怎么测量不确定度

实际上，对于100kΩ，不确定性是无限的，因为阻抗可以取任何值，包括无穷大和负数。
低于0.999 + - （0 - > .02。零到0.02）可以变为1或1.01（负阻）。
所以，是的，你正确地解释了不确定性。
而且你应该知道，对于VNA，通常阻抗范围对于良好的精度大约为5-500欧姆。
对于更高或更低的阻抗，应使用阻抗分析仪。
但是，如果你取得了不错的成绩，很可能是因为8753D硬件的出色RF设计超出了所有人的预期，并且在15年后继续这样做。
;-)



     以下为原文

  In fact, for 100kohms, the uncertaintly is infinite, as the impedance can take on any value including infinity and a negative number.  Becasue 0.999+-(0->.02. zero-to-0.02) can become 1, or 1.01 (negative resistance).

So yes, you interpret uncerainty correctly.  And you should know that normally the range of impedance for good accuracy is on the order of 5-500 ohms for a VNA.  For impedances higher or lower, impedance analyzers should be used.

However, if you are getting good results, it is most likely because the brilliant RF design of the 8753D hardware surpassed all expectations, and continues to do so 15 years later.  ;-)

感谢您的澄清:)鉴于这些惠普产品的性能，您认为规格有点宽吗？



     以下为原文

  Thanks for the clarification :)
Do you think that the specifications are a little wide given the demonstrated performance of these HP units?

附：
我订了你的书的副本！



     以下为原文

  P.S. I ordered a copy of your book!

> {quote：title = Josh写道：} {quote}>谢谢你的澄清:)>你认为这些HP装置的性能表现有点宽吗？
实际上，它们提供了相当大的余量，特别是在低频率。
实际上，对于S11测量，如果VNA硬件稳定，则calkit的质量是主要的决定因素。
我知道，对于负载规定为52 dB的套件，必须在我们的实验室测量58 dB才能保证规格，通常甚至可能是60 dB。
与开路和短路相似：每个的幅度误差几乎可以忽略不计（全反射），但相位存在一定的不确定性（由于物理长度）。
相位误差也归因于幅度误差（以6.6度至1 dB的速率），因此如果相位误差为0.66度，幅度误差将表示为0.1 dB，即使实际上是
非常接近zer。
附：
我希望你喜欢这本书。



     以下为原文

  > {quote:title=Josh wrote:}{quote}
> Thanks for the clarification :)
> Do you think that the specifications are a little wide given the demonstrated performance of these HP units?
Indeed, they provide quite a margin especially at low frequency.

In fact, for S11 measurements, the quality of the calkit is the main deterimining factor, if the VNA hardware is stable.  I know for a fact that for kits where the load is specified at 52 dB, it must be measured at our lab at 58 dB to gaurantee the spec, and typically it might even be 60 dB.  Similarly with the open and short: the magnitude error of each is nearly negligible (being a total reflection) but there is some uncertainty in the phase (being due to the physical length).  The error in phase is also attributed to the error in amplitude (at the rate of 6.6 degrees to 1 dB), so that if the phase error were 0.66 degrees, the amplitude error would be stated as 0.1 dB, even though in fact is is very near zer.

P.S. I hope you enjoy the book.

而不是创建一个新的线程，我想我会扩展这个，因为它仍然相当新。
我使用网络分析仪一点点，所以我知道一般的操作，但是我正在尝试用它们创建不确定性预算，这是我变得非常模糊的地方，而且我对工程方面的知识不是很强。
我将描述我的目标和逻辑，并且如果有人能告诉我我是否走在正确的轨道上，或者我离开基地，我会很感激。
我们有一个8510C系统，带有8515A S参数和一个83630B扫频器（26.5 GHz系统）。
我有一个85052B（我们的主要套件），它具有安捷伦认可的校准，并提供不确定性。
我的目标是验证另一个85052B校准套件，并希望至少达到1：1的TUR。
我的方法如下： - 执行完整的1端口验证，平均设置为1024（验证软件设置相同的设置） - 一旦标准化，我重新测量我们的主要套件的打开/短路的阶段。
- 我用安捷伦提供的读数（在相同点测试）计算读数的误差 - 我从我们的其他试剂盒测量未知数并添加上一步的误差校正 - 对于终止，我直接测量rho现在，
对于我的不确定性预算，这是我试图在合理范围内推动数字的地方。
对于相位测量，我希望通过校正测量误差并加快平均值，我会消除8510中的任何系统性和残留误差。然后，我将使用与安捷伦读取主要套件相关的不确定性，
8510的动态准确度，以及我的A型贡献者来计算我的不确定性。
对于我的回波损耗测量（在rho中），我首先要说明我提供的测量值相对于我们的主要套件中的负载质量（固定和滑动）是有效的方向性，从而也转移了它们的不确定性。
从那里我将使用与相位相同的逻辑。
现在，我希望这是真的： - 我为我们的A型重复性研究运行了20次未知试剂盒。
- 我从GPIB中读取读数，因此显示分辨率不是一个因素。
- 通过将平均值设置为1024，我可以解决任何系统错误。
- 通过计算安捷伦读数的误差来计算相位残差 - 在重复性研究中显示的随机和漂移误差可忽略不计，并且与A型混合使用 - Rho读数表示相对于初级套件中负载的有效方向性所以，是吗？
这个理由浮动，还是我错过了船？
基本上我的问题是，它是否有可能在相同的模型套件上比1：1 TUR更好，但需要注意的是rho表示有效的方向性。
另外，鉴于安捷伦提供的主要套件的不确定性远低于规格，因此在包括我的A型和8510动态精度之后，它们仍然低于我的规格。
我很抱歉这条长篇大论，但我非常感谢那些知识远远超过我自己的人们的帮助。编辑：乌鸦于2013年3月15日下午6:36



     以下为原文

  Instead of creating a new thread, I thought I would expand upon this one since its still fairly new.

I use network analyzers a little bit, so I know the general operation, however I'm trying to create uncertainty budgets with them and this is where I get pretty fuzzy and my knowledge of the engineering side of this is not very strong.  I'll describe my goal and my logic and would appreciate if anyone can tell me if I'm on the right track, or if I'm way off base.

We have an 8510C system with an 8515A S-parameter and an 83630B sweeper (26.5 GHz system).
I have an 85052B (our primary kit) which has an accredited calibration from Agilent with uncertainties provided.

My goal is to verify another 85052B cal kit and hopefully attain at least a 1:1 TUR.

My method is as follows:
- perform full 1 port verification with averaging set to 1024 (same settings the verification software sets)
- once standardized, I re-measure the phase of the opens/shorts of our primary kit.
- I calculate the error of my readings with the readings provided by Agilent (tested at same points)
- I measure the unknown from our other kit and add in the error correction from the previous step
- for the terminations, I measure rho directly

Now, for my uncertainty budget, this is where I'm trying to drive the numbers down as far as I reasonably can.

For the phase measurements, I'm hoping that by correcting for measured error and stepping up the averaging, I'm nullifying any systemic and residual errors in the 8510.  I would then use the uncertainties associated with Agilent's readings of our primary kit, the dynamic accuracy of the 8510, and my type A contributors to calculate my uncertainties.

For my return loss measurements (in rho), I would first declare that the measurements I'm providing are effective directivity relative to the quality of the loads (fixed and sliding) in our primary kit, thereby transferring their uncertainties as well.  From there I would use the same logic as with phase.

Now, here's what I'm hoping is true:
- I ran our unknown kit 20 times for our type A repeatability study.
- I pull the readings from the GPIB, so display resolution isn't a factor.
- By having the averaging set to 1024, I'm accounting for any systemic errors.
- Residual errors in phase are accounted for by calculating the error from Agilent's readings
- Random and Drift errors shown to be negligible during repeatability studies and lumped in with type A
- Rho readings on terms indicate effective directivity relative to loads in primary kit

So, does this rationale float, or am I missing the boat?  Basically my question is if its at all possible to get better than 1:1 TUR on a same model kit with the caveat that rho is indicating effective directivity.  Also given that the uncertainties of our primary kit provided by Agilent are substantially less than the specification so that after including my type A and 8510 Dynamic Accuracy, they still come in at less than my spec.


I'm sorry for the long message, but I would greatly appreciate the assistance here from those with far more knowledge than myself.

Edited by: crow on Mar 15, 2013 6:36 PM

VNA测量不确定度分析非常复杂。
平均仅减少随机错误。
大多数校准误差都是系统的。
校准标准的幅度和相位误差将转换为DUT的幅度和相位误差。
该变换是多个变量的函数 - 校准标准的“实际”响应，误差的性质（磁场或相位）以及DUT的实际响应。
如果使用标准校准套件系数定义的主套件，则精度受系数精度的限制。
您无法测试优于1：1 TAR的其他套件。
您需要使用更好的主要套件或对您的测量进行额外的修正。
由于您从Agilent获得主要试剂盒的校准数据，因此您可以使用校准中的校准数据而不是默认试剂盒系数。
然后，校准数据的不确定性将转移到您的测量值，这应该优于标准规格。
有很多已发表的论文解决了VNA测量的不确定性。



     以下为原文

  VNA measurement uncertainty analysis is quite involve. Averaging only reduces random errors.  Most calibrationtion errors are systematic.  Magnitude and phase errors of calibration standards will transform into magnitude and phase errors of a DUT.  That transformation is a function of multiple variables - the "actual" response of the calibration standards, the nature of the error (mag or phase) and the actual response of the DUT.  If you use your primary kit as defined by the standard calibration kit coefficients, your accuracy is limited by the accuracy of the coefficients. It is not possible for you to test other kits with better than 1:1 TAR.  You will need to used a better primary kit or do additional corrections to your measurements.

Since you get calibration data for your primary kit from Agilent, you may used the calibration data in your calibration instead of the default kit coefficients. The uncertainty of the calibration data will then be transferred to your measurements, which should be better than standard specifications.

There are quite a few published papers that address VNA measurement uncertainties.

谢谢Ken的回复。
至于使用安捷伦提供的标准套件数据，我正在测量相位。
例如，如果我在设备上测量了99度，并且安捷伦提供的数据表明它是100度，那么当我测量未知时，我会纠正1度误差。
所以我将安捷伦测量的不确定性转移到我未知的位置。
但是，我必须在该读数中包括随机，系统和A类不确定性。
我希望能够通过可重复性（使用相同的电缆/适配器进行所有测量）和平均来尽可能多地纠正这些错误。
我将介绍8510C用户手册中提供的信息和错误模型。
此外，停留时间设置是否对测量精度或可重复性有任何重大影响，还是通过平均值相形见绌？
最后，精密航空公司是能够以合理的TUR测量滑动负载的最佳商用标准吗？
（如果这是正确的术语，则使用圆拟合）。
此时我正在严格考虑设备的电气特性，但如果采用物理尺寸测量确实有助于降低不确定性，则可能值得研究。
道歉，因为我仍然非常生气，有很多这样的东西，并且在我去的时候努力学习。
我假设我无法达到1：1，但我希望尽可能接近 - 或者用其他方法或标准来超越它。
我赞美每个人的时间！



     以下为原文

  Thanks Ken for the response.

As for using the data provided by Agilent with the standard kit, that is what I was doing for measuring the phase.  For example, if I measured 99 degrees on a device, and the data provided by Agilent stated that it was 100 degrees, then I would correct for that 1 degree error when I measured my unkown.  So I'm transferring the uncertainty of Agilent's measurement to my unknown.  However, then I must include random, systemic, and Type A uncertainties into that reading as well.  I'm hoping to be able to correct for as much of those errors as possible through repeatability (using same cables/adapters for all measurements) and averaging.  I'm going off of the information and error models provided in the 8510C user's manual.

Also, does the dwell time setting have any major effect on measurement accuracy or repeatability, or is it dwarfed by averaging anyway?

Lastly, is a precision airline the best commercially available standard to be able to measure a sliding load with a reasonable TUR?  (using circle fitting if that's the correct term).  I'm looking at strictly electrical properties of the devices at this point, but if taking physical dimensional measurements will really help drive the uncertainties down, it may be worth looking into.

Apologies as I'm still pretty raw with a lot of this stuff and trying to learn as I go.  I'm assuming I won't be able to hit 1:1, but I'd like to get as close as possible - or perhaps exceed it with alternative methods or standards.

I apprecaite everyone's time here!

通常，使用相对于参考测量的一个器件测量的偏差来校正另一个器件DUT的测量并不是一个好主意，除非器件具有非常相似的特性。
VNA测量的不确定性是测量系统的不确定性和DUT的实际响应的函数。
A类分析可用于确定随机错误。
大多数校准套件相关的误差都是由于加工公差造成的系统误差。
安捷伦的校准标准和验证标准采用相似的公差制造，因此使用验证套件的TUR不会优于1：1。
使用商业上可用的标准不可能做得更好要比1：1更好，人们将需要使用更精确的标准，例如我们为自己的标准制定的标准和许多国家计量的标准。
这些标准非常难以制造，因此非常昂贵。
这些标准可通过尺寸测量来追踪，具有非常小的不确定性。



     以下为原文

  Usually it is not a good idea to use deviations of one device measurement, with respect to a reference measurement, to correct measurements of another device, DUT, unless the devices have very similar characteristics.  Uncertainty of VNA measurements is a function of the measurement system's uncertainties AND the actual response of the DUT.  Type-A analysis is useful to determine random errors.  Most calibration kit related errors are systematic errors as a result of machining tolerances. Agilent's calibration standards and verification standards are fabricated with similar tolerances and therefore TUR using verification kits is no better than 1:1. It is not possible to do any better than that using commercially avaailable standards

To get better than 1:1, one will need to used much more precise standards like the ones that we make for oursleves and many national metrology institudes.  These standards are very difficult to make and therefore very expensive. These standards are traceable through dimensional metrology with very small uncertainties.

肯，是的，我几乎都在猜测。
至于您用来验证自己的标准设备的尺寸测量，您可以在典型的校准套件上做些什么吗？
我猜你可以从那里计算你的校准系数？
是否有关于此流程的论文/资源？
- 再次感谢输入



     以下为原文

  Ken,

Yeah I was pretty much guessing as much.  As far as the dimensional measurements you use to verify your own standard devices, is that something you can do on a typical cal kit as well?  I'm guessing you could calculate your cal coefficients from there?  Are there any papers/resources on this process available?

--thanks again for the input

您可以查看以下参考文献：应用笔记1287-11“为Agilent矢量网络分析仪指定校准标准品和试剂盒”http://cp.literature.agilent.com/litweb/pdf/5989-4840EN.pdf KHWong，“
使用精密同轴空气介质传输线作为校准和验证标准，“Microwave Journal，Dec。1988 pp 88-92 KH
Wong，“物理测量校准标准的表征”，第69届ARFTG会议文摘，1992年6月Ken



     以下为原文

  You can check out the following references:

App Note 1287-11 "Specifying Calibration Standards and Kits for Agilent Vector Network Analyzers"
http://cp.literature.agilent.com/litweb/pdf/5989-4840EN.pdf

K.H.Wong, “Using Precision Coaxial Air Dielectric Transmission Line as Calibration and Verification Standards,” Microwave Journal, Dec. 1988 pp 88-92
K.H. Wong, “Characterization of Calibration Standards by Physical Measurements,” 39th ARFTG Conference Digest,  June 1992

Ken