next up previous contents index
Next: The CCD imager for Up: The ALIS Imager Previous: Some basic concepts   Contents   Index


Selecting an imager for ALIS

The selection process for an ALIS imager includes many considerations of scientific, technical as well as of an economical nature. Naturally this leads to various compromises. A complete ALIS imager consists of optics, filters and a detector with its electronics and supporting systems. The choice of detector determines many parameters of the other imager subsystems, therefore this choice will be discussed first.

Comparison of an ICCD with a CCD imager

An ICCD imager was considered early on in the ALIS project [for example Steen et al., 1990] and indeed many excellent CCD imagers for aurora and airglow, as, for example, the HAARP imager [Lance and Eather, 1993], the Portable Auroral Imager (PAI) [Trondsen, 1998] and the new Finnish all-sky camera [Syrjäsuo, 2001] as well as some of the imagers operated by Kaila [2003a] are based on an image-intensified CCD. Yet, it is important to realise that most CCDs have a higher $ {Q_{E}}$ than that of the photo-cathode in an image-intensifier. For example, the PAI, used for high spatial and temporal resolution auroral imaging, have a $ Q_{E_{CCD}}$ of 67 %, while the quantum efficiency of the photo-cathode is only 28%, both at $ \lambda=5577$ Å [Trondsen, 1998, p. 51]. Scientific grade CCDs might have a $ Q_{E_{CCD}}$ of up to about 90% as demonstrated in Figure 3.1. Despite this a CCD is not always preferable to an ICCD.

Figure 3.2 plots $ \mathit{SNR}$ as a function of column emission rate,

Figure: $ \mathit{SNR}$ vs. column emission, for f/3.9 optics with $ T \approx 0.5$ and 16.7 ms on-chip integration time. (a) Ideal photon detector (Equation 3.29) with $ A_{pix}=576\,\mu m^{2}$ (ALIS). (b) Equation 3.28 for ALIS ccdcam5 (Table 3.2). (c) Same as (a) but $ A_{pix}=310\,\mu m^{2}$ (PAI). (d) Equation 3.38 for the PAI ICCD (Table 3.1). (e) Same as (b) but for the PAI CCD without image intensifier. The horizontal line indicates the threshold of detection at $ \mathit{SNR}=2$.
with $ f_{\char93 }=3.9$ and a fixed integration time of 16.7 ms (NTSC video standard). The CCD data were taken from the technical specifications for the PAI (Table 3.1) as well as from the specifications for an

Table 3.1: ICCD parameters for the portable auroral imager (PAI) obtained from Trondsen [1998, Table 4.1 p.51]. All values at room temperature. Notes: a) Value from data-sheet of the CCD. b) Trondsen has confirmed the value of $ 0.1 \mathrm{mAm}^{-2}$ to be a typo.
Parameter Symbol PAI Unit Notes
MCP mean gain (medium) $ \overline{g}$ 1500
Photo-cathode dark-current $ \overline{n}_{e^{-}_{d,pc}}$ 0.1 $ nAm^{-2}$
Photo-cathode quantum efficiency $ Q_{E_{pc}}$ 28 % at 5577 Å
CCD read noise $ \langle n_{e^{-}_{r}} \rangle $ 80 $ {e^{-}_{RMS}}$
CCD dark current $ \overline{n}_{e^{-}_{d}}$ 0.1 $ \mathrm{nAm}^{-2}$ b) at $ 21^{\circ} C$
CCD full well (anti-blooming off) $ N_{e^{-}_{max}}$ 80 $ ke^{-}$ a)
CCD quantum efficiency $ Q_{E_{CCD}}$ 67 % at 5577 Å
CCD pixel area $ A_{pix}$ $ 11.7\times27.0$ $ \mu m^{2}$
Fibre-optic minification ratio $ M_{FO}$ 1.55

ALIS imager (ccdcam5, Table 3.2). Ideal photon

Table 3.2: Some parameters for the CCD in ALIS-imager ccdcam5 (SI-003AB serial No. 6144GBR10-B2) as measured by the CCD-manufacturer. All values measured at $ -45^{\circ} C$ unless otherwise noted. Notes: a) quadrant with highest value. b) same parameter, as measured by camera manufacturer. c) quadrant A 3% linearity. d) questionable assuming a lower value of about 85% (Figure 3.1) [author's note]. (See also Tables B.1-B.6 in Appendix B)
Parameter Symbol ccdcam5 Unit Notes
Read noise $ \langle n_{e^{-}_{r}} \rangle $ 8.3 $ {e^{-}_{RMS}}$ a) b) $ 7.5 @ 10.5 \mu s/\mathrm{pixel}$
Dark current $ \overline{n}_{e^{-}_{d}}$ 12.4 $ {e^{-}_{RMS}}\ s^{-1}$ at $ -15^{\circ} C$.
Full well $ n_{e^{-}_{max}}$ 316 $ ke^{-}$ c)
Quantum efficiency $ {Q_{E}}$ 89.8 % 4000 Å
Quantum efficiency $ {Q_{E}}$ 98.9 % 5500 Å, at $ -15^{\circ} C$, d).
Quantum efficiency $ {Q_{E}}$ 99.4 % 7000 Å, d).
Quantum efficiency $ {Q_{E}}$ 55.3 % 9000 Å

detectors with the same pixel area as of the ALIS Imager (a) and the PAI (c), have thresholds of detection at approximately 20 kR and 40 kR respectively. The ALIS Imager, curve (b), reaches $ \mathit{SNR}=2$ at about 100 kR which is also the case for the PAI ICCD (d). Curve (e) for the unintensified CCD of the PAI (without image intensifier) does not reach $ \mathit{SNR}=2$ until at approximately 2 MR. Clearly, this CCD is not suitable for low-light observations, without image intensifier. These results are in agreement with Trondsen [1998, Chapter 4].

At low column emission rates, the ALIS Imager and the PAI ICCD appear comparable in $ \mathit{SNR}$. However, note that this is a misleading result, as the PAI ICCD, with its frame-transfer CCD, provides data at NTSC video rates ( $ 30\ \mathrm{frames}/\mathrm{s}$), while the read-out of the ALIS quad read-out full-frame CCD is limited to about 2.8 s at the stated read noise, resulting in a maximum frame rate of about 0.3 frames/s (see Table 3.3 in Section 3.2.2). Therefore, as already noted by Trondsen [1998] it might be concluded that the ICCD is the better choice for temporal resolutions $ \le 1 s$ needed by the high temporal resolution requirements on the PAI.

While the CCD is a linear device, the image-intensifier is an electron-tube exhibiting non-linearities and aging effects making the already non-trivial task of absolute calibration even more complicated. Also image-tubes might bloom, causing the entire image to saturate, if part of the scene within the field-of-view saturates. Exposure to too bright point sources might lead to permanent image retention or intensifier damage [See for example Holst, 1998, and references therein]. For these reasons it was decided to further investigate the feasibility of using an unintensified CCD for the ALIS imagers. However, for such a system to be useful for studies of aurora and airglow, the unintensified CCD must provide acceptable $ \mathit{SNR}$ for column emission rates down to a couple of hundred Rayleigh, for integration times of about 1 s. Figure 3.3

Figure: $ \mathit{SNR}$ versus integration time for (a) 1 MR IBC-IV (b) 100 kR IBC-III, (c) 10 kR IBC-II, (d) 1 kR, IBC-I and (e) 100 R. All at 5577 Å.
displays the $ \mathit{SNR}$ plotted versus integration time for column emissions ranging from 100 R-1 MR. As seen IBC-I-IV have acceptable $ \mathit{SNR}$ for 1 s integration time. Examining Equation 3.28, it is seen that the $ \mathit{SNR}$ can be increased by increasing the number of photoelectrons, i.e. by increasing the pixel area, improving the transmittance, increasing the integration time, or by improving the quantum efficiency (see Equations 3.20 and 3.22). One possibility is to utilise on-chip binning factors (``super-pixels'') to increase the pixel area as demonstrated in Figure 3.4.
Figure 3.4: SNR vs. column exposure rates at 1 s integration time ( $ f_{\char93 }=3.5, T=0.5$) with binning factors: (a) $ 1\times1$, (b) $ 2 \times 2$, (c) $ 4 \times 4$, (d) $ 8\times 8$ and (e) $ 16 \times 16$.
As seen, the threshold of detection improves by the product of the binning factor (i.e. the increase of pixel area). On the other hand, the spatial resolution is decreased by the same factor. This also implies that the frame-rate increases, as there are less pixels to read-out. Thus, for any given measurement situation, application of on-chip binning factors provides a way to optimise a compromise between sensitivity, spatial and temporal resolution.

For high column emission rates the CCDs saturate after reaching their charge well-capacity. It should be noted that pattern noise is not taken into account in the plots. This noise would have decreased the $ \mathit{SNR}$ somewhat at high-signal levels, but is of little interest for photon-limited imaging situations.

Frame rate

The ALIS CCD is a full frame CCD. The read-noise (which increases as the square-root of the pixel-rate, mainly due to the 1/f-noise, see Section 3.1.7) in Table 3.2 was measured at a pixel read-out rate of $ 10.5 \mu s/\mathrm{pixel}$. At this rate it would take 11 s to read $ 1024^{2}$ pixels. One way of getting around this problem is to divide the CCD into sub-arrays. The ALIS CCD is divided into four sub-arrays with identical read-out channels working in parallel (Figure 3.5).
Figure 3.5: At a given pixel rate, a full frame CCD (left) takes four times as long to read-out as compared to an array divided into four sub-arrays (right), each with its own serial read-out register, sense node, amplifiers, S/H circuits and ADC:s.
Table 3.3 illustrates the maximum achievable read-out rates for the ALIS CCDs.

The frame rate, $ \mathit{FR}$, as a function of on-chip integration (``exposure'') time, $ t_{\mathit{int}}$, read-out time, $ t_{\mathit{read}}$, and misc. time before the next image can be read, $ t_{\mathit{misc}}$, (for example caused by the electronics, or the need to flush the CCD etc.) is given by:

$\displaystyle \mathit{FR}=\frac{1}{t_{\mathit{int}}+t_{\mathit{read}}+t_{\mathit{misc}}}$ (3.43)

next up previous contents index
Next: The CCD imager for Up: The ALIS Imager Previous: Some basic concepts   Contents   Index
copyright Urban Brändström