A three-dimensional multifrequency large signal model for helix traveling wave tubes

Citation
D. Chernin et al., A three-dimensional multifrequency large signal model for helix traveling wave tubes, IEEE DEVICE, 48(1), 2001, pp. 3-11
Citations number
20
Categorie Soggetti
Eletrical & Eletronics Engineeing
Journal title
IEEE TRANSACTIONS ON ELECTRON DEVICES
ISSN journal
00189383 → ACNP
Volume
48
Issue
1
Year of publication
2001
Pages
3 - 11
Database
ISI
SICI code
0018-9383(200101)48:1<3:ATMLSM>2.0.ZU;2-8
Abstract
A three-dimensional (3-D) multifrequency large signal model of the beam-wav e interaction in a helix TWT is described. The beam is divided into a set o f discrete rays, or "beamlets", instead of the disks or rings used in one-d imensional (1-D) or two-dimensional (2-D) models, The RF fields supported b y the helix are represented by a tape helix model that uses a modal expansi on including the full (Bessel function) radial dependence of the fields; bo th forward and backward synchronous space harmonics are included in the mod el. RF space charge fields are obtained from solutions of the Helmholtz equ ations for the RF electric and RF magnetic fields, using the beam current a nd charge densities as sources. The de space charge electric field is simil arly obtained from a solution of Poisson's equation. This model has been implemented in a code called CHRISTINE 3D, a generaliza tion of the one dimensional CHRISTINE code. The full three dimensional trea tment permits the accurate computation of large signal gain and efficiency, taking into account the self-consistent variation of beam radius along the interaction space. The code also computes helix interception current and t ransverse beam distributions at the entrance to the collector-important des ign data that are unavailable from a 1-D model. Results from the CHRISTINE 3D code are shown to compare very favorably with measurements of output power, efficiency, and interception current vs. dri ve power. Its predictions for spent beam distributions also compare very we ll with measurements. Run times for the code are problem dependent, but for a single case of inte rest are typically 1 to 5 min on a 450 MHz PC, orders of magnitude shorter than that required for a comparable 3D particle-in cell simulation.