Mike,
This is getting away from the topic even farther but a few people might
find it interesting. When doing a cal on the 8753 family of VNAs from
HP you have to pick a cal type (7mm, N female, n male, 3.5mm). The HP
cal standards are not perfect (at really high frequencies) and the opens
and shorts actually have different reference planes aside from being
offset from the connector plane. After calibration the analyzer
subtracts these offsets to get back to a reference plane that is at the
connector which HP talks about in several documents. In my case,
chapter 5 and particular page 5-9 of the 8753 analyzer book details the
issues clearly.
After doing a cal. and then measuring S11 of either the short or open
standard gives crummy results. This bothered my for a long time till an
HP rep explained the problem. However the 3.5mm standards have almost
exactly the same offset for both the open and short. By telling the
analyzer you are using 3.5mm standards and using them or discrete opens
and shorts and including 33ps of electrical delay after the cal fixes
the majority of the problems when trying to make impedance measurements
far away from 50+j0.
I may not have been super careful when I cal'd the analyzer and 0.5pF
seems about right from the numbers I posted. Everyone is correct about
1pF being deal at 30Mhz when trying to measure high impedances.
Does this make sense?
73, Larry, W0QE
Michael Tope wrote:
> Larry,
>
> If I did the math correct, your fixture (or the 0805 resistor) has
> about 0.5pf of stray C (based on the 30 MHz measurement). What are
> your calibration standards?
>
> 73, Mike W4EF..............
>
> Larry Benko wrote:
>> GM Jim,
>>
>> I certainly hope you or anyone else did not think that since I posted
>> the data to the ohm that any high level of accuracy was implied. If
>> accuracies were better than 10% I would be amazed BUT 10-20% is good
>> enough to tell materials 61, 77, 31, & 43 apart.
>>
>> I have read all the documents you have published this topic and on page
>> 47 of the PowerPoint slide you mention that the VNA doesn't seem to
>> cancel out the stray capacitance of the fixture so I decided to see how
>> my fixture would perform with a couple of resistors. I just measured
>> two 0805 SMT 1% resistors (1.00K & 10.0K) which were soldered across the
>> calibrated text fixture directly. I was doing an S11 measurement.
>>
>> 1.8MHz: 1006 & 10,210 ohms magnitude respectively
>> 30MHz: 984 & 7,100 ohms magnitude respectively (phase for 1k was
>> slightly inductive and for 10k was moderately capacitive)
>>
>> I guess I am not seeing the magnitude of the effect you saw.
>>
>> I have measured many common mode chokes over the years both by doing S11
>> measurements and S21 measurements (in different calibrated fixtures of
>> course). The biggest problem I have seen for chokes of higher common
>> mode impedance is repeatability with the winding placement. Once a
>> choke was picked for a particular use occasional QA checks were done on
>> incoming chokes to prove suppliers were using the correct material.
>>
>> 73, Larry W0QE
>>
>>
>> Jim Brown wrote:
>>
>>> On Sun, 11 Jan 2009 21:51:58 -0700, Larry Benko wrote:
>>>
>>>
>>>
>>>> Ok, you made me fire up the network analyzer (HP8753B).
>>>>
>>>>
>>> Before you place too much confidence in data from your network
>>> analyzer, study my measurement travails, documented in the Coaxial
>>> Chokes Power Point. Ferrite chokes are VERY difficult to measure
>>> accurately, because of the values of impedance you're trying to
>>> measure.
>>>
>>> Here are two elements of the problem: 1) the HF equivalent circuit
>>> is a parallel equivalent circuit with capacitance on the order of
>>> 0.5pF to 4 pF, well within the range of stray capacitance in most
>>> analyzer setups. Yes, some analyzer software attempts to subtract
>>> it out, but most are not very good at it when the unknown DUT is
>>> in this range. 2) The impedance at resonance is typically 500 -
>>> 5,000 ohms. This is far outside the range of unknown impedance
>>> that can be measured with any accuracy by a reflection-based
>>> measurement (that is, S11). That's because the equations that
>>> compute the unknown Z are differences of quantities that are very
>>> nearly equal, so a small difference in any quantity makes a very
>>> large difference in the value of the unknown.
>>>
>>> What you CAN do and get good data is to make measurements of S21
>>> with the unknown Z in the series leg. My Power Point shows how
>>> I've done that, and my data for coaxial chokes was obtained using
>>> that technique.
>>>
>>> But don't trust my measurements of these parameters. Do the curve-
>>> fitting that I've illustrated to find R, L, and C for the parallel
>>> equivalent circuit that produces the impedance curves in the Fair-
>>> Rite data. That is, write the equation for the Z of the parallel
>>> RLC circuit, plot it on the same scale as that for Fair-Rite data.
>>> When the two curves match, you now know R, L, and C.
>>>
>>> 73,
>>>
>>> Jim K9YC
>>>
>>>
>>>
>>> _______________________________________________
>>> RFI mailing list
>>> RFI@contesting.com
>>> http://lists.contesting.com/mailman/listinfo/rfi
>>>
>>>
>>>
>> _______________________________________________
>> RFI mailing list
>> RFI@contesting.com
>> http://lists.contesting.com/mailman/listinfo/rfi
>>
>>
>>
>
_______________________________________________
RFI mailing list
RFI@contesting.com
http://lists.contesting.com/mailman/listinfo/rfi
|