Don Pearce wrote...
It may be you're making the wrong measurement. As we describe in
the low-noise chapter in AoE (pages 428-461), all amplifiers have
two types of noise sources, voltage noise and current noise, and
their design is usually optimized to favor one over the other. In
mic preamps, which deal with low-impedance sources, BJT transistors
are used because they can have much lower voltage noise than JFETs.
It's important to realize that when you connect a microphone, the
current noise is shorted out by the mic's low impedance, and thus
doesn't contribute to the observed noise. But when you leave the
mic input open the source impedance goes way up (to Zdiff = 13.6k
in your case) and the input current noise can become a big factor.
For example, your input transistors Q101 and Q102 are operating at
a high collector current of about 3mA each, as set by Q103 and 104.
A high current is chosen, because it reduces voltage noise density,
perhaps to under 1.0nV/root-Hz, but as a penalty, it increases the
current noise. In your case the Q101 / 102 base current is about
20uA (assuming a beta of 150), which is pretty high. This will
cause an input shot-current noise density of sqrt(2q*Ib) = 1.8pA,
which in turn causes a voltage noise of 12nV across 6.8k resistors,
or sqrt 2 larger = 17nV across the two resistors together. Looking
back in this paragraph, you'll see that this is 17x higher than the
voltage noise alone. By contrast, if a 150-ohm mic was connected,
you can calculate that the current noise would create under 0.4nV,
which is less than the voltage noise, and can be ignored. All this
is explained in our book, if you'd like to understand it better.
The mic inputs are discrete - you need to be looking at Q101 and
102 and so on. I'm presuming the line level stuff is quiet enough.
Right, as John and others have also said. Changing the opamp won't
make any difference for the mic preamp stages.
