.

This leads me to believe that there is a problem with the reflect standard? for TRL will I need to specify a delay if the standard itself is half the length of a zero length thru?

Thanks,
Kevin
Dr_joel


Posts: 2,687
Registered: 12/01/05
Re: N5247A 4-port on-wafer multi-line TRL calibration routines
Posted: Jul 7, 2012 10:11 AM   in response to: KevinK in response to: KevinK
Click to report abuse...   Click to reply to this thread Reply
The frequencies look very low for a TRL calibration. For 100 MHz start frequency, you would need a line that is something like 16 cm in air. Are you using very long lines, or very high dielectric constants.

Is your open longer than 1/2 wavelenght at these frequencies (I would doubt that)? There is a selection in the TRL that provides for using either the center of the thru of the reflect as the phase reference; but this type of error often comes from not resolving the sqrt of phsae properly in the TRL math. Maybe if the open and thru are very different in length, then this could cause such a problem; in such a case it is good to specify the approximate length (delay) of each in the calkit; but I'm not sure if that would affect this problem or not.
KevinK

Posts: 9
Registered: 07/05/12
Re: N5247A 4-port on-wafer multi-line TRL calibration routines
Posted: Jul 10, 2012 5:18 AM   in response to: Dr_joel in response to: Dr_joel
Click to report abuse...   Click to reply to this thread Reply
This is the lowest frequency range for my calibration. I used my longest line and limited my upper freq for quickness of calibration. I have made this calibration regularly with other software.

I am using lines of a couple centimeters, with a relative dielectric of less than 3.

The physical length of the reflect is half the physical length of the through. I typically select that the reference plane is set at the center of the through.

There will be a fringing capacitance that adds to the length of the reflect, and can be seen in the phase of the reflect (above). This is the only difference in length (electrical length).

As seen, there seems to be an issue from 10 MHz to 2.2 GHz. Then, I get the results I expect.

You mentioned that this type of error can occur from not properly resolving the sqr root of phase in the TRL math. What can be done to resolve this.

I was thinking possibly a LRM calibration could help pick up the low frequency end if the calibration routine has a problem with it.

Edit: I was thinking "TRM". It appears that it is possible to do a cal with the load for low freq, and then let lines correct higher frequencies.

Thanks,
Kevin

Edited by: KevinK on Jul 10, 2012 5:34 AM
KevinK

Posts: 9
Registered: 07/05/12
Re: N5247A 4-port on-wafer multi-line TRL calibration routines
Posted: Jul 11, 2012 10:26 AM   in response to: Dr_joel in response to: Dr_joel
Click to report abuse...   Click to reply to this thread Reply
I continued to see the same phase error in all the calibrations. I then tried a cal using the reflect to set the reference planes (given C0) and no length for the through.

This appears to be the best cal so far. However, it is obvious where the algorithm stops using the file based load standard and switches to the longest line (see jpg).

Additionally, this sudden shift at 0.5 Ghz appears in all the data, rendering the cal less than useful.

On a good note, all the phase data was as expected.

Can you help with any tips or pointers with rounding out this transition?

My longest line is approx 22 deg at 500 MHz, so I cannot extend that any further back.

The data based load standard shows 27dB of return loss at 500 MHz, with a max of -43 dB at low freq.
Dr_joel


Posts: 2,687
Registered: 12/01/05
Re: N5247A 4-port on-wafer multi-line TRL calibration routines
Posted: Jul 11, 2012 11:05 PM   in response to: KevinK in response to: KevinK
Click to report abuse...   Click to reply to this thread Reply
This is exactly what you would expect if you used a non-data-based load standard. The load response in the TRM region (below 500 MHz) shows just the connector repeatability because the standard load has a response defined as zero. You must check your calkit definition to be sure that a data based load exists in the cal kit and is defined for those frequencies and is used. Be sure that it has a name that does not get confused with a standard load, and you need to position it properly in the cal kit.

But in thinking about this, I'm not sure TRM supports data-based standards for the load. Dave B or Ken Wong might want to chime in on this, or I'll try it out tomorrow.
bhokkan

Posts: 322
Registered: 12/01/05
Re: N5247A 4-port on-wafer multi-line TRL calibration routines
Posted: Jul 12, 2012 4:04 PM   in response to: Dr_joel in response to: Dr_joel
Click to report abuse...   Click to reply to this thread Reply
FYI it looks to me like we (PNA) do support databased standards for the load in TRM cals.
KevinK

Posts: 9
Registered: 07/05/12
Re: N5247A 4-port on-wafer multi-line TRL calibration routines
Posted: Jul 12, 2012 6:31 PM   in response to: bhokkan in response to: bhokkan
Click to report abuse...   Click to reply to this thread Reply
Dr. Joel,

Thanks for the help.

The good news is that I was able to benchmark a relatively good 4port multiline trl by adjusting certain settings.

To fix the problem with the phase error, I defined the reference plane using the reflect (given a C0), instead of center of the thru, as is typical for TRL. The thru definition remains 0ps (in my case as described above).

To fix the issues related with the low frequency response and Z0 correction for imperfect lines, I used the LRL auto-correction along with ensuring that the data was referenced to system Z0.

I spent some time benchmarking the results, and everything looks as expected. I can now move forward with measurements.

Kudos that the front panel TRL algorithm additionally measures an/bn for each of the lines, and not just the thru.

The bad news is that I am still unable to get a good cal with the match as a data based standard. I am unsure why that is. I ensured that the # of points and frequency lists were identical to my cal, and I also renamed the file something other than load. Same result as I posted previously.

Not sure what is going on with that. I am still open to any new ideas for troubleshooting this.
KevinK

Posts: 9
Registered: 07/05/12
Re: N5247A 4-port on-wafer multi-line TRL calibration routines
Posted: Jul 12, 2012 6:38 PM   in response to: bhokkan in response to: bhokkan
Click to report abuse...   Click to reply to this thread Reply
Dr. Joel,

Thanks for the help.

The good news is that I was able to benchmark a relatively good 4port multiline trl by adjusting certain settings.

To fix the problem with the phase error, I defined the reference plane using the reflect (given a C0), instead of center of the thru, as is typical for TRL. The thru definition remains 0ps (in my case as described above).

To fix the issues related with the low frequency response and Z0 correction for imperfect lines, I used the LRL auto-correction along with ensuring that the data was referenced to system Z0.

I spent some time benchmarking the results, and everything looks as expected. I can now move forward with measurements.

Kudos that the front panel TRL algorithm additionally measures an/bn for each of the lines, and not just the thru.

The bad news is that I am still unable to get a good cal with the match as a data based standard. I am unsure why that is. I ensured that the # of points and frequency lists were identical to my cal, and I also renamed the file something other than load. Same result as I posted previously.

Not sure what is going on with that. I am still open to any new ideas for troubleshooting this.
Dr_joel


Posts: 2,687
Registered: 12/01/05
Re: N5247A 4-port on-wafer multi-line TRL calibration routines
Posted: Jul 12, 2012 8:15 PM   in response to: KevinK in response to: KevinK
Click to report abuse...   Click to reply to this thread Reply
Please post the Citi file that you use for the data based cal standard.
KevinK

Posts: 9
Registered: 07/05/12
Re: N5247A 4-port on-wafer multi-line TRL calibration routines
Posted: Jul 14, 2012 3:05 PM   in response to: Dr_joel in response to: Dr_joel
Click to report abuse...   Click to reply to this thread Reply
Attachment R60.dat (25.5 KB)
As requested
Dr_joel


Posts: 2,687
Registered: 12/01/05
Re: N5247A 4-port on-wafer multi-line TRL calibration routines
Posted: Jul 14, 2012 6:30 PM   in response to: KevinK in response to: KevinK
Click to report abuse...   Click to reply to this thread Reply
The form looks OK to me(I was looking for the uncertainty data), but the Agilent example has a lot more header stuff (from Appendix D of http://cp.literature.agilent.com/litweb/pdf/5989-4840EN.pdf ). You might try adding some of that stuff (I was going to say crap, but that would not be professional). Why we can't read a simple S1P file I'll never know; Brad? (also Brad, maybe you can check the file and see if it looks OK).

CITIFILE A.01.01
#PNA Rev A.01.00
#PNA STDTYPE DATABASED
COMMENT MODEL: 85058-60101
COMMENT SERIAL NUMBER: NOMINAL
#PNA STDREV Rev A.01.00
#PNA STDLABEL “SHORT 1 -M-”
#PNA STDDESC “1.85 mm male SHORT 1
#PNA STDFRQMIN 0
#PNA STDFRQMAX 70000000000
#PNA STDNUMPORTS 1
COMMENT “1.85 mm” known so #PNA DEFINECONNECTOR statement non needed
COMMENT #PNA DEFINECONNECTOR “1.85 mm” 0 70000000000 COAX
#PNA CONNECTOR 1 “1.85 mm” MALE
COMMENT PINDEPTH is optional, only applies to coax devices
#PNA PINDEPTH 1 0.007 0.007
NAME DATA
COMMENT This section describes the s parameter data and weighting
COMMENT factor for the calibration standard
COMMENT COVERAGEFACTOR is used to scale the weighting factor
COMMENT Si,j is sij for the standard. Supported formats: RI
COMMENT Ui,j is the weighting factor for sij.
COMMENT Supported Ui,j formats: RI, MAG
#PNA COVERAGEFACTOR 2
COMMENT note number of points is 509 below
VAR Freq MAG 509
DATA S1,1 RI
DATA U1,1 MAG
VAR_LIST_BEGIN
0
10000000
15000000
...
70000000000
VAR_LIST_END
BEGIN
-

Point your RSS reader here for a feed of the latest messages in all forums