DCS Integrator Capacitors

[roger smith x294]

This is about dielectric absorbtion in integrator capacitors
and I wouldn't expect anyone except die hard analog desingers to
be interested.

I am looking for a source of top quality 500 pF capacitors to to use in
a dual slope integrator with adjustable capacitance.  Previously I have
used 3.3 nF Teflon capacitors from Component Research Corporation
(Metalized, J11 series).  These have proven exceptionally linear
(<0.05% variation) and exhibit low dielectric absorbtion, quoted as
<0.01% maximum voltage recovery.  However, the minimum value in Teflon
made by CRC is 1 nF.  [ I'd rather not put two 1 nF in series to get
500 pF since I am pressed for board space, and they are bulky in spite
of their low value.  IN theory one could generate more capacitance
combinations by switching several capacitors to obtain series as well
as parallel combinations but this leads to using quite a few analog
switches increasing leakage current and reset settling time problems. ]

The only alternatives for 500 pF low dielectric absorbtion capacitors
that I have identified are polystyrene and polypropylene.  CRC quote
the dielectric absorbtion for these materials as "less than 0.02% max
voltage recovery", but they don't make them in sizes smaller than 1
nF.  Other manufacturers make smaller values in both materials (not
teflon unfortunately) but they don't specify dielectric absorbtion at
all.  In fact dielectric absorbtion is rarely specified for any
capacitors, and values vary widely between manufacturers using
nominally similar materials. I have figures from an electronics
handbook which contradict those from a well know textbook which in turn
disagree with what manufacturers data I have found.  As usual
materials withthe same name are not equal.

The text book shows Teflon being 3 to 5 times better than polypropylene
and polyester being mid way between them, whereas when I last measured
dielectric absorbtion (12 years ago!) I found that Teflon and
polypropylene were about equal....

After a long "soak" at 10V:

Material	Maker	Recoil 	Dominant time constant (sec)
________	_____	______	______________________
Polyester/mylar	TRW	0.4 %	60 sec
		Wesco	0.4 %	40 sec
Mica		?	0.3 %	15 sec
Polycarbonate	CRC	0.1 %	35 sec
Polypropylene	TRW	0.025 %	10 sec
Teflon		CRC	0.025 %	10 sec

I guess the good news is that in spite of all the confusion the best
candidates perform at a level where the errors are getting quite subtle.
Does anyone know how noisy the voltage recoil process is?!  This might
affect someone looking at low signal areas in a very contrasty image.

I thought that for such low values the problem should be more tractable
since one doesn't need high dielectric constant material, but I don't see a
better alternative to polypropylene or polystyrene.  Suggestions would be welcome.

Roger Smith, CTIO