CCD-world: Re: Reset Gate Flatband
MYPIXEL at aol.com
MYPIXEL at aol.com
Wed Apr 26 00:08:18 CLT 2000
The following was posted to CCD-world:
************jj
Thanks Tom. . .
Rodger = Roger . . . I made the same mistake (sorry Roger).
We now know more about the physics of reset threshold shifts. Actually the
reference diffusion was OK. The gate (negative precharge) to drain (Vref)
determines where pinchoff occurs. This will be about 8 V for the Cassini (and
Hubble) CCDs. We therefore went beyond pinchoff by 27 V (29 + 6 - 8) in
burn-in !!!! And you know, what happens beyond pinchoff. . . luminescence
and hot electrons! No wonder things charged up the reset gate. . . you would
too. The 21 V Vref "breakdown" we measured is where pinchoff noise really
takes off. . . making it look like diffusion breakdown. That is why
breakdown characteristics "looked" different between Vdd and Vref. Diffusions
should all break down at the same voltage without gate influence (about 40 V
in the case of Cassini). It all comes back now.
It is hard to believe that Roger is experiencing the same problem because his
drive voltages are not that extreme (i.e., nominal). His low reset level is 0
V and Vref is 13.5 V. Gate pinchoff is probably about 8 V taking the device
5.5 V beyond pinchoff. Although the electric field near the drain region will
cause some impact ionization, experience shows there should be no problem
with flat-band. However, oxide/nitride can sometimes act up because it is a
good charging interface (recall EPROMs are based on this interface). Depends
a lot on processing details. It sure sounds like a threshold shift of some
kind. If there was a brief negative transient on the reset pulse, this could
promote hot electrons and charging. There are tests one can perform to
confirm that flat-band is the culprit. . .
Thanks again for the memories. . .
Jim
**************jj
In a message dated 4/25/00 6:11:01 PM Pacific Daylight Time,
stythe.t.elliott at jpl.nasa.gov writes:
<< Jim,
Welcome back. I saw Rodger's message posted on CCD-World a few day ago and
immediately thought of the reset gate burn-in propblem from the Cassini
days. As I recall we clocked the reset gate from -6 to +13v and biased
Vref at +29v. That put a 35V swing across the reset gate. That was
important because you stated in one of your Cassini memos (that I can't
find right now) that the "reset diffusions exhibit low breakdown
characteristics", about 21V.
Tom
<><><><>
From: MYPIXEL at aol.com
Date: Tue, 25 Apr 2000 17:15:34 EDT
Subject: Re: CCD-world: Reset gate threshold shift
The following was posted to CCD-world:
**********************************jj
Roger,
This sounds like a similar problem experienced when burning in flight WF/PC
II and Cassini CCDs. Here the reset threshold dramatically changed during
burn-in. In fact some CCDs were impossible to reset. Several formal reports
were written on the phenomenon experienced. We concluded that the reset gate
insulator had charged during burn-in (i.e., experienced a very large
flat-band shift). The problem was resolved by lowering the burn-in voltages
which were consideralby higher than nominal. I need to look at the reports
again to remember the details and physics involved. CCD manufacturers run
through similar tests to make sure the CCD is reliable for long term use.
There are always surprises when tests like this are performed.
It would be interesting to generate a few transfer curves to see if there
really is a flatband shift involved. Do you think it would worth pulling a
CCD for this purpose?
Jim Janesick
*******************************jj
In a message dated 4/24/00 3:25:26 PM EST, roger at ctios1.ctio.noao.eduwrites:
<< The following was posted to CCD-world:
Dear CCD World,
If you have ever had a new CCD system which gave you a lot of trouble
you will be able to relate to the expereunce I am having now...
The CCD in question was tested by my colleagues at NOAO-Tucson. It was
in fact the best of the batch procured for the second 8Kx8K NOAO Mosaic
(Cerro Tololo's) and was thus set aside for installation on a cold
finger in the camera of a fiber fed spectrograph. After the usual
teething troubles with a new system, most of which were related to the
unusual detector mount, the CCD worked fine. Being under pressure
check out the new spectrograph optics, we took it to Cerro Tololo
immediately, keeping it cold and under vacuum while in transit (not
normally a problem). It has given me trouble ever since.
I have since determined that the reset transistor switching threshold
has increased to a quite remarkable 14.3 V. I've carefully measured
RG_threshold on both outputs of nine identical CCDs, 18 n total, and
find that it lies typically between 4 V and 7 V. I do this by setting
RG_high to ensure that the Reset FET is safely turned on, then plot
RG-feedthrough amplitude as a function of RG_low, then extrapolate
slightly to determine where RG feedthrough is zero, ie the point at
which the FET doesn't turn on any more.
I haven't yet found any way of manipulating the CCD inputs to modify
the RG threshold. We have considerable experience with these devices
having 16 of them in two 8x8K Mosiacs operating with similar voltages,
waveforms and the same design controller, so this is quite a surprise.
Note the gate threshold is higher than the drain voltage instead of
being much lower! The other bias voltages are:
Reset Drain = 13.5 V
Last Gate = -4.0 V
Output Drain = 25.0 V
Output Source = ~20 V
The clocks must be ok since the CCD correctly images everything from
cosmic rays to flat fields, with normal bias levels and normal well
capacity, provided that RG is switched between 8 and 14.8 V instead of
0 and 12 V.
What could have happened? Static discharge is rarely a problem in our
very humid environment. Nonetheless, the CCD has always been handled
on properly grounded antistatic mats (bench and floor) while using a
wrist strap with a continuity alarm. Remember too that I obtained
normal images with normal voltages immediately prior to taking the CCD
to Cerro Tololo and it was kept under vacuum and cold the whoel time
until it ceased to operate.
I wondered if the substrate contact had been lost. Had a bond wire or
CCD contact failed? Inspection under a steroe microscope shows all
bond wires are in contact with their pads. Capacitances from the
clocks to the substrate look normal (measured with CCD unpowered). I
also note that the RG threshold of the other amplifier is normal which
seems to rule out the loss of substrate contact.
-----
As if I didn't have enough problems, amplifier A ceased to operate at
the same time as the RG threshold shifted on amp B. Its symptoms
however are completely different. My hpothesis is that Last Gate is
floating.
All the clocks are working since it images through amp B with a
suitably high RG pulse, and I can verify correct clock reversal when
the amplifier selection is changed. I see perfectly normal video
levels and feedthrough pulses, however no video shifts are seen when LG
is taken too high (as happens on the other amp). Amp A is vaguely
sensitive to light: a vertical overscan can be seen in a flat feild
with more lines than there are in the image area. The vertical edge
response is much pooreer than normal (decays over 5-10 lines instead of
1-2), and the structure along a line is a smooth curve. This behavior
can be induced on the other amp by taking LG extremely negative (-8V)
while raising the low level on SW from -5V to -3V so that the charge
can't get past the barrier presented by LG until the serial register is
overfilled. I hypothesize that a floating LG would behave similarly...
I haven't actually floated a good one to test this!
So I replaced the CCD with a mechanical sample on which I had soldered
the suspect LG pin to its neighbour. I then performed a continuity
test with the system pumped and cold. If LG is open, it as to be in the
CCD or integral pig-tail cable. Bummer!
Although I have not yet found anything wrong with the controller or
dewar, my next step will be to run this CCD in a separate dewar and
controller.
There is a third phenomenon that appeared at the same time: the noise
increased from about 4 e- to 20 e-. I have some doubts that this is
related since it contains significant harmonic content, though the
concidental appearance at the same time as the other two problems is
very suggestive.
I would appreciate any advice or educated guesses as to what might have
happened, or further tests that I could perform. And if you have
nothing to offer, you can at least feel less alone when you get the CCD
blues.
Roger Smith
Senior Electronics Engineer / Manager - Array Controller Projects
Cerro Tololo Inter-American Observatory, Casilla 603, La Serena, Chile
Internet: rsmith at noao.edu Coordinates: 29.54 South, 71.16 West
Phone: 56 (51) 205200 Bilingual receptionist (08:30-21:00)
Fax: 56 (51) 205342/205212 Autoforward via US: 1 (520) 318-8259
Nat. Optical Astronomy Observatories,PO Box 26732,Tucson AZ 85726-6732
- -- 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.
----------------------- Headers --------------------------------
Return-Path: <stythe.t.elliott at jpl.nasa.gov>
Received: from rly-zd01.mx.aol.com (rly-zd01.mail.aol.com [172.31.33.225])
by air-zd04.mail.aol.com (v70.20) with ESMTP; Tue, 25 Apr 2000 21:11:01 -0400
Received: from eis-msg-014.jpl.nasa.gov (eis-msg-014.jpl.nasa.gov
[137.78.160.158]) by rly-zd01.mx.aol.com (v71.10) with ESMTP; Tue, 25 Apr
2000 21:10:25 -0400
Received: from telliott (telliott.jpl.nasa.gov [128.149.33.246])
by eis-msg-014.jpl.nasa.gov (8.9.3/8.9.3) with SMTP id SAA19200
for <mypixel at aol.com>; Tue, 25 Apr 2000 18:10:24 -0700 (PDT)
Message-Id: <3.0.3.32.20000425181736.00941150 at mail1.jpl.nasa.gov>
X-Sender: telliott at mail1.jpl.nasa.gov
X-Mailer: QUALCOMM Windows Eudora Pro Version 3.0.3 (32)
Date: Tue, 25 Apr 2000 18:17:36 -0700
To: <mypixel at aol.com>
From: Tom Elliott <stythe.t.elliott at jpl.nasa.gov>
Subject: Reset gate...
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
>>
- -- 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/
More information about the CCD-world
mailing list