From aho@star.le.ac.uk Thu Nov 15 13:22:13 2001 Received: from artemis.le.ac.uk (artemis.le.ac.uk [143.210.16.126]) by ursa.astro.ku.dk (8.9.3+Sun/8.9.3) with ESMTP id NAA18243 for ; Thu, 15 Nov 2001 13:22:10 +0100 (MET) Received: from xmm7.xra.le.ac.uk ([143.210.40.57] helo=xmm7) by artemis.le.ac.uk with smtp (Exim 3.16 #1) id 164LWW-0000GO-00; Thu, 15 Nov 2001 12:21:40 +0000 Message-ID: <004701c16dd0$7ac2daa0$3928d28f@xmm7.star.le.ac.uk> From: "Andrew Holland" To: "E Hoeg Copenhagen" Cc: "Michael Perryman" , "L. Lindegren" Subject: Re: CCD response Date: Thu, 15 Nov 2001 12:24:28 -0000 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 4.72.3110.1 X-MimeOLE: Produced By Microsoft MimeOLE V4.72.3110.3 Content-Length: 5916 Status: OR I have been looking into the issue of the output node and the "gain switch" which is currently proposed for the Astro CCDs, versus the possibility of calibrating the big signal non-linearity and having a single responsivity output FET. It looks as though the output response goes non-linear for primarily two reasons (neither being the FET's own linearity being exceeded). These are a) the increased charge begining to turn on the reset FET (so that additional charge leaks away through the reset drain) and b) that some excess charge leaks back into the trailing pixel. The relative weights and onset of these effects will depend very much on the CCD output node design and the clock timings and operational voltages. Notwitstanding that I believe there is merit in trying to assess this effect in the lab and we'll undertake to do so. To get you an accurate answer is going to take some further time but it is beginning to look as though there may be some benefit to a more in-depth study for a number of reasons. Firstly if you were to use the CCD in it's non-linear region then the output FET would be optimised differently and if it is to be implamented this should be done before the breadboard CCDs are designed (or 1 node can be switchable, the other node can be designed to saturate). Secondly, on discussing this with the designers at Hirst Research (now part of MAT), they indicate that when the responsivity "switch" is turned on, the resulting CCD responsivity is complex and itself depends upon signal size. It may well be the case that using this approach could add additional complexity to the instrument which has not been fully considered. I'm getting them to send me an internal MAT report on the issue. Having said the above, I will answer the direct questions below. >As you pointed out a realistic assumption would be linearity and then >gradual flattening. The flattening could be tailored to some extent to >cover a larger range of counts. The tailoring would be made at the CCD output node design stage, but I beleive that the principle is correct. > >I wonder which assumption was made for the Fig.3.14 in the GSR ? >The assumptions are not clearly given in the GSR, nor in the GAIA >Concept & Technology Study Report in connection with Fig.3.3/8. > >Perhaps Michael or Lennart could tell us something about the >assumptions ? I do not know either the assumptions. > >The details given in GSR Sect.3.7 CCD Details are as follows. >Sect.3.7.4: maximum charge 330 000 e- for a conversion factor of >3uV/e-. It is also said that the conversion factor could be >switchable, depending on the information from the ASM. I do not >think this is a good idea because we are usually measuring many >stars at a time on each CCD, so which stars should be taken to >control the conversion factor? I beleive that MAT for the breadboard study are proposing to aim for a pixel full well capacity of 400k electrons with a variable responsivity between 2-6 uV/e-. It therefore depends wether in your report you want to use the old figure of 330k or update it to the 400k number which (until the design review is performed) is still liable to change. Personally I would define the Astro full well as 400k for the time being. > >In Fig.8.8 of GSR a saturation limit of 500 000 e- per star for >the BBP is assumed. But this means much fewer counts per pixel. >So this figure is consistent with the 330 000. > >Could you perhaps send me two realistic responses for the 9x27 >um pixel in the Astro focal plane ? One with linear, another with >gradual flattening. Remember that the output node will be linear up to the storage capacity of the pixel. Only if on-chip binning is used with large signals will the signal potentially exceed linearity. For the signals where the full well capacity has been exceeded then it will not be possible to extract photomertic accuracy from the full-well pixels, but should be possible from the wings. The linear part of the response will be linear(?). We are trying to set up a trial in the lab but that may take a few days and also will not be a representative chip. I've also asked MAT if they can provide any data from their testing of astronomical devices. Until I have a bit more information I can't predict when the onset of non-linearity will occur. I'll try to get you something within days. > >Is it correct when I assume that a charge exceeding the above 330 000 >e- in a pixel will flow into the two neighbouring pixels in the >scan direction ? >Will this make high precision astrometry impossible ? >Or could we still use the non-saturated pixel ? >That is all to say: can the charge content in these pixel be understood >quantitatively ? I understand the Astro CCDs will have anti-blooming structures (proposed for the breadboard), therefore rather than spill into the adjacent pixels the charge will just stop increasing at the pixel level and hence cannot be used to assess photometry. The pixels in the wings which are not near full well may possibly be used to assess photometry. > >The neighbouring pixels across scan would *not* be affected (?). > >In case of non-linear response the charge still stay in the pixel until >saturation is reached (?) Just to clarify, the pixel storage capacity will be exceeded at ~400k electrons and will increase no further during the TDI operation. Following exposure, the charge is binned at the serial register (e.g. 1x8 times) for a signal which does not exceed the the full well limit of a pixel, then the maximum binned signal would be 8x400k (in practice less due to the response function of the telescope). > >We may have to divide the present two central patches of 2 pixel >height into 4 patches of only 1 pixel height in order to cope >with the steep slope of the PSF near the centre. >----------------------- I hope to E-mail a curve from MAT later in the afternoon when I receive it. Andrew