This page aims at testing the FLI MaxCam 47-10 CCD camera. This test is made so as to be objective as much as possible and relates facts that can be checked by anybody. The serial number of this camera is 020813-65. FITS or CPA frames files on request (link may be added to this page later on)
Picture 1a : The case containing the parts
Picture 2a : The connection architecture is very straightforward.
|
Item
|
Weight
|
Size
|
| The controller box |
300g
|
140x35x110 mm |
| The camera CCD head |
1.1kg
|
95mm diameter , Length=125mm |
| The AC power supply |
400g
|
140x70x35mm |
The CCD head is small and compact (Weight 1.1kg, size : 95mm diameter , Length=125mm). The cooling is achieved by an air CPU fan (the orange part).
 |
 |
 |
| 1b | 2b | 3b |
 |
 |
 |
| 4b | 5b | 6b |
The shutter size is 25 mm (picture 4b), this one is made by FLI and is located outside the CCD chamber. Nevertheless, the 20mm diameter window located behind the shutter and 15 mm above the CCD does allow a minimum unvignetted F ratio of F12.5 to reach the corner of the CCD (9.4mm from the optical axis), whereas the sides of the CCD (6.65mm from the optical axis) allows F2.25. The corner are vignetted by the 20mm entrance window : this is bug, a 25 mm window would have been much better.
The mechanical interface to attach the camera to a telescope looks well thought, there is three ways to do it : internal threat, 4 screws or external groove.
The CCD is inside the head, without the usual sealed/glued protective window, located at 1mm the silicon surface. This window is usually glued or sealed to the ceramic package. This a gain for quantum efficiency, but dust flying inside the head can reach the CCD surface directly and they are very difficult to flat field out. The CCD is located inside a tight chamber, a 20mm window at 15mm from the CCD is closing the chamber.
If the CCD got ice, the ice goes straight to the delicate silicon CCD surface and leaves spots that are going to be difficult to clean when ice evaporates. FLI says : "our Marconi arrays are mounted without cover slips which yield up to an 8% higher QE value" Yes that's right! But to flat field out shadows from dust dropped over CCD surface is really tough. To me, for an amateur camera, it does not pay off to remove the CCD cover slip, there is too many risks. Have a look here how the dust is spread on the CCD surface (simple flat field image). Time will not help not improve it, dust will fall slowly and build up.
Picture 1d : 253x349 pixels flat field subframe showing sharp dust (stick
on CCD surface), the lens used was a 50mm F4.
A small bag containing desiccant material has been seen just aside the CCD. Long term behavior is unknown, and the user may have to dry it and then open the head : this is a risk again for the dust particles that may fall on the CCD surface. This operation should be done in a clean room.
Picture 7b : The CCD controller box
Controller box is shown in Picture 7b, this one is small and light.
Nevertheless, the panel cuts for the connectors does not look very processional
and the cables can stress the PCB. There is no lock for the cables and they
can detach without notice (it happened to me).
There is an USB cable that goes to the controller box, a 10 pins RJ45 cable that ties the ccd head to the controller box. See in Picture 1c, the CCD head with the RJ45 cable and the power supply cable. Note that the power supply cable can be pull away very easily, no way to lock this cable to the head and secure the connection.
Picture 1c
The conversion factor has been measured with different illumination levels because I have noticed that the conversion factor is changing with illumination.
Plot 1b
At low light level the conversion factor (CVF) is 3e-/ADU, at 150Ke- of signal, this one is 6e-/ADU. The curve should be flat and the conversion factor remain at the same level for the whole dynamic. I recall that the CCD has a peak signal of 100Ke- per pixel (see CCD datasheet). A possible explanation can be found in the section speaking about linearity.
The noise has been measured and is 17.5e- RMS, using a conversion factor of 3.1e-/Adu (4.8 ADU rms) at 142Kpx/sec.
The bias level is around 3680 ADU. This is a little bit too high, it should be better to have it around 200 ADUs.
Picture 1e : Bias frame, this one is clean and free of major defects
Picture 2e : Bias frame : offset jump during readout
Long term bias level stability has not been achieved.
This is interesting to achieve a Fast Fourier Transform (FFT) of a Bias frame. If this bias frame would be pattern noise free (free of noise pick up or parasitic), the FFT image would only show a central peak and nothing else around (Picture 2f).
 |
 |
| Picture 1f : FFT of a bias frame from the Maxcam | Picture 2f : FFT of a perfect bias frame having pure gaussian noise |
Picture 1f shows that the Bias image is contaminated with noise pick up at different frequencies because peaks are visible all over the image. Nevertheless, the noise pattern in the Bias frame is not critical for an amateur CCD camera and the amplitude of the pattern is about 1 to 1.5 Adus.
The linearity has been measured at different level. This could be also a problem with the conversion factor being over estimated at high levels because of charge transfer problem or charge mixing. Since this is difficult to assess, because the conversion factor cannot be measured properly, we have plotted both curves Exposure time versus signal expressed in e- and signal expressed in ADU.
Plot 1g
Plot 2g
If one rely to the 2g curve and assume the conversion factor to be constant the linearity is acceptable. To me it sound more likely a problem of charge transfer or charge mixing for large charge packet.
Also, I have noticed that the CCD data start to be clipped at 37000 ADU, above the signal blooms and smear quickly as shown in image. Considering that the Analog to digital converter is able to go to 65535 ADU, it looks like that the ADC is used up to 15bits and not 16bits. The gain of 3e-/Adu is not the one to use for the CCD47-10, 2e-/Adu would have been better. Again this is not a major issue, just a small waste of the last ADC bit. In that case, assuming a conversion gain of 3e-/ADU over the whole range, the 37000 Adu * 3 e-/Adu =110000e- is consistent with the CCD datasheet that states the pixel full well tho around 100 Ke- per pixel. Again, and more likely the conversion factor is 3e-/Adu over the whole range, but the photon noise measured above 35Ke- is not correct and under estimated.
Picture 1g : How the CCD saturates (center of the frame)
Plot 3g : Diagonal cross section of the previous image : data is clipped
at 37000 adu where the CCD pixels reach the full well limit.
The whole CCD is read (using USB1) in 7sec : this is quick and really fair : 142Kpx/sec.
The CCD temperature has been reported at -21°C by the software, (ambiant
temp is +22°c ). A 120sec exposure shows a mean increase of 202 ADU per
pixel that yield to a 5.1 e-/sec/pix. The additional noise provided in a 120sec
exposure at ambient temp of +22°c is 25e- rms.
Looking to the CCD package temperature versus dark current from the CCD datasheet,
the CCD seems to be indeed at -10°C rather than -21°C. There is here
a discrepancy between the measurement and what the CCD software reports.
Plot 1h : This is the typical dark current curve of an AIMO CCD47-10 versus
the CCD package temperature.
Beware : The light tightness of the CCD head is not perfect, the capture of dark frame in a open light environment lead to an increase of 25 ADU in 120s.
Picture 1h : 120sec dark exposure
|
Pixel type
|
Value/Threshold
|
Cumulative histogram population from ADU=0
|
| coldest pixels | 82 (240e-) | 1% |
| Peak histogram | 186 (576e-) | 40% |
| hot pixels | starts at 429 ADU (1420e-) | 99% |
The dark uniformity is very good, the amount of hot pixels is lower than a Kodak CCD.
Plot 2h : Cross section of a 120s dark frame exposure.
No ice was seen on the CCD surface using a flat field at TCCD = -21.1°C (reported by software).
The middle of the CCD head is really hot (at Troom=23C) about 42°C.
This camera is not so bad, but not excellent. It has very nice pros and also drawback that can be fixed easily : the conversion factor should have been set to 2e-/adu, and the CCD cover slip not removed, the window entrance is too narrow to illuminate properly the edges of the CCD. The size, the system simplicity, the USB interface should attract customers, but they have to be aware of the limitation of this system. The Marconi CCD 47-10 is a superb CCD to be used for astronomy : this is a great value.