1: \documentclass{elsart}
2:
3: \usepackage{epsf}
4:
5: \begin{document}
6: \begin{frontmatter}
7: \title{Faster Background Determination\\
8: {\large \-- a method for gaining time coverage and\\
9: flux measurement accuracy with Cherenkov telescopes}}
10:
11: \author{Dirk Petry}
12: \address{Joint Center for Astrophysics, University of
13: Maryland Baltimore County,\\ 1000 Hilltop Circle, Baltimore,
14: MD 21250,\\ and NASA/GSFC, Code 661, Greenbelt, MD 20771
15: }
16:
17: \maketitle
18:
19: \begin{abstract}
20: An improved way of taking off-source data for background determination
21: in Cherenkov telescope observations is proposed. Generalizing
22: the traditional concept of taking on-source/off-source observations
23: of equal duration (e.g. 30 minutes ON followed by 30 minutes OFF),
24: {\it Faster Background Determination} (FBD) permits an
25: off-source observation with the same zenith angle distribution
26: as the on-source observation to be obtained within less time. The method
27: permits the on-source observation time to be maximized
28: without compromising the quality of the background determination.
29: It also increases the signal significance for strong sources.
30: The only modification necessary in the data acquisition is
31: a small change to the tracking algorithm.
32: The only modification necessary in the data analysis is to introduce
33: a time normalization which does not increase the systematic errors.
34: The method could become the normal observing mode for
35: Cherenkov telescopes when observing strong sources.
36: \end{abstract}
37:
38: \end{frontmatter}
39:
40: \section{Introduction}
41:
42: Since their first successful application in the late 1980s
43: \cite{weekes89}, Imaging Atmospheric Cherenkov Telescopes (CTs)
44: have developed rapidly from pioneer instruments to
45: precision observatories for high-energy gamma radiation
46: with a large user community.
47: With four major new observatories under construction
48: (CANGAROO III \cite{mori}, HESS \cite{hofmann}, MAGIC
49: \cite{lorenz}, VERITAS \cite{weekes2}), it is justified to revisit
50: and optimise the standard CT data taking methods in terms of achieving the
51: best possible scientific output given the limited observation time.
52:
53: CTs can only observe at night (ideally moonless) and during good weather conditions.
54: These constraints typically lead to a total yearly observation
55: time for any one observatory site of roughly 1000 hours. Within this time,
56: the observer must perform two tasks: the observation of the (known or
57: suspected) gamma-ray source (``on-source'' observation) and an auxiliary
58: ``off-source'' observation to determine the background caused by hadronic
59: cosmic rays contained in the on-source observation.
60: Different schemes have been developed to perform the ``off-source''
61: observations. They are described in section \ref{sec-trad}.
62: All schemes have in common that they either achieve less than optimal
63: sensitivity or occupy a large fraction of the
64: total observation time, roughly 50 \%, which reduces the telescope's
65: ability to follow the light curves of rapidly variable sources.
66:
67: In this article, I propose a new method to obtain off-source observations
68: sacrificing a smaller fraction of the total observation time
69: and avoiding increased systematic uncertainties. Section \ref{sec-trad}
70: summarizes the presently used background determination methods, section
71: \ref{sec-new} describes the new method and section \ref{sec-disc}
72: discusses advantages and applications.
73:
74: \section{Traditional background determination}
75: \label{sec-trad}
76: In order to better describe the advantages of the proposed new background
77: measurement method, the methods which have been used so far are briefly
78: summarized here.
79:
80: There are essentially two traditional ways of obtaining an
81: estimate of the number
82: of background events in data from gamma-ray source observations with
83: Cherenkov telescopes: true ON/OFF observations and separately taken OFF data.
84:
85: \subsection{ON/OFF observations}
86: This was the first method ever employed in imaging atmospheric Cherenkov
87: observations \cite{weekes89}.
88: The background is determined by performing a second observation immediately
89: before or after the on-source observation (``ON run''). This second
90: observation, the ``OFF run'', has the same duration as the ON run
91: and is made on a celestial position which is the
92: same as the on-source position except that the Right Ascension is
93: shifted by the duration of the ON-run increased by the slewing time.
94:
95: The method achieves a perfect matching of the zenith angle distributions
96: in ON and OFF run. Also the atmospheric conditions are nearly perfectly
97: matched since the the runs are taken nearly at the same time.
98: The only difference between ON and OFF run are (mostly small) variations
99: in the star field and hence the night sky background noise.
100: This is to a large extent eliminated by software padding, a method
101: which uses added noise from a software random generator to equalize
102: the noise conditions between ON and OFF \cite{cawley}.
103: Field rotation spreads the star field differences over the
104: field of view in the course of observations.
105:
106: Given an observation time (duration of one of the runs) $T$,
107: a gamma event rate $R_{\gamma}$ and a background rate $R_b$
108: (after trigger or arbitrary gamma-hadron separation),
109: the significance $S$ of the event excess $X = R_{\gamma}T$
110: caused by the gamma-ray source in the on-source position is
111: $$
112: S = \frac{R_{\gamma} \sqrt{T}}{\sqrt{2 R_b + R_{\gamma}}}
113: $$
114:
115: In order to obtain the off-source observations, 50 \%
116: of the theoretically available on-source observation time has
117: to be sacrificed. Since some OFF data can be taken while the
118: source under investigation is below the minimum elevation for
119: useful observations, the reduction in on-source observation time
120: {\it for that particular source} may be somewhat less than 50\%.
121: But in any case,
122: determining the hadronic background using ON/OFF observations
123: decreases the {\it total available} observation time by 50\%.
124: Furthermore, it introduces large gaps in the time coverage
125: thereby hampering variability studies.
126:
127: \subsection{Separately taken OFF data}
128:
129: Since the hadronic background is known to be isotropic and
130: time-independent (to a good approximation at energies above
131: several 10 GeV), it is in principle possible to measure the
132: background by taking the OFF run long before or long after
133: the ON run is taken. Also, to decrease the statistical error
134: of the background measurement, more than one OFF run can
135: be used. If the analysis is not testing for the presence of a
136: new (weak) source, the same OFF run can also be used several
137: times for different ON data. Hence, less than 50\% of the total
138: observation time has to be sacrificed for OFF data.
139:
140: If a given observatory were to create a library of OFF source runs
141: on a grid of all different declinations and zenith angles
142: of interest, it could
143: in principle \-- after the completion of the library \-- stop
144: taking OFF data and re-use the data in the library for all future
145: analysis (for new source discoveries, new OFF data may have to be
146: taken for statistical reasons).
147:
148: In reality, however, modifications and aging of the
149: telescope hardware and the ever-changing general atmospheric
150: conditions cause changes in the characteristics of the data
151: which make an off-source data library obsolete within a few years \--
152: roughly the same time than it takes to compile the library.
153:
154: In case the time coverage for a particular source is to be increased,
155: exactly matching OFF data can theoretically be taken at a different time
156: by observing at exactly the same Declination and zenith angle.
157: This means sacrificing observation time of other sources and
158: has the additional drawback that,
159: due to the fact that the atmospheric conditions change on a timescale
160: of a few hours, it is never possible to reach the near-perfect matching
161: of the atmospheric conditions obtained in true ON/OFF
162: observations.
163:
164: To correct for the differences in atmospheric conditions between
165: ON data and separately taken OFF data, one can use the fraction
166: of the data at large ALPHA values (which certainly does not come
167: from the source direction) to normalize the background rate
168: $R_b$.\footnote{There are different names and prescriptions for this
169: method, but all of them are equivalent.}
170: This can be done in an integral way for the whole gamma signal
171: (e.g. \cite{aharonian99b} and references therein, \cite{catanese98}) or
172: separately for different gamma energy ranges when a spectrum has
173: to be derived (with or without assuming a correlation between the bins
174: \cite{petry02}).
175: In both cases, larger systematic errors on the integral flux
176: or spectral parameters respectively are the price for not taking
177: true ON/OFF data.
178:
179: \subsection{``Wobble mode''}
180:
181: If the diameter of the CT's field of view exceeds $4^\circ$, on-source
182: and off-source observations can be taken at the same time by observing
183: the known or suspected source position off-axis by $\approx 0.5^\circ$
184: and deriving the background
185: from the analysis of events coming from the mirror position in the other
186: half of the camera. This so-called
187: ``wobble mode'' (see e.g. \cite{aharonian99a}) gives maximum time coverage
188: but reduces the effective collection area for gamma-rays by truncating part of the
189: field of view where the air-shower images from the source direction are expected.
190: It also shifts the images of the shower maxima of the events of interest into
191: a region of the field of view which has worse optical properties leading
192: to a deterioration in the gamma-hadron separation.
193:
194: The truncation effects become negligible when a camera with a field of
195: view larger than
196: $5^\circ$ becomes available. Equipping telescopes with such large cameras, however,
197: is often financially impossible.
198:
199: In any case, the mirror position in the other half of
200: the field of view does not provide the exact same zenith angle distribution as
201: the on-source position leading to additional systematic errors in spectra.
202: One way to compensate for this, at least approximately, is to alternate the angle by
203: which the source position is shifted off-axis between
204: e.g. $+0.5^\circ$ and $-0.5^\circ$.
205:
206: \section{The new method}
207:
208: \label{sec-new}
209:
210: \subsection{Description}
211:
212: The new background determination method proposed here is a
213: generalized version of the traditional ON/OFF observations described above.
214: Instead of having ON and OFF runs of equal duration, the observer
215: chooses two parameters: the total duration $T_t$ of the two runs together and
216: the fraction $f$ of $T_t$ which is used for the OFF observation.
217: The ON observation time is then
218: \begin{equation}
219: T_{\mathrm{on}} = (1 - f) T_t
220: \end{equation}
221: while the OFF observation time is
222: \begin{equation}
223: T_{\mathrm{off}} = f T_t
224: \end{equation}
225: The ON observation is then performed with duration $T_{\mathrm{on}}$ as
226: usual. The OFF observation, however, since it has in the general case a
227: duration different from the ON observation, has to be performed
228: at a {\it different tracking speed} in order to cover the same
229: zenith angle range. The tracking speed is scaled by the
230: ratio of the run durations $c$ where
231: \begin{equation}
232: c = \frac{1-f}{f}
233: \end{equation}
234: The correction of the tracking speed can simply be achieved by
235: {\it substituting the absolute time $t$ in the tracking calculations
236: by a modified absolute time $t'$} given by
237: \begin{equation}
238: t' = t_{\mathrm{start}} + c \cdot (t - t_{\mathrm{start}})
239: \end{equation}
240: where $t_{\mathrm{start}}$ is the (unmodified) absolute
241: time at the beginning of the OFF run.
242:
243:
244: \begin{figure}
245: \leavevmode
246: \centering
247: \epsfxsize=15cm
248: \epsffile{fig1.eps}
249: \caption{\label{fig-example} Example of the application of FBD to an
250: observation of the Blazar Mkn 421 with $T_t = 60$~min, $f = 0.25$
251: ($\Rightarrow T_{\mathrm{on}} = 45$~min, $T_{\mathrm{off}} = 15$~min),
252: slewing time between runs = 1~min. {\bf Left:} The ON and OFF run seen in the
253: local Alt-Az coordinate system. {\bf Right:} The same seen on the RA-DEC
254: coordinate system. The circles indicate the typical telescope acceptance
255: for gamma-like airshowers and have 2.4$^\circ$ diameter.
256: }
257: \end{figure}
258:
259: This substitution has the effect that if $f \neq 0.5$, the tracking speed
260: is faster ($f < 0.5$) or slower ($f > 0.5$) than normal.
261: For example, if $f$ is chosen to be $0.25$ and $T_t = 60$~minutes,
262: then $T_{\mathrm{on}} = 45$~minutes, $T_{\mathrm{off}} = 15$~minutes
263: and $c = 45/15 = 3$. In this case, the telescope would be tracking
264: three times faster during the OFF run.
265: Figure \ref{fig-example} illustrates this example.
266:
267: {\it The RA/DEC coordinates used for the OFF run are the same
268: as for the traditional ON/OFF case.}
269: If $f$ is chosen to be $0.5$, the observation is a traditional
270: ON/OFF observation.
271:
272: Given the number of events after arbitrary analysis stages for
273: the ON and OFF run, $N_\mathrm{on}$ and $N_\mathrm{off}$,
274: the number of excess events (``gammas'') $X$ is calculated
275: as
276: \begin{equation}
277: X = N_\mathrm{on} - c N_\mathrm{off}
278: \end{equation}
279: and the significance $S$ of this signal is
280: \begin{equation}
281: \label{equ-sig1}
282: S = \frac{X}{\Delta X} =
283: \frac{ N_\mathrm{on} - c N_\mathrm{off}}{\sqrt{N_\mathrm{on} + c^2 N_\mathrm{off}}}
284: \end{equation}
285: The error of $c$ is negligible since it is implemented
286: by a comparatively very accurate time measurement.
287:
288: As will be shown further below, $f = 1/(c+1)$ should always be chosen
289: to be $\geq 0.5$. The modified time $t'$ is therefore always
290: faster than normal time. Hence the new method is named
291: {\it Faster Background Determination} (FBD).
292:
293: {\it Note that the fact that the telescope is not tracking a fixed point
294: in the sky during the OFF run (because it is moving faster than
295: the Earth's rotation) does not compromise the data quality.
296: Field rotation leads to a changing starfield configuration anyway,
297: also for the traditional ON/OFF case. A superimposed drift of the
298: starfield (a few degrees within 15 minutes in the typical case) will
299: not change this situation\footnote{Modern approaches to dealing with bright
300: stars in the field of view either take the affected photomultiplier tubes out
301: of the trigger logic or lower the high voltage on them but do not switch them
302: off. This happens in a computer-controlled, reproducible fashion. Therefore
303: there is no additional new precaution necessary to deal with a
304: drifting starfield as opposed to a purely rotating one. The drift of
305: the OFF starfield leads to an increase of the probability of having a
306: bright star in the field of view by about a factor up to 3. See the
307: discussion in section \protect\ref{sec-disc}.}.}
308:
309:
310: Depending on the angular diameter $d_{\gamma}$ of the part of the
311: telescope's field of view from which gamma-like shower images are
312: accepted in the data analysis,
313: a {\it minimum
314: OFF run duration has to be required} to avoid overlap of the ON
315: and OFF regions. The minimum duration $T_{\mathrm{off,min}}$ is given by
316: \begin{equation}
317: \label{equ-fov}
318: T_{\mathrm{off,min}} = \frac{d_{\gamma}}{\cos(\mathrm{DEC})} \cdot 4\ \mathrm{minutes/degree}
319: \end{equation}
320: where $\mathrm{DEC}$ is the Declination of the ON source position.
321: For a typical CT with $d_{\gamma} = 2.4^\circ$, $T_{\mathrm{off,min}}$
322: would be 9.6~min$/\cos(\mathrm{DEC})$, i.e. between 9.6~min and 15~ min for
323: $|\mathrm{DEC}| < 50^\circ$.
324:
325:
326: \subsection{Optimization}
327:
328: The new method (FBD) has two parameters which have to be chosen
329: by the observer: total observation time $T_t$ for one ON/OFF run pair
330: and the time fraction $f$ used for the OFF run.
331:
332: The choice of $T_t$ is dictated by practical considerations,
333: the timescale of changes in the atmospheric conditions and
334: the field of view of the camera (Equation \ref{equ-fov}).
335:
336: \begin{figure}[t]
337: \leavevmode
338: \centering
339: \epsfxsize=11.5cm
340: \epsffile{fig2.eps}
341: \caption{\label{fig-sigvsf} \small Assuming a background rate $R_b = 10$
342: and a run pair duration $T_t = 60$ (both in arbitrary units), the
343: significance $S$ of a gamma signal with gamma rate to background ratio
344: $r = R_{\gamma}/R_b = 0.01, 0.1, 1.0, 10.0$ is plotted versus
345: the OFF-run time fraction $f$. The curve labeled ``Locus of Maximum''
346: connects the maxima of all significance curves.}
347: \end{figure}
348:
349:
350: The optimal choice of $f$ is obviously independent of $T_t$,
351: but it depends \-- as it turns out \-- on the intensity of the
352: gamma-ray source. In order to see this, Equation \ref{equ-sig1}
353: for the significance $S$ of the gamma signal is rewritten
354: substituting
355: $$
356: c = \frac{1-f}{f},\ N_\mathrm{on} = (R_\gamma + R_b)(1-f)T_t,\
357: N_\mathrm{off} = R_b f T_t
358: $$
359: where $R_\gamma$ is the gamma event rate and $R_b$ is the
360: background event rate.
361: This gives
362: \begin{equation}
363: S = \frac{(R_\gamma + R_b)(1-f)T_t - R_b f T_t}
364: {\sqrt{(R_\gamma + R_b)(1-f)T_t + (\frac{1-f}{f})^2 R_b f T_t}}\\
365: = \frac{R_{\gamma}\sqrt{(1-f)T_t}}{\sqrt{R_\gamma + R_b + \frac{1-f}{f} R_b}}
366: \end{equation}
367: Choosing example values for $R_\gamma$, $R_b$ and $T_t$ , the
368: significance $S$ can be plotted versus $f$ in order to investigate the
369: dependence. This was done in Figure \ref{fig-sigvsf} for $T_t = 60$ with
370: $R_b = 10$ and $r = R_{\gamma}/R_b = 0.01, 0.1, 1.0, 10.0$.
371:
372:
373: The remarkable result which becomes visible in Figure \ref{fig-sigvsf}
374: is that {\it as the signal to background ratio increases, the position $f_0$
375: of the maximum in the significance curve decreases}.
376:
377: \begin{figure}[b]
378: \leavevmode
379: \centering
380: \epsfxsize=11.5cm
381: \epsffile{fig3.eps}
382: \caption{\label{fig-f0vsr} \small The optimal value of $f$ where maximum
383: gamma signal significance is obtained as a function of the ratio $r$
384: of gamma event rate and background event rate (see Equation \protect\ref{equ-f0}).
385: }
386: \end{figure}
387:
388:
389: The value $f_0$ for which the maximum significance is obtained, can be
390: analytically calculated by determining the zeros of the derivative
391: d$S$/d$f$:
392: \begin{equation}
393: \frac{\mathrm{d}S}{\mathrm{d}f} =
394: \frac{R_\gamma ( R_b + 2 R_b f + f^2 R_\gamma ) T_t}
395: {2 f \sqrt{R_b/f + R_\gamma} ( R_b + f R_\gamma ) \sqrt{(1-f)T_t}} = 0
396: \end{equation}
397: This is essentially a quadratic equation in f. One finds that one of the zeros
398: is always negative and therefore not physical in this context. The
399: remaining zero is
400: \begin{eqnarray}
401: f_0 &=& \sqrt{(R_b/R_\gamma)^2 + R_b/R_\gamma} - R_b/R_\gamma\\
402: \label{equ-f0}
403: &=& \sqrt{\frac{1}{r^2} + \frac{1}{r}} - \frac{1}{r}
404: \end{eqnarray}
405: where $r = R_{\gamma}/R_b$ is the signal/background ratio
406: (not signal/noise!) as above. Figure \ref{fig-f0vsr} shows $f_0$ as a function of $r$.
407: The figure and Equation \ref{equ-f0} show an interesting property of the
408: FBD scheme:
409:
410: {\it For each observation, there is a single optimal value of $f$ which
411: depends only on the ratio $r$ of the gamma event rate
412: and the background event rate. This value is always less than 0.5
413: but approaches 0.5 asymptotically with decreasing $r$.}
414:
415: In other words, the traditional value $f = 0.5$ is only optimal for
416: weak gamma sources. For stronger sources like the Blazar Mkn 421 during
417: a flare, one obtains a more significant signal if one
418: devotes more time to the ON than to the OFF observation,
419: i.e. chooses $f<0.5$.
420:
421:
422: \section{Discussion}
423:
424: \label{sec-disc}
425:
426: The Faster Background Determination method will bring three major advantages:
427:
428: \begin{enumerate}
429:
430: \item Increased time coverage without increased systematic errors
431: in the determination of flux and spectrum compared to normal ON/OFF
432: observations.
433:
434: \item Reduction of systematic errors and simplification of data analysis
435: compared to methods using separately taken OFF data.
436:
437: \item Moderate improvement of the statistical accuracy of flux and spectral
438: measurements since the significance of the gamma signals is maximized for
439: a given total observation time.
440:
441: \end{enumerate}
442: The most important are points 1 and 2.
443:
444: Concerning point 3, one can show that the maximum possible increase in significance
445: when using FBD instead
446: of normal ON/OFF observations is only dependent on the signal to
447: background ratio $r = R_{\gamma}/R_b$ and is described by
448: the following formula:
449: \begin{equation}
450: \label{equ-sincvsr}
451: \frac{S_{\mathrm{FBD}}}{S_{\mathrm{norm}}}
452: = \sqrt{2 (1 + \frac{2}{r}) \frac{1 + r - \sqrt{r + 1}}{1 + r + \sqrt{r + 1}}}
453: \end{equation}
454: where $S_{\mathrm{FBD}}$ is the significance obtained by making an
455: ON/OFF observation in FBD mode using the optimal OFF time fraction $f_0$
456: (Equation \ref{equ-f0}) and $S_{\mathrm{norm}}$ is the significance
457: when the OFF time fraction $f = 0.5$ is used instead (normal ON/OFF run).
458:
459: \begin{figure}[htb]
460: \leavevmode
461: \centering
462: \epsfxsize=10cm
463: \epsffile{fig4.eps}
464: \caption{\label{fig-sincvsr} \small The improvement of the gamma signal significance
465: when using FBD instead of normal ON/OFF runs. (See Equation \protect\ref{equ-sincvsr}.)
466: }
467: \end{figure}
468:
469: Figure \ref{fig-sincvsr} shows the ratio described by Equation \ref{equ-sincvsr}.
470: It is always larger than unity, i.e. FBD is always better than normal ON/OFF
471: if the optimal OFF time fraction $f_0$ is used.
472:
473: \begin{table}[hb]
474: \caption{\label{tab-comp} \small Properties of the background determination
475: methods under discussion.}
476: {\tiny
477: \begin{tabular}{l|c|c|c|c}
478: & Traditional ON/OFF & Separate OFF & Wobble$^*$ & FBD \\
479: \hline
480: Max. possible & $\approx$ 60\% & 100\% & 100\% & $\approx$ 80\% \\
481: \ one-source time & & & \\
482: \ coverage & & & \\
483: \hline
484: Max. possible & 50\% & $\approx$ 66\% & 100\% & $\approx$ 80\% \\
485: \ all-source time & & & \\
486: \ coverage & & & \\
487: \hline
488: Statistical Errors & standard & smaller & smaller & smaller \\
489: \ for same overall & & by $\leq$ 30 \% & by $\leq$ 30 \% & by $\leq$ 15 \%\\
490: \ obs. time & & & \\
491: \hline
492: Systematic Errors & minimal & larger & larger & minimal \\
493: \hline
494: Application & precision & new source & new source & precision \\
495: & measurements & search & search, & measurements \\
496: & of weak sources & & multi-$\lambda$ & any source strength, \\
497: & & & campaigns & multi-$\lambda$ campaigns \\
498: \hline
499: \end{tabular}
500:
501: $^*$only possible if diameter of camera field of view included in trigger $> 4^\circ$
502: }
503: \end{table}
504:
505:
506: Of course, for unknown sources, the signal to background ratio is unknown and
507: hence $f_0$ cannot be determined. However, as one can see from Figure \ref{fig-sigvsf},
508: the maxima of the significance curves are broad and an approximate value
509: for $f$ already gives good results.
510: This also means that using FBD, the time coverage can be increased significantly
511: in exchange for only a small decrease in sensitivity.
512:
513: Generally, when in discovery mode where accuracy of flux measurements is not
514: the primary concern, observations with separately taken OFF data or
515: wobble mode observations may be an
516: equally good way to find a new source. But as soon as the presence of the source is
517: established and accurate flux and spectral measurements are of interest,
518: FBD is the method of choice.
519:
520: Due to the increase in the sky area covered by the OFF observation using FBD,
521: the likelihood that bright stars occur in the OFF region increases
522: by a factor up to $\approx$ 3 depending on the choice of the OFF
523: run time fraction $f$. Figure \ref{fig-example} already
524: shows the mildly extreme case with $f=0.25$. Near the
525: galactic plane, CTs have traditionally
526: had problems with their background determination due to the
527: presence of many bright stars. The FBD
528: method will only slightly worsen an already difficult problem.
529: CTs have to work with low photomultiplier gain and the above mentioned
530: dynamical lowering of high voltage values and modification of the
531: trigger map to make progress here.
532:
533: The value of the signal to background ratio $r$ depends on the gamma-hadron
534: separation capability of the telescope and the state of the source.
535: FBD seems to be particularly helpful for the observation of Blazars
536: because (a) they reach the highest values of $r$, (b) they have
537: unproblematic starfields surrounding them as most of them are sufficiently far
538: away from the galactic plane, and (c) due to their variability, time-coverage
539: is of interest.
540: The observation schedules of all CT observatories have always made
541: a special effort to dedicate large fractions of the observation time to
542: flaring Blazars. For example, in 1997, when Mkn 501 showed an unprecedented
543: flaring state of several months duration, observatories were dedicating
544: more than 50 \% of the available time to this source.
545:
546: Finally, for very much the same reasons that FBD is beneficial for
547: Blazar observations, it will also be beneficial for gamma-ray burst
548: follow-up. Choosing $f$ very low (possibly as low as $0.1$), it will be
549: possible to maximize the on-source time without compromising the quality
550: of the background determination.
551:
552: Table \ref{tab-comp} summarizes the properties of FBD and the traditional
553: background determination methods.
554:
555:
556: \begin{thebibliography}{99}
557:
558: \bibitem{aharonian99a} Aharonian, F., et al., 1999, A\&{}A, 342, 69
559: \bibitem{aharonian99b} Aharonian, F., et al., 1999, A\&{}A, 349, 29
560: \bibitem{catanese98} Catanese, M., et al., 1998, ApJ, 501, 616
561: \bibitem{cawley} Cawley, M. F., 1993, in Lamb, R.C. (ed.) ``Towards a Major
562: Cherenkov Detector II (Calgary, Canada)'', 176
563: \bibitem{hofmann} Hofmann, W., et al., 2001, Proc. 27th ICRC (Hamburg), OG 2.05, 2785
564: \bibitem{lorenz} Lorenz, E., et al., 2001, Proc. 27th ICRC (Hamburg), OG 2.05,
565: 2789
566: \bibitem{mori} Mori, M., et al, 2001, Proc. 27th ICRC (Hamburg), OG 2.05,
567: 2831
568: \bibitem{petry02} Petry, D., et al., 2002, ApJ, 580, 104
569: \bibitem{weekes89}
570: Weekes, T.C., et al., 1989, ApJ, 342, 379
571: \bibitem{weekes2}
572: Weekes, T.C., et al., 2002, Astropart. Phys., 17, 221
573:
574: \end{thebibliography}
575:
576: \end{document}
577: