🚀 go-pugleaf

In article <1108@rtech.UUCP> jchan@rtech.UUCP (Jeff Chan) writes:
>
>Here's a question for you old farts, audiophiles, and EEs:
>
>What are the appropriate uses and benefits/disadvantages of all the
>various types of capacitors?  For example, electrolytics are often
>used in power supplies.  I suspect that polycarbonates are thermally
>quite stable...

Back before I got involved with microelectronics, I had some design
jobs involving critical capacitors (all the capacitors I deal with now 
are flat, very small, and "custom"!). Rather than give opinions about
all the myriad dielectric types available in discrete capacitors, I'll
mention those with salient good features that stood out. The trade
mags like Electronic Design often show tables of this data in feature
articles, every year or two.

My comments are relevant to high-accuracy applications like signal paths
and instrumentation (I will leave power-supply filtering to someone 
-- evidently everyone -- else). Electrical properties important to these
applications include linearity, stability and low dielectric relaxation.
The last of these is especially important in pulsed applications like 
sample/hold circuits, and also the hardest to get hard facts on; you 
pretty much have to measure.  An example of the effect is for an 
abruptly discharged capacitor to exhibit some "memory" of its charging 
voltage after the discharge, due to residual polarization of the 
dielectric. In many dielectrics the magnitude can be 0.01% or more, 
which is bad news in some applications.  I regret that I don't have any 
DR figures handy.

1. Polycarbonate (Lexan (tm)) caps have some near-ideal properties,
especially their phenomenal temperature stability, which can be as
good as 0.1% over the military temperature range. They are also very
expensive (dollars apiece in quantity, instead of cents). While that
fact alone would no doubt commend them to some audiophiles ;-),
audio is not one of the applications that needs their principal
attribute of temperature stability. Instruments like precision 
oscillators exploit that property. Also, I believe TFE (Teflon (tm))
caps share similar properties with polycarbonate, but are even more
expensive, thus making them even more suitable for one-upmanship.
Only kidding, of course.

2. Polystyrene caps have temperature stability approaching that of
polycarbonate, but are manufactured to a coarser initial tolerance
and are MUCH less expensive (order of few cents in quantity 100).
They are useful for applications where stability may be important but
initial value is not (a trimmed function generator, for example). 
They are also clean in other respects. A drawback is physical size;
they are typically more bulbous than, say, Mylar (tm) caps of the same
value and voltage, though the Mylar units are much less stable.
Mylar caps are more common by far than polystyrene and denser, as I've
said, but no cheaper, and their second-order dielectric properties
as I recall are largely inferior to polystyrene.

3. Ceramic caps, especially the monolithic versions developed for
digital-IC decoupling service, are relatively dense in the large values
(0.1 uF and larger), and their high-frequency response is superior to
those with plastic dielectrics (polywhatever), which typically contain
a spooled sandwich structure with larger series inductance. The
ceramic caps also have poorly controlled value, usually rated GMV
(guaranteed-minimum) or -20, +80% and often, in my experience, show
enough leakage conductance to be a problem in high-impedance
instrumentation circuits, and their temperature stability is among
the worst. They are therefore most useful in wideband and power-supply
decoupling applications.

The foregoing dielectric types are about all you need for low-value
(1 uF and smaller) applications. A rub comes if you need something
larger, where these types all get bulky, in signal-path applications.

A large divide exists between electrolytic and non-electrolytic types,
so much so that it is unfortunate they are both called capacitors;
this causes lots of confusion among the inexperienced, who don't
realize the fundamental difference. Electrolytics work of course by
forming a (temporary) insulating layer between a metal surface and
an electrolytic gel or liquid, in response to an applied DC voltage.
The ONLY reason they are used is to gain the extremely high capacitance
density possible from the very thin dimension of the electrochemically-
formed insulating layer. They are inherently polar devices; must have
a steady DC bias (and are, simply from that, nonlinear); and show, in
most models, limited lifetime. Because of these three drawbacks they
must be used with great care and awareness, and they are very 
successful in certain applications, like the low-frequency filter
capacitors in DC power supplies.

*************  Anti-electrolytic flame follows  ****************

Unfortunately, designers have been sticking electrolytics in
everywhere they needed a large-value capacitor. The result?
Audio distortion, for one thing (as already noted here and in
rec.audio); but also, in-circuit failures where the electrolytics
experience no or inadequate DC bias (they need a "forming" voltage
equal to a fraction of the "working" rating) or simply from their
limited lifetime. 

I have been repairing audio equipment since about 1970 and I cannot
count the times I have seen "right channel is out, or nearly out,
and (maybe) comes back with loud music." (If the right channel is
completely out, it may well be due to an electrolytic interstage
coupling capacitor that failed; if the volume is way down, quite
possibly it's an emitter- or source- or opamp-resistor bypass 
capacitor, also electrolytic.) Often these capacitors fail because
they are rated for 25 volts and are operated with a two-volt DC bias.

Anyway, it's worth avoiding electrolytics of all kinds in analog
signal paths and relegating them to power supplies where they are
fat and happy facing DC and some 60-hertz ripple. It's just hard to
do this if you need 100 or 1000 microfarads in your signal path.
(One place you generally *don't* need capacitors, at least fundamentally,
is in series with the speaker at an audio output stage. With proper
design it's not hard for modern amp circuits to maintain stable 
zero-DC output voltages if the power supply is roughly symmetric.)

If you can figure out a way to pack, say, 100 uF at 50 volts into a
cubic centimeter or so, using cheap, dry, stable, voltage-independent
materials, the world will beat a path to your door. Unfortunately, with
readily manufactured insulators, physics is against you.

Max Hauser, UC Berkeley EECS Department, IC Design Group

   UUCP: ...{!decvax}!ucbvax!eros!max
   Internet (old style): max%eros@berkeley
   Internet (domain style): max@eros.berkeley.edu

(PS: for those who asked, I will post answers to the old-fart quiz soon.)