[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: MMC dissipation factor measurement



Original poster: "Jim Lux by way of Terry Fritz <twftesla-at-qwest-dot-net>" <jimlux-at-earthlink-dot-net>

You can accurately measure 0.1 degrees without too much trouble.. And,
especially if you use the DC current through a resistor replacement
approach, you don't actually care what the exact temperature is, as long as
it is the same. (that's how precision RF measurements are made with
bolometers and thermistors)

You DO need to control the conductive losses... keep the outside temp
reasonably constant, etc.

An air capacitor of 0.15 uF is going to be pretty large!  As for details of
lossyness of that air cap (for which you will need to allow for the epsilon
of 1.0005..) it's covered in lots of detail in the NBS circular.. Along
with precision capacitor design information (remember to use fused silica
to support those plates!)

I think, having looked over the various literature on the RLC meters, that
the best approach will be some form of precision voltage measurement.
Getting 22-24 bit measurements (6-7 digits) is plausible.  Using a scope,
with its 10-12 bit A/D, just isn't going to hack it (although, the
decrement measuring approach given in NBS circular is applicable to the
scope.. and there, you are combining 10,000 measurements, so the accuracy
should be on the order of sqrt(10,000)=100 times better than the raw
measurement (i.e. 1 part in 409600... )


Tesla list wrote:
> 
> Original poster: "Paul Nicholson by way of Terry Fritz
<twftesla-at-qwest-dot-net>" <paul-at-abelian.demon.co.uk>
> 
> Terry wrote:
> > measuring the dissipation
> > factor of a poly MMC cap?  Like <0.001 dissipation factor...
> 
> Interesting.  And you'd want it to be the dissipation at the
> TC operating frequency...
> 
> How about:  Put the 'test' MMC into a calorimeter and incorporate it
> into a cap bank attached to a primary to obtain the normal resonance.
> 
> Couple in a couple of hundred watts of CW at Fres.  The temperature
> inside the calorimeter will rise over a period of time to settle at
> some value.  Measure the calorimeter temperature, the room temperature,
> and the RF input VA *to the test MMC*.
> 
> Then replace the test MMC in the calorimeter with a resistance carrying
> DC.  Adjust the DC to obtain the same temperature difference above
> ambient as you got from the MMC dissipation.
> 
> Then the DC power equals the RF power that was dissipated in the MMC
> and you work can out the loss factor from that.
> 
> If you can measure to 1 degC and obtain a temperature rise of say
> 10degC above ambient in the calorimeter, then you'll get 10% accuracy.
> You can probably measure the MMC volts and current to 1% accuracy so
> the temperature measurements will be the dominant source of error.
> 
> The difficulty might be to actually get enough power dissipated in the
> MMC to get a decent temperature rise.  You might need to put a whole
> cap bank's work of MMC's into the thing, rather than the one!
> 
> > I have a TEK 3012 scope.
> 
> The box it came in could provide the polystyrene for the
> calorimeter :))
> 
> OK, you don't like that plan?  Alright, another then:
> 
> Make up an air spaced parallel plate capacitor to the same value as
> the test MMC.  Resonate each with a large inductor to the TC frequency
> and obtain the two Q values with the pinger and tcma.  Use the
> difference in Q factor to calculate the extra ESR introduced by the
> MMC.  This method has two problems: you have to assume the air spaced
> cap has negligible loss, and the result depends on measuring the small
> difference between two Q factors.
> 
> And a third method that you could try is to measure that phase angle.
> Capture the V and I of the test cap at the TC frequency in some
> convenient setup and measure the time delay between the zero crossing
> points of the two waveforms.  Calibrate out the phase error of the
> scope by repeating the measurement again with the roles of the two
> Y amplifiers exchanged.  This last I think would be the least
> accurate with only 10,000 points per sweep.
> --
> Paul Nicholson
> --