校准后S21纹波无法消除

0.05 dB纹波实际上非常小，校准过程中电缆和/或电缆的移动很容易成为问题的根源。
您使用的是高质量的相稳定电缆吗？
您是否将它们固定在工作台上以最大程度地减少校准步骤期间以及之后连接到DUT时的移动？
这些快速噪音般的涟漪还是类似于驻波？
消除噪声是罪魁祸首的可能性，在校准期间打开10-20扫描平均值和/或降低IFBW。



     以下为原文

  0.05 dB ripple is actually very small and bad cable and/or cable movement during calibration can easily be the source of the problem.  Are you using high quality phase stable cables?  are you securing them to the work surface to minimize movement during calibration steps and afterwards when you connect to the DUT?  Are these fast noise-like ripples or do they resemble a standing wave?  to eliminate the possibility that noise is the culprit, turn on 10-20 sweep averages and/or reduce the IFBW during the calibration.

我们使用的是测试级Mega Phase电缆，连接校准标准之间似乎没有超过6英寸的移动。
校准期间和DUT的测量范围在3英寸直径范围内。
IFBW在1kHz时非常低，但我们可以尝试进一步降低它。
由于纹波在位置上相当稳定，所以平均似乎没有多大意义。
波纹看起来不像驻波，它更像锯齿。



     以下为原文

  We are using test grade Mega Phase cables and there doesn't seem to be more than 6 inches of movement between connecting cal standards.  The measurement during calibration and of the DUT are within a 3in diameter area.  The IFBW is quite low at 1kHz, but we can try lowering it further.  Averaging doesn't seem to do much which makes sense since the ripple is quite stable in position.  The ripple doesn't look like a standing wave, it's more saw tooth.

我可以做的另一个建议是使用PNA的引导校准和执行和未知直通校准。
这将消除适配器移除所需的所有额外步骤。
您基本上会在每个端口上执行OSL，然后将两根电缆连接到任何直通适配器（只要它的损耗小于40 dB）。
如果您有一个2端口PNA-L型号，那么您可以直接执行此操作，但如果您有4端口型号，则必须执行宽带第1层校准，我们称之为“全球三角洲匹配”（更多信息_
*这里*_ ）。



     以下为原文

  Another suggestion I can make is to use the PNA's Guided cal and perform and Unknown Thru calibration.  this will eliminate all the additional steps needed for the adapter removal.  you'll essentially do a OSL on each port and then connect the two cables with any thru adapter (as long as it has less than 40 dB of loss).  If you have a 2-port PNA-L model then you can do this directly, but if you have 4-port model you'll have to perform a broadband 1st tier calibration that we call the "Global Delta Match" (more info _*HERE*_ ).

> {quote：title = daras写道：} {quote}> 0.05 dB纹波实际上非常小，校准过程中电缆和/或电缆的移动很容易成为问题的根源。
Em，有趣的daras，你说+/- 0.05 dB很小，因为Joel表示我的S21纹波在+/- 0.02 dB处于良好的极限。
这看起来是否过度波动？
什么值得注意的？
我使用的设备比Dave_11要差一些，后者有PNA。
我有* 8720D VNA * 85032B N校准套件* Agilent 85131F VNA电缆，带3.5 mm到N的适配器。
（但我只工作到6 GHz，而不是8.5到11.5 GHz）使用我的8720D VNA，85052B 3.5 mm校准套件和Agilent 85131F VNA电缆，我似乎无法获得比+/- 0.02 dB更好的效果，
我试图尽可能保持。
事实上，0.02 dB似乎是一个好日子，但我通常可以做到+/- 0.05 dB。
我希望我能够更好地使用85052B中的滑动负载。
当我使用该套件作为光荣的85052D，这是原始海报（Dave_11）。
*（我很乐意看到安捷伦或Maury Microwave制作一些关于使用滑动载荷和验证套件的视频）*。
也许有滑动负载我可以做得更好。
戴夫



     以下为原文

  > {quote:title=daras wrote:}{quote}
> 0.05 dB ripple is actually very small and bad cable and/or cable movement during calibration can easily be the source of the problem. 

Em, 
interesting daras that you say +/- 0.05 dB is small, since Joel indicated my S21 ripple at +/- 0.02 dB was on the limits of being good. 

Does this look to be excessive ripple? What caues it? 

I was using somewhat more antiquated equipment than Dave_11, who has a PNA. I had

* 8720D VNA
* 85032B N calibration kit
* Agilent 85131F VNA cables with 3.5 mm to N adapters. 

(But I was only working up to 6 GHz, not 8.5 to 11.5 GHz)

I can't ever seem to get any better than about +/- 0.02 dB with my 8720D VNA, 85052B 3.5 mm cal kit and Agilent 85131F VNA cables, which I try to keep as still as possible. In fact, 0.02 dB seems a good day, but I can usually do better than +/- 0.05 dB. 

I wish I was more confident of being able to use the sliding loads in my 85052B properly. At the minute I'm using the kit as a glorified 85052D, which is what the original poster (Dave_11) has. *(I would love to see Agilent or Maury Microwave produce some videos about using sliding loads and verification kits)*. Maybe with sliding loads I could do better. 

Dave

我已经看到使用未知通过cal引起的纹波，并且发现适配器移除cal可以提供更好的结果。
因此，我认为，消除额外的步骤可以降低因电缆移动而增加误差的风险，但添加一种似乎产生更多纹波的校准方法似乎是一种清洗。
老实说，我不了解全球校准或我为什么需要它。
不会有更多的校准步骤，未知的直通/全局校准与单个适配器移除校准？
我会尝试更换电缆或夹紧电缆。



     以下为原文

  I have seen ripple caused by using the unknown thru cal and have found that the adapter removal cal provides better results.  So eliminating extra steps to, I assume, reduce risk of adding error due to cable movement, but adding a calibration method that seems to produce more ripple seems a wash.  Honestly, I don't understand the global cal or why I would need it.  Wont there be more calibration steps with the unknown thru/global cal verses a single adapter removal cal?  I'll try replacing the cables or clamping them.

> {quote：title = Dave_11写道：} {quote}>我已经看到使用未知通过cal引起的涟漪，并且发现适配器移除cal提供了更好的结果。
因此，我认为，消除额外的步骤可以降低因电缆移动而增加误差的风险，但添加一种似乎产生更多纹波的校准方法似乎是一种清洗。
老实说，我不了解全球校准或我为什么需要它。
不会有更多的校准步骤，未知的直通/全局校准与单个适配器移除校准？
我会尝试更换电缆或夹紧电缆。
如果你正在使用良好的电缆并坚持所有良好的计量实践，那么我认为未知的通过cal不会给你带来太大的改进，虽然我没有亲自观察到这一点（我使用ECAL几乎所有的测量，
如果你已经看到了未知通过的额外波纹，那么你不应该使用它。
然而，在我在非计量设置中进行的大多数客户交互中，由于校准步骤的减少，切换到未知通过导致了很大的改进。
至于执行全局增量匹配的需要，它仅限于某些PNA-L型号，这些型号没有足够的接收器用于完整的双反射计配置，这是未知通道所需的。
在这些情况下，GDM校准是一次宽带校准，在测试端口（无电缆）完成，并补偿了双反射计的缺乏，通过这样做，您仍然可以简化日常的s参数校准。



     以下为原文

  > {quote:title=Dave_11 wrote:}{quote}
> I have seen ripple caused by using the unknown thru cal and have found that the adapter removal cal provides better results.  So eliminating extra steps to, I assume, reduce risk of adding error due to cable movement, but adding a calibration method that seems to produce more ripple seems a wash.  Honestly, I don't understand the global cal or why I would need it.  Wont there be more calibration steps with the unknown thru/global cal verses a single adapter removal cal?  I'll try replacing the cables or clamping them.

If you are using good cables and adhering to all the good metrology practices, then I suppose the unknown thru cal is not going to give you much improvement, and while I haven't personally observed this (I use ECals for almost all the measurements that I make), if you have seen additional ripple with the unknown thru, then you shouldn't use it.  However, in the majority of the customer interactions I've had in non-metrology settings, switching to unknown thru has resulted in great improvement due to the reduction of calibration steps.

As for the need to perform the global delta match, it is limited to certain PNA-L models that don't have enough receivers for a full dual reflectometer configuration, which is required for an unknown thru cal.  in those cases the GDM cal is a one time, broadband calibration that is done at the testports (no cables) and compensates for the lack of dual reflectometer and by doing that you still simplify the day-to-day s-parameter calibrations.

> {quote：title = drkirkby写道：} {quote}> Em，>有趣的daras，你说+/- 0.05 dB很小，因为Joel表示我的S21纹波在+/- 0.02 dB处于良好的极限。
>实际上，我认为joel所暗示的是，考虑到系统剩余误差，你不能做到+/- 0.02 dB，但总是做得最差。
当我们计算残余误差时，我们不考虑用户的电缆和其他不利的环境条件，我们还假设所使用的任何校准标准都处于最佳性能。



     以下为原文

  > {quote:title=drkirkby wrote:}{quote}
> Em, 
> interesting daras that you say +/- 0.05 dB is small, since Joel indicated my S21 ripple at +/- 0.02 dB was on the limits of being good. 
> 

Actually, I think what joel was suggesting is that given the system residual errors, you could not do better than +/- 0.02 dB, but you can always do worst.  When we do computations of residual errors, we don't take into account users' cables and other adverse environmental conditions and we also assume any calibration standards being used are at their optimal performance.

> {quote：title = daras写道：} {quote} >> {quote：title = drkirkby写道：} {quote} >> Em >>>有趣的daras，你说+/- 0.05 dB很小，因为Joel表示我的
在+/- 0.02 dB处的S21纹波处于良好的极限。
实际上，我认为joel建议的是，考虑到系统残留误差，你不能做到+/- 0.02 dB，但总是做得最差。
啊，我显然完全误解了乔尔。
你让我感觉好多了daras，在看到英格兰队在世界杯上输给乌拉圭之后，我需要这些比赛！
如果我的涟漪略高于+/- 0.02 dB，我就不会太担心，这就是我在“美好的一天”所获得的。
正如你所说，人们总是会变得最糟糕。
戴夫



     以下为原文

  > {quote:title=daras wrote:}{quote}
> > {quote:title=drkirkby wrote:}{quote}
> > Em, 
> > interesting daras that you say +/- 0.05 dB is small, since Joel indicated my S21 ripple at +/- 0.02 dB was on the limits of being good. 
> > 
> 
> Actually, I think what joel was suggesting is that given the system residual errors, you could not do better than +/- 0.02 dB, but you can always do worst.  

Ah, I obviously totally misunderstood Joel. You make me feel a lot better daras, which I need after just watching England lose 2-1 to Uruguay in the World cup! 

I wont worry quite as much if my ripples are a bit more than +/- 0.02 dB, which is what I get on a "good day". As you say, one can always get worst. 

Dave

我同情;
这是一个艰难的损失。



     以下为原文

  I sympathize; that was a tough loss.

我一直在使用安捷伦不确定度计算器试图确定我应该看到的最佳值，而且我必须做错了，因为它的传输速度约为0.3dB。
我想测量0.25dB。
使用N5230C，245，带有合理匹配的适配器移除校准和85052D套件。
甚至没有考虑到电缆挠曲。



     以下为原文

  I've been using the Agilent Uncertainty Calculator trying to determine the best I should expect to see and I must be doing something wrong because it's about 0.3dB in transmission.  I'm trying to measure 0.25dB. Using N5230C, 245 with an adapter removal calibration with reasonable matches and an 85052D kit.  Not even accounting for cable flexture.

事实上，你可能没有做错任何事。
大多数人低估了与测量相关的误差，并且几乎总是规格不确定性比预期更糟。
使用无珠航空公司的非常好的校准方法是0.047更糟糕的情况，使用85052C短线TRL校准（在本例中在20 GHz的PNA-X上）。
RSS是0.029。
如果您在PNA上使用我们的计量选项，您可以将其降低到约0.02 dB。
对于PNA（非PNA-L）上的85052D套件，最坏的情况是0.192;
我可以想象PNA-L会有点差，所以0.3 dB的不确定性并非不合理。
问题当然是负载，没有滑动负载套件或TRL套件，不确定性不会低于0.1 dB。
如果您的套件完全磨损，您的负载可能会降低并且会损害您的测量结果。
0.05 dB对您的系统来说非常有用，实际上只比上面显示的TRL性能略差。
这是“你得到你付出的，而不是你所希望的”的一个例子;
事实上，你比预期的最坏情况不确定性要好6倍，根据我的经验，不确定性计算器给出的结果比你在实践中看到的结果差2x-3x（记住，它是基于最坏情况的负载，
最坏情况源匹配，最坏情况负载匹配，所有频率相同）。
在实践中，这些从不以相同的频率排列，这就是我们RSS一般的原因。
如果你真的想用0.02测量0.25，你可能需要一个更好的盒子。



     以下为原文

  in fact, you are probably not doing anything wrong.  Most people underestimate the errors associated with a measurement, and almost always the spec uncertainty is worse that expected.

The VERY BEST calibration methods, using beadless airlines, is 0.047 worse case, using an 85052C short-line TRL cal (on a PNA-X at 20 GHz, in this example).  RSS is 0.029.  If you use our metrology option on the PNA you can push that down to about 0.02 dB. 

For an 85052D kit on an PNA (not PNA-L), the worst case is 0.192; I can imagine that the PNA-L would be a little worse so 0.3 dB uncertainty is not unreasonable.

The problem of course is the load, and without a sliding load kit or TRL kit, the uncertainty won't drop below 0.1 dB.

If you kit is worn at all, your load can be degraded and it will hurt your measurements.  0.05 dB is actaully pretty good for your system and infact only slightly worse than the TRL performance shown above.

This is an example of "you get what you pay for, not what you hope for"; in fact you are getting 6 times better than the expected worst case uncertainty, and in my experience, the uncertainty calculator gives a result that is about 2x-3x worse than what you see in practice (remember, it is based on worst case load, worst case source match, worst case load match, all at the same frequency).   In practice, these never line up at the same frequency so that is why we RSS in general.

If you are really trying to measure 0.25 with 0.02, you might need a better box.

在向任何电子设备论坛发布有关“如何”问题的消息时，将有助于提供型号和软件修订。
我在Agilent E5071C软件版本11.31上看到过与此类似的内容，但我在跟踪上显示S21相位，我注意到如果标记相位值下降到小于1度，则显示的标记值将以毫秒显示。
显示dB时不会发生这种情况，因为dB的正常格式始终为0.0000或10.000或100.00。
在这种情况下，提供显示的比例值可能会很好。



     以下为原文

  When posting messages to any electronic equipment forum concerning "how to" questions it would help to give the Model Number and Software Revision. I have seen something similar to this on an Agilent E5071C Software Revision 11.31 but I was displaying S21 Phase on the trace and I noticed that if the Marker Phase value dropped to less that 1 Degree the displayed Marker Value would be displayed in millidegrees. This does not happen when displaying dB in that the normal format for dB is always 0.0000 or 10.000 or 100.00. It would probably be good in this instance to also provide the displayed Scale Value.