[time-nuts] how to find low noise transistors

Richard (Rick) Karlquist richard at karlquist.com
Sun Jul 19 11:24:09 EDT 2015


On 7/18/2015 2:16 AM, Attila Kinali wrote:
>
> I always wonder how you figure out whether a transistor is low noise
> or not. What part of the datasheet hints at which transistors have low
> noise and which have not? Even if it's just try and measure, how
> do you find good candidates to measure?
>
> 			Attila Kinali
>

For a BJT operating above the 1/f noise corner, and at non-microwave
frequencies, the noise properties depend only on RF current gain and
base spreading resistance.  See "Low noise electronic design" by
Motchenbacher and Fitchen.  RF (not DC) current gain can be measured the
usual ways, but base spreading resistance has to be inferred from
noise figure measurements made with low source resistance.  The
RF current gain is the real fundamental noise property of the device
that you cannot change.  Fortunately, it can be determined from the
data sheet, if not directly, then by calculating it from DC current
gain and F-sub-t, based on the operating frequency.  The low frequency
current noise (above the 1/f corner) is simply equal to the shot
noise of the DC base current.  The low frequency voltage noise is
the sum of the Johnson noise that a resistor would have if its
value were the sum of the base spreading resistance and half of
r-sub-e.  Where r-sub-e is the "emitter resistance",IE the effective on 
resistance of the transistor.  Base spreading resistance can be
overcome by using a sufficiently high source impedance and/or
paralleling devices (if you can tolerate the additional capacitance).

At frequencies such as 100 kHz and 10 MHz, it is very easy to get
a noise figure well below 1 dB with a BJT, so it should be no great
problem to find a suitable device.

Even lower noise figures are available with JFET's, which have
noise current equal to the shot noise of gate current, which is
specified.  The resulting noise current is negligible for most
devices.  This leaves the noise voltage, which is just the
Johnson noise of a resistor equal to the channel resistance.
By scaling to larger devices and/or paralleling devices, this
can be reduced to arbitrarily low values.  The limiting factor
is the substantial capacitance of JFET's.  This limits them
to about 1 to 10 MHz, before high beta BJT's dominate.  I have
observed noise figure of below 0.2 dB in JFET's at 2 MHz.

Below 50 to 100 MHz, MOSFET's and ePHEMT's have excessive 1/f
noise and are a non starter.  Above the 1/f corner, it is easy
to get noise figures of a few tenths of a dB with ePHEMT's.

All of this discussion doesn't address 1/f noise, which could
be an issue in oscillators and low phase noise amplifiers.
For that purpose, you are back to characterizing devices yourself.
Putting negative feedback around the transistor can alleviate
this by reducing upconversion of noise.

Rick Karlquist N6RK


More information about the time-nuts mailing list