\documentstyle[12pt,a4,psfig,charissatech]{article}

\setlength{\parindent}{0in}
\setlength{\parskip}{1ex}

\newcommand{\us}{~$\mu$s}
\newcommand{\um}{~$\mu$m}
\newcommand{\vct}{~V~cm$^{-1}$~$\tau^{-1}$}
\newcommand{\DE}{$\Delta$E}

\docref{CH/Tech/95/02}
\shorttitle{Hybrid Tests}
\subtitle{Report on beam tests at the ANU}
\title{Gas Hybrid Detector Tests}
\author{Steven Singer}
\date{1 July 1995}

\begin{document}

\maketitle

\section{Introduction}

A series of experiments have been performed to explore the behaviour
of the gas cell in the CHARISSA gas-silicon hybrid detectors
\cite{NC01}. The hybrids comprise a 50~mm deep longitudinal \DE\
section and a 50~$\times$~50~mm$^2$ Si strip detector with 16 strips
(each 3~mm wide) orientated parallel to the reaction plane. A
cross-section of a detector is shown in figure~\ref{hybrid}. The
experiments were performed with a view to finding ways of improving
the performance of the detectors.

\begin{figure}[t]
\centerline{\psfig{file=hybrid.eps,width=.9\textwidth}}
\caption{Cross-section of the interior of a detector. All measurements are
in mm unless otherwise stated.\label{hybrid}}
\end{figure}

The effects of all operating parameters have been quantitatively studied and
recommendations made for their optimal values. A position dependence of the
\DE\ system has been identified. Its dependence of different particle types
and grid configurations has been measured.

\section{Experiment and analysis description}

The experiments involved a series of beams elastically scattered from a gold
target. Elastic scattering was chosen to reduce the energy spread of
particles entering the detector and the scattered energy was essentially
constant over the whole detector surface.

A systematic variation of operating parameters used an $^{16}$O beam at
31.1, 48.6 and 70~MeV to study the effect they had on resolution and
efficiency.  Other runs analysed involved a $^{12}$C beam at 45~MeV (April
1994), and a $^{40}$Ca beam at 190~MeV (September 1993). Data taken by Neil
Curtis \cite{NC01,NC02} for the scattering of $^{12}$C at 84~MeV (September
1993) were also analysed for comparison.

Known problems with the design were :

\begin{itemize}
\item	The grid wires shadowed the detector, reducing its efficiency by 10\%
	per grid plane. This gave an overall reduction of 19\% for a single
	detector. With 2 detectors used in coincidence, the overall reduction
	was 34\%.
\item	The \DE\ signal appeared to vary with position across the detector.
\item	The grid wire shadowing led to dips in angular distribution
	measurements (particularly in singles data) since the grid wires
	were orientated parallel to the reaction plane.
\end{itemize}

Alterations tried were :

\begin{itemize}
\item	Using thinner grid wires.
\item	Removing the shield grid.
\item	Using a metalised PET sheet instead of a grid.
\item	Altering grid wire orientation to be parallel to the strips.
\item	Altering the operating bias.
\item	Altering the amplifier shaping time.
\end{itemize}

In the main experiment, three sets of runs were performed using different
grid configurations. The configurations are summarised in
table~\ref{tabgrid}.  For each configuration, three beam energies were used,
namely 31.1, 48.6 and 70~MeV. Shaping times of 1, 2 and 3\us\ and bias
voltages of 190, 380, 570 and 760~V corresponding 1, 2, 3 and 4\vct\ were
employed. 760~V measurements were only performed for run set 2 since this
was at the verge of electrical breakdown, and the preamplifier was placed at
risk. At 48.6~MeV all bias voltage and shaping time combinations were
tried and the optimum values inferred. For the other beam energies the two
parameters were varied individually with the other held at its inferred
optimum.

\begin{table}[t]
\begin{center}
\begin{tabular}{|c|c|c|c|c|c|}
\hline
&&\multicolumn{2}{c|}{Anode grid}&\multicolumn{2}{c|}{Shield grid}\\\cline{3-6}
Set & Detector & Wire      &             & Wire      &            \\
    &          & Thickness & Orientation & Thickness & Orientation\\\hline
1   & 1        & 100\um    & Horizontal  & 100\um    & Vertical   \\\cline{2-6}
    & 3        & 100\um    & Vertical    & 20\um     & Horizontal \\\hline
2   & 1        & 1\um\ PET & ---         & blank     & ---        \\\cline{2-6}
    & 3        & 20\um     & Vertical    & 20\um     & Vertical   \\\hline
3   & 1        & 20\um     & Horizontal  & blank     & ---        \\\cline{2-6}
    & 3        & 20\um     & Vertical    & blank     & ---        \\\hline
\end{tabular}
\end{center}
\caption{Combinations of grids used in detector tests.\label{tabgrid}}
\end{table}

\section{Analysis}

For the off-line analysis, the program divided the detector into pixels. The
pixel size was 3~mm in the Y direction (determined by the silicon strip
width) and 0.75~mm in the X direction (giving 64~$\times$~16 pixels).

For particles incident in each pixel of the silicon detector, the
corresponding \DE\ signals were averaged, and the width (standard deviation)
of the peak was extracted. In this procedure, small signals attributable to
noise were discarded. The range of channels defining the peak was determined
by inspection of the sum spectrum of all the pixels and then the individual
pixels were automatically analysed within this window.

For cases when the X-Y position dependence was not being specifically
studied, the analysis selected a range of pixels at the centre of the
detector acceptance. For analysis of signal heights, this area was
4~$\times$~1 pixels (3~$\times$~3~mm$^2$), for analysis of signal widths the
area was 20~$\times$~5 pixels (15~$\times$~15~mm$^2$). A small area was
used for the signal height averaging to reduce the effect of the systematic
variation of the \DE\ signal with position. A larger area was used for the
width as the width was roughly constant with detector position so an
averaging could be performed on many points to reduce statistical
effects. The error in the width was estimated, as in the case of the
\DE\ itself, by calculating the standard deviation of the individual
pixel measurements and dividing by the square root of the number of pixels.

\section{Effect of the diameters of the grid wires}

The two grid planes in the detectors served different purposes and could
have placed different requirements on the wire diameters. The best choice
for the shield grid was complicated because it might have operated as
either an electrostatic shield plane or as a physical barrier to positive
ions reaching the silicon detector.

The two main unwanted effects of the grid wires were their reduction of
efficiency, and their effect on angular distributions.  Four main ways of
overcoming the reduction of efficiency and the shadowing were tried :

\begin{itemize}
\item	Using thinner grid wires. Going from 100~$\mu$m wires to 20~$\mu$m
	wires would reduce the effect per grid plane from 10\% to 2\%. This
	would reduce, but not eliminate, the shadowing.
\item	Removing the shield grid. This would halve the efficiency loss in the
	detector, and also halve the number of shadows.
\item	Using an metalised PET sheet instead of wires. This would remove the
	shadowing entirely, but at a possible cost to resolution due to
	straggling in the PET sheet.
\item	Altering the wire orientation to be horizontal (parallel to the
	strips). This would have no effect on the efficiency, but since the
	strips were 3~mm wide, any shadowing should be undetectable.
\end{itemize}

Each of these changes had the anticipated effect on the shadowing, and
presumably on the efficiency. The shadowing from the 20\um\ wires when
orientated vertically was barely detectable, but still at a level that affects
angular distribution measurements. Dips in the position spectrum were observed
to be approximately 7\% for the whole, gain-matched detector. With horizontal
wires or the PET anode no shadowing was seen.

\section{Linearity}

The measured signal height was plotted against the predicted energy loss
calculated with {\it dedx} and found to show excellent linearity (see
figure~\ref{DEcalib}).

\begin{figure}[t]
\centerline{\psfig{file=set3-calib.eps,width=.9\textwidth}}
\caption{Comparison of predicted and measured \DE\ loss for run set 3.
\label{DEcalib}}
\end{figure}

The offsets for the two telescopes were 60.0 and 42.2~channels respectively.
These correspond to 0.98 and 0.58~MeV. The offsets were attributed to either
offsets in the ADCs, or base line effects in the amplifiers.

\section{Shaping time and bias voltage effects}

The effects of the shaping time and bias voltage on the height and
resolution of the \DE\ signal were investigated. Only the central portion of
the detector was used for this analysis. Since the calibration of the spectrum
was dependent on the operating parameters, the peak widths obtained in the
analysis were expressed as a percentage of the signal height. Note that this
quantity is, apart from a constant factor, the resolution in keV. Sample
results are shown in figure~\ref{shaping}.

\begin{figure}[t]
\centerline{
	\parbox[t]{.48\textwidth}{
		\centerline{\psfig{file=height.eps,width=.478\textwidth}}
		Figure~\ref{shaping}a: Signal height variation.
	}
	\hspace{.01\textwidth}
	\parbox[t]{.48\textwidth}{
		\centerline{\psfig{file=width.eps,width=.478\textwidth}}
		Figure~\ref{shaping}b: Signal width variation.
	}
}
\caption{Variation of signal height (a)  and width (b) with shaping time and
bias voltage. The values for the width in the figure are standard deviations.
\label{shaping}}
\end{figure}

It was found that the signal height at 1\us\ tended to be lower than at the
other shaping times. Under these circumstances the peak width (measured as
a percentage of signal height) also increased, but by more than would be
expected simply from the reduction in height. That is, the width measured in
channels got worse.


Similarly a bias of 1\vct\ performed poorly compared to the other biases.
The tendency towards degraded resolution was particularly evident for
1\us\ and 1\vct used together.

It is deduced that, particularly at 1\vct, the charge collection time for
the signal was becoming comparable to 1\us. This meant that not all the
charge was being collected, directly giving the reduction in signal
height. The increase in peak width is also explained if not all the charge
was collected as statistical fluctuations would be relatively more important.

The resolution performance does not distinguish between 2 and 3\us, or
between 2, 3 and 4\vct. However, the higher biases are fraught with other
dangers. During one run (in run set 2) at 4\vct, a preamplifier was damaged
by electrical breakdown and had to be replaced.  On a separate experiment,
running at approximately 3\vct, the \DE\ detector was seen to fire
continuously until its bias was reduced. This was attributed to
micro-discharges. With regard to shaping times, the choice of a 3\us\
shaping time, was rejected as there is a larger probability of pileup than
at 2\us.

It may be noted that when the preamplifier was replaced in the experiment, as
mentioned above, the signal height was seen to increase and the relative
signal width decrease by the same factor. This meant that the width in volts
stayed constant when the preamplifier was replaced.

The recommended operating parameters are the lowest bias voltage and shaping
time that give no reduction in the performance. On the basis of the recent
observations, these are 2\us\ shaping time and 2\vct\ bias.

\section{Position variation}
The variation of the \DE\ signal with position was investigated. The effect
of bias voltage, shaping time and grid configuration was analysed. The
signal height was measured as a function of detector position and these
measurements were used to build up a 3 dimensional surface showing the
variation of signal height with position. This surface had a characteristic
``armchair'' shape with ridges at 3 edges of the detector, and a sharp
fall-off towards the side of the detector where the grids were supported.
An example of this surface is shown in figure~\ref{figarmchair}. This
surface has been smoothed for display purposes so that statistical
fluctuations do not obscure trends.

\begin{figure}[t]
\centerline{\psfig{file=armchair.eps,width=.9\textwidth}}
\centerline{\psfig{file=armchair-slices.eps,width=.9\textwidth}}
\caption{Example signal height surface showing the characteristic armchair
shape (2\us\ and 2\vct\ with a 20\um\ anode grid and no shield grid).
\label{figarmchair}}
\end{figure}

The main features of these surfaces were studied conveniently by examining
orthogonal slices through the centre of the detector. The slices were either
across the strips (vertical slices) or along a strip (horizontal
slices). Vertical slices showed a `U' shape across the centre of the
detector with evidence of a ridge and subsequent fall-off at the edges. The
ridges were within 3~mm of the edges of the detector. Horizontal slices
showed a similar ridge at one edge of the detector, with the other end
showing a monotonic fall-off at the extreme edge, this latter edge being
the one closest to the grid supports. Examples of these slices are
also shown in figure~\ref{figarmchair}.

In order for a comparison to be made between different runs, the surfaces
were normalised by dividing by the height of the central point of the
detector. The major effect seen was that for runs without a shield grid the
height of the ridge was about 8\%, for those with a shield grid it was closer
to 4\%. A more detailed summary is given in table~\ref{DEpos}. It is worth
noting that for both runs in run set 1 with a bias voltage of 1\vct\ the
variation was much closer to the values of reference \cite{NC02} than the
other runs.  These measurements indicated that 20\um\ grids were as effective
as 100\um\ grids in reducing the ridge height. Slices showing the variations
are in included figure~\ref{slices} and are discussed below.

\begin{figure}[p]
\centerline{
	\psfig{file=set1-t1.eps,width=.45\textwidth}
	\hspace{.03\textwidth}
	\psfig{file=set1-t2.eps,width=.45\textwidth}}
\centerline{
	\psfig{file=set2-t1.eps,width=.45\textwidth}
	\hspace{.03\textwidth}
	\psfig{file=set2-t2.eps,width=.45\textwidth}}
\centerline{
	\psfig{file=set3-t1.eps,width=.45\textwidth}
	\hspace{.03\textwidth}
	\psfig{file=set3-t2.eps,width=.45\textwidth}}
\centerline{
	\psfig{file=calcium.eps,width=.45\textwidth}
	\hspace{.03\textwidth}
	\psfig{file=carbon.eps,width=.45\textwidth}}
\caption{Armchair slices for various grid configurations and bias voltages.
Additional information can be found in table~\protect\ref{DEpos}. Curves
labelled ``Neil's data'' refer to reference~\protect\cite{NC02}.
\label{slices}}
\end{figure}

The height of the ridge was independent of shaping time and bias voltage,
depending only on the grid configuration. The fall-off outside the ridge
additionally depended on bias voltage. For the same grid configurations,
the 2 detectors showed nominally the same height `U' shape, but showed
different fall-offs outside the ridge.

It should be pointed out, that there was an expected geometrical effect
due to path angle through the gas that would have a `U' shape. This effect
was easily calculated and was 1\% if the detector was 170~mm from the
target, or 0.5\% if the detector was 240~mm from the target.

The main experiment was performed with an $^{16}$O beam, 76~$\tau$ of propane
and the detector 170~mm from the target.

Additional measurements of the armchair have been performed on other
experiments. These are summarised on table~\ref{DEpos} along with
the results from the main experiment. Briefly, data for $^{40}$Ca ions
(5 times the $^{16}$O energy loss) and $^{12}$C ions (0.5 times the energy
loss) were consistent with the results already discussed, taking into account
the appropriate grid configurations.

\begin{table}[t]
\begin{center}
\begin{tabular}{|c|c|c|c|c|c|c|c|c|}
\hline
         &           &       &Pres-&Dist-&Shield&      &     &  \\
Date     & Beam      &Energy &sure &ance &grid  &Bias  &\DE  & Effect \\
/Set     &           & MeV   &$\tau$&mm  &\um   &V/cm/$\tau$&MeV&\% \\
\hline
Apr 94/1 & $^{16}$O  & 48.6  & 76  & 170 & 100  & 2--3 & 7.7 & 4 \\
Apr 94/1 & $^{16}$O  & 48.6  & 76  & 170 & 100  & 1    & 7.7 & 1--2 \\
Apr 94/2 & $^{16}$O  & 48.6  & 76  & 170 & 20   & 1--4 & 7.7 & 4--5 \\
Sep 93   & $^{40}$Ca & 190.0 & 120 & 240 & 100  & 1.7  & 38  & 3--4 \\
Sep 93   & $^{12}$C  & 84.0  & 120 & 240 & 100  & 1.7  & 4   & 1--2 \\
Apr 94/3 & $^{16}$O  & 48.6  & 76  & 170 & None & 1--4 & 7.7 & 8 \\
Apr 94   & $^{12}$C  & 45.0  & 60  & 170 & None & 2.5  & 3   & 8 \\
\hline
\end{tabular}
\end{center}
\caption{Summary of \DE\ position variation results.\label{DEpos}}
\end{table}

The smallest armchair effects observed were 1--2\%. There is no clear
association with any operating parameter, although both runs were with
a relatively reduced electric field. In particular, the $^{40}$Ca data (3--4\%
effect) were taken under identical conditions as the data of reference
\cite{NC02} (1--2\% effect). Note that the results shown for the Curtis data
\cite{NC02} are for a re-analysis using codes from the present work and
agree with the original analysis.

It is worth noting that, as shown in figure~\ref{shaping}, the intrinsic
resolution of the \DE\ signal (that is, the width with the systematic
position variation removed) varied between 3\% and 6\% standard deviation,
or 7\% and 14\% FWHM. For the 2\us, 2\vct\ settings on run set 3 (see
table~\ref{tabgrid}), the intrinsic widths were 12\% for
telescope 1 and 7\% for telescope 3. The overall widths, which include the
position variation were 15\% and 13\%. These are roughly consistent with what
would be expected for adding the $\pm$4\% position spread to the widths.
These figures show the degree of improvement in \DE\ resolution that can be
obtained if the position variation is corrected in the analysis.

No satisfactory explanation has yet been found for the position dependence
of the \DE\ signal.

The observed widths were also compared to those calculated for straggling of
ions in the gas. Using the program {\it strag} the straggling of $^{16}$O
ions in 76~$\tau$ propane was calculated to be 0.18~MeV~FWHM out of an
overall energy loss of 8.3~MeV (note that this disagreed with the {\it dedx}
calculation of 7.7~MeV). This FWHM was 2\% of the total energy loss which is
much smaller than the 7\% and 12\% intrinsic FWHM seen. This suggests that
the \DE\ width was dominated by electronic noise. The electronic noise was
deduced to be at least 0.5~MeV and generally in the range 0.5--0.8~MeV~FWHM
depending on the preamplifier and cabling. These results were obtained with
preamplifiers outside the 2~m vacuum chamber and approximately 2~m of
cable between the detector and preamp.

\section{Summary}

A summary of the main conclusions drawn is :

\begin{itemize}
\item	The \DE\ signal is linear with energy deposited in the gas cell.
\item	From the intrinsic width measurements, the optimum operating
	parameters for the gas cell were 2\vct\ bias and 2\us\ shaping time.
\item	There is a variation in the \DE\ signal with position in X and Y
	showing a characteristic armchair shape.
\item	The \DE\ variation with position is comparable in magnitude with
	the intrinsic resolution (that is, the resolution with the systematic
	variation removed).
\item	Grid configuration has no effect on intrinsic resolution or signal
	magnitude.
\item	Shaping time and bias have little effect on \DE\ variation with
	position.
\item	Removing the shield grid roughly doubles the \DE\ variation with
	position.
\item	Isolated examples of data showing a very small position dependence for
	\DE\ tend to correspond to reduced field strengths ($<$ 2\vct).
\item	The \DE\ intrinsic width is dominated by electronic noise.
\end{itemize}

\begin{thebibliography}{99}
\bibitem{NC01} N. Curtis {\it et al}, NIM {\bf 351A} (1994) 359--370
\bibitem{NC02} N. Curtis, private communication
\end{thebibliography}

\end{document}