From: "Malcolm Watts"
Date: Tue, 26 Jun 2001 09:30:46 +1200
Subject: Re: [TSSP] Some Considerations
Hi Paul,
On 23 Jun 01, at 15:00, Paul wrote:
> Malcolm wrote:
>
> > any msmt must be "central" to the resonator axis to avoid distorting
> > the electric fields
>
> Would also distort the field when on the axis of the coil. If removed to
> a distance, then the lead-out wire would add its distortion instead. We
> will have to accept some field distortion somewhere.
Quite right. I wasn't concise enough. I meant to avoid off-axis
distortion.
> > ...may be checked for disturbance by F and Q msmts before and after.
> > I don't see why doing such checks at low levels would be a problem as
> > long as there is no corona n the probe at normal running voltages.
>
> Agreed. A couple of small signal measurements would provide sufficient
> data to model the probes' effect. By incorporating the probe into the
> model in this way, we make its presence 'official'. However, if we want
> to use the topvolts probe to confirm that a particular toroid breaks out
> at the voltage predicted by Terry's program, then the probe itself must
> not introduce a local field gradient anywhere exceeding the max surface
> gradient of the toroid. Hmm. The advantage, if any, to a axial probe
> (either above the coil, or inside it) would be the relative ease at
> which the modified field can be calculated if necessary.
It might not be necessary. A small probe positioned far enough above
the top terminal would effectively melt into the background yet still
give a reading at TC amplitudes. I think an approach suggested by
Greg Leyh has a lot to commend it. It needs to be fleshed out. The
basic idea is to suspend a disk above the top terminal and calibrate
it by feeding the terminal (at measurement height) with a known
voltage that could be verified by direct connection (from underneath
say). However, I think a toroidal topload might present some
problems. For one thing, the e-field drops off near its centre. A
spherical topload would give a more consistent field distribution I
think. The probe itself could then be quite small. A calibration
check must be done at the resonant frequency of the system so
negating the effects of reactance in the actual measuring scope and
its lead. I also think probe current is an important factor that
should be taken into account. It might actually be possible to do a
visual check for field distortion by taking advantage of cold
streamer formation as recorded on film using a bulb shutter.
Please understand that I am throwing ideas around here. We need
to decide on exactly how to do this and look for any possible
problems before actually setting a system up. It is a difficult
exercise and I think would need great clearance i.e. a *really large
room*. I might be able to second such a room towards the end of the
academic year but the exercise would have to be "ready to go".
> Dealing with brush discharge impedance measurements would presumably be
> much easier because the properties of the discharge would be dominated
> by the local field around the discharge point and the probe would be
> just an extra constant RC load.
>
> > Top voltage vs breakout ... single shot mode ...
>
> The predicted breakout voltage, derived on the basis of 26kV/cm surface
> field, would apply only to the first shot of a run (and so would be
> independent of BPS), so I agree that single shot measurements are
> necessary. Given a satisfactory result from the single shots, it would
> be interesting to measure by how much the breakout voltage is reduced
> on subsequent shots of a run. Presumably after sufficient shots, the
> breakout voltage will have fallen as low as it's going to go, which
> means that a given topload is characterised by two numbers - the initial
> breakout voltage, and a lower, 'continuous run' breakout voltage.
> Associated with this latter figure would be a max streamer length.
This might lead us into defining streamer dependences on top volts as
well as energy storage. I would be most interested in seeing these
figures. An example: what is the streamer length dependence on
(a) an increased breakrate (more power by virtue of BPS) with a
constant Ep? Where does it top out? What is happening to top voltage?
(b) same breakrate and initial output voltage but more Ep (more power
by virtue of Ep)? This implies increasing Ctop to match Ep so that
initial voltage remains constant. Hence, charge/energy available from
the topload.
No formal analysis to my knowledge has ever said anything about
TC streamer length dependence on breakrate. In fact it has been
completely ignored in some rather prominent previous analyses which
suggests that the authors of those analyses weren't even aware of how
their coils were operating. Some give a set figure of x kV/inch,
others state bluntly that y feet = 1MV. Some coilers experience
recognize that there is a power dependence without separating the
various factors which might be causal. For example, increasing Ep
with the same topload increases initial Vtop but also causes an
increase in repetitive power. Others (myself included) have noted
that putting a much larger topload on a coil but keeping Ep and BPS
constant have resulted in brighter and longer streamers despite a
drop in Vtop. A hidden factor in there: Lp is increased to maintain
tune and of course unloaded Qp goes up as a result (as does transfer
efficiency by proxy). It is not difficult in principle to engineer
ways around these objections so that only a single factor at a time
is changing. The difficulties are more practical: To leave Lp
constant in the situation just outlined, one should actually decrease
Ls which, if all else is to be left the same implies stripping the
wire off the secondary and winding fewer turns in the same height and
diameter as Ctop is increased.
> Digressing slightly, has anyone ever illuminated a breaking-out topload
> with a projector lamp, casting the light onto a screen. Streamers would
> be rendered totally invisible but the hot, low density pathways which
> accumulate over a series of bangs might be rendered visible as shadows
> on the screen.
That is a really interesting idea.
>
> Marc wrote:
>
> > how does a poly tube filled with salt water or a conducting
> > fluid and graphite plugs in the ends sound?
>
> Sounds like a water divider. I made one of these a couple of weeks
> back, but I've not got around to trying it yet. Used de-ionised water
> in mine - a 1.5 metre by 1/4" I.D. polyethylene tube, with 1/4" jack
> plugs a tight fit in each end to form electrodes. It's DC resistance is
> around 50 Meg but I've yet to measure the AC response. 50 Meg is quite
> a low resistance to hook onto the top of a TC, but if it is stable
> enough then, as Marco says:
>
> > we can measure it with a KNOWN load and then "model out" its effect.
>
> Malcolm wrote:
>
> > It seems to me that this is a chicken and egg situation. We want to
> > verify models so incorporating a model as part of the verification
> > process seems self-defeating.
>
> We can manage so long as the topvolts probe is stable and linear, ie
> no breakout from the probe itself as the voltage is turned up.
> This is because we are directly measuring the topvolts, rather than
> deducing it from say, a base current measurement. If all we had was
> a base current reading, then we could not deduce the top volts without
> already knowing the load impedance - that's the chicken and egg loop -
> a loop which is broken by going for the topvolts directly.
Agreed.
> I think that to strive for a non-perturbative topvolts probe would raise
> problems in trying to accurately register the minute trickle of current
> through the probe. At the other extreme, a divider which draws too
> much current from the coil top (either reactive or lossy) may prevent
> the coil from reaching the topload breakout voltage that we want to
> measure.
>
> The output Z of the coil is Q/(2.pi.f.Cee) which for a 100kHz coil with
> a Q of 200 and a Cee of 40pF (a fairly large coil, eg 10" by 30") then
> Zout is around 8 Meg. If we could aim for a probe resistance of at least
> 10 times this value, then we could obtain 10% accuracy even if we didn't
> bother to 'model out' the probe load.
I have been thinking in terms of 10x this again. The shunt
capacitance of a resistive probe is a troublesome mechanism though
perhaps not impossible to deal with using calibration procedures at
Fr.
> Malcolm wrote:
>
> > I'd run with the capacitive divider idea as being cleaner than a
> > resistive chain (which would also have some capacitance). Question
> > then is: how much additional capacitive loading can we tolerate and
> > still claim a good measurement?
>
> >From the modeling point of view, I think we can accomodate some quite
> large probe capacitances. The potential difficulty comes when the
> charge stored by the probe becomes a noticeable fraction of the total
> and therefore would begin to have an effect on streamer formation.
>
> In connection with this, Greg Leyh wrote, (forwarded by Bert):
>
> > The capacitor stack that I placed in my coil was not for measurement
> > purposes, but rather to see if increasing the top capacitance improves
> > the arc performance.
>
> Would be interesting to know the outcome of that experiment.
Greg decided on the basis of that expt to go for large toploads (an
idea strongly touted by Richard Hull of TCBOR through expt).
> The idea that Greg proposes, ie a capacitive divider formed by the
> toroid and some kind of pickup plate suspended above it, is basically
> a suspended field probe as was suggested earlier. Would only work with
> toroids wide enough so that the pickup plate cannot see the coil charge.
Again I think a spherical topload is called for. Also a probe that is
quite small. We could see whether probe shape has any substantial
influence on measurements.
> To summarise my views:
>
> a) We can accomodate a modeled-out probe resistance of resistance
> greater than say, 10 times the output resistance, and 1/10th the total
> topload capacitance, without any worries, even if it is a little non-
> linear. We can probably take account of a probe which applies a heaver
> load than this, providing it is linear, stable, and does not prevent
> the topload reaching breakout.
>
> b) Capacitive dividers seem to be high up on everyones list of
> suggestions. Maybe a different approach is required depending on
> whether we're measuring the breakout voltage of a certain toroid, or
> the voltage behind a CW brush discharge. The E-field pickup of the
> toroid field is one extreme kind of capacitive divider, but its
> calibration may be affected by streamer formation in a way that a
> remotely placed divider column would not be.
For a suspended probe, streamers could be encouraged to form radially
away from the topload rather than vertically I think.
> c) For determining whether Terry's program will correctly predict top-
> load breakout voltages, I think a calibrated accuracy of 5% would be
> enough to give a good confidence in the predictions. For determining
> brush discharge impedances, the important thing is linearity, ie freedom
> from internal corona loss within the divider, since the impedance is
> arrived at by considering the departure from linearity of the top volts
> as the power is raised.
>
> Cheers All,
> --
> Paul Nicholson,
> Manchester, UK.
I am excited by the possibility of getting some real figures on a
number of facets of operation. I believe we can get there alright.
Regards,
malcolm
Maintainer Paul Nicholson, paul@abelian.demon.co.uk.