[time-nuts] how to find low noise transistors
Charles Steinmetz
csteinmetz at yandex.com
Sat Jul 25 13:03:35 UTC 2015
Rick wrote:
>OTOH, the other cure for high base spreading resistance is
>to simply parallel multiple devices. This avoids the bad
>side effects you mention.
Yes, if you can operate each of the transistors at the same current
you would have used for a solo transistor.
Typically, though, you need to reduce the current through each
transistor so that the total current remains approximately the same
as you would use for one transistor, in which case you do not escape
the side effects. There is a practical limit to increased total
current in circuits that use collector resistors to transforn the
devices' transconductance into voltage gain -- as you increase the
device current, you need to reduce the collector resistors (and,
therefore, stage gain) or increase the power supply voltage (which
gets out of hand in a hurry). In some cases, this can be avoided by
using current source or current mirror collector loads. Also, when
paralleling transistors, you need to allow for the noise of any
emitter ballasting resistors that may be necessary to equalize
currents between the transistors.
All is not lost, however -- the parallel device strategy encourages
the designer to use multiple, lower current devices to replace a
single, higher current device (for example, using multiple 2N5179s to
replace a single 2N5109).
Some '80s hi-fi preamps went to absurd lengths with parallel input
devices. A few had 20 or so devices in parallel for each of the four
input devices of a complementary differential pair (x2 for stereo, so
160 input transistors for just one stereo stage). But in order for
the input bias and offset currents to remain somewhat reasonable, and
for other practical reasons, the current per device had to be reduced
to the point that the bad side effects were glaringly apparent. (One
more all-too-common example of a designer being blindly committed to
his or her Good Idea and pursuing it to the detriment of the product.)
Best regards,
Charles
More information about the Time-nuts_lists.febo.com
mailing list