AD549: Noise problem

Document created by analog-archivist Employee on Feb 23, 2016
Version 1Show Document
  • View in full screen mode

in the datasheet of the AD549(J) there is no "worst case data of noise levels"
specified, only the _typical_ values. In conjuction with some strange
measurements in a test cirquit (which was much less noisy in the past, but
containing the same type of chip), we want to know, if there are some reference
values of the possible deviation in noise level. Would be the increase of noise
in a range of 30% (@10kHz-100kHz) classifiable as "normal" ?

is there a general reference value (i.e. an approximate percentage by rule of
thumb of the typical value), which can be used for values in a data book, when
there is no value specified ?


In my experience if an AD549 device becomes more noisy
during its life this is often the result of electrostatic damage
(ESD). These devices, because of their very low bias
current specifications, have sensitive inputs and are quite
vulnerable to ESD - you should take every possible
precaution when handling them.

An increase of 30% in the noise of a particular device is
quite likely to be the result of ESD or other overload.

A variation of 30% from the "typical" value is not
unexpected in a randomly-selected device, though.
I should expect variation of -50%/+100% at 2 sigma
(note that this is an educated guess - NOT a production

The current noise of s simple FET (and a simple bipolar
junction transistor, if it comes to that) is the Schottky
noise of its bias current - so if you can measure Ib you
can calculate the current noise. The current noise
spectral density is SQUARE ROOT(q . Ib) Amps/Root Hz
where q is the charge on an electron (1.6E-19 Coulomb).
ESD in a FET usually increases Ib - and thus noise current.

The expected variation of a "typical" parameter depends
quite a lot on what parameter we are discussing. 2:1
variation of offset voltage is quite common, 2:1 variation
of quiescent current is much less likely. But do remember
that over its lifetime the actual "typical" value of a parameter
may change - but the data sheet probably will not.