点与扫描平均的相对优势/劣势是什么？

为降低噪音，您首选的应该是减少IFBW。
这与点平均非常相似。
但是，在某些情况下，IFBW减少将无效。
例如，当脉冲宽度太窄而无法适应较低的IFBW时，请考虑脉冲情况。
在这些情况下，您需要进行平均。
在两种平均类型中，您应首先使用点平均值。
它通常要快得多，尤其是当您的IFBW很宽时的脉冲情况。
那是因为你避免了RF和LO源的大部分移动。
通过点平均和宽带宽可以轻松实现10,000平均值（100kHz带宽，平均值为10k，每点采集时间为100ms）。
但是，扫描平均值的10,000平均值将花费不切实际的时间。
尽管如此，扫描平均还有几个优点。
首先，您可以看到更新以递增方式发生。
其次，扫描时间是一致的。
因此，如果您的测量具有一些受扫描时间长度影响的扫描动态，那么您将需要使用扫描平均，以便您可以保持测量的扫描时间一致。
SMC + Phase测量各个接收器的原始相位，并且随着时间的推移会逐渐漂移，因此您只应使用SMC + Phase进行扫描平均。
最后，IFBW减少和平均之间还有一个有趣的区别。
IFBW减少是在各个接收器上进行的，而平均值是在SParameters进行了比例后进行的。



     以下为原文

  For noise reduction, your first preference should be to just reduce the IFBW. This is very similar to point averaging.

However, there are some cases where the IFBW reduction won't work. For example, consider a pulsed scenario when the pulse width is too narrow to fit in a lower IFBW.  In those cases, you would need averaging.

Of the 2 types of averaging, you should first prefer to use point averaging. It is usually much faster, especially for that pulsed scenario when your IFBW is wide. That's because you avoid most of the moving of the RF and LO source. You can easily do 10,000 averages with point averaging and a wide bandwidth (100kHz BW with 10k averages is 100ms acquisition time per point). But 10,000 averages with sweep averaging would take an impractical amount of time.

Still, sweep averaging has a couple of advantages. For one, you can see the updates happen incrementally. Secondly, the sweep time is consistent. So, if you have a measurement that has some sweep dynamics that are affected by the length of time of the sweep, then you will want to use sweep averaging so that you can keep the sweep time consistent for the measurement. SMC+Phase measures raw phase on individual receivers and these drift a lot over time, so you should only use sweep averaging with SMC+Phase.

Lastly, there is one more interesting difference between IFBW reduction and averaging. IFBW reduction is performed on the individual receivers, whereas averaging is performed on the SParameters after they have been ratioed.

> {quote：title = johanericsson写道：} {quote}>为了减少噪音，你的第一个偏好应该是减少IFBW。
这与点平均非常相似。
为什么IF降噪更可取？
我只在8720D或8753ES VNA上使用这种方法，因为平均值似乎是一个无限脉冲响应滤波器，而且我不知道何时可以安全地考虑测量有效。
改变IF带宽似乎是最简单的方法，并且工作正常。
我从来没有发现任何需要低于最小IF带宽，我认为是10赫兹。
在我的HP 4291B射频阻抗/材料分析仪上，我认为没有办法改变IF带宽。
但是有扫描和点平均，我注意到它也支持在PNA上。
>但是，在某些情况下，IFBW减少将无效。
例如，当脉冲宽度太窄而无法适应较低的IFBW时，请考虑脉冲情况。
在这些情况下，您需要进行平均。
是的，我可以看到。
>在两种平均类型中，首先应该使用点平均值。
它通常要快得多，尤其是当您的IFBW很宽时的脉冲情况。
那是因为你避免了RF和LO源的大部分移动。
通过点平均和宽带宽可以轻松实现10,000平均值（100kHz带宽，平均值为10k，每点采集时间为100ms）。
但是，扫描平均值的10,000平均值将花费不切实际的时间。
经常让我想知道平均值比IF带宽减少更有用的一点是，扫描平均值可能对微小的温度变化不太敏感。
如果为真，这可能会建议扫描平均值优于点平均值，但我可以看到点平均值应该快得多，因为您根本不更改振荡器频率。
>尽管如此，扫描平均还有几个优点。
首先，您可以看到更新以递增方式发生。
其次，扫描时间是一致的。
因此，如果您的测量具有一些受扫描时间长度影响的扫描动态，那么您将需要使用扫描平均，以便您可以保持测量的扫描时间一致。
SMC + Phase测量各个接收器的原始相位，并且随着时间的推移会逐渐漂移，因此您只应使用SMC + Phase进行扫描平均。
关于温度变化的观点基本上是这样，尽管世界上最好的意愿，即使可以将室温保持在0.00000000000001摄氏度，仪器内的温度也会有微小的变化。
>最后，IFBW减少和平均之间还有一个有趣的区别。
IFBW减少是在各个接收器上进行的，而平均值是在SParameters进行了比例后进行的。
也许平均大量噪声S参数测量可能比仅从接收器中的一组干净（噪声较小）测量计算S参数一样糟糕。
我无法理解这一点，但我相信这两种方法可能导致不同的结果，这可能就是为什么你说降低IF带宽通常比平均更好。
感谢您的帮助。
戴夫



     以下为原文

  > {quote:title=johanericsson wrote:}{quote}
> For noise reduction, your first preference should be to just reduce the IFBW. This is very similar to point averaging.

Why is IF noise reduction preferable? 

I only use this method on my 8720D or 8753ES VNAs, since the averaging appears to be an infinite impulse response filter, and I'm never quite comfortable of knowing when its safe to consider a measurement valid. Changing the IF bandwidth seems the easy way out, and works fine. I've never found any need to go below the minimum IF bandwidth on that, which I think is 10 Hz. 

On my HP 4291B RF impedance/material analyzer I don't think there's any way to change the IF bandwidth. But there is both sweep and point averaging, which I noticed is supported on the PNA too. 

> However, there are some cases where the IFBW reduction won't work. For example, consider a pulsed scenario when the pulse width is too narrow to fit in a lower IFBW.  In those cases, you would need averaging.

Yes, I can see that. 

> Of the 2 types of averaging, you should first prefer to use point averaging. It is usually much faster, especially for that pulsed scenario when your IFBW is wide. That's because you avoid most of the moving of the RF and LO source. You can easily do 10,000 averages with point averaging and a wide bandwidth (100kHz BW with 10k averages is 100ms acquisition time per point). But 10,000 averages with sweep averaging would take an impractical amount of time.

Something that has often made me wonder if averaging is more useful than IF bandwidth reduction, is that sweep averaging is likely to be less sensitive to minute changes in temperature. If true, this would have suggested sweep averaging preferable to point averaging, but I can see that point averaging should be a lot faster, as you are not changing the oscillator frequencies at all. 

> Still, sweep averaging has a couple of advantages. For one, you can see the updates happen incrementally. Secondly, the sweep time is consistent. So, if you have a measurement that has some sweep dynamics that are affected by the length of time of the sweep, then you will want to use sweep averaging so that you can keep the sweep time consistent for the measurement. SMC+Phase measures raw phase on individual receivers and these drift a lot over time, so you should only use sweep averaging with SMC+Phase.

My point about temperature changes was basically this, as despite the best will in the world, even if one could keep the room temperature stable to 0.00000000000001 deg C, there are going to be minor changes of temperature inside an instrument. 

> Lastly, there is one more interesting difference between IFBW reduction and averaging. IFBW reduction is performed on the individual receivers, whereas averaging is performed on the SParameters after they have been ratioed.

Maybe averaging a lot of noisy S-parameter measurements might be worst than just computing the S-parameters once from a single set of clean (less noisy) measurements out of the receiver. I can't get my head around that, but I can believe those two methods might lead to different results, which maybe is why you say reducing the IF bandwidth is generally preferable to averaging. 

Thank you for your help. 

Dave