junction noise (really)

[Ian Fritz]

List people --

The recent discussions about noise prompted me to have another look at my
noise source (the old EN design). I recorded the white noise output into
CoolEdit and analyzed the spectrum. Although I had choosen the transistor
for its smooth sound, the spectrum showed a fair rise at low frequencies,
indicating a significant 1/f contribution.

Since the CoolEdit procedure is quite fast (30 sec or so), I decided to go
through a couple of bags of transistors I had around. Mostly I looked at
2N3904/6's. Out of about 50 transistors, I only found one hero that showed a
perfectly flat response down to 20Hz. Three or four others had only a small
amount of 1/f noise and the rest were quite poor. The differences were
apparent audibly, but an audible comparison is somewhat difficult because of
the different output levels of the various units and because of the
difficulty in remembering a sound over time.

The hero unit had the lowest ouptup of the whole lot, and the largest
reverse breakdown voltage. So I think the old idea of trying out all your
old junker transistors for the largest output is not a good idea at all if
you want good white noise.

  Ian