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

Re: Variation of secondary Q



Original poster: "Paul Nicholson by way of Terry Fritz <twftesla-at-qwest-dot-net>" <paul-at-abelian.demon.co.uk>

Terry wrote:
> I assume Paul's data can tolerate of few hundred PPM.

Indeed.  The only real point in extracting F from the trace as well
as Q is so that when we see variation in Q, we can look to see if
there was an associated variation in F too.  This might help us to
isolate the causes of Q variation.  I'd be quite happy with 0.1%
error on F, which you can achieve without calibrating the scope
timebase.  But then if the scope just happened to pick up another
signal...

Jim wrote:
> there is a huge transmitter of signals at convenient
> frequencies accurate to 1 part in 1E13 fairly close to you....
> One transmission is at 60 kHz,  

...I'd use it - it is accuracy for free. I use a similar UK
transmission to calibrate the period space of a pulsar receiver
(to around 9 decimal places) by injecting the Rugby MSF signal into
the receiver front end, which filters through to the reduction
software.

Terry wrote:
> Do you think Fo will vary? 

Sure to, with temperature at least!  Unless you are tuned to the
peak response of the coil, the formula

 Vtop/Vbase ~= Q

or more accurately,

 Vtop/Vbase = Q * sqrt(Les * Ces) / sqrt(Lee * Cee)

will not apply. Only a slight offset of the sig gen from the coil's
Fres will cause a large error in Q by this method.  Your only hope
then is to arrange a self-oscillating driver for the experiment,
so that Fres is constantly tracked, but even that leaves behind a
significant tracking error.

> I could probably buy the RS-232 interface for the 3012 scope for
> the same price.

Yup, advantage: you can use it to automate other experiments;
disadvantage: ties up the 3012 and you'll have to make up some code
to command the scope and retrieve data.

> I pulled the data for the last on and put it here:
> http://hot-streamer-dot-com/temp/Qvar/TEK00000.CSV

Very clean waveform - far cleaner than necessary.

I've put the FT in 
 http://www.abelian.demon.co.uk/tmp/tek00000.ft.gif

No sign of any 60kHz breaking in.  If you can resist the temptation
to average multiple traces, we'll still get enough signal for the
F and Q extraction, and maybe we will also see a whiff of 60kHz
coming through too. When you average over 128 shots, the 60kHz adds
incoherently so its amplitude is attenuated by sqrt(128), a factor
of 11.  Don't worry about the trace being noisy.  The least-squares
process by which we match A * cos( w*t+p) * exp( -r*t) to the
waveform ensures that the noise (and vertical quantisation error)
is reduced by a factor
  sqrt(samples_per_trace) = sqrt(10,000) = 100.
Further, I can chop out any FFT components outside of the range
Fres +/- BW, thus removing the low frequency RC pedestal and most
of the other bits of noise.

> Here is a LabCam pic of the setup.
> http://hot-streamer-dot-com/temp/Qvar/Image1.jpg

By comparison with the ringdown analysis, this circuit introduces a
lot of sources of error: Sensitivity of capacitive pickup, gain of
the x100 amp, gain of the x1 amp, tuning of 33120 to fres, 
variation of Tek S10D probe Z and C.  With the ringdown analysis,
the only significant source of error is the waveform noise, and
this is known, very small, and more or less constant.

Can we have a CSV of a single shot ringdown?
--
Paul Nicholson
--