PHYSICAL REVIEW ACCELERATORS AND BEAMS 23, 082803 (2020)

Measuring vacuum component impedance for the Argonne

Advanced Photon Source upgrade
*
Medani P. Sangroula
Physics Department, Illinois Institute of Technology, Chicago, Illinois 60616 USA

Ryan R. Lindberg, Robert M. Lill, and Randall Zabel

Advanced Photon Source, Argonne National Laboratory, Argonne, Illinois 60439 USA

(Received 7 July 2020; accepted 27 July 2020; published 19 August 2020)

The Advanced Photon Source plans to upgrade to a multibend achromat (MBA) lattice that will
dramatically decrease the electron beam emittance, thereby enhancing the x-ray brightness by two to three
orders of magnitude. Electron beam focusing in the MBA requires small-aperture vacuum components that
must also have a small impedance so as to minimize rf-heating and collective instabilities. As part of this
effort, this paper focuses on coupling impedance measurements and analysis of certain critical Advanced
Photon Source Upgrade vacuum components. Impedance measurements of accelerator components have
traditionally been done with the coaxial wire method, which is based on the fact that the Transverse Electro-
Magnetic (TEM) mode of the coaxial cable can mimic the Coulomb field of a particle beam; however this
measurement technique has various limitations. This paper describes our approach to measure the coupling
impedance using a Goubau line (G-line), which is essentially a single wire transmission line designed to
propagate Sommerfeld-like surface waves whose fundamental Transverse Magnetic (TM) mode mimics
the Coulomb field of a relativistic particle beam. We describe in detail the measurement procedure that we
have developed for the G-line, including the measurement setup and proper definition of a reference,
measurement procedure and advantages, and our experience regarding how to reduce systematic
experimental error that we learned over the course of the measurements. Starting with our initial suite
of measurements and simulations designed to benchmark and validate the novel G-line based measurement
technique, we present the measured results for several Advanced Photon Source Upgrade vacuum
components, including those of two rf-gasket designs and the beam position monitor-bellows assembly.

DOI: 10.1103/PhysRevAccelBeams.23.082803

I. INTRODUCTION typically characterized by the coupling impedance that can

be associated with each vacuum component.
Ultralow emittance storage rings based on multibend
Although the impedance is a well-established concept,
achromat lattices can provide a generational leap in x-ray
predicting it in a real accelerator is somewhat challenging.
performance by increasing the brightness and coherent flux For example, the recent work by Smaluk et al. [5]
by two to three orders in magnitude [1–3]. The compact compared the simulated impedance budget with beam-
and strong magnets for such lattices demand small gap based measurements for 15 storage rings, and reported a
apertures, which in turn may lead to an increased inter- significant discrepancy between these two results. One
action between the electron beam and the chamber walls. possible explanation could be that the even the best
This strong interaction between the beam and the vacuum impedance predictions using the most sophisticated codes
chamber can lead to various deleterious collective effects are only as good as the underlying model of the compo-
including an increase in energy spread, a degradation of nent. Typically, the simulated structure represents an ideal
beam quality or even beam loss, overheating of vacuum design which may miss subtle changes made to ease its
components, etc., [4]. The strength of this interaction is production, or fail to adequately account for manufactur-
ing tolerances. To try and minimize these issues the
Advanced Photon Source Upgrade (APS-U) project
*
msangroul@bnl.gov has chosen to assess the impedance of various components
using a combination of sophisticated electromagnetic
Published by the American Physical Society under the terms of simulation and rf bench measurements. In this paper
the Creative Commons Attribution 4.0 International license.
Further distribution of this work must maintain attribution to we describe our experimental efforts to measure the
the author(s) and the published article’s title, journal citation, longitudinal impedance of a variety of APS-U vacuum
and DOI. components.

2469-9888=20=23(8)=082803(15) 082803-1 Published by the American Physical Society

SANGROULA, LINDBERG, LILL, and ZABEL PHYS. REV. ACCEL. BEAMS 23, 082803 (2020)

The conventional coaxial wire method, in which the

TEM mode of a coaxial wire simulates the Coulomb field
of a relativistic beam, has a long established history of
measuring impedance [6–10]. One major drawback of this
technique is its need for a matching network of resistors
whose frequency dependence limits the bandwidth of the
measurement. If not accounted for, this can lead to higher
noise and false signals at higher frequencies due to the
appearance of higher order modes (HOMs). Some of the FIG. 1. Cartoon picture showing the schematic of the G-line
challenges associated with these HOMs can be reduced by, setup.
for example, using an appropriate combination of resistive
matching outside and rf-absorbing foam inside the device brief description of our measurements of a gate valve liner
under test (DUT) [11,12], but this is very challenging if the and pumping cross.
DUT is a closed system.
Another issue with the traditional method is that it
typically requires a relatively large-diameter central con- II. GOUBAU-LINE (G-LINE)
ductor to obtain a suitable characteristic impedance of the The Goubau line [14,16,17] is a dielectric coated, single
DUT. The large inner conductor results in larger perturba- wire transmission line that transmits signals via surface
tions to the DUT, and furthermore leads to greater uncer- waves. Early analysis of electromagnetic surface waves was
tainties in the position of the chamber axis [13]. While one done by Sommerfeld [20] and Zenneck [21], who showed
can consider improving this situation by using a thin central that an electromagnetic wave could be supported at the
conductor, the associated errors can be substantial [8], and cylindrical interface between a metallic wire and air.
the resulting large characteristic impedance will require a Although Sommerfeld waves have a theoretical attenuation
suitably designed matching network. much smaller than that in coaxial cables or rigid wave-
The recently developed Goubau line based test fixture guides, their practical application is limited due to the fact
[13–15] can overcome most of the issues that arise in the that in low-loss systems the physical extent of the electro-
traditional method, especially at higher frequencies. The magnetic fields becomes very large [16,22].
Goubau line [14–17] is a dielectric coated single wire Nonradiating surface waves of bounded extent can be
transmission line in which the fundamental TM mode of supported by modifying the surface of the central con-
the surface wave has an electric field that mimics the ductor [23]; for example, a perfectly conducting wire that is
properties of the Coulomb field of a relativistic particle coated with a dielectric material can support TM surface
beam. Once an appropriate launching and receiving waves of low loss and small extent [23,24]. In 1950 George
design (that we describe later) has been made, the G- Goubau [16] showed how to employ these waves for
line setup does not require any additional matching signaling applications, and we call the resulting trans-
networks, and can be used with a wide range of DUTs mission line a Goubau line (G-line). Importantly, Goubau
with no change in design. While the G-line has been used described tapered structures that can efficiently couple the
as a transmission line for a long time, its application as a quasi-TEM mode of a coaxial cable to the TM mode of the
diagnostic device in the accelerator community began guided surface wave.
about a decade ago, and its application to impedance In the vicinity of the wire, the G-line’s fundamental TM
measurements began a few years ago [13,18]. In what mode mimics the electromagnetic field of a relativistic
follows we describe the design and use of this technique electron beam. This led Ref. [14] to suggest that a G-line
to evaluate the impedance of some critical APS-U vacuum could be used by the accelerator community to experi-
components. mentally test how a relativistic electron beam would
This paper is organized as follows. In Sec. II we describe interact with a variety of accelerator vacuum components.
the novel G-line method, starting with a brief mathematical The Goubau line is superior to the traditional coaxial wire
description based on that found in Ref. [15,19], and method in numerous respects: the G-line setup is simple
proceeding to a detailed description of the experimental since it does not require a complicated matching network; it
design and our measurement procedures. Section III perturbs the boundary conditions less due to the micron
describes how we benchmarked the G-line setup using a sized wire; it provides more accurate impedance matching;
relatively simple and well-known pillbox-type cavity, and and it enables broad-band measurement [17,25]. A simple
compares impedance results between measurement and schematic cartoon picture of the G-line setup is shown in
simulation. Finally, we show in Sec. IV the measured Fig. 1, where the blue cones facing each other represent a
results for APS-U vacuum chamber components including surface wave launcher and a receiver (also called horns), the
detailed results for two rf-gasket designs and the beam dashed black line indicates the thin dielectric coated
position monitor (BPM)-bellows assembly, along with a metallic wire which connects the launcher and receiver,

082803-2
MEASURING VACUUM COMPONENT IMPEDANCE … PHYS. REV. ACCEL. BEAMS 23, 082803 (2020)

and each of these horns are connected to a vector network γ 2d − γ 20 ¼ k2d − k20 ¼ k20 ðϵr − 1Þ: ð3Þ
analyzer using two coaxial cables.
The G-line has been used by several groups to character- The guided wave solutions are fully specified by
ize beam position monitors (BPMs) and current transform- enforcing the continuity of μEz =Bϕ at both interfaces.
ers [15,17,19,26]. Simulations showing the G-line’s Inside the perfectly conducting wire we have Ez ¼ 0, so
potential to assess vacuum component impedance were that the boundary matching to (1c) at r ¼ a implies that
described in Refs. [13,25], while the first experimental m ¼ −J0 ðγ d aÞ=Y 0 ðγ d aÞ. Outside the wire the bounded,
impedance measurements from a G-line were reported in cylindrically symmetric solutions are given by the Hankel
Ref. [18]. The latter experimental setup differs from ours ð1Þ ð1Þ
functions of first kind H0 and H 1 , and we have
in that Ref. [18] replaced the receiver cone in Fig. 1 with
rf-absorbers, and determined the impedance from the S11 ð1Þ
Ez μ ωγ H 0 ðγ 0 rÞ
reflection coefficient; typically, measurements based on the ¼ 0 2 0 ð1Þ if r ≥ b: ð4Þ
S11 is not as accurate as those that employ the S21 Bϕ =μ k0 H ðγ 0 rÞ
1
transmission coefficient [8].
Equating (4) at r ¼ b to the same ratio from (1) results in
A. Brief theory of G-line surface waves  
γ d J0 ðγ d bÞY 0 ðγ d aÞ − J 0 ðγ d aÞY 0 ðγ d bÞ
The mathematical theory describing surface waves on a ϵr J1 ðγ d bÞY 0 ðγ d aÞ − J 0 ðγ d aÞY 1 ðγ d bÞ
dielectric-coated wire can be found in many places (see,   ð1Þ
e.g., [15,16,19,22]). In this section we will review that part γ 2d b b H0 ðγ 0 bÞ
≈− ln ¼ γ 0 ð1Þ : ð5Þ
of the theory that is relevant to understanding the design of ϵr a H1 ðγ 0 bÞ
our experimental apparatus, paying particular attention to
the extent to which the fields mimic a relativistic electron The second line employs a useful approximation to the
beam and fall off ∼1=r, where r is the radial distance from Bessel functions which is valid when the surface wave
beam axis. After a certain radial distance, the fields decay phase velocity is very close to the speed of light c. This
exponentially as illustrated in the Ref. [15]. condition is fulfilled when either the dielectric thickness is
We consider a metallic wire with radius a that is coated much smaller than the wire radius, d ≪ a, or when d ≲ a
with a thin dielectric material of thickness d ¼ b − a, so and the radius a is very small compared with the wave
that the outer radius of the coated wire is b. We adopt length [16]; both of these criteria are satisfied for our
cylindrical coordinates and assume periodic solutions in z. G-line setup.
Inside the dielectric, the radial solution is given by a sum of Solving Eqs. (3) and (5) will give the surface wave
the Bessel function J n ðxÞ and the Neumann function Y n ðxÞ, parameters γ 0 and γ d as a function of frequency ω, which
so that for a ≤ r ≤ b the electromagnetic fields are [16] we can then use to study the free-field solutions Bϕ ∼ Er ∼
ð1Þ
H1 ðγ 0 rÞ. In particular, we would like to compare these
l fields in the vicinity of the wire to the known 1=r behavior
Er ¼ A ½J1 ðγ d rÞ þ mY 1 ðγ d rÞe−iðωt−lzÞ ð1aÞ
γd of a relativistic electron beam. We did this for our G-line
design by numerically solving (3) and (5) in Mathematica
k2d for a wire with inner radius a ¼ 143.5 microns that is
Bϕ ¼ A ½J ðγ rÞ þ mY 1 ðγ d rÞe−iðωt−lzÞ ð1bÞ coated with 20 microns of a ϵr ¼ 3.5 dielectric film (so that
ωγ d 1 d
b ¼ 163.5 microns). We found that at 10 GHz the deviation
of Er and Bϕ from the desired 1=r behavior was less than
Ez ¼ iA½J0 ðγ d rÞ þ mY 0 ðγ d rÞe−iðωt−lzÞ ; ð1cÞ
10% out to a distance of 12.3 mm, while at 5 GHz the
discrepancy at 11 mm was ∼2.5%. Since the nominal
where A is an amplitude factor and m is a function that will radius of the tested APS-U vacuum chamber components is
be determined by the boundary conditions. Equation (1) 11 mm, the surface waves appropriately model the 1=r
defines the solutions in terms of the frequency ω and properties of the electron beam over the 1 to 10 GHz
propagation constant l, whose relation to other quantities is frequency range of our measurement.
given by We also verified this 1=r behavior of the G-line with CST
simulations. Figure 2 shows a contour plot of the magni-
γ 2d ¼ k2d − l2 ¼ μd ϵd ω2 − l2 ¼ ϵr k20 − l2 ; ð2Þ tude of Ey at the plane x ¼ 0 for the frequency of 2 GHz,
while Fig. 3 plots the radial electric field extracted from this
where the relative dielectric constant ϵr ¼ ϵd =ϵ0 , the contour plot on a log-log scale, showing that the field falls
dielectric permeability μd ¼ μ0 for a nonmagnetic material, off as 1=r out to the radial distance of about 65 mm, which
and k0 ¼ ω=c. In free space outside the wire we similarly agrees quite well with the semi-analytic result of 66.6 mm.
define γ 20 ¼ k20 − l2 , which we see results in Note that 65 mm also happens to be the size of the

082803-3
SANGROULA, LINDBERG, LILL, and ZABEL PHYS. REV. ACCEL. BEAMS 23, 082803 (2020)

FIG. 2. Contour plot obtained from CST simulation showing the localized electric fields, y-component, at 2.0 GHz in the G-line.

launching and receiving horns, and that at 2 GHz ∼ 95% of mode size that one wants to transmit (i.e., the mode size at
the surface wave energy is contained within a cylinder of the minimum frequency of interest), while the ratio of the
the cone radius. At 1 GHz about 80% of the field energy is diameter to length should be small enough for smooth
in this same distance, giving sufficient launching and matching, but is ultimately limited by reasonable length
receiving efficiency over the entire 1 to 10 GHz range. constraints; previous works have chosen this ratio in the
range between 0.4 and 1.0 [14,27].
B. Description of the G-line setup designed The G-line system designed at Argonne National Lab is
at Argonne National Lab shown in Fig. 4. The launching and receiving horns are
The previous section described steady-state surface made of aluminum with an opening diameter of 130 mm
waves that could in principle be used for impedance that allows for good coupling to surface waves above
measurements. Goubau [27] showed that these waves 1 GHz, and a tapered length of 210 mm to provide good
can be efficiently excited using launching and receiving impedance matching; these dimensions are similar to those
“horns” that are tapered, conical structures which serve to described by Goubau in his seminal paper [16] on surface
minimize the impedance mismatch between the coaxial wave propagation. Within each horn is a central brass taper
cables and the single-wire transmission line. In other that connects the coaxial cable to the central wire. To
words, the horns effectively convert the TEM mode of smoothly match the impedance, the brass taper is transi-
the input coaxial cable into the fundamental TM surface tioned from its maximum diameter of 7.35 mm at the
wave mode of the dielectric coated wire and back again coaxial line to its final diameter of 1.05 mm at the wire over
[14]. These horns are actually transmission line tapers a total of six 1.05 mm steps. The brass tapers are covered in
matching the coaxial cable impedance (50 Ω) to the wave Fig. 4 by teflon spacers (the white circles at the opening of
impedance of the Goubau line [28,29]. A properly designed each horn), which serve both to position the brass taper
horn will limit coupling to higher-order modes that can lead along the central axis and to help to damp higher order
to deviations in the mode wave-front [22,27]. The diameter modes. Finally, the brass tapers are connected to the single
of the horn opening is determined by the largest effective wire transmission line, which we have chosen to be a
927 mm long magnet wire with dimension of 29 American
Wire Gauge (diameter ∼287 microns) that has a 20 μm
polyimide coating with dielectric constant of 3.5.
Theoretically, the G-line surface waves have a lower

FIG. 3. Graph showing Ey from contour plot with radial FIG. 4. Experimental G-line setup designed at the Argonne
distance (red dots), and a fitted line (black dots) to the field National Laboratory to measure the coupling impedance
data displaying 1=r characteristic of a relativistic particle beam. of a DUT.

082803-4
MEASURING VACUUM COMPONENT IMPEDANCE … PHYS. REV. ACCEL. BEAMS 23, 082803 (2020)

smaller than that of the first one due to reflections on

the return path at the first step. Hence, the inferred
impedance of the second transition would be greater than
50 Ω, and similar (but more complicated) effects will occur
with the smooth impedance transitions to and from the
horns. We plan to investigate this further.
Theoretically, the impedance is a function of frequency
so that impedance mismatch cannot be fully avoided for all
measurements, especially in the traditional coaxial method,
where it increases due to additional capacitive behavior
observed in the resistive circuit used for impedance
matching at higher frequencies; nevertheless, it can be
minimized using the G-line setup as the reference sleeve
normalized out this mismatch.

FIG. 5. A screen-shot image showing the TDR measurement C. Impedance measurement procedure
(S11 plot with respect to time) at the single ended mode by placing using the G-line system
a DUT in the middle of the G-line setup and terminating the Experimentally determining the impedance of any vac-
receiver horn with a 50 Ω load. uum component requires inserting the device under test
(DUT) into the G-line test apparatus, calibrating the net-
work analyzer with an appropriate reference, and measur-
cut-off frequency of zero, but the launcher geometry ing the response of the DUT. For the first step, we place the
introduces certain practical limitations that for our geom- DUT directly between the two horns, thread through the
etry set a lower frequency of around 0.5 GHz. Simulations magnet wire, and solder the wire to the horns. We then
have shown that this should be sufficient for the APS-U position the DUT vertically with knobs and horizontally
structures, while the 10 GHz maximum, which is set by the with positioning screws so that the wire passes directly
1=r characteristics of the chosen magnet wire, is well above down the chamber axis, and finally fix the DUT to the
the 3 sigma frequency f ≈ 6.8 GHz set by the APS-U optics board with holding clamps and screws.
minimum bunch length of 50 ps. In addition, the quality Once the DUT is properly positioned, we set the
of coaxial cables and connectors that we have are not good reference by inserting a flexible, 0.1 mm thin brass sleeve,
enough to measure above 10 GHz due to very high shown in Fig. 6(a). This sleeve forms an effective reference
attenuation of rf-signals. chamber by virtue of its precise fit and good rf contact with
Our first test of the G-line performance was to measure the nominal 22 mm APS-U chamber as shown in Fig. 6(b).
the time domain reflectometry (TDR) to evaluate the Furthermore, the sleeve can be easily inserted and removed
impedance matching from the coaxial cable to the wire without disconnecting the central wire or disturbing the
and through a DUT. TDR measures the reflection coef- experimental setup. Having established the reference, we
ficient (S11 -parameter) with respect to time, from which then calibrate the network analyzer so that the S21 -signal
one can determine the degree of impedance mismatch at becomes completely flat, gently remove the sleeve, and
any point along a transmission line. Our TDR measurement then record the S21 parameter of the DUT. This measure-
is in Fig. 5, which shows a smooth impedance transition ment yields the normalized response SN21 ¼ SDUT REF
21 =S21 that
from the 50 Ω coaxial cable to that of the dielectric coated we can then convert to an impedance using appropriate
wire, and then to the characteristic impedance of the DUT at formulas, as we discuss later. Our measurements will focus
the middle of the G-line. There are small variations where
the DUT is located which we believe can be attributed to
mode conversion from TM to TEM modes at the entrance
and back again at the exit. As we discuss in the next section,
use of the reference sleeve should normalize out these
effects. We do not completely understand why the TDR
measurement indicates an impedance above 50 Ohm after
the receiver cone. One potential explanation is related to the
fact that multiple changes in impedance along a line result
in multiple reflections that need to be taken into account
[30]. For example, if the impedance changes from, say, FIG. 6. Brass sleeve; (a) without inserting into DUT, and
50 Ω to 100 Ω and then back to 50 Ω, the measured (b) partially inserted into the DUT of a nominal APS-U diameter
reflection from the second impedance step would be of 22.0 mm.

082803-5
SANGROULA, LINDBERG, LILL, and ZABEL PHYS. REV. ACCEL. BEAMS 23, 082803 (2020)

FIG. 8. Comparison between the measured data for a DUT for

two different gap spacing between G-line cones; the blue curve
corresponds to 417 mm, and the red curve corresponds to
927 mm.

(green line) to that when the sleeve length matched that of

the gate valve (black). We see that choosing the appropriate
reference sleeve eliminates the unphysical overshooting
response, resulting in an accurate measurement of the S21 .
In addition, we found that the systematic error of the
G-line measurement becomes undesirably large if the
launcher cones are too close to one another. These findings
are consistent with the simulation results of Ref. [19], who
found that the wire needs to be sufficiently long to insure
good launching and receiving efficiency. This is because
the G-line needs sufficient space between the cones and the
DUT for the surface waves to fully develop. Increasing the
wire length further can in principle lead to additional
improvements, but any benefit must be weighed against
FIG. 7. (a) A prototype gate valve for the APS-U with a brass
sleeve (yellowish cylinder) longer than its length. (b) Comparison
other sources of error and how much lab space can be
between the APS-U gate valve response for two different sleeve occupied. Figure 8 illustrates this point by comparing the
lengths; green curve represents a response for a longer sleeve, and measured response from a flange and gasket when the two
the black curve indicates the same response while using an exact G-line cones are 417 mm apart (blue line) to that when the
sleeve. cones are placed 927 mm apart (red line). In this case we
find that increasing the length between the cones by a factor
of 2 decreases the systemic error by a factor of 10. While
on the longitudinal impedance, both because we are
the reduction in noise was not as dramatic for other cases,
particularly concerned with the potential difficulties asso-
we did find at least a factor of 2 noise reduction for all other
ciated with rf heating, and because our impedance model- APS-U vacuum components.
ing [31] predicts that the primary contribution of these
vacuum components is to the longitudinal microwave
instability in the APS-U. III. BENCHMARKING OF THE G-LINE METHOD
During the course of our measurements we found that We are among the first groups to use a two-horn Goubau
their accuracy is significantly impacted by the choice of line setup to measure the coupling impedance of accelerator
reference sleeve. In particular, we discovered that using a vacuum components. Hence, we believed it was important to
sleeve whose length precisely matches that of the DUT is first test its performance using a relatively simple and well-
imperative to obtain reliable results. Our initial thought was understood cylindrically symmetric cavity. The dimensions
to use a common sleeve for all chambers, but we observed were chosen to match those of the cavity formed when two
sharp, unphysical positive peaks (sometimes referred to as quick con-flat (QCF) flanges are joined without any gasket:
“overshooting”) of the S21 -parameter at low frequencies the input and output cylindrical pipes have 11 mm radii and
when the sleeve protruded outside the DUT. Figure 7(a) are 90 mm long each, while the cavity formed by the flanges
shows a picture of one case where the sleeve was too long, is 2.54 mm long with a radius of 24.2 mm. The CAD model
while Fig. 7(b) compares measurement with this setup of this benchmarking cavity is shown in Fig. 9(a).

082803-6
MEASURING VACUUM COMPONENT IMPEDANCE … PHYS. REV. ACCEL. BEAMS 23, 082803 (2020)

FIG. 10. Simulated plots comparison to benchmark the G-line

system for a 2.54 mm wide stainless steel 316L pillbox type
cavity with and without G-line setup, where the red curve
represents normalized S21 -data with G-line setup, and the black
FIG. 9. Two separate CAD models to simulate S21 -parameter of curve represents the normalized data without the G-line system.
a 2.54 mm wide pillbox type cavity: (a) without the G-line
system, where the central black line is the dielectric coated copper circumference with copper tape to ensure good rf contact.
wire, and (b) the same cavity in the G-line system. We compare the measured and simulated S21 parameter in
Fig. 11(b), where the black curve represents the simulated
A. Benchmarking the S21 parameter response of the cavity, and the green curve is the measured
one with G-line setup.
Our first step was to determine whether the launching
and receiving horns introduced any artifacts in the meas-
urement. To do this, we compared the simulated S21 -
parameter from a simulation of the entire G-line test fixture
to that of a simplified simulation that only included the
central wire inside the DUT. A 2.54 mm wide pillbox type
cavity was chosen as a DUT and is formed by joining two
flanges together without putting any gasket in between.
Both simulations were performed using the transient solver
of CST Microwave Studio, and Fig. 9 shows the CAD
designs used for the simulation without (a) and with (b) the
G-line system. Both simulations include the dielectric
coated wire, and Fig. 9 also indicates the positions of
the input and output waveguide ports by the red planes.
We compare the normalized S21 -parameters obtained
from the two simulations in Fig. 10. The red curve plots the
normalized data obtained from the simulation of the full
G-line system, while the black curve plots the same using
the simplified DUT model. We see that the agreement is
quite good, and significantly that the position of the
resonance peaks in both cases lies at the same frequency.
This shows that the G-line launching and receiving setup
does not introduce any artifacts, and that we can in the
future compare experimental results to the simplified
simulations that only include the DUT. This allows for
simulations that have improved numerical resolution for a
reasonable computational time.
The next step was to compare the benchmarked simu- FIG. 11. (a) Chain-clamp along with two QCF-flanges to form
lation results to those obtained from experiment. As a bench-marking cavity. (b) Measured data comparison with
mentioned previously, we made the experimental cavity simulation, where the green curve represents normalized S21 -data
by joining two QCF-flanges together with a chain-clamp as with the G-line setup, and the black curve represents simulated
shown in Fig. 11(a), after which we wrapped the outer S21 -data without the G-line.

082803-7
SANGROULA, LINDBERG, LILL, and ZABEL PHYS. REV. ACCEL. BEAMS 23, 082803 (2020)

FIG. 12. Resonance peak positions found from CST simulation

for a 2.54 mm wide perfectly conducting pillbox cavity with
different radii. The green, dark blue, magenta, black, purple, and
red curves represent the radii of 24.2 mm, 24.0 mm, 23.8 mm,
23.6 mm, 23.4 mm, and 23.2 mm respectively.

We observe that the measured and simulated resonance

peaks are very similar, except that the frequency of
measured peak is about 0.1 GHz higher (∼1.7%). Since
the basic theory of pillbox cavities predict that the
frequency of the fundamental TM010 mode varies inversely
with the radial size of the cavity, we hypothesized that the
difference in the position of the resonance could be
explained by small geometric differences between the
simulated and measured cavity. We then tested the sensi-
FIG. 13. Comparison between the measured and simulated
tivity of the frequency to the geometry by varying the radial
impedances: (a) real part, and (b) the imaginary part for a bench
dimension of the cavity in our CST simulations; the marking cavity. The black curve represents the simulated imped-
resulting S21 -parameters are shown in Fig. 12, which ance using the CST wakefield solver, while the green curve
clearly depicts the 1=R dependence of the resonance peak indicates the measured impedance calculated from the HP
position. From these simulations we see that a ∼0.2 mm formula.
difference in the cavity radius could account for the
discrepancy between measured and simulated frequency
position of the cavity resonance. After these simulations we using the loss-less formula in air, Zc ¼ Z2π0 lnðd1 =d2 Þ, where
did remeasure the cavity radius with Vernier calipers, but Z0 ≈ 120π Ω ≈ 377 Ω is the impedance of free space, d1 is
did not find any such discrepancy in the dimensions. We the inner diameter of the beam tube, and d2 is the outer
therefore consider the 0.2 mm (or ∼0.1 GHz) to quantify diameter of the central conductor. There are some concerns
how accurate the G-line is, and how much confidence we regarding the validity of the lumped impedance formula
should attribute to our measurements. Eq. (6), since the presence of the wire perturbs cavity fields
and leads to a certain detuning, but for the micron sized
B. Impedance of the benchmark cavity wire that we have used it works relatively well [34].
There are several formulas [8,9,32,33] of differing We also compared this calculated impedance with
domains of applicability that relate the S21 -parameter to simulation results obtained using CST Wakefield solver.
impedance. The benchmarking cavity can be approximated The wakefield was simulated for 5000 mm behind the
as a lumped impedance structure since its width is much exciting charge to insure sufficient spectral resolution of the
less than the diameter of the beam pipe, so that the Hahn- corresponding impedance. We used CST Wakefield solver
Pedersen (HP) formula [8] applies: to directly obtain the impedance partly to improve accuracy
  and partly to verify our conversion of S21 to impedance
1 − SN21 Eq. (6). Figure 13 shows the comparison between real and
k ðωÞ
ZHP ¼ 2Zc : ð6Þ
SN21 imaginary parts of the measured and simulated impedances,
where the green curve represents the measured response
Here, Zc is the characteristic impedance of the coaxial and the black curve represents the simulated response. The
cable (reference pipe in our case), which we approximate comparison shows that the real and imaginary parts of the

082803-8
MEASURING VACUUM COMPONENT IMPEDANCE … PHYS. REV. ACCEL. BEAMS 23, 082803 (2020)

measured and simulated impedances looks very similar

except for a slight shift in resonance peak position which
accounts for ∼0.4 mm discrepancy in the radial size of the
cavity in this case. We had observed ∼0.2 mm shift while
comparing measured and simulated S21 -parameter, and the
additional 0.2 mm shift may be due to the perturbation
produced by the central conductor to represent particle
beam in the measured data since the wakefield simulation
does not have this central conductor.
It is clear from the above comparison between the
measured and simulated data that the Goubau line provides
a novel impedance measurement technique that is powerful FIG. 14. Gasket options for the APS-U lattice; A beryllium-
and accurate. The impedance formula given by Eq. (6) is copper comb rf-gasket (left), and a single conflat copper gasket
derived for a loss-less transmission line. Although the (right).
G-line system has some losses, the distance between the
cones is sufficiently small such that we can approximate it for these gaskets including their analysis in the following
as a loss-less transmission line and the impedance formula subsections.
that we used are valid [8,18].
1. Beryllium-copper comb rf-gasket
IV. MEASUREMENTS OF CRITICAL VACUUM This gasket achieves its rf-seal using thin and elastic
CHAMBER COMPONENTS beryllium-copper fingers which we refer to as a comb in
this paper. While this gasket requires an additional outer
In this section we describe the G-line measurements gasket for the vacuum seal, installation is simplified
and impedance analysis of several critical APS-U vacuum by screwing the gasket into the flange wall as shown in
components. These components were selected from the Fig. 15(a). Furthermore, the compressible fingers of this
APS-U impedance model [31], which in turn was devel-
oped by identifying and simulating impedance contribu-
tions from engineering drawings. The primary goal of our
bench measurements is to ensure that all components will
be subject to small levels of rf heating in the APS-U ring,
while at the same time verifying our impedance calcula-
tions as best as possible.

A. rf gaskets for QCF flange

Because the APS-U vacuum system calls for ∼2000
flanges, their design must be reliable and contribute as little
impedance as possible. Standard vacuum gaskets leave a
large cavitylike structure between the 22-mm diameter
chamber and the vacuum-sealing knife edge as discussed in
the section of benchmarking, which would lead to an
undesirably low threshold current for the longitudinal
microwave instability and potentially problematic levels
of rf heating. Therefore, another design is required that can
economically balance good vacuum performance with ease
of installation and impedance cost. Of three initial designs
one was rejected due to the difficulty of its installation,
while the remaining two are shown in Fig. 14. The first
design on the left is composed of two parts: an outer ring
into which the knife edge bites for the vacuum seal (not
shown), and an inner ring whose beryllium-copper comb
provide the rf shield. The second gasket on the right is a
single piece of annealed copper whose outer area allows for
the vacuum seal at the flange knife edge, while having an
inner ring that is designed to provide rf contact at (or near) FIG. 15. (a) A QCF flange with the beryllium-copper comb
the nominal chamber radius. We present measured results rf-gasket. (b) Initially measured S21 -data for this gasket design.

082803-9
SANGROULA, LINDBERG, LILL, and ZABEL PHYS. REV. ACCEL. BEAMS 23, 082803 (2020)

FIG. 16. (a) Plane view of a CST CAD model showing a FIG. 17. (a) A typical brass washer used to fill the gap
0.4 mm longitudinal gap (cyan color) between the Be-Cu comb between rf-fingers and the flange wall. (b) Comparison
rf-gasket and the side wall of QCF flange. (b) Comparison between the measured responses of the beryllium copper comb
between measured (red curve) and simulated response (black gasket with good rf-contact (dark blue curve), and poor rf-contact
curve) due to this 0.4 mm gap. (red curve).

design alleviates fears of a poor vacuum seal and/or the rf-gap with the brass washers, having the same
introduction of virtual leaks. internal diameter as that of nominal APS-U beam pipe
Surprisingly, our initial S21 measurements of the beryl- of 22 mm, one of which is shown in Fig. 17(a) and
lium-copper rf-gasket showed a resonance peak between remeasured the transmission coefficient. In this case, we
4 and 5 GHz as shown in Fig. 15(b), which was in contrast found a nearly flat S21 -parameter shown as the dark blue
to our simulation predictions. After a careful investiga- line in Fig. 17(b).
tion of this gasket, we found that the comb fingers were not Finally, we proceed to compute the complex impedance
fabricated according to specifications: the fingers were from the measured S21 -parameter for Be-Cu comb gasket
shorter than designed, which led to a ∼0.4 mm gap when there was a gap between the gasket and the flange
between the fingers and the flange wall. wall, which we compare to simulation results from CST
To verify that this lack of rf-contact was a main cause wakefield solver in Fig 18. The red line is computed from
behind the observed resonance peak, we added a 0.4 mm the experimental data using the Hahn-Peterson formula for
gap into the CST model as shown by the cyan colored area the lumped impedance structure, while the black line is
in Fig. 16(a), and compared the new simulation result with simulated impedance obtained from CST. The measured
the measured one. Figure 16(b) shows this comparison, and simulated results are in good agreement with each
where the red curve represents the measured data and the other, and all show an impedance characteristic of a cavity
black curve represents the simulated data. Comparison that is qualitatively similar to the benchmark cavity of the
shows both resonance peaks at the same frequency though previous section.
simulation overestimates the peak amplitude. We have seen that the comb gasket could be suitable
We then wanted to show that the observed resonance from an impedance perspective provided it has proper
peak could be eliminated by making a good rf-contact tolerances to ensure good rf-contact. On the other hand, the
between the rf-fingers and the flange. For this, we filled the APS-U is looking to limit its reliance on beryllium-copper,

082803-10
MEASURING VACUUM COMPONENT IMPEDANCE … PHYS. REV. ACCEL. BEAMS 23, 082803 (2020)

FIG. 18. Comparison between measured and simulated imped- FIG. 19. (a) A QCF flange fabricated without proper tolerance
ances for the Be-Cu rf-gasket: (a) Real part, and (b) Imaginary showing a circular groove at the joint surface between flange and
part. The red curve represents measured plot using the HP beam pipe. (b) Comparison between measured (red curve) and
formula, and the black curve represents simulated plot using simulated (black curve) responses for the single conflat gasket
CST wakefield solver in both (a) and (b). with flanges having grooves.

and there are some lingering concerns regarding its ease of as shown in Fig. 19(a). These two grooves prevented the
installation. Hence, at present this gasket is not the first gasket from making a good rf-contact with the flange,
choice. which in turn led to electromagnetic coupling between the
beam pipe and the flange cavity.
2. Single conflat gasket We measured the radial dimension of the circular groove
The single conflat gasket is designed to provide both to be 0.5 mm, and estimated its width to be approximately
rf-and vacuum seals. This gasket has two triangular-shaped the same. We then included this cavitylike groove in the
edges at its inner diameter, called A-joints or lips, which original CAD design and reran the simulation to cross
provide an rf-contact between gasket and flange. The check our previously measured result. In Fig. 19(b) we
circular knife-edge on the inner surface of the flange bites show that the measured resonance peak (red) is similar to
the gasket near its outer diameter and hence provides that observed in the simulation (black), except the ampli-
vacuum seal. The A-joint contains a couple of small tude of the simulated peak is smaller and at a slightly lower
notches (holes), which allow vacuum pumping between frequency than that measured. Additional simulations
the inner gasket ring and the outer knife edge. showed that a 0.5 mm difference in the radial position
The initial measurement of transmission coefficient for of knife edge can explain this discrepancy; we observed a
the single conflat gasket also showed a resonance peak similar situation for the benchmarking cavity in Section III.
between 8 GHz and 9 GHz plotted with the red curve in Later on, we tested the performance of the single piece
Fig. 19(b), which was not predicted by CST simulation. gasket using the properly designed flange without circular
After disassembly we found a small circular groove at the groove shown in Fig. 20(a). The first test used a gasket
inner welding surface between the flange and the beam pipe whose rf lip (A-joint) was chosen to be 2.7 mm wide.

082803-11
SANGROULA, LINDBERG, LILL, and ZABEL PHYS. REV. ACCEL. BEAMS 23, 082803 (2020)

FIG. 21. A real picture of a prototype BPM-bellows assembly

for the APS-U.

impedance cost of the BPM-bellows assembly to under-

stand all sources of heating including that of beam-induced
rf-heating.
Figure 21 shows a real image of a prototype BPM-
bellows assembly designed for the APS-U. The BPM is
comprised of four signal pick up buttons arranged sym-
metrically at a 12 mm distance from the beam axis. Each
button is 2 mm thick and 8 mm in diameter, while the BPM
pin dimensions have been tailored to best match the
characteristic impedance to 50 Ω. The lowest resonance
FIG. 20. (a) An appropriately toleranced flange. (b) Measured
mode due to the button geometry is expected to have a
responses of the conflat gasket with a slightly small rf lip (red and
green), and a suitably sized rf lip (blue and magenta). frequency of about 11 GHz.
We plot the measured response for BPM-bellows
assembly as the green curve in Fig. 22(a), which shows
Unfortunately, measurements for the G-line showed that no resonance peaks over the entire 10 GHz range. This
this was insufficient to provide good rf contact as shown by agrees with the CST simulations plotted as the red curve in
the red and green lines in Fig. 20(b). The rf lip width was the same Fig. 22(a). The measured response is within the
then increased to 2.8 mm, which we found both did not noise level of the Network Analyzer in the frequency range
interfere with the vacuum seal while adequately shielding of 1 GHz to 6 GHz, however it increases by nearly a factor
the flange joint as seen by the blue and magenta lines in of 2 in the frequency range of 6 GHz to 10 GHz. The RMS
Fig. 20(b). In addition, we found that increasing the chain value of the measured data is found to be 0.07 and 0.23 in
clamp torque from 13 Nm to 18 Nm eliminated a small, the frequency range of 1 GHz to 6 GHz and 6 GHz to
low-Q resonace near 5.5 GHz. 10 GHz, respectively.
The single conflat gasket has since been adopted as the To figure out the possible cause behind the increased in
flange gasket of choice, because we have found it to amplitude of the measured response beyond 6 GHz, we
provide suitable vacuum and impedance performance, remeasured the BPM-bellows response after shifting its
while being somewhat easier to install and maintain. longitudinal position by 10 mm with respect to its initial
central position. The resulting comparison is shown in
Fig. 22(b), where the black curve is the response when the
B. BPM-bellows assembly BPM-bellows is placed at the center between the two G-line
The APS-U has stringent demands on the mechanical horns, and the red curve is the response when it is shifted
stability of the beam-position monitors (BPMs) so that they longitudinally by 10 mm. The comparison shows that the
can meet tolerances of ≲1 micron rms drift per week. Any response is shifted by ∼1800 in phase above 5 GHz. In
local heating can potentially displace the buttons and another words, the frequency position of the peaks and
overwhelm these tolerances. In addition, the BPM-bellows valleys of the S21 -parameter change only when we make
assemblies contribute the second largest longitudinal change in DUT’s longitudinal position along the G-line by
impedance for the MBA lattice. Therefore, we tested the a small amount. Since this does not change any of the

082803-12
MEASURING VACUUM COMPONENT IMPEDANCE … PHYS. REV. ACCEL. BEAMS 23, 082803 (2020)

body of the gate valve is made out of stainless steel 316L.

Simulation of the impedance caused by these rf-liners is
challenging due to their intricate geometry, and our
experience at the present APS has shown that gate valves
can experience significant rf heating. Hence, careful design
and rf measurements for future gate vale liners are
important. For the prototype APS-U liner, we found that
the measured transmission coefficient showed no observ-
able resonance peaks below 10 GHz as we anticipated from
CST simulation.
The APS-U pumping cross will use a total of six vented
slots, three along the top and three along the bottom, which
will act as a beam screen from the ion pumps. The pumping
cross has the same 22 mm internal diameter as that of
nominal APS-U beam pipe, and it will be welded into a
57 mm inner diameter cross body which is again welded to
the 316 L stainless steel flanges. Both the pumping cross
and the cross body are made from 6063 aluminum. An ion
pump will be connected to one of these flanges to maintain
vacuum pressure. The measured response of the pumping
cross also showed a flat S21 over the entire 10 GHz range as
predicted by computer simulation.

V. CONCLUSION
We have described how the novel G-line can be used to
measure the coupling impedance of a wide array of vacuum
components. After a brief theoretical background we have
FIG. 22. (a) Comparison between measured, and simulated explained our measurement procedure in some detail,
S21 -parameter of a prototype BPM-bellows assembly, where the including the measurement setup and proper definition
green curve represents the average data of three measured trials, of a reference, advantages over the traditional coaxial wire
and the red curve represents the simulated data. (b) Comparison method, and our experience regarding how to reduce
between the measured responses of the same BPM-bellows systematic experimental error that we learned over the
assembly; initial central position (black curve), and 10 mm off course of the measurements. After benchmarking the
longitudinally with respect to initial central position (red curve). G-line measurement technique both from simulation
and experiment, we presented the measured results for
internals of the DUT, the signal is not purely noise, and the APS-U rf-gaskets and BPM-bellows assembly, and
seems to indicate that it is related to reflections and gave a brief summary of impedance considerations and
distortions associated with the placement of the DUT itself. measurements for the gate valve liner and pumping cross.
These effects may ultimately limit the sensitivity of the The measured results showed that the APS-U BPM-bellows
G-line impedance measurement, and warrant further analy- assembly, gate valve liner, and the pumping cross have
sis in the future. been properly designed and manufactured to specifications,
with no observable resonance peaks. On the other hand,
C. Other components impedance evaluations of several flange designs have
displayed resonances that we subsequently attributed to
We have also measured the impedance of other APS-U improper machining and/or poor tolerance control, and we
vacuum chamber components including a gate valve liner have worked to ensure future designs can be made to
and a pumping cross. The details of these components and specifications. Finally, we showed that the G-line is a
their measurement results can be found in Ref. [35]. Here, relatively simple tool to measure vacuum component
we merely give an overview of the components and impedance over a broad frequency range.
summarize the results.
An important part of the APS-U gate valve is its rf-liners,
ACKNOWLEDGMENTS
which ideally serve to electromagnetically shield the
electron beam from the large opening that houses the gate The authors would like to thank X. Sun for CST help,
valve doors. The gate valve liners have a 20 micron-thick B. Stillwell and J. Carter for providing the .stl simulation
silver coating on top of stainless steel liners. The whole files and for ordering the measurement and test equipment,

082803-13
SANGROULA, LINDBERG, LILL, and ZABEL PHYS. REV. ACCEL. BEAMS 23, 082803 (2020)

A. Brill for TDR measurement help, and H. Cease for [14] J. Musson, K. Cole, and S. Rubin, Application of Goubau
providing important guidance and support throughout this surface wave transmission line for improved bench testing
work. Mr. Sangroula is very thankful to his advisor of diagnostic beamline elements, Technical report, Thomas
Prof. Carlo Segre at Illinois Institute of Technology for Jefferson National Accelerator Facility, Newport News,
VA (United States), 2009, https://accelconf.web.cern.ch/
providing this research opportunity, and invaluable
pac2009/papers/th6rep047.pdf.
mentoring during his Ph.D. program. This work was
[15] F. Stulle and J. Bergoz, Surface waves for testing of beam
supported by the U.S. Department of Energy, Office of instrumentation, in Proceedings of the 3rd International
Science by Argonne National Laboratory under Contract Particle Accelerator Conference, New Orleans, LA,
No. DE-AC02-06CH11357. 2012 (IEEE, Piscataway, NJ, 2012), p. 10, https://pdfs
.semanticscholar.org/0473/c7cde5574ea1c5ddf53b22b6cce-
5246a2709.pdf?_ga=2.199789933.234981343.1597091577-
835814721.1596201448.
[16] G. Goubau, Surface waves and their application to trans-
[1] M. Borland, Ultra-low-emittance light sources, Synchro- mission lines, J. Appl. Phys. 21, 1128 (1950).
tron Radiat. News 27, 2 (2014). [17] F. Stulle, J. Musson, and J. Bergoz, Goubau line beam
[2] R. Hettel, Dlsr design and plans: an international overview, instrumentation testing, the benefits, Proc. IPAC 2014
J. Synchrotron Radiat. 21, 843 (2014). (2014), https://accelconf.web.cern.ch/IPAC2014/papers/
[3] M. Eriksson, J. F. van der Veen, and C. Quitmann, thpme096.pdf.
Diffraction-limited storage rings–a window to the science [18] F. Stulle, J. Bergoz, and H. W. Glock, Measurement
of tomorrow, J. Synchrotron Radiat. 21, 837 (2014). of coupling impedances using a Goubau line, in Proceed-
[4] A. W. Chao, Physics of Collective Beam Instabilities in ings of 5th Int. Beam Instrumentation Conf.(IBIC’16),
High-Energy Accelerators (Wiley, New York, 1993). Barcelona, Spain, Sept. 13-18, 2016 (JACOW, Geneva,
[5] V. Smaluk, Impedance computations and beam-based Switzerland, 2017), pp. 720–723, https://accelconf.web
measurements: A problem of discrepancy, Nucl. Instrum. .cern.ch/ibic2016/papers/wepg41.pdf.
Methods Phys. Res., Sect. A 888, 22 (2018). [19] S. Y. Kim, F. Stulle, C. K. Sung, K. H. Yoo, J. Seok, K. J.
[6] A. Faltens, E. C. Hartvig, D. Mohl, and A. M. Sessler, Moon, C. U. Choi, Y. Chung, G. Kim, H. J. Woo et al.,
An analog method for measuring the longitudinal Characterization of the Goubau line for testing
coupling impedance of a relativistic particle beam with beam diagnostic instruments, J. Instrum. 12, P12016
its environment, Technical report, Lawrence Berkeley (2017).
National Lab. (LBNL), Berkeley, CA (United States), [20] A. Sommerfeld, Über die fortpflanzung elektrodyna-
1971. mischer wellen längs eines drahtes, Ann. der Physik und
[7] M. Sands and J. Rees, Slac-Report No. pep-95, 1974. Chemie 67, 233 (1899).
[8] H. Hahn and F. Pedersen, Coaxial Wire Measurements of [21] J. Zenneck, Über die fortpflanzung ebener elektromagne-
the Longitudinal Coupling Impedance (No. BNL-50870) tischer wellen längs einer ebenen leiterfläche und ihre
(Brookhaven National Lab., Upton, NY, 1978), https:// beziehung zur drahtlosen telegraphie, Ann. Phys. (Berlin)
www.osti.gov/servlets/purl/6591101. 328, 846 (1907).
[9] L. S. Walling, D. E. McMurry, D. V. Neuffer, and H. A [22] G. Goubau, Single-conductor surface-wave transmission
Thiessen, Transmission-line impedance measurements for lines, Proceedings of the IRE 39, 619 (1951).
an advanced hadron facility, Nucl. Instrum. Methods Phys. [23] F. Harms, Elektromagnetische wellen an einem draht mit
Res., Sect. A 281, 433 (1989). isolierender zylindrischer hülle, Ann. Phys. (Berlin) 328,
[10] U. Niedermayer, L. Eidam, and O. Boine-Frankenheim, 44 (1907).
Analytic modeling, simulation and interpretation of broad- [24] J. Stratton, Electromagnetic Theory (John Wiley & Sons,
band beam coupling impedance bench measurements, Nucl. New York, 2007).
Instrum. Methods Phys. Res., Sect. A 776, 129 (2015). [25] M. Sangroula, R. Lindberg, R. Lill, and R. Zabel, Meas-
[11] F. Caspers, Impedance determination from bench mea- uring the coupling impedance of vacuum components
surements, CERN Technical Report No. CERN-PS-2000- for the APS upgrade using a Goubau line, in 9th Int.
004-RF, 2000, https://cds.cern.ch/record/437306/files/ps- Particle Accelerator Conf.(IPAC’18), Vancouver, Canada,
2000-004.pdf. 29 Apr—4 May, 2018 (JACOW, Geneva, Switzerland,
[12] G. R. Lambertson, A. F. Jacob, R. A. Rimmer, and F. 2018), pp. 3211–3214, http://accelconf.web.cern.ch/
Voelker, Techniques for beam impedance measurements ipac2018/papers/thpak005.pdf.
above cutoff, Technical report, Lawrence Berkeley Lab., [26] M. Spata, T. L. Allison, K. E. Cole, J. Musson, and J. Yan,
1990. Evaluation and correction of the non-linear distortion of
[13] M. Sangroula, R. Lill, R. Lindberg, and X. Sun, Impedance cebaf beam position monitors, Technical report, Thomas
measurement of vacuum chamber components for the Jefferson National Accelerator Facility, Newport News, VA
advance photon source (aps) upgrade, in 8th Int. Particle (United States), 2011.
Accelerator Conf.(IPAC’17), Copenhagen, Denmark, [27] G. Goubau, Surface wave transmission line, US Patent
14-19 May, 2017 (JACOW, Geneva, Switzerland, 2017), 2,685,068 A, July 27 1954.
pp. 3583–3586, https://accelconf.web.cern.ch/ipac2017/ [28] R. W. Klopfenstein, A transmission line taper of improved
papers/wepva134.pdf. design, Proceedings of the IRE 44, 31 (1956).

082803-14
MEASURING VACUUM COMPONENT IMPEDANCE … PHYS. REV. ACCEL. BEAMS 23, 082803 (2020)

[29] R. P. Hecken, A near-optimum matching section without [32] V. G. Vaccaro, Coupling impedance measurements,
discontinuities, IEEE Trans. Microwave Theory Techn. 20, Technical Report No. SCAN-9502087, 1994.
734 (1972). [33] E. Jensen, An improved log-formula for homogeneously
[30] J. A. Strickland, H. A. Zimmerman, G. Long, and G. Frye, distributed impedance, Report No. CERN-PS-RF-NOTE-
Time-Domain Reflectometry Measurements, Measurement 2000-001, 2000, https://cds.cern.ch/record/960162/files/
Concepts, 1st ed. (tektronix. Inc., Beaverton, Oregon, cer-002626446.pdf.
USA, 1970). [34] H. Hahn, Validity of coupling impedance bench measure-
[31] R. Lindberg et al., Advanced photon source upgrade ments, Phys Rev. Accel. Beams 3, 122001 (2000).
project preliminary design report, Technical Report [35] M. P. Sangroula, Coupling impedance measurement and
No. APSU-2.01-RPT-002, Argonne National Lab, analysis of critical vacuum chamber components for the
September 2017, https://www.aps.anl.gov/APS-Upgrade/ Advanced Photon Source Upgrade, Ph.D. thesis, Illinois
Documents. Institute of Technology, 2018.

082803-15