是否可以使用N8973A NoisefigureAnalyser测量0.1dB噪声系数

许多RF人员不太了解的一个问题，当他们这样做时，不喜欢，是噪声系数准确性的问题。
（不喜欢，因为答案并不像他们喜欢和复杂的那么好）。
首先，您必须了解噪声系数测量的不确定性取决于许多因素，它不是一个简单的测量。
以下是主要贡献者：1）噪声源不确定性（如果使用噪声源 - 实际上除了使用噪声源测量噪声系数之外还有其他方法）。
对于大多数常见噪声源，这通常约为0.2 dB，无需特殊校准。
2）通过DUT的噪声参数匹配测试系统及其对DUT的影响。
作为LNA设计人员，您肯定必须知道DUT的噪声系数取决于它的噪声参数，如果输入的匹配不是最佳匹配（gamma opt），则噪声系数不会最小。
大多数设计试图在50欧姆下实现最小噪声系数。
大多数噪声源不是50欧姆，但通常为45-55欧姆。
因此，根据噪声参数中的“r因子”，噪声源的失配误差可以拉动DUT的输入匹配并改变其明显的噪声系数。
这就是为什么，如果你尝试不同的噪声源，你会得到不同的噪声系数结果。
最低的结果可能不正确，因为它可能是DUT最初设计和测试的45欧姆实际噪声源，如果您还使用45欧姆噪声源进行验证，您将获得低读数。
但如果您碰巧使用的电压恰好为50欧姆，则读数会略高，但这个更高的读数可能是正确的50欧姆噪声系数。
失配效应可能非常高，大约为0.2-0.5 dB。
3）DUT的增益（或者说，输出端的噪声过大）。
所有噪声测量系统都有自己的内部噪声。
如果DUT的增益+噪声系数在放大器输出端产生的噪声（噪声高于kTB电平）类似于或低于NFA系统的噪声系数，则会出现明显的误差和抖动。
测量。
如果DUT是高增益，则放大器输出的噪声功率明显高于NFA输入噪声，这种误差源不是很大的影响。
为获得最佳测量精度，实际上您应该使用基于噪声参数的噪声系数测量系统。
PNA-X上的噪声系数应用实际上执行了大约50欧姆的噪声参数提取，并从中计算出50欧姆的噪声系数。
这大大消除了上面的错误＃2。
对于错误＃1，PNA-X支持两种校准方法：使用噪声源或使用功率计。
功率计方法的理论误差略低，因为最佳功率传感器的不确定性略好于最佳噪声源的不确定性。
对于错误＃3，PNA-X具有内置LNA以实现低噪声（实际上基于SA和基于NFA的系统也可以），但只有PNA-X使用其自身接收器的噪声参数才能正确地去嵌入
接收器的噪声功率来自整体噪声测量。
基于SA和NFA的系统测量其接收器的50欧姆噪声功率，并从总噪声中减去该功率，但如果被测量的DUT的输出匹配不完全是50欧姆，那么根据SA的噪声参数的性质
或NFA，连接DUT时产生的噪声功率与校准期间的噪声功率不同，因此内部噪声的减法将不正确。
此错误可以增加或减少整体噪声系数。
总之，在我的观点中，PNA-X噪声系数测量应用是确定DUT 50欧姆噪声系数的最佳市售噪声系数测量系统，并且与国家劳动力使用的最佳方法（如NPL）相比较
和NIST）。
这是一篇参考文献（由NIST和Keysight共同撰写，包括我自己），描述了这种关联工作：http：//ieeexplore.ieee.org/xpls/abs_all.jsp ?arnumber = 5755200＆amp; tag = 1



     以下为原文

  One issue that a lot of RF people don't understand well, and when they do, don't like, is the issue of noise figure accuracy. (don't like because the answer isn't as good as they like and complicated).

First and foremost, you must understand that the uncertainty of a noise figure measurement depends on many things, it is not a simple measurement.  Here are the key contributors:

1) Noise source uncertainty (if you are using a noise source -- in fact there are other methods than using a noise source to measure noise figure).

This is typically on the order of 0.2 dB for most common noise sources, without a special calibration.

2) Match of the test system, and it's effect on the DUT, through the DUT's noise parameters.

As an LNA designer you surely must know that the noise figure of the DUT depends upon it's noise parameters, and if the match of at the input is not the optimum match (gamma opt) then the noise figure will not be minimum.  Most designs try to acheive minimum noise figure at 50 ohms.  Most noise sources are NOT 50 ohms, but typically range from 45-55 ohms.  So, depending upon the "r-factor" in the noise parameters, the mismatch error of the noise source can pull the input match of the DUT and change it's apparent noise figure.  This is one reason why, if you try different noise sources, you will get different noise figure results.  The lowest result may not be correct, in that it might be the DUT was initially designed and tested wtih a 45 ohm actual noise source, and if you also use a 45 ohm noise source for verification you will get a low reading.  But if you happen to use a noise source that is exactly 50 ohms, it will read slightly higher, but this higher reading might be the correct 50 ohm noise figure.  Mismatch effect can be quite high, on the order of 0.2-0.5 dB.

3) Gain of the DUT (or rather, excess noise at the output).  

All noise measurements system have their own internal noise.  If the the gain+noise figure of the DUT produces excess noise (noise above the kTB level) at the output of the amplfier that is similar to or lower than the noise figure of the NFA system, then there will be significant error and jitter in the measurement.  If the DUT is high gain, then the noise power coming out of the amplifier is significantly higher than the NFA input noise and this source of error is not a big effect.


For the best measurement accuracy, in fact you should use a noise parameter based noise figure measurement system.  The noise figure application on the PNA-X does in fact perform noise parameter extraction, around 50 ohms, and computes the 50 ohm noise figure from this.  This eliminates to a great extend the error #2 above.  

For error #1, the PNA-X supports two calibration methods: using a noise source or using a power meter.  The power meter method has slightly lower theoretical error based on the fact that the uncertainty of the best power sensors is a little better than the uncertainty of the best noise source.

For error #3, the PNA-X has a built in LNA for low noise (in fact the SA based and NFA based systems do as well) but only the PNA-X uses the noise parameters of it's own receiver to properly de-embed the receiver's noise power from the overall noise measurement.  SA and NFA based systems measure the 50 ohm noise power of their receivers, and subtract that from the overall noise, but if the output match of the DUT being measured is not exactly 50 ohms, then by the nature of the noise parameters of the SA or NFA, the noise power created when the DUT is connected is NOT the same as the noise power during calibration, and so the subtraction of the internal noise will not be correct. This error can either add or subtract from the overall noise figure.

In conclusion, in my opition, the PNA-X noise figure measurement application is the best commercially available noise figure measuerment system to determine the 50 ohms noise figure of a DUT, and is comparible to the best methods used at national laboritories (such as NPL and NIST).  Here is a reference paper (co-authored by NIST and Keysight including myself) which describes this correlation effort:
http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=5755200&tag=1

>总而言之，在我看来，PNA-X噪声系数测量应用是商用的最佳噪声系数测量系统，用于确定DUT的50欧姆噪声系数，并且与国家劳动力使用的最佳方法（如
NPL和NIST）。
这是一篇参考文献（由NIST和Keysight共同撰写，包括我自己），描述了这种关联工作：> http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=5755200&tag=1亲爱的先生，As
你说，我们可以使用PNA-X噪声系数测量精确测量低噪声放大器的0.1 dB，0.2 dB和0.3 dB噪声系数。
您能否详细介绍一下安捷伦的PNA-X系列以及PDF格式和我们需要的噪声源以及标称噪声比（ENR），我们需要测量0.1 dB，0.2 dB，0.3 dB噪声
低噪声放大器（LNA）图。
你能澄清一下吗？你能否详细说明。



     以下为原文

  > In conclusion, in my opition, the PNA-X noise figure measurement application is the best commercially available noise figure measuerment system to determine the 50 ohms noise figure of a DUT, and is comparible to the best methods used at national laboritories (such as NPL and NIST).  Here is a reference paper (co-authored by NIST and Keysight including myself) which describes this correlation effort:
> http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=5755200&tag=1

Dear Sir,

             As you said, we can measure 0.1 dB, 0.2 dB, and 0.3 dB noise figure of the low noise amplifier using PNA - X Noise figure measurement accurately.

             Can you give full details,  Which series Agilent's  PNA-X and details in the form of PDF and what Noise Source we need and What nominal excessive noise ratio (ENR), we need to have measurement of 0.1 dB, 0.2 dB, 0.3 dB noise figure of low noise amplifier (LNA).
           Can you clarify it and can you give details.

对于迟到的回复感到抱歉：您可能希望将PNA-X网络分析仪与选项029一起使用。校准使用噪声源（例如346C）或功率计。
如果您有功率计，则校准或测量不需要噪声源。
PNA-X使用与NFA不同的方法，几乎​​在所有情况下都更准确。
实际上，据我们所知，它是使用国家实验室噪声接收器测量噪声系数的最准确的仪器（并且它与它们紧密匹配）。



     以下为原文

  Sorry for the late reply:  You would want to use the PNA-X network analyzer with option 029. Calibration uses either noise source (such as 346C) or a power meter.  A noise source is not needed for the calibraiton or the measurement if you have a power meter.  PNA-X uses a different method than an NFA, and in almost all cases is more accurate.  In fact, as far was we know it is the most accurate instrument for measuring noise figure outside using national labs noise receivers (and it closesly matches them).

亲爱的先生，感谢您的回复确认。
我们想要使用PNA-X系列N5249A（10 MHz - 8.5 GHz）网络分析仪。
我对PNA-X系列N5249A网络分析仪有一些说明（1）当我读到PNA-X系列N5249A网络分析仪的数据表时。
在测量低噪声放大器（LNA）的噪声系数时，我们不需要连接噪声源346B（10 MHz-18 GHz）和标称的过高噪声比ENR 15 dB，就像我们使用安捷伦噪声系数分析仪时一样
8973A（10 MHz-3 GHz）用于测量噪声系数（NF）。
我请你确认上述说明。
（2）从这个PNA-X系列N5249A网络分析仪，我们可以测量LNA的散射参数S11，S21，S12和S22，还可以测量三阶输入截取点（IIP3）和P1 dB压缩点。
如果可能，我们是否需要使用组合器组合两个信号来测量IIP3和P1 dB压缩点。



     以下为原文

  Dear Sir,
                 Thank you for your kind reply for confirmation. We want to use  PNA- X Series N5249A( 10 MHz - 8.5 GHz) network analyzer. I have some clarification about PNA-X series N5249A network analyzer 

(1) As I read datasheet of PNA- X series N5249A network analyzer. While measuring noise figure of low noise amplifier (LNA), we don't require to connect the Noise source 346B (10 MHz- 18 GHz) and nominal Excessive Noise Ratio ENR 15 dB as like when we are using Agilent Noise Figure Analyzer N-8973A( 10 MHz- 3 GHz) for measuring Noise Figure (NF). 

I request you to confirm the above clarification.

(2) From this PNA-X Series N5249A network analyzer, can we measure Scattering parameters S11, S21,S12, and S22 of LNA and also can we measure third order input intercept point (IIP3) and P1 dB compression point. If it is possible, do we need to use the combiner to combine two signals for measuring IIP3 and P1 dB compression point.

> {quote：title = nimbargi写道：} {quote}>亲爱的先生，>谢谢您的回复确认。
我们想要使用PNA-X系列N5249A（10 MHz - 8.5 GHz）网络分析仪。
我对PNA-X系列N5249A网络分析仪有一些说明>>（1）当我读到PNA-X系列N5249A网络分析仪的数据表时。
在测量低噪声放大器（LNA）的噪声系数时，我们不需要连接噪声源346B（10 MHz-18 GHz）和标称的过高噪声比ENR 15 dB，就像我们使用安捷伦噪声系数分析仪时一样
8973A（10 MHz-3 GHz）用于测量噪声系数（NF）。
>>我要求您确认上述说明。
在PNA-X上进行噪声系数测量时，您不需要噪声源。
在噪声系数测量的校准过程中，甚至可以选择使用噪声源（正如Dr_Joel先前指出的那样）。
>>（2）从这个PNA-X系列N5249A网络分析仪，我们可以测量LNA的散射参数S11，S21，S12和S22，还可以测量三阶输入截取点（IIP3）和P1 dB压缩点。
如果可能，我们是否需要使用组合器组合两个信号来测量IIP3和P1 dB压缩点。
您可以在与噪声系数测量相同的通道中进行LNA的S参数测量（如果希望S参数与噪声系数测量的频率相同）。
您还可以使用PNA-X在LNA上进行IIP3和P1 dB压缩测量，但是您需要额外的应用选项（IIP3为087，增益压缩为086）。
为了能够进行双音IIP3测量，你需要一个带有组合器选项的PNA-X（4端口盒上的423或2端口2源PNA-X上的223）。



     以下为原文

  > {quote:title=nimbargi wrote:}{quote}
> Dear Sir,
>                  Thank you for your kind reply for confirmation. We want to use  PNA- X Series N5249A( 10 MHz - 8.5 GHz) network analyzer. I have some clarification about PNA-X series N5249A network analyzer 
> 
> (1) As I read datasheet of PNA- X series N5249A network analyzer. While measuring noise figure of low noise amplifier (LNA), we don't require to connect the Noise source 346B (10 MHz- 18 GHz) and nominal Excessive Noise Ratio ENR 15 dB as like when we are using Agilent Noise Figure Analyzer N-8973A( 10 MHz- 3 GHz) for measuring Noise Figure (NF). 
> 
> I request you to confirm the above clarification.

You do not need a noise source during a Noise Figure measurement on the PNA-X.  The use of a Noise Source is even optional during the calibration for a Noise Figure measurement (as Dr_Joel pointed out earlier).

> 
> (2) From this PNA-X Series N5249A network analyzer, can we measure Scattering parameters S11, S21,S12, and S22 of LNA and also can we measure third order input intercept point (IIP3) and P1 dB compression point. If it is possible, do we need to use the combiner to combine two signals for measuring IIP3 and P1 dB compression point.

You can make the S-Parameter measurements of the LNA in the same channel as your Noise Figure measurement (If you want the S-Parameters at the same frequencies as the Noise Figure measurement).  You can also make the IIP3 and the P1 dB compression measurements on your LNA with a PNA-X, but you will need additional application options (087 for IIP3 and 086 for Gain Compression).  to be able to do the 2-tone IIP3 measurements, you will need a PNA-X with the combiner option (423 on a 4 port box or 223 on a 2-port 2-source PNA-X).

> {quote：title = nimbargi写道：} {quote}>在我们的组织中，我们有以下规格NOISE Figure ANALYZER >> * N8973A（10MHz - 3GHz）NOISE图形分析仪和噪声源是346B（10MHz-18GHz）标称ENR
15dB。* >>我的问题是，从上面的规格NOISE FIGURE ANALYZE R和NOISE SOURCE。
我们能够测量0.1 dB或0.2 dB或0.3 dB或0.4 dB甚至小于0.2 dB的噪声系数。
因为我们设计了低噪声放大器（LNA），噪声系数为0.2 dB。
>>如果不可能，使用哪个数字NOISE Figure ANALYZER和NOISE SOURCE来测量小于0.2 dB的噪声系数。
虽然他的方法并不适用于生产线，但瑞典无线电火腿Leif A***rink（SM5BSZ）在测量低噪声系数值方面做了大量工作，且不确定性很小。
http://www.sm5bsz.com/lir/nfprec/nfprec.htm（但请务必先阅读他首先链接的页面）顺便说一下，有谁知道为什么术语“精度”在测量仪器的规格中被如此多地使用
？
如果DVM的准确度为0.1％，是否意味着99.9％不准确！
戴夫



     以下为原文

  > {quote:title=nimbargi wrote:}{quote}
> In our organization,  we have the following specification NOISE FIGURE ANALYSER
> 
> *N8973A (10MHz - 3GHz) NOISE FIGURE ANALYSER and NOISE SOURCE is 346B (10MHz-18GHz) nominal ENR 15dB.*
> 
> My question is, from the above specification NOISE FIGURE ANALYSE R and NOISE SOURCE. We are able to measure the 0.1 dB or 0.2 dB or 0.3 dB or 0.4 dB or even less than  0.2 dB noise figure. Because We have designed the Low Noise Amplifier (LNA) with 0.2 dB noise figure.
> 
> If it is not possible, which number NOISE FIGURE ANALYSER and NOISE SOURCE is used for measuring the  Noise figure less than 0.2 dB.

Although his methods are not really applicable to a production line, the Swedish radio ham Leif A***rink (SM5BSZ) has done a lot of work on measuring low values of noise figure with small uncertainties. 

http://www.sm5bsz.com/lir/nfprec/nfprec.htm
(but make sure to read the pages he links first) 

BTW, does anyone know why the term "accuracy" is used in so much in specifications for measurement instruments? If a DVM has an accuracy of 0.1%, does it mean it is 99.9% inaccurate!  

Dave

> {quote：title = drkirkby写道：} {quote}> BTW，有没有人知道为什么术语“准确度”在测量仪器的规格中被如此多地使用？
如果DVM的准确度为0.1％，是否意味着99.9％不准确！
>> Dave事实上，除了白话之外，我们在描述规范时尽量不要使用“准确性”这个词。
我们使用“不确定性”来定义测量的准确性。
因此，准确性是一种属性，类似于美，而数字方面的“不确定性”就是衡量标准，类似于“漂亮，非常漂亮，有点普通，不幸”。此外，我们经常以统计的方式提及不确定性，
例如“该测量的不确定度，95％置信度，为0.02 dB” - 观察点：如果不确定性包含真随机噪声，并且我的意思是噪声可以取任何值，那么就不能真正设置
任何绝对限制，因为嘈杂的信号可以承担任何值，包括非常接近无限，即使只是远程可能。
因此，如果您询问“绝对最大误差是多少”并且误差分析包含噪声，那么每个答案必须是“它接近无限”。



     以下为原文

  > {quote:title=drkirkby wrote:}{quote}
> BTW, does anyone know why the term "accuracy" is used in so much in specifications for measurement instruments? If a DVM has an accuracy of 0.1%, does it mean it is 99.9% inaccurate!  
> 
> Dave

In fact, except in the vernacular, we try to never use the word "accuracy" when describing specifications.  We use the "uncertainty" to define the accuracy of a measurement.

Thus, Accuracy is the attribute, akin to Beauty, and "uncertainty" in terms of numbers is the measure, akin to "pretty, very pretty, a little plain, and unfortunate"

Further, we often refer to uncertainty in a statiscial way, such as "the uncertainty of that measurement, with 95% confidence, is 0.02 dB"

--

observational point: If an uncertainty contains true-random noise, and by that I mean that noise can take on any value, then one cannot really set any absolute limits, because a noisy signal can take on any value including quite near infinite, even if only remotely possible.  Thus if you ask "what is the absolute maximum error" and the error analysis includes noise, then every answer must be "it approaches infinite".

要迁移，sourceId：118650



     以下为原文

  to be migrated, sourceId: 118650

> {quote：title = drkirkby写道：} {quote}>>事实上，除了白话之外，我们在描述规范时尽量不要使用“准确性”这个词。
我们使用“不确定性”来定义测量的准确性。
>> Keysight可能会*在描述规范时尝试*永远不要使用“准确度”这个词，但它们在这方面不会成功。
正是我说“我们 - 从不使用”而不是“我们从不使用”的原因....为星星拍摄，如果你没有到达它们，你仍然可能登陆月球。
但是，在一个非常大的公司中，有许多不同的部门和产品，这些部门和产品是在几十年内开发出来的，并且在不同的行业中使用，通常的用法可能会有所不同。
只要含义清楚，我就没有看到任何问题。
> {quote：title = drkirkby写道：} {quote}>我知道你的意思，但只是为了非常挑剔，我可以说，鉴于大多数乐器本质上是数字的，它们的显示器可以显示的是有限的。
回到3458A，只有这么多的电压可以显示8.5位数。
>> Dave不要将准确性与分辨率混淆。
我们的分辨率通常比我们的准确度更高，而VNA就是一个很好的例子。
我们的分辨率为0.0001 dB，但不确定度为0.02。
另一方面，很难想象我们可能具有比分辨率更高的准确度的情况。
在DVM的情况下，准确度似乎被认为是阅读和范围的重复。
而且我猜分辨率也随着范围而变化。



     以下为原文

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

> > In fact, except in the vernacular, we try to never use the word "accuracy" when describing specifications.  We use the "uncertainty" to define the accuracy of a measurement.
> 
> Keysight might *try* never to use the word "accuracy" when describing specifications, but they don't succeed too well at it. 

Exactly the reason I said "we -try- to never use" rather than "we never use"....  Shoot for the stars, and if you don't reach them you still might land on the moon.  But in a very large company, with a lot of different divisions and products that were developed over many decades, and used in different industries, the common usage can be different.  I don't see any issue as long as the meaning is clear.

> {quote:title=drkirkby wrote:}{quote}
> I know what you mean, but just to be very picky, I could say that given most instruments are digital in nature, there is a limit to what their display can show. Going back to the 3458A, there is only so many volts you can display with 8.5 digits. 
> 
> Dave
Don't confuse Accuracy with Resolution.  We often have better resolution than we have accuracy, and VNA's are an excellent example.  We have resolution of 0.0001 dB but uncertainty of 0.02.  On the other hand, it's hard to think of a case where we might have better accuracy than resolution.  In the DVM case, accuracy seems to be spec'd as a perecentage of reading and range.  And I guess the resolution also changes with range.