On Sat, Jul 14, 2012 at 9:28 PM, Richard (Rick) Karlquist <
richard@karlquist.com> wrote:
>
>
> On 7/14/2012 2:15 PM, George Dubovsky wrote:
>
>> All,
>>
>> I guess it's time for me to put up my "common mode choke" test technique
>> for all to shoot at. I have a HP 4815A Vector Z meter, a N2PK VNA, and a
>> HP 8753C VNA. I have all the probe adapters for the Z meter, and I have
>> been bitten by the small amount of stray C that you just can't eliminate
>> (but you should be able to calculate out). I have finally decided that
>> my standard test is based on the following:
>>
>> I made up a board out of .062 FR-4 that is approximately 6" across (and
>> about 4" wide). I cut three 50 Ohm microstrip lines full width across
>> the 6" dimension. I soldered an N connector on each end (6 connectors
>>
>
> SINCE YOU ASKED, there are a number of unresolved issues here from
> a strict metrological viewpoint:
>
> 1. Your fixture is not an enclosed box. A defined Faraday cage is
> recommended. The results might be affected by lead dress or whether
> the test bench is conductive or not. Any fixture imposes an arbitrary
> and finite spacing between the DUT and the ground plane. In general,
> the measured impedance will depend on this. There is no "correct"
> answer.
>
Agreed. In all cases, I placed the unit under test on a wood workbench. In
addition, after I had decided on the final balun design for each band, and
packaged it in its final pvc enclosure, I attached it to a short section of
2' aluminum tubing to simulate mounting on the yagi boom
>
> 2. The N connectors should be male and female and you should use
> a 2 port type N calibration kit.
>
Agreed. The 2 females is a luxury I allowed myself at HF, but it is a known
source of inaccuracy.
>
> 3. You used the common short cut of a through cal instead of a
> full 2 port cal. This means that you are not calibrating out source
> and load mismatch. A common short cut to fix this is to put pads
> around the DUT. This can work fairly well.
>
I always used either 6 dB or 10 dB pads on both ports (right at the test
fixture) for these tests.
>
> The designer of the network analyzer can recognize that the through cal
> shortcut is popular and accomodate the user by going to a lot of
> trouble to provide well matched ports. Or not. It turns out that
> getting well matched ports translates into massive padding. This
> has the disadvantage of loss of dynamic range. Then there is the
> question of allowing the user to turn the internal padding on and
> off. What if he turns it off and then proceeds to do a through
> cal anyway? There is no simple answer to this discussion.
>
> If the environment seen by the DUT in the test fixture is not
> the same as it sees in the application, then the measurement
> is just an approximation.
>
I did attempt to simulate mounting on the boom in my testing, including all
of the mechanical hardware that would be involved.
>
> It is entirely possible that you will get results accurate enough
> for ham radio since you are working at HF. But the technique
> would not be deemed completely rigorous by metrologists.
>
Ahh, what do I care about the weather... oh, metrologists ;-)
I do appreciate your comments, Rick, and I respect your professional
credentials. My goal going in to this exercise was to provide a family of
CM chokes that would not screw up my YO-optimized yagis and, after
measuring some commercial stuff in the junk box, I concluded that I could
do better.
73,
geo - n4ua
_______________________________________________
UR RST IS ... ... ..9 QSB QSB - hw? BK
|