CCD-world: Red-sensitive CCDs with skinny pixels

[Sutherland, Scott]

The following was posted to CCD-world:

Whew!  Okay, your description is beyond my understanding of
CCD technology, so if you don't mind, I'll ask a few targeted 
questions.  (Hey, I'm only a chemist 8^))

	What do you mean by pixel height? a (5-10) x 200 um^2 pixel, or a
	depletion depth of 200 um?

I was referring to the physical height of the pixels.  The current
CCD I have been using (already mounted into a spectrograph)
is a SONY ILX511 with pixels 14 um wide and 200 um tall.  I have
found 4000 pixel (and higher) linear arrays that have pixels that
are square (e.g. 9 um x 9 um), but I need a larger pixel height
to get more pixel area and collect more light.

	You can have considerably better IR response with 200 um
depletion--above
	50% at 1000 um, for example.  See our SPIE 3649, 80-90 (1999) paper
on the
	QE of CCDs of various thicknesses. It's also on our web page
ccd.lbl.gov

I was not aware that such a high QE at 1000 nm was possible.  The best
commercial product I have seen is 40% (Princeton Instruments marketed it
a couple of years ago) and the one I am using now has a QE of ~ 20% at
1000 nm (Hamamatsu 7031).  Most of the front illuminated, non-thinned
CCD's used in spectroscopy have 3-5% QE at 1000 nm.  However, based
on the sparse information released by SONY on the ILX511, it has very low
QE at 1000 nm (QE peaks at ~450 nm).  I can get a SONY ILX526, which
has 3000 7 um pixels (200 um in height), but it has the same NIR sensitivity
problem as the ILX511.  I am looking for an analog of the ILX526 that has
enhanced IR sensitivity (I think 5-10 percent is sufficient, but I am open
to exploring higher QE options.

Are CCD's with such high NIR response commercially available?

	There are a lot of tradeoffs. To take advantage of small pixels, you
need
	a small PSF contribution from lateral diffusion in the CCD. With a
200 um
	thick CCD depleted all the way to the back and with a high bias
voltage,
	diffusion itself will contribute 6--8 um rms. If there is a
field-free
	region, then it will probably dominate the PSF, contributing an rms
width
	exactly equal to the field-free thickness. Back-illuminated thin
CCDs
	usually show fringing in the red.

Okay, this is where my knowledge ends.  First, what is PSF?  I assume
lateral diffusion is similar to blooming, but not necessarily at saturation,
correct?  What does it mean when you say 'diffusion will contribute
6-8 um rms'?  And what is meant by a field-free region? and  field-free
thickness?  What do you mean by fringing?


	For red and NIR I don't think it matters much if it is back
illuminated,
	except that the IR coating can then be optimized. But if it is
thinned to
	the usual 20 um the NIR reaches the numbers you target, and I know
of no
	case in which the ff region does not contribute enough to the PSF
that
	there is no point in 5 or even 10 um pixels. As far as I know the
new SITe
	UV-B optimized CCDs beat the PSF problem (10 um thick) but lose on
the QE.
	The Lincoln Labs high-resistivity chips (40 um sensitive region)
	significantly extend the red response, but I would expect a
diffusion PSF
	contribution large enough that there is no point your small pixel
width
	(Barry Burke can correct me.) Our developmental CCDs have a 300 um
	depleted region and no ff region, but also have a diffusion
contribution
	of 6--8 um to the PSF, depending on the bias voltage.  (See our ESO
ODT
	99 paper, also on the web page. Steve Holland has recently
significantly
	extended this analysis, to be published.)

>From what I have read in spectroscopy-based articles, back illumination
allows
for increased IR sensitivity, as does back thinned.  When I see QE curves
for these devices, they seem to corroborate this.

If I understand correctly, what you are saying is that if I want a deep
depletion 
device, or a back thinned device, small pixels will not work.  I assume that
is why most of these devices have 25+ um pixel width.  Is taht correct?

Okay, what I really need is a linear CCD with similar pixel dimensions to
the
SONY ILX526 (7 um wide, 200 um tall) that has BETTER QE in the NIR
than the SONY device.  5-10% at 1000 nm would be sufficient.  I have been
unable to find a source of such a device.  If I can find one that is 25-30
um
wide, that would be even better (e.g. ~ 4100 pixels).  

Thanks for the information.  As you may have figured out, I am an end
user of CCD systems, not a manufacturer of CCD chips, nor am I
doing basic R&D on CCD technology.  I base my selection of CCD's on
dark current, QE, pixel dimensions, cost, and a few other 'high end' 
parameters without going into the intricacies of how the different 
designs of CCD's affect their performance.  However, I do appreciate
your insight.  You have shown me that I have simply learned the tip
of the iceburg when it comes to CCD's.

BTW, what is your take on CMOS sensors?  I am trying to get a 
good handle on how well they perform with respect to low end
CCD's like the Texas Instruments TC241, the Kodak KAF0400,
and other front illuminated CCD's.  Comparison parameters such
as QE, dark current, photo response non-uniformity, and dark response
non-uniformity are of concern.

Sincerely,

Scott Sutherland
Senior Scientist
GAMMA-METRICS
- -- CCD-world -- --
CCD-world is fully moderated. Send posts to CCD-world at astro.ku.dk
Standard replies will go to the list; address personal replies manually.
For more information, please go to:  http://www.not.iac.es/CCD-world/