From: Paul
Date: Tue, 27 Mar 2001 15:37:44 +0100
Subject: Re: [TSSP] Remaing CW problems & next stage of the project
Hi Boris, All, Boris wrote: > Maybe this will be just repeating but I don't seem to > remember we got straigthened completely out the > following: You're right. > 1) Problems of the coils at high elevation. > Why so noticable discrepancy between mesured and > simulated reso freqs. Still an open question. Kurt's measurements on his small coil showed that small features of the real coil, un-accounted for by the model, can introduce errors large enough to account for the discrepancies, but we don't know whether this kind of thing accounts for all the problems. > Is the only reason suspicious external capacities of > elevated system with attached wire to it and signal > generator or there's perhaps more behind it? There may still be something wrong with the model, either theoretical, eg not accounting for dielectrics, or in implementation, eg a coding error which only shows up at small radius. I'm hoping that someone will supply Fres measurements on a small coil with an additional tube inside, in order to see the rough size of dielectric effects. > 2) > I've noticed possibility for the difference between > reso > frequencies of the coil structure excited by voltage > source (typical msm with signal generator),and > corresponding "current source" frequency (theoretical > case of the current wave being pushed in a bottom). > In the first case the bottom of the coil is at small > potential but I haven't registered that in voltage > graphs displays on tssp web pages. The voltage is too small to see on the graphs. The profile graphs in all the documentation apply to isolated resonators driven by a perfect source. I have an 'extended' version of tsim which models the coil plus the additional interactions of a realistic external circuit, and this model reproduces the observed features, eg splitting of the resonance into a more complicated response, eg for Terry's probed small coil, I get 311.850 Max Ibase (Measured) 312.920~ Ibase 90 deg to Vtop (Predicted) 313.020 Max Vtop (Measured) 314.500~ Ibase in phase with Vbase (Predicted) which is an altogether more complicated picture. The predicted values above are based on guesses of various circuit reactances, so they're not likely to be too accurate. > 3) > Next stage of TSSP: When and how to approach to > transient simulations modelling (time domain). > More alternaives.. I've been busy revising documentation to match the latest code changes, and working on the next installment of 'theory notes'. Nothing much to report yet. A few items...the latest simulator code will model base feed, primary feed (incorporating acmi's code), center feed, and top feed (this latter is useful for looking at output admittances). This is for v0.9, to be released shortly. A major revision of the tech notes, pn1205, is necessary to describe all these changes, and I'm still plodding on with that tedious job. As a bit of light entertainment I managed to combine the (real) functions Cext(x), Cint(x,y), and M(x,y) into a single (complex) function F(x,y) which completely describes the physical coupling of the solenoid. It acts as a linear operator on the voltage profile 'vector'. I've yet to compute the eigenvectors of F but I think these will turn out to be the voltage profiles for each of the overtones, with the corresponding eigenvalues being inversely proportional to the frequencies. It remains to be seen where this leads to (if anywhere), but it might provide a more direct way to obtain the resonant frequencies (as opposed to the heuristics and trial solutions done by tsim). Presumably the eigenvectors of F(x,y) form a complete orthogonal set, so any particular voltage profile is then a vector in the space spanned by these basis vectors. It might be possible to obtain from F(x,y) an operator which is an infinitesimal generator of the time evolution of the basis vectors, which means that we could explore the time domain behaviour by just computing the amplitude of each basis vector, rather than having to compute the entire V(x) for each delta t. All falls apart of course when the load impedance is a function of top voltage. More realistically, I have a time domain model which is just a basic brute force iterative method. It has a tendency to explode after a few tens of cycles, and it uses too few elements to be considered quantitative, but it will cope very nicely with non-linear elements in the circuit. It's a couple of years since I last played with it, so there's probably quite a bit to bring up to date. Cheers, -- Paul Nicholson, Manchester, UK. --
Maintainer Paul Nicholson, paul@abelian.demon.co.uk.