I'm designing a low-noise, high-bandwidth (100 MHz) pre-amp using JFETs. The choice for JFETs was because it needs to operate with large source impedances (>10 kΩ).

I spent a while looking for a suitable JFET (high g, low input C, good noise-figure). I came across ones from CEL (for example, the CE3520K3 (datasheet, home page) with g = 50 mS , NF = 0.5). They don't specify input capacitance, but it must be sub-pF given the operational frequencies. What's the catch? The datasheet does specify large gate currents, so current shot noise will be high. Does anyone have experience working with such devices?

  • \$\begingroup\$ Datasheet says "s-parameters are available on the website " and indeed they are. Instead of citing an input capacitance at specific test conditions they provide scattering parameters so you can model the input impedance for your application. \$\endgroup\$
    – crasic
    Commented Apr 11, 2019 at 19:19
  • \$\begingroup\$ s-parms start at 2GHz no use \$\endgroup\$ Commented Apr 11, 2019 at 20:25
  • \$\begingroup\$ Use the 2GHz s-param value, and on the Smith chart proceed to decompose the Z into real and imaginary (capacitive) component values. \$\endgroup\$ Commented Apr 12, 2019 at 4:07
  • \$\begingroup\$ What is the BW at 100MHz \$\endgroup\$ Commented Apr 12, 2019 at 4:22
  • 1
    \$\begingroup\$ I've used JFETs in stuff, not my favorite. For a linear amplifiers they have a narrow linear range ~1-2v. Also their depletion mode so you need to bias the gate negatively. If it were me I would search high and dry for a high BW opamp to use to buffer the impedance. \$\endgroup\$
    – Tony
    Commented Sep 13, 2019 at 1:09

1 Answer 1


There are no catches (lol) Only Specs

The s-parameters start at 2GHz

! CE3520K3
! N-channel HJ-FET
! Vds = 2 V Id = 6 mA
# Hz S  MA   R 50
!      f            S11               S21               S12               S22
!    Hz        MAG       ANG       MAG       ANG        MAG        ANG       MAG       ANG
2000000000 0.98035944 -29.159311 4.3293347 149.29152 0.02862794 67.64743 0.65920687 -24.487707

What is the purpose?

In the meantime, WHAT's ALL THIS STUFF ABOUT ...?


enter image description here

It's just balancing act

enter image description here

A delicate balancing act.

enter image description here

Just start with good Design Specs, including DFM, DFT , then perform DVT and if passes all tests. Congrats, Its Perfect. enter image description here

or startover until it passes your design specs.

enter image description here

Or do it right the 1st time with better specs.


20 Years ago HP R&D came up with this

To address the growing handset power amplifier needs for the emerging Personal Communications Services (PCS) markets, a 3-volt, single-supply, enhancement-mode pseudomorphic high-electron-mobility transistor (E-PHEMT) has been developed.

The device exhibits +33dBm output power and 65% drain efficiency at 1.88 GHz.

Consider a state of the art FET buffered Differential Probe specs enter image description here

enter image description here

Using MGF4919G enter image description here enter image description here REF

  • 1
    \$\begingroup\$ Thanks for the info, it very helpful. The design already has a resonant front end, with pretty much the exact the values you suggest. The task is to make a resonant photodiode amplifier which can be shot-noise limited <pA/sqrt(Hz) level to detect a very weak AM light source. Initially we tried using low-noise FET op-amps, but found they had too much current noise at high frequencies. We are testing some designs to see if we can do better with just a common source or cascode. We will see how I get on... \$\endgroup\$ Commented Apr 11, 2019 at 20:00
  • \$\begingroup\$ Where are your specs? Is this something fancy like a Interferometric Gravitational wave Detector? \$\endgroup\$ Commented Apr 11, 2019 at 20:02
  • 1
    \$\begingroup\$ It is a one off devices for research (sadly not LIGO, but they have some nice papers on it), so cost/size/power constraints don't come in to it. All I care about is being photon shot noise limited for powers about 10 uW. My photodiode has a sensitivity of 0.35 A/W (I have to work @ 450 nm as the end application is spectroscopy), so if I want the Johnson–Nyquist current noise to be below the photon-shot noise I need a source resistance to be >10 kΩ. The modulation frequency has to >50 MHz. \$\endgroup\$ Commented Apr 11, 2019 at 21:22
  • \$\begingroup\$ that's doable in Ga-As BJT's \$\endgroup\$ Commented Apr 11, 2019 at 22:11
  • \$\begingroup\$ How does 0.65dB Noise figure sound at 100MHz ? \$\endgroup\$ Commented Apr 11, 2019 at 22:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.