From: Bert Hickman
Date: Mon, 21 Oct 2002 23:39:18 -0500
Subject: Re: [TSSP] Top Voltage
Hi Paul, Comments interspersed below... Paul wrote: > Summary > ======= > > Assumptions, Definitions, Hypotheses > ------------------------------------ > 1) Assume that corona onset begins when the surface field > reaches 30kV/cm (low frequency or DC), as per the North report. > Lets call this the corona onset field, COF. > > 1a) The COF may fall, say to 26kV/cm above 30kHz. > > 1b) The COF may rise, say by a factor of 2 or 3 for > short fast signals, perhaps due to inertia at molecular > level. Perhaps its a lower probability that free electrons will be in the "right" places so that the e-field will be strong enough to initiate avalanches during a brief voltage pulse. > > 1c) We expect the COF to be proportional to air density, > specifically particle number density. Hence from this a > fairly straightforward variation with temp and pressure. > > 1d) For a given electrode geometry, a given COF leads to a > corona onset voltage, COV, based on quasistatic E-field > calcs in the usual way (eg north, tssp). > > 1e) Below the COV, the field is loss-free and described > by the quasistatic approximation. > > 1f) Only local physics is involved in the corona onset, > no long range effects, so COV should be in good agreement > with calcs. > > 1g) Variations to COF due to frequency, temperature, and > pressure, can be treated as adjustments to the 'standard' > value of 30kV/cm. The above assumptions sound reasonable. > > 2) > As the voltage rises above the COV, corona forms into > recognisable leaders. Let's call this the breakout > voltage, BV. > > 2a) For short gaps, we expect BV to be close, perhaps > indistinguishable, from the COV. > > 2b) In some cases the leaders will initiate a full discharge > arc. We'll not bother to draw distinction between leaders > in air and arc discharges, whichever happens first counts > as the BV. Thus BV can stand for breakout or breakdown volts. > > 2c) For long gaps, and TC terminals, we expect BV to be > distinctly higher than the COV. > > 2d) The interval between BV and COV is likely to depend on > frequency and air particle number density. > > 2e) Long range effects become significant when considering > BV, as opposed to only local for COF. > > 2f) COV-BV interval would likely be reduced by memory > (channels of low air density) of previous shots. > > 2g) As voltage increases beyond COV towards BV, we expect a > fairly smooth and continuous increase in corona, therefore > corona-related effects should smoothly increase: Q drop, > terminal current, DC offset, etc, all should increase steadily. Sounds reasonable if averaged over enough shots. However, the statistical nature of corona-to-streamer transitions will result in significant variability from shot to shot. > > Some comments > ------------- > We'd like to be able to calculate the COV and BV for > typical TC geometries. > > Lets stick to signals with slow risetimes, say > 1uS, > to avoid effects (1b), so that would be fres < 250kHz, > to begin with. Big, low frequency coils, line frequency, > or DC. > > Lets stick to slow BPS to avoid effects (2f), to begin with. > > Things to do > ------------ > Let's concentrate on recognising and measuring COV and BV. > We must learn to be good at predicting and measuring COV, and > get good agreement, before we can go any further. > > Firm up the geometry around OLTC topload to better define > the E field. > > Look for signs of corona onset occuring earlier than the > observed breakout:- > a) Onset of decline of Q factor; > b) Appearance of DC base current component; > c) Low level light output; > d) Increase in apparent terminal capacitance; > e) Slight drop of Fres; > f) Increase of HF noise output from the system. > > Measure breakout thresholds for various arrangements of > terminals:- > a) Various TCs; > b) Various gaps of a few cm; > c) Must be well defined geometries, prefer one terminal > grounded, rather than balanced. > > Some specific expts: > -------------------- > For OLTC... > > Plot Q as function of Vfire, for small sphere and large > sphere (at same height). Look to see if Q starts to > decline later with large sphere. > Test also with no rod/sphere. Q should not decline at > all until much later. That would prove Q drop is due > to corona onset. > > Plot BV (and COV if detectable) as function of adjustable > ceiling height. Ceiling in 2 or 3 positions only. Just > small sphere and large sphere. > > Capture rod/sphere terminal current in the voltage regime > below the obvious BV. Analyse to look for signs of corona > onset. > > If corona onset is discernable by Q or rod/sphere current, > see if DC base current appears in region between COV and BV. > If so, then space charge is an issue in determining COV-BV > interval. If DC base current delays until BV reached, then > we can ignore space charge and treat the initial corona as > bulk neutral. I suspect it's more a matter of scale. Initially we should get negative corona which almost immediately self limits due to nearby space charge formation (this decays and the corona begins anew, forming a relaxation oscillator - Trichel pulses). Higher voltage oscillations may result in both positive and negative corona, tending to neutralize resonator base voltage. Once we get a true streamer breakout, a significant chunk of charge is transferred quite suddenly. This should be an easily detectable event (via a large jump in DC base voltage). Long spark propagation should preferentially be with positive electrode polarity. > > For other coils... > > Try to observe effects of corona onset prior to obvious > leader formation. There's a stage in between, sometimes called corona flashes or corona-streamers. These are only visible in a darkened room, and can be heard as "pops" in other (less noisy) HV equipment such as HVDC supplies or VDG generators. Further increases in voltage finally result in formation of hot leader root, and actual leader propagation (the observable "streamers" we normally see when coiling). > > Try to measure peak base current at COV and BV for well > defined TC geometries. OLTC is probably the most controlled environment we have... > > Also, > > Find published data on gap breakdowns. Abundantly available for sphere-sphere. Less common is sphere-plane and rod-plane gap breakdown... > > Find published estimates of equilibrium between charged > particle density and prevailing E-field. > > Attempt to model virtual enlargement of terminal due to > (hopefully neutral) initial corona. Estimate radial > current in the extra 'virtual' region and see if breakout > coincides with a certain total or max current or current > density. > -- > Paul Nicholson, > -- > Best regards, -- Bert -- -- Bert Hickman Stoneridge Engineering "Electromagically" (TM) Shrunken Coins! http://www.teslamania.com
Maintainer Paul Nicholson, paul@abelian.demon.co.uk.