Message	Date	From	Subject
2192	2008-10-06 11:05:27	Glen Leinweber	LM386 noise
Some comments about LM386 being noisy prompted me to pull out app-notes for this chip. Wide-band noise in 10KHz. bandwidth, with 600 ohm input Z yields equivalent input noise of 3uV. A preamp chip touted as "very low noise preamp" has 0.5uV equivalent input noise operated in a similar way. So 3uV isn't too bad by comparison. Many op-amps are in the same ballpark. I think that works out to a 20dB noise figure, so for really low noise work, a single stage transistor ahead of a LM386 would help. A common input circuit is a 10K volume pot with wiper going directly to a LM386 input pin. In this case, ground the other input to the bottom (grounded) end of the pot. For higher source impedance, AC-couple the unused input to ground with a capacitor. Leaving the unused input pin open invites current noise from the 50K input bias resistor (internal to the chip). On another note, the 0.05uf capacitor in series with a 10 ohm resistor to ground at the output pin is there to take care of oscillations of the "pull-down" PNP-equivalent transistor of the totem-pole class-AB output stage. While sinking current, this output stage can be a bit unstable, whereas sourcing current (from the "pull-up" NPN) is much more stable. In their app-note, National Semi calls this kind of oscillation the "bottom-side- fuzzies", since it occurs only on the negative swing of the output waveform ;-)
2195	2008-10-06 11:24:54	Leon	Re: LM386 noise
----- Original Message -----
2218	2008-10-09 18:12:06	jr_dakota	Re: LM386 noise
The cap in series with the resistor

Some comments about LM386 being noisy prompted me
to pull out app-notes for this chip. Wide-band noise in
10KHz. bandwidth, with 600 ohm input Z yields
equivalent input noise of 3uV. A preamp chip touted as
"very low noise preamp" has 0.5uV equivalent input
noise operated in a similar way.
So 3uV isn't too bad by comparison. Many op-amps
are in the same ballpark. I think that works out to a 20dB
noise figure, so for really low noise work, a single stage
transistor ahead of a LM386 would help.
A common input circuit is a 10K volume pot with wiper
going directly to a LM386 input pin. In this case, ground
the other input to the bottom (grounded) end of the pot.
For higher source impedance, AC-couple the unused
input to ground with a capacitor. Leaving the unused input
pin open invites current noise from the 50K input bias
resistor (internal to the chip).

On another note, the 0.05uf capacitor in series with a 10
ohm resistor to ground at the output pin is there to take
care of oscillations of the "pull-down" PNP-equivalent
transistor of the totem-pole class-AB output stage.
While sinking current, this output stage can be a bit
unstable, whereas sourcing current (from the "pull-up"
NPN) is much more stable. In their app-note, National
Semi calls this kind of oscillation the "bottom-side-
fuzzies", since it occurs only on the negative swing
of the output waveform ;-)