CMOS-testing and configuration
Abstract
The observa+on and recording of long sequences of high-quality images required for Exoclock measurements
of exoplanet transits put major demands on the observers and therefore on the instrumenta+on used. The
availability of high-quality and low-cost CMOS (or sCMOS) cameras has opened wide fields of astronomical
imaging to many amateur astronomers. However, it was recognised that the complex features now available
with these cameras presented a veritable zoo of possibili+es, some of which might well interfere with the ob-
jec+ve of high-quality photometric observa+ons, as demanded by the Exoclock project and other fields, includ-
ing variable star measurements and DSO observa+ons. The top-end CMOS and CCD cameras currently being
used by amateur and professional observers within Exoclock have thus been reviewed. In par+cular, user-con-
trolled features such as gain / contrast etc. which can interfere with high-quality photometric imaging and also
be confusing to the amateur have been assessed and explained. It is hoped to obtain four top-end cameras
from QHY in the near future that will be tested in some detail under both laboratory condi+ons and on various
telescopes of 20 – 40 cm aperture that are readily available to the Exoclock observers. The Report is primarily
aimed at the Exoclock observing community. However, it is expected that the calibra+on procedures discussed
and presented, as well as the general conclusions, will be of use to the broader amateur astronomy communi-
ty.
Introduction
This document is intended primarily to aid the tes+ng and configura+on of modern CMOS cameras. Much, if
not all, of the tes+ng can also be applied to CCD sensors but some of the aTributes and uses of CMOS sensors
discussed at the end may not be easily applicable to CCD sensors.
Whilst also relevant to colour CMOS cameras (DSLR and the like) intended for general photography, it is aimed
primarily at monochroma+c, astronomy cameras using either CMOS or sCMOS sensors. For the purposes of
this document the name CMOS will be used to indicate both CMOS and sCMOS sensors, unless otherwise indi-
cated. For readers who are unclear as to the main differences between CMOS and CCD technology it is sug-
gested that they consult the web, which contains many references on this.
The document is not intended to promote CMOS over CCD sensors since CCDs are more than suitable for Exo-
clock-type work but, since CMOS is a rela+vely new technology for more complex astronomical use, it was felt
necessary to look at this new technology in more detail with a par+cular emphasis on photometry.
Glossary of terms
1 Well The collec+on area of the silicon which converts photons to electrons.
2 Full Well The maximum number of electrons that can be contained in the Well.
3 Gain Amplifica+on applied by the electronics to the analogue signal contained in the
Well
4 ADC Analogue to digital conversion of the signal into machine-readable units
5 ADU Analogue (to) digital units. This expresses the value of the well per single digital
unit when the analogue signal is digi+sed in an analogue to digital converter
(ADC). This value depends on the digi+sa+on used-normally 12, 14 or 16 bits
(4096, 16384 or 65536 units).
6 eGain The number of electrons corresponding to a single ADU. This is Gain dependent.
7 Gain Index A number through which the manufacturer expresses Gain in dimensionless units
(e.g. as a percentage or a number).
8 Read Noise the noise added by the electronics when the analogue signal is read
9 Bias current A current applied by the electronics to each pixel to bring the value of each signal
slightly above zero
10 Dynamic range general term deno+ng the ra+o between a minimum and maximum signal
11 Noise Part of the signal which represents the uncertainty within measuring a signal
12 Offset/Black level This is a charge applied to the sensor by the electronics to ensure that no pixels
are at or below zero. This ensures that all incoming photons are converted to elec-
trons and are thus measured by the electronics
Overview and General Assumptions
Tes+ng methods and tes+ng-related theory are reviewed, with emphasis on several key areas related to CMOS
technology. However, the document is not intended to discuss basic sensor opera+on, except to the extent
that it is related to actual tes+ng or the specific configura+on of a sensor during test.
The analysis provided applies only to sensors which are linear. That is, the output increases linearly with the
number of incoming photons. These analyses are not, for example, generally applicable to intensified or to
electron-mul+plying cameras (EM-CCD).
The analyses also assume that only the dark current of the device is temperature dependent.
Several essen+al work areas are considered:
• Equipment required for Tes+ng.
• Standard Calibra+on Techniques.
• SeZng of Gain.
• SeZng of Offset/Black Level in rela+on to Gain Indices (GI).
• Tes+ng for Linearity.
• Measurement of Electronic Gain (eGain, eG), Read Noise (RN) and Full Well (FW).
• Effects of eG, RN and FW on Dynamic Range (DR).
• Prac+cal implica+ons of CMOS proper+es.
The cornerstone of the approach discussed is that it is en+rely possible to characterise the basic performance
of a camera using three simple tests-flat fields, bias frames and flat fields at varying light intensity or exposure
+me.
Please note that the test results discussed here are all based on the use of a Finger Lakes Instrumenta+on Ke-
pler 400 TVISB. It should not be assumed that all of the results outlined here will apply to all CMOS cameras.
The only way to understand the performance of other sensors/cameras is to conduct tests. The main tests are
discussed here but, given the pace at which CMOS development is taking place and the diversity of CMOS sen-
sors, it is en+rely possible that other sensors may respond differently.
A series of rather more diverse tests is planned on QHY sensors being loaned for test and evalua+on.
Equipment Required for Testing
Most tests can be conducted either on a test bench or with the camera on a telescope.
The ability to obtain accurate Flat Fields is also essen+al and this will be further discussed later on in the doc-
ument.
Tes+ng some of the prac+cal implica+ons will require that the camera be installed on a telescope to image and
measure star fields.
It is en+rely possible to conduct all of this work with the camera on a telescope, although more accurate and
reproduceable results would most likely be obtained on a test bench. The methodology discussed assumes
that the tester has an appropriate way of obtaining good dark fields, flat fields and bias frames on the tele-
scope. This is discussed below in slightly more detail.
Standard Calibration Techniques
Dark Frames
Dark frames are frames which measure the signal level which arises purely from the electronics. Some signals
contained within the dark frames are not +me dependent (mainly bias signal and read noise) but some are
+me dependent (thermal signal). Calibra+on is conducted to remove as much of these signals as possible,
leaving “behind” only those por+ons of the signal which are sta+s+cally intrinsic to the uncertainty of the sig-
nal (noise).
Flat Fields
Flat fields are exposures taken of a flat, evenly illuminated area. These flats are then used to reduce the pixel
by pixel varia+on due to response differences between individual pixels as well as dust or other light obscuring
elements which affect the response of pixels.
Bias Frames
Bias frames are frames which measure the response of a sensor at, or very close to, zero exposure +me. They
are part of a dark frame and, apart from using them to measure Read Noise, they are not used here since it is
suggested that full dark frames be taken at every +me and temperature used.
Tes+ng requires the use of calibra+on and measurement techniques which are typical of what is required in
these circumstances:
• Dark frames.
Dark frames should only be taken under the same condi+ons under which final “science” exposures will be
taken. This implies principally that no scaling should be conducted-either with +me or with temperature. This
is to avoid possible non-linear varia+ons in single pixel response, even if average values taken over many pixels
may show linearity. Indeed, these single pixel varia+ons are largely taken into account by measuring values
over many pixels (e.g. over a 500x500 matrix).
• Bias frames.
Bias Frames are used only to measure Read Noise. It is not suggested that separate bias frames be subtracted
from uncalibrated images. Rather, full dark frames (which also contain the bias signal) should be subtracted as
above.
• Flat Frames
Flat Frames are perhaps the most difficult to produce accurately and require a flat, evenly-illuminated light
source or the use of the sky just aoer sunset or just before sunrise. If ar+ficial illumina+on is used, it is proba-
bly best to use con+nuous-spectrum, incandescent sources (such as tungsten halogen) with some form of cur-
rent stabilisa+on and allow these sources to reach opera+ng temperature before beginning to use them.
Broad-spectrum LED sources can also be used but, since these s+ll do not produce fully con+nuous spectra,
some tes+ng may be required to determine their suitability.
The use of standard fluorescent tubes or compact fluorescent tubes (CFL) is not recommended.
Setting of Gain Index/eGain
Most CMOS cameras have the ability to apply different amplifica+on to the charge contained in the well, prior
to digitalisa+on. The way this is implemented depends on the approach taken by the manufacturer but, in
general terms, it is possible to apply several Gain Indices which will result in differing eGains.
Digitalisa+on varies depending on the ADC used but is typically 12, 14 or 16 bits.
This implies that ADU values can be 0-4095, 0-16383 or 0-65535.
Gain Indices are largely related to nomenclature used by the camera manufacturer but all relate a specific
number (GI), which can be expressed as a percentage or some other dimensionless number, to a specific eGain
which is expressed in e-/ADU.
For the purposes of this discussion, an increase in GI indicates an increase in amplifica+on and this implies a
reduc+on of the eG value, when expressed in e-/ADU.
For the CMOS sensor tested, several proper+es emerge which are relevant to this discussion:
• As amplifica+on increases (eG in e-/ADU decreases), RN decreases.
• As amplifica+on increases (eG in e-/ADU decreases), pixel to pixel varia1on increases.
• As amplifica+on increases (eG in e-/ADU decreases), Full Well decreases.
For some cameras, eG (via GI) can be selected from a con+nuous spectrum whilst for other cameras only dis-
crete eG values can be selected. It is suggested that eG values be selected which cover the spectrum of ampli-
fica+on available for the camera. It is also suggested that not more than 5 values be selected, purely to enable
tes+ng within reasonable +me frames. Subsequently, should the user desire it and the camera allow it, other
eG values can be considered.
Offset/Black Level Settings (OS/BL)
For each of the eG values being inves+gated, an appropriate OS/BL needs to be determined. This value should
be established at typical opera+ng condi+ons (predominantly temperature). For this work, only bias frames
need be used.
The essen+al principle here is that no bias frame should show zero values since a zero value would affect the
linearity of the sensor’s response when the bias frame is subtracted. Since a bias frame is taken with a very
short exposure, it will contain very liTle thermal current, thus dark noise can be effec+vely ignored.
At each of the eG seZngs, OS/BL needs to be adjusted so that no pixel has a zero value.
Take a bias frame with varying OS/BL values un+l no pixel shows a zero value within the en+re frame. The his-
togram should be a Gaussian looking curve just to the right of zero. Bear in mind, however, that excessive OS/
BL levels will increase the bias current and will take up space in the CMOS well.
Note these values for every eG being evaluated. From then on, all frames should be taken using these values
under the same opera+ng parameters.
Testing for Linearity
Not all eG values show the same linearity, that is a linear response with photon absorp+on. For many astro-
nomical applica+ons, such as photometry, this is an essen+al requirement and it is strongly suggested that lin-
earity tests be conducted even if the manufacturer supplies data.
For the purposes of this work, it is sufficient to take a number of dark-calibrated flat fields with varying expo-
sure +mes to cover the lowest ADU levels to the highest.
However, there are several important requirements:
• The light source should be stable; hence some on-+me must be allowed to ensure that this is so. Typ-
ically, broad spectrum LED sources can be used since the primary aspect being considered is the relat-
ive response difference between frames. Ideally, the illumina+ng light source would be a tungsten
halogen lamp which is allowed to stabilise. Even more stable light sources for scien+fic uses can obvi-
ously be employed, if available.
• Since the light intensity needs to remain constant over +me, it is essen+ally not possible to use sky
flats taken at dawn or sunset, since the light varies quickly over +me.
• At every eG (and OS/BL) seZng, dark frames will need to be subtracted. Dark frames should not be
scaled either with temperature or +me so every exposure +me will require its own calibra+on frames.
Use a full dark frame (a dark frame which also contains a bias signal) to avoid possible non lineari+es.
In those instances where the sooware used allows the subtrac+on of just a single dark frame, this is
probably sufficient. Slightly beTer results would be obtained by combining several dark frames but
this can most likely be avoided at first and may not be necessary.
• A range of exposure +mes needs to be used which allows at least 15 data points, going from low to
high ADU levels. The actual ADU range will vary depending on digi+sa+on (12, 14 or 16 bit). In most
instances it will be possible to use the same set of exposure +mes for all eG values.
• The exposure +mes can be scaled so that at the lowest amplifica+on factor (highest e-/ADU) the
frames go to satura+on at the longest exposures. As eG/amplifica+on increases (e-/ADU decreases)
the curves will simply reach satura+on at lower exposure +mes. However, it is important that the re-
sponse be adequately sampled so some trial and error may be necessary.
• To minimise thermal signal, it is suggested that exposure +mes be kept rela+vely short, going perhaps
from a few tenths of a second to a few seconds. This also has the desirable effect of shortening over-
all tes+ng +me.
• It is best to measure a central por+on of the frame to ensure that amplifica+on noise (and other edge
effects) are minimised. A 500x500 central pixel area would be sufficient.
At every combina+on of eG and exposure +mes the following parameters need to be noted/measured on the
dark calibrated frame using a suitable sooware measuring tool:
• Exposure +me.
• ADU maximum, minimum, mean, median and standard devia+on.
Note that, as eG/amplifica+on increases, pixel to pixel varia+on may also increase resul+ng in higher standard
devia+ons.
For every eG, mean ADU level can then be graphically ploTed versus exposure +me.
The resul+ng graph should look something like the graph below.
From the graph, it should be possible to determine the eG levels which show acceptable linearity. By way of
example from the plot, which is structured for 16 bit images (0-65535 ADU), it can be seen that the two lowest
gain indices taper off well before the maximum ADU values for 16 bit images.
At the same +me, two of the eGains (12,69 and 17,36 e-/ADU) show an impressive linearity almost to the satu-
ra+on limit, while the rest show lower but acceptable linearity.
The eGain values used in the above graph are taken from data supplied by the manufacturer but eGain and RN
can also be determined experimentally. This is the subject of the following sec+on.
eG 4,85 e-/ADU Mean (ADU) eG 6,58 e-/ADU Mean (ADU)
eG 8,87 e-/ADU Mean (ADU) eG 8,61 e-/ADU Mean (ADU)
eG 12,69 e-/ADU Mean (ADU) eG 17,36 e-/ADU Mean (ADU)
eG 24,94 e-/ADU Mean (ADU) eG 47,85 e-/ADU Mean (ADU)
70000
52500
ADU
35000
17500
0
0 1500 3000 4500 6000
Exposure Time (mSec)
Measurement of Electronic Gain (eGain, eG), Read
Noise (RN) and Full Well (FW).
Some manufacturers of CMOS cameras supply data on eG (versus a Gain Index, such as a number or percent-
age), RN (at a par+cular GI or eG) and Full Well.
However, it is possible to determine these values experimentally either to confirm the data supplied by the
manufacturer or to establish a data set, in the event that these data are not supplied.
The data can then be used to determine op+mum opera+ng parameters, in rela+on to the proper+es/perfor-
mance being sought.
Testing
It is suggested that, following the above tests for linearity, one should select an ample range of eGains to cover
all the expected opera+ng condi+ons under which the camera will be used. It is also suggested that eGains be
selected which will allow acceptable linearity (as a very rough rule of thumb, so that the sensor is linear out to
85% of its satura+on limit but tes1ng is strongly suggested). It is possible to use other eGains with lower lin-
earity for specific purposes but, generally, this should only be done when one is aware of these limita+ons.
Based on the typical proper+es of commercial CMOS sensor Gain , it is likely that a maximum of 5 eG (e-/ADU)
values will be sufficient to cover the majority of requirements.
Tes+ng is based on the Janesick method, who is considered to be the foremost expert on this topic since the
90s. His book, “Photon Transfer”, is available from SPIE. Current price in the USA is $66.
The test procedure is based on documents produced by Arne Henden in rela+on to the tes+ng of the QHY600
camera. For those interested in doing so, a search on “AAVSO QHY600” should give access to what is essen+al-
ly an exchange of communica+ons between various AAVSO members. In any event, the link follows: hTps://
www.aavso.org/qhy600-tests
Note that the method of altering exposure in the AAVSO document is different to that suggested for the test in
this document. However, for the purposes of the accuracy required here, the method suggested in this docu-
ment seems to have proven sufficiently accurate.)
Some use has also been made of the latest European Machine Vision Standard 1288 on the characterisa+on of
image sensors and cameras. The website is:
hTps://www.emva.org
For those wishing to look at the func+oning of sensors and cameras in much greater detail, it is suggested that
the EMVA publica+on be studied. Note that there are two publica+ons, one of which is applicable to sensors
with a linear response (the case being considered here) and those with a non-linear response.
For the par+cular eG being inves+gated, use an exposure +me which will allow a flat field exposure to achieve
values (in ADU terms) of about half of the maximum value corresponding to the digi+sa+on being used. This is
purely a rule of thumb which will apply to most sensors. Other exposures can be used if the user is certain that
the exposure is in the linear por+on of the response curve. Normally opera+ng at 50% of maximum ADU will
be in the linear part of the curve. For example, at 16 bits (0-65535) using a flat exposure +me giving about
30000 ADU will normally be in the linear por+on of the response curve.
Take 2 flat exposures at this level (flat1 and flat2) and two bias exposures (zero1 and zero2).
The formulae used to calculate the desired parameters are:
1. flatdif=flat1-flat2
2. zerodif=zero1-zero2
3. eG=(((mean(flat1)+mean(flat2))-(mean(zero1)+mean(zero2)))/(sigma(flatdif))⌃2-(sigma(zerodif))⌃2 )
4. RN= eG *sigma(zerodif)/sqrt(2)
Firstly, flat1 and flat2 are subtracted and the standard devia+on (sigma(flatdif)) of this difference frame is mea-
sured. Note that if the difference results in zero or nega+ve numbers a constant may need to be added to
properly characterise the curve of which the standard devia+on is measured. Most measuring sooware allows
this to be an op+on.
Similarly, zero1 and zero2 are subtracted and the standard devia+on (sigma(zerodif)) of this difference frame is
measured. Again, note that if the difference results in zero or nega+ve numbers a constant may need to be
added to properly characterise the curve of which the standard devia+on is measured.
Following is a table of some measured values for the Kepler400 opera+ng at 6 different eGains, along with a
test conducted on an SBIG ST-8 CCD:
Camera/Opera*ng eGain (e-/ Read Noise (e-) Effec*ve Full Dynamic Range (dB)
Mode ADU) Well (e-)
Kepler400 LG 1,85 14,94 41,20 61164 63
Kepler400 LG 2,49 11,10 37,52 45451 62
Kepler400 LG 3,70 7,56 33,43 30974 59
Kepler400 HG 1,85 2,34 4,37 9579 67
Kepler400 HG 2,49 1,59 3,51 6507 65
Kepler400 HG 3,70 1,08 3,05 4413 63
SBIG ST-8 Bin 1x1* 2,86 22,28 100000* 73
SBIG ST-8 Bin 2x2** 2,69 26,08 100000** 72
Notes:
* Nominally, the ST-8 is a 16-bit camera but if one applies the FW formula, the calculated FW would be
in excess of 180000 e-. The FW specified by SBIG is 100000 e-, sugges+ng that the true digi+sa+on is probably
15 bits.
** In binning 2x2, eG remains very similar to 1x1 sugges+ng that the FW remains essen+ally the same.
Indeed, the DR is basically unvaried. In some CCD cameras, the eG would increase meaning an effec+ve in-
crease in FW. This does not appear to be the case for the ST-8.
Effects of eG, RN and FW on Dynamic Range (DR)
To take a prac+cal example from the above table, for the Kepler400 opera+ng at a measured eG of 2,34 e-/ADU
(compared to the manufacturer supplied value of 2,57 e-/ADU), the RN was measured to be 4,37 e-. Compare
this with CCD sensors which typically have a read noise of 3 or 4 +mes this value. At a measured eG of 1,08 e-/
ADU (manufacturer supplied value 1,10 e-/ADU), RN was measured to be 3,05 e-.
From these numbers, at a given digi+sa+on, it is possible to calculate both the FW and the DR. In the case of
the Kepler, 12 bit digi+sa+on is applied (0-4095) so at a gain of 2,34 e-/ADU, the FW is calculated to be:
1. FW=Max ADU*eG
2. FW=4095*2,34
3. FW=9579 e-
As eG decreases so does FW, but the effect on DR is interes+ng, bearing in mind that DR=20log(FW/RN).
Even though FW decreases, DR can increase as eG decreases. This is because RN decreases. The effect can be
significant. Indeed, in this set of data, DR is maximised at an eG of 2,34 e-/ADU despite having a small FW of
9579 e-.
However, the interes+ng effect comes from stacking (summing) frames, for example at an eG of 2,34 e-ADU
with a stacking of 16 frames.
Camera/Opera*ng Mode Read Noise (e-) Effec*ve Full Dynamic Range (dB)
Well (e-)
Kepler400 HG 1,85 2,34 17,48* 153264** 79
Notes:
* Assuming a RN of 4,37 e- for a single frame, since RN adds in quadrature, the RN of 16 frames is:
4,37*sqrt(16)=17,48 e-.
** Since the FW of a single frame is 9579 e-, the effec+ve FW (assuming summa+on) of 16 frames is:
9579*16=153264.
This results in a significantly higher effec+ve DR of 79 dB. So, if DR is an important considera+on, stacking ex-
posures is a good way of increasing DR to levels which are comparable with or beTer than CCD cameras. In-
deed, some sooware allows this to happen essen+ally transparently by dividing a nominal integra+on into n
number of subframes.
Some of the results of doing this will be discussed in the next sec+on.
Some practical aspects and uses
CMOS sensors have some characteris+cs which can add extra func+onality, compared to CCD sen-
sors. A few are discussed here.
Stacking subframes
Following on from the above theore+cal calcula+on of the effects of stacking on DR, below are 2 examples of
the effect in prac+cal terms.
Managing Satura+on
Two exposures were taken of a star field (NGC1708), both of the same total dura+on (32 secs). One exposure
was a single exposure of 32 secs whilst the second was divided into 16x2 sec subframes, added on the fly (as
the exposure was progressing) by sooware. The single exposure showed satura+on in the star being measured
while the mul+ple subframe exposure shows the same star without satura+on. Nominally, integra+on +me
remains the same.
Since satura+on introduces a non-linearity, clearly, for photometry, it is important that both target and compar-
ison stars not be saturated and so this represents one possible way of u+lising exposures long enough to aver-
age out scin+lla+on effects whilst maintaining the target and comparison stars unsaturated.
It should be borne in mind that the single exposure invariably will show dimmer stars but the useful DR (mea-
sured as a ra+o between the dimmest star in the field with an acceptable SNR (say, greater than 10) and the
brightest unsaturated star in the field) will be higher in the sub-exposure frame.
32 sec single exposure, saturated star
16x2 sec exposures stacked-on-the-fly, star not saturated
Of course, satura+on can also be managed by reducing exposure +me and defocusing. Perhaps more impor-
tantly, the effect on useful DR is discussed more fully in the next sec+on.
Improving Useful DR
The same star field discussed above was analysed in the two different exposure modes as outlined above-a
single 32 sec exposure and a stacked-on-the-fly image consis+ng of 16x2 sec exposures.
The dimmest star in the 2 fields with an SNR a liTle over 10 was measured in terms of ADU intensity (sum of all
pixels in the measuring aperture) and the brightest, unsaturated star was also measured in the same way.
Note that the stars measured were not the same in the two fields since selec+on was made based on SNR for
the dimmest stars and highest brightness just short of satura+on for the brightest stars. The measuring aper-
ture used was 2xFWHM.
It should be borne in mind that these numbers are purely representa+ve of the specific camera and tes+ng
condi+ons used. Other cameras and different tes+ng condi+ons (e.g. the use of filters, binning, differing sky
background and star eleva+on) may well show different results.
Exposure length Intensity (ADU) SNR Dynamic range Dynamic range
(sec) (ra*o) (dB)
32 3495 13,8
32 127441 487 36,5 31
16x2 6118 13,6
16x2 2639905 3492 431 52,7
Nonetheless, it is felt that these results will be broadly typical of CMOS sensors which can use stack-on-the-fly
methods, which have low read noise and fast file read-out. Obviously, if sufficient storage capacity exists,
stacking post capture of very short exposures into a suitable sub-frame, will also work
The same approach can be used to establish the magnitude range which can be captured in a single exposure
and which maintain linearity. If done as prepara+on for a photometric session, it will ensure that target and
comparison stars remain in linear parts of the response curve. A magnitude range can be established for the
various opera+ng modes (eGain, black level) the user would normally employ.
By varying the opera+ng mode, exposure +me and the number (if any) of sub-exposures, the user can accomo-
date a number of situa+ons, ranging from very bright stars to dim.
Lucky Imaging and the Effect on FWHM
1366 images of NGC 1708 were taken at an eG 0,54 e-/ADU and a RN of 2,54 e- and at an exposure +me of 0,05
sec. These images were then classified by average FWHM over the field.
The fast read-out allows taking a large number of images in a rela+vely short +me. However, this depends on
the type of camera and opera+ng sooware and the ability of the system to capture images (capturing a stream
of images depends on many factors such as processor speed and type, type of connec+on used (e.g. USB2 or
USB3), type of media used (e.g. SSD or normal hard disk), frame rate and frame size). The Kepler 400 allows
capturing full frames at about 24 FPS in one of the opera+ng modes. In this case the effec+ve capture rate was
about 2 FPS, due to some sooware issues but the principles remain the same.
FWHM
2,9
300
Number of frames
225
150
75
0
0,0 1,0 2,0 3,0 4,0 5,0
FWHM (arcsec)
Nearly 16% of the frames showed average FWHM values beTer than or equal to 2,9 arcsec. The average value
over the rest was about 3,4 arcsec so an effec+ve improvement of 0,5 arcsec was seen for this limited number
of frames.
Several points should be made:
• The more significant effects of Lucky Imaging are normally seen over fields of only a few arc minutes. The
field of these images was about 26 arc minutes square and an average over this field was taken. Over smaller
fields it would be reasonable to expect beTer results. The test conducted here was purely to look quickly at
a broad effect and further work should be conducted.
• Using only 16% of the exposure series obviously means that a large number of exposures are not used. This
implies a high overhead. Hence, this approach will be of use where such a high overhead can be accepted.
• Clearly, the very short exposure +mes used imply that there will be a limit to how faint poten+al targets can
be. In this par+cular case, the faintest detectable stars (SNR greater than 10) were at about GMag 15,5. The
above exposures were taken with a 37 cm RC telescope.
Conclusions
This document is not exhaus+ve but hopefully it will provide enough s+mulus for a user to conduct her or his
own tes+ng and experiment with the characteris+cs of the camera being used. However, it is hoped that it
will facilitate a degree of uniformity within the Exoclock group, in par+cular amongst CMOS users.
Do look at the lis+ng of cameras being used by the ExoClock group and do add your own camera if it’s not
present: hTps://docs.google.com/spreadsheets/d/1-J0yXYvPMAilVUu5v4hXsPFr4HIZPuZe8kmo3kxh-
S0s/edit?usp=sharing
This should be seen as a live document so comments, cri+ques and sugges+ons for further work are welcome.
To facilitate receiving comments please use either the CMOS-WG Slack channel or write to the following:
Roland Casali mrcas62@gmail.com;
David Rees walnut1@easynet.co.uk;
Leon Bewersdorf bewersdorff@pm.me
