The PS directional coupler

General The coupler consists of a 700mm long 140mm i.d. tube with four tapered strips mounted between pairs of vacuum feedthroughs. The strips are 375mm long and taper from 30mm down to 3mm. The taper appears to be somewhere half way between exponential and hyperbolic. (At 100mm from the wide end, the width is 14mm.) The design is apparently based on the SPS directional coupler [1]. The monitor has been sitting in PS SS98 until March 1999, when it was removed to make room for Andreas Jansson's quadrupole PU. It hasn't been reinstalled since, although we came close on several occasions. The coupler is not particularly clean. Whether it is clean enough for vacuum remains to be seen. The flange surfaces appear to be undamaged. The N-connectors on the feedthroughs are ordinary PTFE-insulated 50Ω N sockets. They are badly corroded and must be replaced. In fact, the measurements were plagued by incessant problems due to poor contacts. All following measurements have been made using an HP8753D network analyser NA).

Coupler cross section

The coupler as installed in SS98 until March 1999

Strip impedance

The NA is connected to the narrow end of the strip, with the wide end terminated into 50Ω. The plot represents a reflectometric step response, calculated from the S22 frequency domain data. The positive reflection of about 0.08 implies Zstrip=59Ω, and relatively flat, despite the strip's taper. (Its distance to the wall decreases in proportion to its width.) The marker indicates the location of the feedtrough connector. The measured strip length is about 2.5ns. That matches the calculated value for a 375mm strip suspended in vacuum very well (2l/c).

Wire impedance In order to simulate the passage of a beam, a wire (actually a 6mm rod) is strung through the PU. The NA is connected to the end nearest to the wide end of the strip and we examine the reflections in order to judge its adaptation. Again, the plot is a step response calculated from the frequency domain reflection data. The wire impedance is approximately 160Ω. The near end is matched using a parallel 73Ω resistor, and the far end is terminated with a 50Ω N terminator in series with a 110&Omega resistor. The matching in the PU volume is very good (between points 2 & 3). The impedance is higher in the PU than in the adapters, because the PU is a 140mm diam. cylinder, while the adapters are sections of standard PS elliptical vacuum chamber. The wire sags a little in the middle, which is visible as the slight slope in the adapter area (between points 1 & 2). The reflection at point 4 shows that the terminator is slighty inductive.

Transfer function (frequency domain) With the NA connected as before, we observe the signal on the strip. The red curve is for the forward direction, i.e., with the signal taken from the wide end of the strip. The green curve is the backward direction, from the narrow end of the strip. The opposite end is terminated into 50Ω. Note the linear frequency scale. The LF cut-off frequency is about 110MHz. Beyond 1.2GHz, the response becomes rather bumpy and what little directivity the coupler had fades away. In the area in between, the transfer ratio varies around -32dB, corresponding to a transfer resistance Rt=4Ω (160 * 10^(-32/20)). Since the wire sagged in the general direction of the strip under test, that value is probably slightly on the high side. The directivity is in the 20dB ballpark.

Transfer function (time domain) The setup is exactly the same as for the frequency domain plot. The red curve is for the forward direction, i.e., with the signal taken from the wide end of the strip. The green curve is the backward direction, from the narrow end of the strip. The opposite end is terminated into 50Ω. This is a step response calculated from the frequency domain S21 data. The time scale is 3ns/div.

References