LTTechNote1 TelescopeThroughput
LTTechNote1 TelescopeThroughput
Abstract: We measure the absolute photon efficiency of the Liverpool Telescope using ob-
servations of photometric standard stars and a throughput model of the IO:O instrument.
Over a period of 1 year following mirror re-coating the telescope transmission is in the
range 73–79% over 4000–10000Å. Transmission in the u0 -band (∼ 3500Å) is slightly lower
at 65%. Immediately following mirror re-coating the throughput is shown to be consistent
within 4% to predictions for two reflections from bare-aluminium-coated optics. The tele-
scope throughput (two reflections) degrades at a rate ≈ 0.0002 mag day−1 . Also presented
are the effective wavelengths and widths for the u0 , g 0 , r0 , i0 , z 0 , B, V broad-band filters, tak-
ing into account all optical components of the system.
1 Summary of Results
The telescope throughput was measured using observations of standard stars with the IO:O camera, tak-
ing into account the characteristics of the CCD quantum efficiency, filter transmission, cryostat window
transmission and the telescope effective area. The results are summarized in Table 1.
2 Methodology
The sensitivity of a telescope/instrument combination is generally measured in terms of the natural instru-
mental Zero Point. This is the magnitude of a star which provides one detected photo-electron per second.
In order to convert this to an absolute throughput as a per cent of the incoming flux, we need to know the
absolute incoming photon flux and make allowances for all other photon losses; atmosphere, filters, detector
QE etc. Since the photometric measurements are performed by the complete system (telescope and instru-
ment), all these components need to be included in the analysis. For this study we use the IO:O camera.
This means rather than just the telescope efficiency, this document must cover both telescope and IO:O
efficiency. The results are presented both as a measured throughput for the complete telescope-plus-IO:O
system and for the telescope alone once a model for losses in the IO:O instrument have been added back in.
The full system zero point (including telescope, atmosphere, instrument) is determined from observa-
tions of standard stars. We used the secondary standard star network established by the SDSS Calibration
Telescope[1]. Since this is based on an AB magnitude system, all the u0 , g 0 , r0 , i0 , z 0 filters by definition
give a flux density of 3631 Jy for a magnitude zero star at the top of the atmosphere. This means a zero
magnitude star provides
∆λ
F0 = 3631 (1.51 × 107 )
λ0
photons per second incident on one square metre above the atmosphere where λ0 is the effective wavelength
of the filter and ∆λ is the pass-band width.1
The strategy we follow is to:
1
Practical details in the initial establishment of the SDSS reference frame[2] mean small corrections (< 0.04 mag) need to
be applied and the zero points for u0 and z 0 are not quite 3631 Jy.
Figure 1: Upper panel compares raw instrumental magnitudes from the two extraction
strategies. Compared to large, fixed diameter apertures, the automatic isophotal ellipses
do a good job of including the entire flux of the object with an obviously much smaller
scatter. The slope of the linear fit to the data reveals < 5% non-linearity to the isophotal
photometry extractions. This is however unimportant because observations of many stars
are averaged to give an effective overall zero point for the system and corrected via an aper-
ture correction to match the large aperture extractions. The lower panel is a histogram of
the magnitude differences between the two methods, used to derive the aperture correction,
matching the mean zero point from the two methods.
• measure the one photo-electron per second zero point (ZP) for each filter by comparison to the SDSS
standards,
• derive the photon count rate (airmass = 0) which we would observe through our optical system for a
zero magnitude star,
• convert to flux density,
• compare to the expected value of 3631 Jy.
This gives the end-to-end efficiency of the entire instrument and telescope optical system and can be broken
down further to investigate the contribution of each component. The following factors are all spectrally
sensitive and were directly measured by their respective manufacturers as a function of wavelength:
• Detector QE ,
• Filter pass-bands,
• Cryostat window transmission.
All remaining throughput losses may then be attributed to reflectivity of the telescope mirrors.
3 Photometry
The IO:O instrument has a full set of SDSS style u0 , g 0 , r0 , i0 , z 0 filters and an incomplete set of Bessell filters
(B and V ). Five fields of photometric standards (Table 2) are automatically observed by the LT scheduler
Figure 2: Derivation of the colour, extinction and mirror degradation corrections which
were performed independently for each filter. Here the g 0 -band data are illustrated.
on every night expected to be photometric. All fields were drawn from the SDSS secondary standards
network. Three are also covered by the Landolt equatorial fields and are thus appropriate for use with the
Bessell filters.
Reference magnitudes in the SDSS system (msdss ) were taken from SDSS[1, 3, 4]. For Bessell B, V
filters, we use the magnitudes from the Stetson observations[5] of the Landolt equatorial fields[6]. Stetson
provides magnitudes for many more stars per field rather than the small subset selected by Landolt. Where
both authors provide a magnitude, they typically agree to better than 0.002 mag. Other than this, the B, V
data were handled the same as u0 , g 0 , r0 , i0 , z 0 and will not generally be discussed as a separate case. The
zero magnitude flux densities for the Bessell filters were taken as 4130 Jy for B and 3781 Jy for V [7, 8].
For this study we selected all observations over nine months from 2015 July 01 (mirror recoating) until
2016 March 10. Data were inspected by eye and rejected if any of the frames on that night showed obvious
cloud.
The photometry is based on SExtractor automatic isophotal ellipse counts plus an aperture correc-
tion derived from large (10 arcsec diameter) circular apertures. This combines the advantages of capturing
the total flux of large apertures with the lower noise of small apertures while being adaptive enough to handle
varying seeing conditions and PSF. As illustrated in Figure 1 there is a small difference between the circular
and isophotal apertures as a function of target brightness. This effect is relatively small (≈ 0.05 mag) and
not important as we use the ensemble average zero point from all the stars observed on any given night.
Raw instrumental magnitude is defined as
ADU × GAIN
minst = −2.5 log10 − 0.04
EXPTIME
where ADU is the SExtractor AUTO isophotal ellipse counts, GAIN = 1.755 electrons per count (mea-
sured by the photon transfer curve method on 2014 January 24), EXPTIME is the exposure integration time
in seconds and 0.04 is the aperture correction that matches the isophotal apertures to the 10 arcsec circular
aperture.
Atmospheric extinction corrections were applied on a per image basis according to the airmass of each
observation. A single typical extinction value was derived for each filter by a linear fit to a plot of msdss −
minst vs. airmass for our entire data sample. The top-left panel of Figure 2 illustrates the measured g 0 -band
extinction of 0.16 magnitudes per airmass.
Colour correction transformations were also derived from our data (top-right panel of Figure 2) and
applied in order to correct between our filter bands and the reference catalogues’ optical systems. For the
SDSS-style filters these corrections are small except z 0 for which the red cutoff is defined by the CCD, not
the filter2 .
By applying all the above corrections, for every individual observation we calculate an inferred zero
2
IO:O uses an e2V deep depletion CCD compared to the thinned SITe chip on the SDSS calibration telescope.
-0.04
-0.06
-0.08
-0.1
-0.12
-0.14
LT, this analysis
-0.16 CFHT, uncleaned
CFHT, CO2 cleaned
-0.18
3000 4000 5000 6000 7000 8000 9000 10000
Wavelength (Angstrom)
Figure 3: Measured rate at which the IO:O system zero point decays over time. In-
cluded for comparison are direct reflectometer measurements of two CFHT coating witness
samples[10] which were stored exposed in the enclosure with and without routine CO2
snow cleaning. The CFHT reflectance data have here been scaled to show the effect of two
refections.
point. This zero point is the magnitude of a star that would yield 1 photo-electron per second in the detector
above the atmosphere (i.e., airmass = 0).
All observations were then combined into a final system zero point through a two-step process. Firstly
a nightly average was obtained as a sigma-clipped mean of all the measurements on each single night. This
allows detection and rejection of poor nights which were not previously identified. These could for example
have been nights of severely dusty atmosphere. The nightly average zero points are illustrated as large red
dots in the lower panel of Figure 2. Secondly, after rejecting the most severe outliers, the final zero point
was taken as the median of the set of nightly averages.
Since the observations cover a substantial time range, it is also possible to detect a small temporal
decline in total system throughput. We attribute this primarily to tarnishing and dust accumulation on
the mirrors’ aluminium coating, but could in principle also include contributions from dust accumulation
on, or degradation of, any optical surface. This is illustrated in the lower panel of Figure 2 which shows
the inferred g 0 zero point decays at a rate of 0.00016 magnitudes per night. Figure 3 shows how the mirror
surface degradation varies with wavelength. The calibration presented here represents the average telescope
throughput over the nine months analysed.
Figure 4: View of the primary mirror with a pinhole camera at the instrument plane, ob-
tained Nov 2007 after the baffles were redesigned and updated. The cross of the secondary
support vanes is visible and the rest of the structure visible in the mirror is reflection of
steelwork on the interior of the enclosure dome. The inner and outer diameters of the active
collecting area were defined as the point where the image brightness was 50% of the mean
mirror surface brightness and are marked with red rings. The ratio of inner to outer diameter
is 0.423, precisely as defined in the optical design.
is obstructed by the secondary mirror structure; both the baffle around the mirror itself and the spider vanes
that support it. The optical design includes a central obstruction with outer diameter of 846mm which
is a factor 846/2000 = 0.423 of the diameter or 17.9% of the area of the primary. Including the support
vanes increases this to 18.5%. There is also a small amount of obstruction from thick cables (such as the
earthing strap that runs along the spider vanes and the SkycamZ camera, giving a total estimate of aperture
obstruction as 20% and the effective telescope collecting area is 2.5133 m2 .
As a check we can measure the central obstruction directly from images taken of the interior of the
telescope with a pinhole camera mounted at the instrument plane. Because a pinhole camera has effectively
infinite depth of field, the entire interior structure of the telescope is visible and the angular diameters of
the primary mirror and the secondary mirror baffle can be measured directly. They precisely match the
optical design of 17.9% in area. The pinhole camera did not offer sufficient angular resolution to measure
the smaller obstructions which contribute to our total estimated obstruction of 20%.
where S is the response curve of the filter alone. Similarly the effective wavelength (λ0 ) of the entire
instrument is given by R
S Q W λ dλ
λ0 = R
S Q W dλ
where Q and W are the response curves of the CCD and cryostat windows respectively.
The effective width (∆λ) is the width of a hypothetical perfect top-hat filter with 100% transmission and
passing the same total flux. This is simply the divisor of the above formulae for λ0 .
Z
∆λ = S Q W dλ
As previously stated a zero AB magnitude star will by definition provide Fref0 =3631 Jy at the top of the
atmosphere. The apparent flux density for a zero magnitude star in the LT observations is given by
100.4ZP λ0
FLT
0 =
(1.51 × 107 )(Collecting Area)(∆λ)
with the telescope collecting area expressed in square metres and the pass band descriptors (λ0 , ∆λ) in
Angstroms.
90
80
70
60
QE (%)
50
40
30
20
Typical CCD231-84, BI, NIMO, Astro ER1
10 Measurements of the IO:O device
Model used for IO:O analysis
0
3000 4000 5000 6000 7000 8000 9000 10000 11000
Wavelength (A)
Figure 5: The IO:O detector QE curve showing monochromatic measurements and contin-
uum model.
For the analysis here we created a model of the full QE curve for our detector by taking the ‘typical’
curve for the type of device and adjusting it to pass through the five spot measurements of our actual device.
The resulting model (Figure 5) required only a linear tilt of the QE curve by a few per cent towards the blue
to fit the data.
3230 – 10460 Å. The measured transmission is shown in Figure 6 and may be seen to have almost no effect
on the spectral response of the instrument.
Figure 6: Absolute transmission of the IOO croystat window measured by the vendor after
application of the anti-reflection coating.
4.2.3 Filters
High quality spectrophotometer measurements of all our filters were provided by the manufacturers and
shown in Figure 7. Each filter is multiplied by the CCD QE and cryostat window transmission to give the
total SED sensitivity of the instrument. The spectral response function of the telescope itself is then the
only unknown and can be derived from the photometry.
In Table 4 we present effective wavelengths and pass-band widths for each filter alone and that when
convolved with the CCD QE and cryostat window transmission.
100
90
80
70
CCD QE
Throughput (%)
60 Window
u'
50 g'
r'
i'
40
z'
30
20
10
u' g' r' i' z'
0
3000 4000 5000 6000 7000 8000 9000 10000
Wavelength (A)
100
90
80
70
Throughput (%)
60
V
B
50
Window
CCD QE
40
30
20
10
B V
0
3000 3500 4000 4500 5000 5500 6000 6500
Wavelength (A)
Figure 7: Transmission curves for the filters in IO:O. Thin lines are the measured trans-
mission of the filters alone. The magenta line, essentially flat at 98% transmission, is a
measurement of the cryostat window. The black curve is our model of the CCD QE. Thick
lined filter traces are the product of all the above terms to give the effective spectral response
of the entire instrument. The tick marks above each filter curve show the effective central
wavelengths. The effective central wavelength of z 0 alone without the CCD is ill-defined
and not shown.
In Table 5 and Figure 8 we present the measured telescope throughput and modeled throughputs of each
of the subsystem components numerically and graphically. Geometric collecting area and atmospheric
extinction are accounted for in the measurement of the instrumental zero point. CCD QE, filter and cryostat
window throughputs all contribute to the system’s effective wavelength in the conversion from photon count
to flux density. The residual discrepancy between the reference flux and the LT’s natural zero point is the
mirror reflectivity. The mean value of 76.4% in Table 5 implies two reflections each of 87.4%. Across the
visible band, the theoretical maximum reflection for bare aluminium is in the range 89–92%.
For the CCD and cryostat window the throughput is reported as the ratio of the measured count rate to the
count rate that would have been observed without that component’s losses. This was derived by numerically
integrating the models in Figure 7 over wavelength with and without each component included.
As a geometric description of the filter throughput and how it compares to a hypothetical, perfect top-hat
filter, we quote the ratio of the effective width to the full width at half maximum. The ‘Filter’ and ‘Total’
data points for Bessell B, V fall substantially below the interpolation of the SDSS filters because of these
filters’ broader shape and lower peak transmission.
Correcting the measured telescope throughput for B, V, g 0 , r0 , i0 , z 0 by the measured secular degradation,
the results parallel the prediction for perfectly polished solid aluminium, offset by −4%. This small offset
is unsurprising and may reflect either uncertainties in our measurement or a lower than theoretical coating
efficiency.
Fref
0 FLT
0 Telescope Filter CCD Window Total Atmos.
(Jy) (Jy) (%) (%) (%) (%) (%) (%)
u0 3767 2461 65.3 81.3 46.6 97.8 24.2 60.0
g0 3631 2850 78.5 86.9 76.1 98.2 51.0 86.0
r0 3631 2807 77.3 94.6 95.8 98.2 68.8 91.2
i0 3631 2642 72.8 94.7 92.1 98.4 62.4 95.5
z0 3565 2715 76.2 95.9 40.5 98.5 29.1 96.4
B 4130 3235 78.3 70.9 65.6 98.2 35.8 81.5
V 3781 2844 75.2 86.4 88.0 98.2 56.2 88.5
Table 5: Fref
0 is the reference zero magnitude flux above the atmosphere. This defines the AB
magnitude scale for the Sloan-like filters and is a derived quantity for the B, V filters. ‘Telescope’
is the measured telescope throughput. Contributions of each component in the model are also
given. Column ‘Total’ gives the end-to-end efficiency of LT-plus-IO:O compared to a 100% QE
detector on a perfect 2.513 m2 telescope with hypothetical 100% transmission top-hat filters of
the same FWHM as our actual filters. Zenith (airmass = 1) transmission through the atmosphere
is included for comparison, but not included in the end-to-end system throughput estimate.
100
90
80
70
Throughput (%)
60
50
40
Telescope (average)
30 Telescope (new)
Filter
CCD QE
20
Cryostat window
Total, All the above
10 Atmosphere
Al, two reflections
0
3000 4000 5000 6000 7000 8000 9000
Wavelength (Angstrom)
Figure 8: This plot shows the values tabulated in Table 5, the measured and modeled
throughput of each system component. ‘Telescope (average)’ is derived from hZPi in
Table 3 whereas ‘Telescope (new)’ estimates the throughput immediately after mirror re-
coating, based on ZP0 .
References
[1] Smith, Tucker, Kent et al., 2002, AJ, 123, 2121
[2] Eisenstein et al., 2006, ApJ, 167, 40
[3] J.A. Smith, private communication
[4] http://www-star.fnal.gov/NorthEqExtension ugriz/index.html
[5] http://www.cadc-ccda.hia-iha.nrc-cnrc.gc.ca/en/community/STETSON/standards/
[6] Landolt, 1992, AJ, 104, 1
[7] Allen’s Astrophysical Quantities, 4th Edition, 2001
[8] Bessell, 1979, PASP, 91, 589
[9] Bertin & Arnouts, 1996, AAPS, 117, 393
[10] Magrath, 1997, PASP, 109, 303