无法获得增强响应校正的方程式

好吧，我仔细检查了我的数学，然后我让我们的软件人员挖掘代码并拉出ERC数学，他们都匹配，所以我仔细看了你的笔记，我发现你说：“将两个端口连接在一起
冲洗并测量传输跟踪：直接来自S21的ETF。“
BZZT！
不，对不起，但是感谢您玩“制作自己的错误纠正方法”当您连接直通时，您正在测量ETF Plus与ESF和ELF的交互的组合。
ETF术语必须通过完整的2端口校准来确定。
所以我会问：为什么不进行全2端口校准？
为什么要写自己的？
并且，如果你试图编写自己的，并且你的DUT不是m-f（因此，可插入），那么你还必须处理非零通过的影响。
我们使用adpater删除数学在8753中执行此操作。
为避免使用内置校准，似乎很麻烦。
事实上，对于你的第二点，如果8753的ESF（源匹配）为零，则S11在S21数学中无关紧要，因为反射的所有东西都被端口1终端吸收（ESF = 0意味着反射信号的完美吸收）
并且人们期望S11的效果随着ESF的改善而减弱。



     以下为原文

  Well I double checked my math, then I had our software guys dig into the code and pull out the ERC math, and they all match, so I looked more carefully at you notes and I found you said:

"Connect both ports together with a flush through and measure Transmission Tracking:ETF directly from S21."

BZZT!  No, sorry, but thanks for playing 'make up your own error correction methods'


When you connect the thru, you are measuring the combination of ETF Plus the interaction of ESF and ELF.

The ETF term must be determined by a full 2 port calibration.

So I'll ask: why not do a full 2 port calibration?  Why write your own?  And, if you try to write your own, and your DUT is not m-f (hence, insertable) then you also have to deal with accounting for the effect of the non-zero thru.  We do that in the 8753 with the adpater removal math. Seems like a lot of trouble to avoid using the built-in calibration.

If fact, to your second point, if ESF (source match) of the 8753 is zero, then S11 does NOT matter in the S21 math, as everything reflected is absorbed by the port 1 termination (ESF=0 implies perfect absorption of reflected signal) and one expects the effects of S11 to diminish with improved ESF.

感谢Joel博士，这个星球上没有多少人能够理解我的问题，更不用说以这么好的幽默来回答它了。
在回答您的问题时，我们并没有完全避免使用内置的错误纠正 - 我们试图了解它是如何工作的。
如果在看起来完全取消RF课程之后在大学进行射频测量，那么人们越来越感兴趣，有利于数字射频课程。
高速数字似乎正在完成一个完整的圆形回到RF测量和原理。
无论如何 - 我离题了。
我以为你已经解决了我们的问题但唉，我只是把我们的程序写错了。
我们按照Doug Rytting关于12纠错的应用笔记的第18页正确执行了它：http：//www-ee.uta.edu/online/adavis/ee5349/NA_Error_Models_and_Cal_Methods.pdf我们将端口连接在一起，并通过刷新和
测量负载匹配ESF作为应用1端口纠错的S11测量。
然后我们测量/计算传输跟踪为ETF =（S21M-EXF）（1-ESF.ELF）所以我们似乎仍然坚持原始S11测量似乎对S21结果产生不利影响的原始问题。
我的想法是关于端口1和2的加载，而不是反射信号。
如果BPF的输入阻抗变为0，则分析器在端口1处看到的电压将被加载，而端口2处的电压则不会。
（这是一个涉及un-un变换器的复杂滤波器，因此端口匹配可能不是倒数）这种想法，以及教科书方法不起作用的事实是导致我们发明我们自己的等式的原因。
真正困扰我的事实是我们的方程式有效而另一方面没有？



     以下为原文

  Thanks Dr Joel,
There aren't many people on the planet that could understand my question, let alone answer it with such good humor.

In answer to your questions, we're not avoiding using the in built error correction at all - we're trying to gain an understanding of how it works.
There's a growing interest if RF measurements at the uni after what seemed to be a total ousting of RF courses, in favor of digital ones.
Hi speed digital seems to be completing a full circle back into RF measurements and principles. Anyway - I digress.

I thought you'd solved our problem but alas, I just wrote down our procedure wrong. 
We performed it right, as per page 18 of Doug Rytting's app note on 12 error correction: 
http://www-ee.uta.edu/online/adavis/ee5349/NA_Error_Models_and_Cal_Methods.pdf

We connected the ports together with a flush through and measured Load Match ESF as the S11 measurement with 1 port error correction applied. 

Then we measured/calculated Transmission Tracking as ETF=(S21M-EXF)(1-ESF.ELF)

So it seems we're still stuck with the original problem that the raw S11 measurement seems to have an adverse effect on the S21 result.

My thinking was about the loading of ports 1 and 2, not the reflected signals.
If the input impedance to the BPF goes to 0, the voltage seen by the analyzer at port 1 is loaded down while the voltage seen at port 2 is not.
(it's a complex filter involving an un-un transformer so port matching is probably not reciprocal)

This thinking, and the fact that the text book approach wasn't working is what lead us to invent our own equation.
The fact that really puzzles me is that our equation works and the other does not?

> {quote：title = Dr_joel写道：} {quote}>所以我会问：为什么不做一个完整的2端口校准？
为什么要写自己的？
并且，如果你试图编写自己的，并且你的DUT不是m-f（因此，可插入），那么你还必须处理非零通过的影响。
我们使用adpater删除数学在8753中执行此操作。
为避免使用内置校准，似乎很麻烦。
Joel和/或Ken，您认为添加到8753的最有用的校准方法是什么？在已发表的科学文献中可以获得这样的信息？
我假设后来的8753系列，如后来的8720系列，可能在其固件中实现了增强的响应，但不是未知的校准威廉希尔官方网站 。
+ A。
Ferrero，U。Pisani，使用未知'Thru'进行双端口网络分析仪校准，IEEE Microwave和Guided Wave Letters，Vol。
2，No。12，pp 505-507，1992年12月+我从未使用过未知的自己，但我怀疑未知 - 通过本来是一个更好的项目，或者如果时间允许，值得将工作扩展到。
但也许你可以认为更有用的东西，可以用非专有的材料来实现。
Dave编辑：drkirkby于2013年10月16日上午5:35编辑：drkirkby于2013年10月16日上午5:41



     以下为原文

  > {quote:title=Dr_joel wrote:}{quote}

> So I'll ask: why not do a full 2 port calibration?  Why write your own?  And, if you try to write your own, and your DUT is not m-f (hence, insertable) then you also have to deal with accounting for the effect of the non-zero thru.  We do that in the 8753 with the adpater removal math. Seems like a lot of trouble to avoid using the built-in calibration.

Joel and/or Ken,

What in your opinion would be the single most useful calibration method to add to an 8753 for which the information to do so is available in the published scientific literature? 

I assume the later 8753 series, like the later 8720 series, probably implemented enhanced response in their firmware, but not the unknown-thru calibration technique.

+A. Ferrero, U. Pisani , Two-Port Network Analyzer Calibration Using an Unknown ‘Thru’, IEEE Microwave and Guided Wave Letters, Vol. 2, No. 12, pp 505-507, Dec. 1992+

I've never used the unknown thru myself, but I suspect unknown-thru would have been a better project, or if time permits one worth expanding the work to. But perhaps you can think something even more useful, which could be implemented from material that's not proprietary. 

Dave

Edited by: drkirkby on Oct 16, 2013 5:35 AM

Edited by: drkirkby on Oct 16, 2013 5:41 AM

附件

• Two-port_network_analyzer_calibration_using_an_unknown-thru.pdf148.5 KB
  

是的，这将是最有用的。
但是，人们需要首先应用全局 - 增量匹配校准，然后才能应用未知的应用程序。
相当多的步骤。



     以下为原文

  yes, that would be the most useful.  But one would need to apply first a global-delta-match calibration and then a appy unknown through. Quite a lot of steps that is.

ERC的等式是正确的，并且在所有支持ERC的安捷伦网络分析仪中使用了很长时间，并且没有我所知道的问题。
它来自单一流图，类似于Doug Ryttings论文的第16页。
由于未测量DUT的S12和S22，因此丢弃了S12和S22的所有项。
通过一些代数简化，可以推导出S21 ERC方程。
你不能假设Esf为零。
您提出的等式与信号流图模型和理论不一致。
S21方程适用于收发和非互易设备，因为负载匹配效应是在S11m项中考虑的。
对于互惠设备，只有S22术语退出。
低损耗互易设备的S11（实际）需要考虑负载匹配项。
您是否在等式中使用原始测量数据？
S11m和S21m是未校正的测量数据。
肯



     以下为原文

  The equation for ERC is correct and is used in all the Agilent Network analyzers that support ERC for a long time and had no issues that I am aware of.  It was derived from the singal flow graph, similar to page 16 of Doug Ryttings paper.  Since S12 and S22 of the DUT are not measured, all the terms with S12 and S22 are dropped.  With some algebraic simplification, the S21 ERC equation can be derived.  You can not assume Esf to be zero.  The equation that you came up with does not agree with the signal flow graph model and theory.  The S21 equation is good for receiprocal and non-reciprocal devices since the load match effect is account for in the S11m term.  For reciprocal devices, only the S22 term drops out.  S11(actual) of low loss reciprocal devices needs  to account for the load match term.
  
Did you use raw measured data in the equation?  S11m and S21m are uncorrected measured data.  

Ken

感谢Ken的回复，我们无法再访问HP8753分析仪，但我认为在这个阶段我们应该重复实验。
我同意你的观点，它在所有已知理论的面前飞逝。
这通常表明我们在某处收集原始数据时发生了错误。
我尽快，我会安排再做一次实验。
如果情况相同，我会发布原始数据和我们用来纠正它的matlab代码。



     以下为原文

  Thanks for the reply Ken,
We no longer have access to the HP8753 analyzer but I think at this stage we should repeat the experiment.
I agree with you that it flies in the face of all known theory.
That generally points to the fact that we've made an error somewhere gathering the raw data.

As soon as I can, I'll arrange to do the experiment again.
If things go the same way, I'll post the raw data and the matlab code we use to correct it.