AnsweredAssumed Answered

Gating Function on 8720ES

Question asked by dhamilton on Oct 30, 2013
I am using my 6 foot cables with either the Log periodic antennas or two horn antennas. If I use the transform in broadband mode, (S21), I see a nice peak just where I expect it. At about 5.2 ns. If I set the gate center at 5.2 ns, and a width of (it does not matter here) of about 500 ps, then in the time domain I see the nice peak I expect to see. Now if I turn the gate on, and the transform off I get a frequency trace on the screen. Now this is where my problem arises. If I go into "display" and store that trace into memory, and then select data/memory, I should see a line at 0 db. Instead, I see a wiggly line that makes no sense at all. In fact it is tens of db off. Now I have no direct evidence that an 8720b does this aspect correctly or not, but I also have observed the following. If I have my set up running, and then press "cal", then hit response, then thru, I should see a nice line at 0 db. If the gate is off, I do. If the gate is on, I get garbage just like data/memory above. My friends with the 8720b's can run a response thru cal all day with the gate on and get the expected results. I am pulling out what little of my hair I have left over this.

 

Today in frustration, I tried anything and everything. I checked S11 the same way and got the same type of actions using either data/memory or response thru cal. However by chance I came upon a set of circumstances where I could get BOTH (normalization) situations to work as expected. Looking at S11, with the gate on with the display in data/memory, if I set the gate center at -5.2 ns!! +/-500 ps the gate would normalize to the expected results. If the gate center went outside that tolerance it would give garbage again. I also looked at S21, this time if the gate center was 0 s +/- 500 ps both response thou cal and data/memory would normalize. This makes no sense to me.

ran a 1 port cal (s11) with no cable connected to the left port of the VNA. A Smith chart with a load confirmed the calibration as being believable.

 

Then I hooked my six foot cable (open ended) to the port and looked at s11. I saw line at 0 db in log mag as expected.

 

I looked at the time domain response, and saw a small reflection at the cable port connection, and a much larger one from the end of the cable. I set the gate to be centered at the end reflection, but did not turn it on yet.

 

Then I turned off the transform.

 

I then went back to log mag display. I stored the data into memory. I then looked at "data/memory". I saw a straight line at 0 db as expected. 

 

When I turn the gate on, the coefficients are calculated, and now the display/memory is over -75db across the entire frequency band. If you turn off the gate the display renormalizes to 0 db. 

 

This non-normalization of the display with the gate on in "data/memory" but also seen in "cal", "response", "thru" (not checked with "open" for s11), is not what I expect to see. This is the root of my problem. I believe this problem is also what I see in the normalization peculiarities in the time domain as well. Am I missing something?  

Outcomes