AnsweredAssumed Answered

Reflection coefficient over 1

Question asked by bar28 on Feb 5, 2016
Latest reply on Feb 11, 2016 by bar28

Like • Show 0 Likes0
Comment • 7

I have done 1-port calibration (50MHz to 18GHz) with 8510C using 85054D end of the test port cable. When I'm measuring another 85054D calkit Open and Shorts reflection coefficient following quite nicely calculated Open and shorts polynomical reflection coefficients. If I'm repeating the same calibration and measurements with 8720 series analyzers, Open and short results showing slightly over 1 reflection coefficient values on some part of the sweep (standard definitions are exatly the same on 8510C and 8720). Where this difference might come? When I'm comparing 8510C and 8720 open and short results together, phase values showing almost the same value. Only the magnitude values has difference. All used analyzers are well factory calibrated and meet spesifications, so instruments are good shape.

No one else has this question

Outcomes

7 Replies

Dr_joel Feb 9, 2016 12:11 PM
Mark Correct
Correct Answer
For precision type measurements, you have to have the 8720 in stepped sweep mode. See if that makes a difference.

Also, when you say greater than 1, is it 1.0001 or 1.1?, in other words, how far off is it. (best just look at S11 in dB). And did you test the stability of the test cables?
Like • Show 0 Likes0
Actions
- bar28 @ Dr_joel on Feb 10, 2016 7:45 AM
  Mark Correct
  Correct Answer
  Test port cable is always locked carefully on one position. That way I always trying to minimize errors what test port cable can cause. We speaking about 1.001 to 1.003 level. I found interesting thing today, that if doing VNA 1-port measurement calibration an order Load, Open and Short. After that when I'm measuring DUT's open and Short results are below 1 and following nicely polynomical values with 8510 and 8720. If measurement calibration is done on order Open, Short and Load then Open/Short measurements looks worse. Interesting thing, but standards connection order really matters on DUT's results..at least with 8720 series analyzers.
  
  Edited by: bar28 on Feb 10, 2016 8:43 AM
  Like • Show 0 Likes0
  Actions
  - Dr_joel @ bar28 on Feb 10, 2016 4:00 PM
    Mark Correct
    Correct Answer
    I can't really believe the connection order causes a difference, except that perhaps there is some slight heating or other effect from taking the data and so the different order has different heating effect. Another cause could be connector repeatability. Perhaps there is a difference in the connection repeatability of the standards. For testing this: apply a standard, set data into memory and data minus memory. The data should drop to something like -60 to -80 dB. Then disconnect and reconnect the standard. What you see will be the connector repeatability. Also rotate the standard by 90,180,270 degrees from the first connection to see if there is some off-center effect. See if one standard is worse than the others.
    Like • Show 0 Likes0
    Actions
drkirkby Feb 10, 2016 9:02 AM
Mark Correct
Correct Answer
You are not by any chance confusing the sex (or should I say gender?) of the calibration standards are you? See picture attached, which is a short from an 85054B cal kit. Used on an 8720 or 8510, you would select SHORT (F) during calibration, despite the standard is male.

A reflection coefficient of 1.001 to 1.003 is a return loss of -0.0009 dB to -0.026 dB. What's the uncertainty specification of the 8720? I'm certain it is greater than 0.0009 dB under any conditions, and probably 0.026 dB, especially with cables around.

There's going to be some error on the 8720 as it can't take into account the inductance of the short. According to the PNA data for the 85054D, the female short has a delay of 27.99 ps and the male short 63.078 ps My 8720D's version 7.74 firmware shows the delay for the 85054B/D female short as 28.003 ps, so I assume the odd 28.003-27.99=0.013 ps = 13 fs is an attempt to compensate for the inductance of the short, which is not going to be correct except at one frequency. For the male short, the firmware shows 63.106 ps, which is 28 fs more than the PNA value.

Dave
Attachments
- connector-sex.jpg576.6 KB
Like • Show 0 Likes0
Actions
- bar28 @ drkirkby on Feb 10, 2016 9:43 AM
  Mark Correct
  Correct Answer
  Hi,
  
  I'm not using firmware calibration at all. All calibration standards are individualy characerized by Swiss METAS and polynomical data is replaced by more accurate database values. So each standards (Open, Short and Load) I have accreditated/traceable results where has taken account connector effetcs too. VNA-Tools II is immune to analyzer itself, because it just measuring the Raw-data. Error calculation happening inside the program and all is done correctly on that side. Even testport adapters pin-deptsh are adjusted by Keysight enought to recression side avoiding the near field effect. All is done very carefully. So my assume on this case is , that this tiny error on high reflect components magnitude must come now mostly from my VNA box. I have heard before too, that someone else has mentioned that calibration standards connection orders can have own effect on some VNA boxes behavings. Maybe today's ENA's and PNA's has better architecture where order not have any effects
  
  Here is specs from 8722ES. All errors goes still nicely inside the uncertainty.
  
  Edited by: bar28 on Feb 10, 2016 10:42 AM
  Attachments
  - Sieppaa.PNG103.9 KB
  Like • Show 0 Likes0
  Actions
  - Dr_joel @ bar28 on Feb 10, 2016 4:03 PM
    Mark Correct
    Correct Answer
    In thinking about this, perhaps it is an off-center conductor for example, if the load makes the conductor on-center, and then you do open and short, and the open makes it off-center, then the short might have a different measurement. But I never heard of standard order making any difference if there is not high power, or bias attached.
    Like • Show 0 Likes0
    Actions
    - bar28 @ Dr_joel on Feb 11, 2016 6:43 AM
      Mark Correct
      Correct Answer
      I agree with you and I thought totally same things too. Well, I can't find now true fact why It behaves such a way. I repeated measurements using the 7mm 85050B calkit and same effect can see it too. I'm happy, that all working well now on Load-Open-Short order. In genetally I always trying to measure similar type of components at the same time and not mixing them during the measurement process.
      Like • Show 0 Likes0
      Actions

Reflection coefficient over 1

Outcomes

Attachments

Attachments

Related Content

Recommended Content