通过VNA进行频谱分析-平均？

S21是比率测量值（B / R1）。
根据定义，任何s参数都是比率测量。
对于你想要做的事情，你需要将参数更改为B.现在872X比我的时间早一点，所以我不知道更改参数的确切前面板按键顺序，但我知道那个框
允许您进行非比率测量。
还有几件事需要考虑：1-如果你不想看漂移，请将两个盒子的10MHz参考锁定在一起。
2- a VNA没有像频谱分析仪那样的无图像前端。
当你扫描B接收器时，你不仅会看到主信号，它的谐波和其他可能的杂散，而且你也会看到所有这些信号的图像。
根据您要完成的任务，可能会也可能不会。
较新的VNA，如Keysight PNA-X（N52xx系列），可以选择进行软件镜像抑制的频谱模式，并为SA提供基本相似的结果。
3-平均将有助于切换到非比率B测量，因为您已在另一篇文章中阅读过。
实际上，对于您尝试使用这种旧硬件而言，最好使用更宽的IFBW，然后使用更多的平均值来提高SNR。
因为另一个问题是传统VAN中使用的带宽滤波器不是高斯滤波器，并且如果您没有精确调整到感兴趣的信号，则会有大的旁瓣可能会使功率读数失真。



     以下为原文

  S21 is a ratio measurement (B/R1).  any s-parameter by definition is a ratio measurement.  for what you are trying to do, you will need to change the parameter to B.  Now the 872X is a little before my time, so I don't know the exact front panel key sequence for changing a parameter, but I know that box does allow you to make non-ratio measurements.

A couple of more things to consider:

1- if you don't want to see drift, lock the 10MHz references of the two boxes together.
2- a VNA does not have an image-free front end like a spectrum analyzer.  when you sweep the B receiver, you will not only see the main signal, its harmonics and other possible spurs, but you will also see the images of all of these signals as well.  Depending on what you are trying to accomplish, that may or may not be ok.  Newer VNAs, like the Keysight PNA-X (N52xx series) do have an option for a spectrum mode that does software image rejection and will give essentially similar results to an SA.
3- Averaging will help when you switch to a non-ratio B measurement as you have already read in the other post.  in fact for what you are trying to do with this older hardware, you will be better off using a wider IFBW and then using more averaging to improve SNR. because the other problem is that the bandwidth filters used in a traditional VAN are not gaussian and have large sidelobes that may distort the power reading if you are not tuned exactly to the signal of interest.

非常感谢！
我看到S参数是比率测量，复杂平均（实部和虚部）是在这种情况下降低本底噪声的正确方法。
对于非比例信号，另一篇文章建议平均磁场/相位。
在我的情况下，这似乎更好。
只是为了更好地理解这一点，是因为非比率测量中的扫描不是连贯的（不是锁相的）？
在比率测量中，多次扫描是相干的，因此平均可以将SNR提高n ^ 2？
实际上，我正在通过GPIB收集每次扫描的所有实/虚数据，并通过PC运行后处理，例如平均。
如果我测量S21并将端口1设置为0dBm，那么这会给出与非比率测量相同的答案（仅限B）吗？
10 MHz参考是一个很好的建议。
后面板连接器上的手册有以下两种：1）外部参考频率输入（EXT REF INPUT）2）高稳定性频率参考输出（选项1D5）有哪些建议适合使用？
我该如何连接它们？



     以下为原文

  thank you so much!

I see that S-parameters are ratio measurements, and the complex averaging (real and imaginary parts) is the proper way to reduce noise floor in that case. For non-ratioed signal, the other post suggests averaging mag/phase. That does seem to work better in my case. Just to understand this better, is it because the sweeps in non-ratio measurements are NOT coherent (not phase locked)? While in ratio measurements, multiple sweeps are coherent so averaging can improve SNR by n^2? 

In fact, I am collecting all the real/imaginary data for each sweep through GPIB and run post processing, such as averaging, by a PC. If I measure S21 and set port 1 as 0dBm, would that give me the same answer as a non-ratio measurement (B only)?

The 10 MHz reference is a good suggestion. The manual on rear panel connectors have the following two:
1) external reference frequency input (EXT REF INPUT)
2) high-stability frequency reference output (Option 1D5)
Any suggestions on which is the proper one to use? How should I connect them?

至于你的测量，如果你试图测量某个独立光源产生的信号，为什么还要坚持测量S21，这是一个通过DUT测量“传输”的参数。
在您的情况下平均不起作用，因为您正在进行B / R1的比率测量，其中在B上测量的信号来自与在R1上测量的信号不同的源。
这两个源具有完全不同的非相干相位，因此，您基本上随机化了矢量B / R1，当您平均一系列随机矢量时，您最终将幅度平均到测量系统的本底噪声，甚至
如果你正在测量一个真实的信号。
> {quote：title = meta5718写道：} {quote}>实际上，我正在通过GPIB收集每次扫描的所有实/虚数据，并通过PC运行后处理，例如平均。
如果我测量S21并将端口1设置为0dBm，那么这会给出与非比率测量相同的答案（仅限B）吗？
>不 - 它没有。
只有你只是获取幅度数据，并且如果你的跨度中的所有频率的R1幅度完全为0 dBm（它不会是），那么这只会起作用。
因为两个接收器（R1和B）正在测量两个不同的非相干源，所以不能使用矢量数据（实/虚），但为什么要使用S21开始？
进行此测量的唯一正确方法是仅测量B接收器。编辑：daras于2014年8月26日上午9:55



     以下为原文

  as for your measurement, if you are trying to measure a signal generated by some independent source, why are you insisting on measuring S21, which is a parameter designed to measure "Transmission" thru a DUT.  The averaging in your case does not work, because you are making a ratio measurement of B/R1, where the signal measured on B is from a different source than the signal measured on R1.  these two sources have completely different and non-coherent phases, and therefore, you are essentially randomizing the vector B/R1 and when you average a series of random vectors, you will eventually average the magnitude into the noise floor of the measurement system, even if you are measuring a real signal.

> {quote:title=meta5718 wrote:}{quote}
> In fact, I am collecting all the real/imaginary data for each sweep through GPIB and run post processing, such as averaging, by a PC. If I measure S21 and set port 1 as 0dBm, would that give me the same answer as a non-ratio measurement (B only)?
> 

NO - it does not.  that would only work if you were just taking the magnitude data and if the R1 magnitude is perfectly 0 dBm at all frequencies in your span (which it won't be).  Because the two receivers (R1 and B) are measuring two different non-coherent sources, you cannot use vector data (real/imaginary), but again why use an S21 to begin with??


The only correct way of making this measurement is to measure just the B receiver.

Edited by: daras on Aug 26, 2014 9:55 AM

再次感谢您的投入！
我同意仅关于测量B接收器的评论。
猜猜我试图解决它，因为我从未做过接收器测量。
您能否确认对于非比率（仅B接收器）测量，正确的平均方法应该是mag / phase而不是real / imag？
或者至少那是VNA在内部如何做到的？



     以下为原文

  Thanks again for the input! 
I agree with the comments on measuring B receiver only. Guess I was trying to work around it as I have never done receiver only measurements. 

Could you please confirm that for non-ratio (B receiver only) measurement, the correct averaging method should be on mag/phase rather than real/imag? Or at least that's how VNA would do it internally?

新的VNA在无比率测量中进行磁场/相位平均。
我无法确定8720是如何做到的，因为就像我提到的那样，这是在我的业务时间之前。
如果您在PC中执行所有这些作为后期处理，那么您甚至不需要从VNA读取实/虚数据。
您正在测量的内容中没有有用的相位信息，因此您也要注意，只需阅读LogMag或LinMag格式的数据，然后在程序中对其进行平均。
但是如果你想继续阅读实/虚对，那么只需计算幅度，平均值并忘记相位。



     以下为原文

  the new VNA's do mag/phase averaging on none-ratio measurements.  I can't say for certain how the 8720 does it, since like I mentioned, it was before my time in the business.

if you are doing all this as a post process in the PC, then you really don't even need to read real/imaginary data from the VNA.  what you are measuring has no useful phase information in it, so you mind as well, just read the LogMag or LinMag formatted data and just average that in your program.  but if you want to keep reading the real/imaginary pairs, then just compute the magnitude, average and forget about the phase.

这完全有道理。
谢谢！



     以下为原文

  That totally makes sense. Thanks!