From: Paul
Date: Thu, 28 Mar 2002 17:51:57 +0000
Subject: Re: [TSSP] H/D stuff?
Hi Marco, All
I'm afraid this thread began without any introduction...
This is really a convergence of three separate issues,
1) Obtaining a measured current profile which demonstrates the
elevated current max due to internal capacitance. As per fig 6.1
of pn2511 and the associated text.
2) In order to achieve (1) we need accurate Q measurements so that
we can employ a resistance insertion method.
3) Given accurate Q measurements, we can also begin to investigate
variations of Q due to temperature, humidity, variability of
ground circuit losses, etc.
The accurate Q measurements are obtained by setting the coil
ringing through a step voltage transient applied to the base.
http://hot-streamer.com/temp/Paul/3-24/PingSchem.jpg
The coil mode spectrum is extracted from a base current sample
captured using a scope. As a further issue, there is the prospect
of the extraction program (tcma) evolving into a general purpose
tool for analysing TC waveforms to report F, Q, k, and maybe
efficiency.
Terry has wound a special coil,
http://hot-streamer.com/temp/Paul/3-24/P3240005.jpg
with 11 tapping points into which resistances can be inserted.
By moving a resistance along the coil and measuring the Q at each
step, we obtain the current profiles
http://www.abelian.demon.co.uk/tssp/tmp/f1-cmp.gif
http://www.abelian.demon.co.uk/tssp/tmp/f3-cmp.gif
http://www.abelian.demon.co.uk/tssp/tmp/f5-cmp.gif
in a way which is independent of the resistor's value and does not
require any absolute voltage or current measurements.
As yet, there is neither a web page for the test coil and
experiment, the curve comparison results, or for the analysis
program. Hopefully the long weekend ahead will give me a chance
to put these up.
Terry, will you be moving the coil images and CSV data to a more
permanent directory on hot-streamer.com? If so, I can deep link
to them. The quality of these experiments is such that they
deserve a decent write up.
Since this morning, I've been able to spend an hour on tcma and
have removed some of the nastier problems. Eg we get for the
repeat set of (avg 16) data taken on 26/03/02:
TEK00000.CSV
PK FREQ kHz (Error +/-) Q FACTOR (Error +/-) LEVEL
1 229.956 (0.01%,21Hz) 465.70 (0.14%, 0.7) -0.0dB
2 578.100 (0.06%,375Hz) 184.74 (1.43%, 2.6) -27.1dB
3 904.140 (48.70%,440287Hz) 106.58 (16.55%,17.6) -42.8dB
Accounted for 100.0% of input signal
TEK00001.CSV
PK FREQ kHz (Error +/-) Q FACTOR (Error +/-) LEVEL
1 229.939 (0.02%,35Hz) 283.93 (0.12%, 0.4) -0.0dB
2 577.969 (0.05%,281Hz) 167.34 (0.93%, 1.6) -22.3dB
3 903.913 (0.49%,4402Hz) 102.18 (6.98%, 7.1) -37.0dB
Accounted for 100.0% of input signal
TEK00002.CSV
PK FREQ kHz (Error +/-) Q FACTOR (Error +/-) LEVEL
1 229.951 (0.02%,42Hz) 234.09 (0.11%, 0.3) -0.0dB
2 578.093 (0.04%,211Hz) 134.76 (0.93%, 1.3) -22.7dB
3 904.229 (48.70%,440330Hz) 87.39 (8.06%, 7.0) -36.7dB
Accounted for 100.0% of input signal
TEK00003.CSV
PK FREQ kHz (Error +/-) Q FACTOR (Error +/-) LEVEL
1 229.952 (0.02%,44Hz) 219.48 (0.17%, 0.4) -0.0dB
2 578.137 (0.09%,501Hz) 135.53 (0.81%, 1.1) -22.0dB
3 904.088 (0.87%,7829Hz) 98.44 (5.23%, 5.2) -34.5dB
Accounted for 100.0% of input signal
and so on. Performance is pretty solid now at f1/4 and f3/4, but
todays alteration (preconditioning by multiplying the trace by
a sine) has reduced the sensitivity to the higher modes. Ignore
the LEVEL columns above. The neat thing about this method is that
it can gauge its own error very well. Q accuracy is now
consistently better than 0.2% at f1/4 on all measured traces but
we may have to lose f5/4 and above, at least for the time being.
Marco wrote:
> Precisely, how this new measurement of a coil Q is practically
> conducted (setup, jumpers, etc.)?
I think we'll have to defer thorough explanation to a web page.
Terry, are you thinking of doing a write-up?
> What are its advantages?
Very accurate and repeatable: there are no absolute analogue
measurements required. The only item subject to calibration is the
scope timebase.
> How is it that each mode frequency has got a Q factor of its own?
Due to the frequency dependence of most (all?) of the factors
affecting Q.
--
Paul Nicholson,
--
Maintainer Paul Nicholson, paul@abelian.demon.co.uk.