幅度/相位瞬态测量-高速/高精度

大家好，这可能不是最好的论坛，但我开始在网络分析器应用论坛没有回复，所以我在这里转发。
+ _Task_ +我的任务是将数字衰减器的开关时间测量到0.1 dB以内，最终值为1度，预期时间为30 ns。
频率为1至3 GHz。
衰减器是2位6 dB步长（总共18 dB）。
在1GHz时，相位差小于1度，而在3GHz时，相位差大约为5度/步。
PNA-X我对此进行了研究，但即使在CW模式下IF BW全开，我也无法获得时间分辨率。
这将是完美的，因为我可以进行响应校准并同时获得幅度和相位。
+幅度+ - 功率传感器_（U2021XA）不好，因为它们没有时间分辨率。
_Crystal Detector_看到大约55 mV，18 dB，或者我需要一个0.03 mV步长的示波器。
不幸的是，我可以访问的示波器已经使用了10年（非安捷伦），并且产生的误差是误差项的10倍。
_Spectrum Analyzer_（非安捷伦）具有50 MHz的分辨率BW，但时机已经过时了。
_Oscilloscope_我不得不直接使用示波器，因为它是数字的并且输入为50欧姆。
数据以10GS / s进行采样。
在参考状态（0 dB）下，我最大化了屏幕上的电压。
它没有包络功能，所以我只使用最大功能。
我也正在对最大数据进行400点运行平均以“平滑”它。
这似乎是合理的，客户很高兴，但我不是因为我通常会在平滑数据上看到0.1 dB的纹波。
当我处于18 dB状态时，情况更糟。
+阶段+非常有限。
我把一个鉴别器放在一起。
我把源分开了。
一条路径有线路担架，进入双平衡混合器的LO端口。
另一个端口进入可变衰减器，被测器件和混频器的RF端口。
IF端口具有低通滤波器，在感兴趣的频率下具有至少80dB的抑制。
在这里，我一直在寻找各种各样的小问题。
我正在研究从Holzworth Instrument获得一个相位检测器，看看这是否有助于解决我的一些问题。
一直在考虑VSA和新的Infiniium S系列（10位A / D）。
（想法？）另外，考虑直接使用示波器并取出混频器。
取两个信号，直接进入示波器。
调整线条担架以获得相同的过零点。
由于频率已知，我可以计算相位偏移（时间）。
振幅也可以。
所以我正在寻找任何输入，以获得高精度的价值和时间。
在此先感谢您的任何建议。
Michael Virostko集成设备技术



     以下为原文

  Hello All,

This might not be the best forum, but I started in the network analyzer application forum with no replies so I am reposting here.

+_Task_+

I have been task with measuring the switching time of a digital attenuator to within 0.1 dB and 1 degree of the final value and the expected time is 30 ns. The frequencies are 1 to 3 GHz. The attenuator is a 2 bit 6 dB step (18 dB total). At 1 GHz the phase delta is less than 1 degree and at 3 GHz it is about 5 degrees / step.

PNA-X
I looked into this, but even with the IF BW wide open in CW mode, I can not get the time resolution. This would be perfect because I could do a response calibration and get both amplitude and phase simultaneously.  

+Amplitude+

-Power Sensor_ (U2021XA) are not good since they do not the time resolution.
_Crystal Detector_ saw about 55 mV for 18 dB or I would need a step of 0.03 mV step for the oscilloscope. Unfortunately the oscilloscope I have access to is over 10 years old (non Agilent) and yields 10 times that for a error term.
_Spectrum Analyzer_ (Non Agilent) had 50 MHz resolution BW but the timing was way off.
_Oscilloscope_ I had to settle on using the oscilloscope directly since it was digital and had a 50 ohm input. Data was sampled at 10 GS/s. In the reference state (0 dB) I maximized the voltage on the screen. It did not have a envelope function, so I just use the maximum function. I also am doing a 400 point running averaging on the maximum data to "smooth" it. It seems reasonable and the customer is happy, but I am not since I typically see 0.1 dB ripple on the smoothed data.  It is even worse when I am in the 18 dB state.

+Phase+
Very limited here. I put together a discriminator. I split the source. One path has the line stretcher and goes into the LO port of a double balance mixer. The other port goes into a variable attenuator, the device under test and into the RF port of the mixer. The IF port has a low pass filter with at least 80 dB of reject at the frequency of interest. Here I have been finding all sorts of little issues that are much to long to go into. I am looking into a getting a phase detector from Holzworth Instrument to see if this helps with some of my issues.

Been thinking about VSA and the new Infiniium S-Series (10 bit A/D). (Ideas?)

Also, thinking of using a oscilloscope directly and taking the mixer out.  Take the two signals and go directly into the oscilloscope.  Adjust the line stretcher for identical zero crossing. Since the frequency is known I could calculated the phase offset (time).  Amplitude can be done also.

So I am looking for any inputs to get the high accuracy in value and timing.

Thanks for any advice in advance.

Michael Virostko
Integrated Device Technology