Boki said:
What a professional comments!
It seems that my noise is tick-tick-tick at 5 to 50 ticks/second, when
I play sound, the noise become louder, when I stop sound, the noise
become smaller gradually.
If this is not the ripple problem ( I can't sure yet ), I don't know
why capacitor solves this problem.
Like Graham said: this is NOT ripple, it is a supply decoupling problem.
If it were ripple, the noise would be at exactly the same frequency as the
AC supply line, e.g., 60Hz or 50Hz depending on where you live, or else at
the switching frequency of your power supply if you are using a switching
regulator or converter (which would probably be supersonic).
You're familiar with putting a 0.1uF capacitor across the power supply pins
for your digital ICs, right? This is the same idea. The issue is that
power supply lines have nonzero resistance and inductance, and that
interacts with the amplifier to form an oscillator. The details vary
depending on the circuit. The LM386 is, as Graham says, notorious for
low-frequency oscillation, also known as "motorboating".
Want a more modern amplifier chip? Go to National Semiconductor's web site,
and search for "audio power amplifier". Take a look at their "Boomer"
series of chips. (Or, you could go to TI, or Zetex, etc. I just happen to
be a little more familiar with National.) There are dozens of them, many
aimed at exactly your requirements. You have your choice of linear or
switching ("class D").
I have measured my DC power supply ripple, it is about 40~80mv(pp), how
is this value? safe or not? or, what is the safe value for audio
amplifier ? ( not Hi-Fi quality, MP3 AV output is enough )
It depends on how much noise you want on the output, and the power supply
rejection ratio (PSRR) of the amplifier. For instance, if you have 60dB
PSRR, and you have 80mV of supply ripple, then you will have 80uV of ripple
on the output. The LM386 is specified at 50dB PSRR. Most modern amps are a
lot better than 60dB, so they can tolerate quite a lot of ripple. But you
have to decide for yourself how much you can accept.
If you were driving headphones, with typical 95dB SPL/mW efficiency and 64
ohm impedance, then 80uV of ripple would be 25dB of sound pressure level,
which is just barely audible in a quiet room. The noise output of the
amplifier would probably dominate, especially since ripple is usually at low
frequencies where our ears are less sensitive anyway.
