Welcome to the fray!!!
Capacitors are a hot topic in audio forums. The good and bad sound of certain types of caps. Rules of thumb, guidelines, superstition...
As a lifelong EE who has designed hundreds of circuits with
thousands of capacitors, no capacitor data sheet specifies
any audio characteristics. They deal with measurable
physical properties like voltage, current, resistance,
capacitance, inductance, non-linearity, temperature, size,
package....
The type of capacitor used in each of these cases is highly
dependent on the circuit specifics. A good coupling capacitor
might make a terrible bypass cap.
In frequency dependent circuits, capacitors provide filtering of
certain frequencies. Tone circuits, Equalizers, Phono Preamps,
loudness circuits, amplifier compensation capacitors are all
examples. In these circuits, capacitors provides a frequency
dependent impedance per the equation Xc = 1/(2*pi*F*C). By
combining with resistors and sometimes inductors, High pass and
Low pass network are built.
The important parameters here are the capacitance, temperature
drift, long-term drift, long therm reliability, and stability with
frequency and voltage. The resistors in these circuits have the
same needs, but resistors are typically much better in these
respects.
Fortunately, we have film capacitors to meet these needs. Film
caps are readily available from about 100pF up to 10uF. Larger
value Caps are available, but pricey and large. If you look at the
many 10uF film caps at Digikey, they about $3 or more (Qty 1).
In smaller values, we have NPO / COG ceramics and mica. Typically
10nF or less. These are quite stable with temperature, voltage and
frequency.
There are several different dielectrics available. The more
pricey ones (teflon, poly are often for specialized applications
such as precision sample-and-holds. Metalized Polyester,
Polyethylene, Polypropylene all have similar electrical
characteristics.
A special case of Frequency Dependent Circuits is speaker
crossovers. Here, the capacitors must handle high currents, up to
several amps. Fortunately there are Film caps designed for this.
Most coupling caps are there to block DC and pass the entire
audio frequency range from 10-20Hz up to 20KHz or more. They can
be low level signals like mic or phono, line level ~1V, or speaker
level. With modern amplifiers, speaker coupling caps are pretty
rare.
Ideally a coupling circuit will drive the cap with a low
impedance (0 to 1000 ohms) and receive into a high impedance
(5-100K ohms). This is a simple High-pass filter, with a cutoff
frequency of 1/(2*pi*R*C). To pass 20Hz with less than 0.1dB of
loss, the cutoff frequency must be about The critical spec
is
As an example, lets say we want to select a cap to feed a 10K
line-level load, and to have -0.1db loss at 20Hz. Typical
calculation for high-end audio. Don't forget that every coupling
cap causes a 20Hz loss, and there may be many in an audio system.
68uF is extreme for a film cap, so an electrolytic is needed. Or
increase the load resistor, or relax the -0.1dB requirement. In
any case, it's a fairly large cap.
An even more difficult situation is the capacitor in the feedback
network of an audio amp. Typically the feedback resistor divider
is about 40K / 2K (gain of x21), so the capacitor needs to be
significantly lower than 2K ohms at 20Hz. If you aim for the
-0.1dB spec, that means the cap needs to be about 5x the previous
69uF example (10K / 2K = 5.0) or 350uF.
Here is an example Audio Amp circuit from the TI LM3886 data
sheet. Ci is shown as 22uf, which would cause the low
frequency response cutoff to be F = 1/(2*pi*1K*22uf) =
7.24Hz. So about - 4db down at 20 Hz. Not great. And they
don't show the polarity. Chickens!
Oh, and these coupling caps may need to be bipolar. In the case of
the amp feedback cap, you may be able to predict the bias current
polarity, and therefore the bias voltage polarity. What is the
distortion performance of a 330uF electrolytic at audio
frequencies, low AC currents, and about +50mV of DC bias? Good
question. This is why comprehensive amplifier distortion tests are
important.
There are a few ways to address the problem of the large,
perfect, bipolar cap. You can build a DIY bipolar cap by using 2
caps of twice the value, in series, back-to-back (+ to +, or - to
-).
One way is to replace it with a short circuit. But this
requires that the amplifier now have very good DC performance as
well as excellent AC performance. Instead of building an excellent
DC amplifier, is to use a DC servo on the amplifier. A DC servo is
basically an integrator in the feedback loop that actively
regulates the amplifier DC output to 0V. This effectively
simulates the feedback capacitor. The integrator can use a large R
and therefore a smaller C, such as a small, well-behaved film
cap.And an Opamp. Replacing a difficult capacitor with an
integrator in the feedback loop is one of my favorite analog
design tricks.
You could build a DC coupled feedback, and add an offset voltage
trim. Requires a manual operation, and you would need to insure
that the DC won't drift much.
Bypass caps typically filter the power supply rails. These
perform several benefits:
In analog audio, noise reduction is key. Transistor amps
typically have poor PSRR (power supply rejection ratio), so noise
on the power supply can get into the audio. R-C filters are used
for the most sensitive stages. Op-amps have better PSRR due to
their use of differential stages and current sources. But PSRR
falls off at high frequencies.
One thing that irks me about 'Op-amp rolling', is that an op-amp
is just part of a circuit. The overall circuit (loop) gain affects
the frequency response and distortion. The op-amps frequency
response and slew-rate affect the performance of the circuit. The
source and feedback resistors need to be tuned to the specific
op-mp to optimize noise. Saying that a certain op-amp sounds
better depends totally on hte circuit. And the subtle effects:
CMRR, PSRR, output drive, FET vs Bipolar, temperature, and phase
of the moon all affect the overall response. Changing an op-amp is
just changing