How do BJT vs JFET vs HBT compare in terms of \$1/f\$ noise amplitude and corner freqency? I know these transistor types are very different, used for different frequencies and powers, and each have many different variations all with different flicker noise, but on average is one type better than another?
\$\begingroup\$
\$\endgroup\$
1
asked May 17, 2017 at 14:14
-
\$\begingroup\$ I've read Pchannel MOSFETs have lower 1/F noise, because the channel is "lower" or deeper, thus less upset by surface charge traps emptying and refilling. \$\endgroup\$– analogsystemsrfCommented May 17, 2017 at 14:46
Add a comment
|
1 Answer
\$\begingroup\$
\$\endgroup\$
9
It depends on the crystalline structure of the epitaxial wafer and junction geometry and square of the current needed for bias in both conductor and dielectric. Flicker Noise is random pink noise usually measured <100Hz in \$A^2/Hz \$ as 1/f noise, but contributes to phase noise in RF.
- GaAs can be much better or worse than Si.
- carbon resistors are worse than metal film which are worse than copper resistors which are worse than "NP0" manganin conductors considered flicker noise-free.
The explanation is simple but hard to visualize.
Imagine a small low leakage cap on a unijunction gate or a DIAC with a small bias current and a breakdown voltage for the semiconductor shorting out the cap and then charging up again. This is a fixed f relaxation oscillator. Now imagine that certain semiconductor crystals have a higher leakage current (Early Effect) with greater BDV generating bigger partial discharges between charged molecules before they break down under the electric field (in a nano scale ) . Then imagine millions of relaxation oscillators of random pulse rate low pass filtered by the dielectric between the charged conductor atoms. So this RC time contant affects the relaxation rate while the high series leakage and shunt capacitance low pass filters these broadband pulses or "flickers".
So in my theory, it is the device with the highest RC leakage time constant and highest BDV where the product produces a corner frequency at some A^2/Hz with applied bias voltage. I call this flicker, random (PD) or partial (nano crystalline) discharge , for your understanding and consideration.
Which device?. depends on conduction power, leakage input bias current, dielectric doping levels, crystalline épiwafer nano-structure and meta structure such as FET, BJT or HJT.
But we know for sure how to rank conductors as I did above so we generally use MF for lower noise exclusively and carbon or WW where high current noise does not matter.
I don't know how to tell you which device has lowest random flicker noise for ANY random design.
... but you may look at GaAs FETs and compare.
-
\$\begingroup\$ Is it true that the bigger the transistor the lower the 1/F corner frequency? So something like 3V 10mA transistor is going to have the corner frequency high and 1200V 1A transistor is going to have the corner frequency low. \$\endgroup\$ Commented May 17, 2017 at 15:57
-
\$\begingroup\$ high V results in lower C and lower 1/f but all the other variables must be considered. \$\endgroup\$– D.A.S.Commented May 17, 2017 at 16:57
-
\$\begingroup\$ you mean that if I send 10 watt through transistor as 100 volt and 0.1 ampere it will have less 1/f noise than 10 volts,1 ampere? So high current = high 1/f noise but high voltage doesnt cause high 1/f noise if the current isnt increased as the result of high voltage? \$\endgroup\$ Commented May 17, 2017 at 18:40
-
\$\begingroup\$ nope, flicker noise increases with \$I^2\$ \$\endgroup\$– D.A.S.Commented May 17, 2017 at 19:47
-