From: "Terrell W. Fritz"
Date: Mon, 25 Mar 2002 12:20:10 -0700
Subject: Re: Ready :-)) - Re: [TSSP] short H/D and stuff
Hi Paul,
Tonight I will take another run (it takes about 15 minutes). I'll watch the
temperature and let the scope average over say 16 waveforms. I will also
change the timing to help get higher frequency data. I also have some big
chunks of Delrin (12 x 12 x 1 inch) so I'll try sticking one on top of the
bucket and rerun to see if that shows up in the data. Maybe adding a big
dielectric mass like that will give clues as to the Q effects. The bucket also
has a bottom and a heavy support near the top which add odd dielectric masses.
I could probably cut them off without killing the coil to remove them as a
source of error. The coil sits on a Styrofoam box. It could be suspended by
string too.
If your dielectric models are symmetrical around the Z axis, I can try to find
round things that will match cylindrical coordinates that will be easy to
model.
The resistor is 100.57 ohms and I think the jumper clips add a tiny bit of
resistance too. Looks like that value does well.
I'll order some parts to make a much nicer pinger that can operate outside and
such. The Bertan HV supply is a little awkward for portable use. I also need
to spend a few minutes with MathCad to do FFTs here...
I will try reducing the size of the CSV files too. Maybe put them in one file
with only one set of timing numbers and ZIP them. I don't know if you have
high speed or modem internet.
Looks like this is all working great and it opens up many new ways to test
coils!
Cheers,
Terry
At 12:14 PM 3/25/2002 +0000, you wrote:
>I wrote:
>
>> I should have enough info in the waveforms, and from the model,
>> to determine the value of the resistance that you use.
>
>Starting with
>
> 2*pi*f*Lee/R = Q0 * Qs/(Qs - Q0)
>
>derived in a previous post, in which Q0 is the measured Q with
>R in the base connection, and Qs is the normal Q of the coil
>without added R.
>
>From the data, we have for each frequency,
>
> f1/4 f3/4 f5/4 Origin
> Qs: 464.44 185.29 106.80 (3-24/TEK00000.CSV)
> Q0: 283.93 168.62 102.70 (3-24/TEK00001.CSV)
>
> f: 229.91 kHz 578.09 kHz 904.58 kHz (3-24/TEK00000.CSV)
> f: 231.61 kHz 583.92 kHz 939.46 kHz (tssp tfcp1.in)
>
> Lee: 51.77 mH 50.05 mH 38.07 mH (tssp tfcp1.in)
>
>Applying R = 2*pi*f*Lee*(Qs - Q0)/(Q0 * Qs) and using the model's
>value for f, rather than the measured (so that it is consistent
>with the modeled Lee), we get
>
> f1/4 f3/4 f5/4
> R: 103.1 ohm 98.0 ohm 84.0 ohm
>
>Since the formula for R is most sensitive to the difference Qs-Q0,
>the error in the estimate of R will be at least the fraction
>
> sqrt(2) * Eq * sqrt( Qs*Q0)/(Qs - Q0)
>
>where Eq is the fractional error in the Q determination. Eq is
>typically around 0.3%, so an error estimate from measurement error
>alone is
>
> f1/4 f3/4 f5/4
>Q err: +/-0.9 % +/-4.5 % +/- 11 %
>
>Since we are using the model's predicted values for the product
>Lee * f, and f at least has the known errors
>
> f1/4 f3/4 f5/4
>F err: +0.7% +1.0% +4.0%
>
>and if Lee has a similar error, then we can put forward the final
>result for R:
>
> f1/4 f3/4 f5/4
> R: 103.1 ohm 98.0 ohm 84.0 ohm
> +/-2.3 % +/-6.5 % +/-19 %
>
>For this short coil, the ratio Les/Ldc is greater than unity:
>
> f1/4 f3/4 f5/4
> Ldc: 39.18 mH 39.18 mH 39.18 mH
> Les: 44.34 mH 54.00 mH 58.61 mH
>
>Les/Ldc: 1.13 1.38 1.50
>
>and the measured current profile now stands as a convincing
>demonstration of why the effective series inductance at resonance
>usually differs from the DC value.
>
>If you were to stand another secondary on top of this short coil,
>say 10" diam by 30" long, then you would have a 39"x10" secondary
>with the lower 25% available for current profiling. With this
>arrangement you could demonstrate the elevated current max of a
>normal h/d TC secondary.
>
>This trace analysis method of extracting Q values seems to be
>working very well. I can now think of about a million experiments
>that can be done now that we can measure Q accurately.
>
>One interesting point is that the measured Q is rather higher than
>the predicted Q for this coil. I'm not used to seeing this -
>usually its the other way around. The Q measurements are now
>precise enough for me to tackle this part of the software.
>
>The calculated winding resistances are
>
> DC: 32.91 ohms
> AC: 37.83 ohms (DC + skin effect)
>Proxy: 155.45 ohms (DC + skin effect + proximity loss)
>
>Obviously, finding a way to gauge the proximity loss properly
>is going to make a big difference.
>
>As regards the errors in the current profile comparisons - up to
>20% in the current peaks at f5/4, this looks like it is due to
>poor determination of internal capacitance by the model. Cint is
>being under-estimated at short range by about 10%, enough to give the
>frequency error we see at f5, and this accounts for most of the f5
>current profile error. I'll re-run the model with coil former
>dielectric compensation turned on to see if that fixes it.
>
>How long did it take to capture the 12 CSV files? Was it likely that
>the temperature in the room changed by more than a degree during
>the run. In future I think we may have to ask for temperature
>readings, since the Q will change by about 0.4% per degree C, which
>is more than the precision of the Q extraction.
>--
>Paul Nicholson,
>--
Maintainer Paul Nicholson, paul@abelian.demon.co.uk.