I studied electronics a long time ago (and forgot almost everything) but now I would like to create a FET amplifier up to 100 MHz of extremely small signals (e.g. 1nV or even smaller). Signals are so small that an opamp (I believe) is not possible to use. I would like to simulate it first (ngspice or LTspice) and then actually build it.

I know the noise level will also be around some nV...but please lets assume I work with extremely low temperatures and this is not an issue.

Power efficiency is not an issue (could be 1% if needed) and power supply could be at 20-30 V if needed.

I though about using the cascade amplifier with one FET and one BJT NPN with a current source driven by 2x BJT PNP transistors (using a current mirror). This works well until ~1 MHz, but not to 100 MHz. I believe my problem is finding a good FET transistor (maybe a GaAsFET?)

A good starting place was this circuit (not using a current source...but close enough): https://www.designworldonline.com/cascode-amplifiers-2/

However the FET (https://pt.mouser.com/ProductDetail/Broadcom-Avago/ATF-34143-BLKG?qs=RuhU64sK2%252Bt35SXiAk9qwg%3D%3D) is no longer for sale.

I actually don't understand this circuit in the R5 resistance: since the transistor is a FET, isn't the gate current zero? Why is R5 there? I simulated this circuit (using the linear spice model) and I see that R5 does make quite a difference in the spectrum...hence I can only "conclude" that its because we are using the linear spice model and not an accurate model (doesn't seem to have V_th for example). But I think I'm missing something.

cascade circuit with passive

So I guess my main problem is how to find a good FET transistor that:

  • works from 100 kHz to 100 MHz (with a good gain). If this is not possible, then at least from 1 MHz to 80 MHz would be nice.
  • has a "full" spice model (not just the linear model since I want to see the transient response)
  • Is currently being sold (not deprecated)

But of course, if there is a better way to design this amplifier please let me know.

Many thanks in advance for your help!


  • 3
    \$\begingroup\$ Crucial questions at these frequencies and levels. What is the output impedance of your source? What is the input impedance of your load? Do you really want to build - perhaps for learning, or for fun, or for value engineering if you need to make commercial quantities of them, or can you buy? \$\endgroup\$
    – Neil_UK
    Dec 22, 2022 at 8:26
  • 4
    \$\begingroup\$ Where are you getting the terms "linear model" and "full model" from? I've never encountered them before. If you can, link us (or paste as a code block in the question) the model you used for the ATF-34143 so we can get a better idea of what you're trying to say. \$\endgroup\$
    – Ste Kulov
    Dec 22, 2022 at 8:32
  • 2
    \$\begingroup\$ What do you want to do with the amplified signal? Because if you have 80 MHz (that's 69 dBHz) of bandwidth, then good ole Johnson-Nyquist Thermal Noise tells us you'll have -174 dBm/Hz + 69 dBHz = -105 dBm as noise input power (at 20°C). A 1nV amplitude signal, depending on over which impedance that Nanovolt exists, will have less power. Which is fine if you can deal with it, but just so you know that what you're building comes with signal-theoretical challenges, not only practical challenges \$\endgroup\$ Dec 22, 2022 at 9:22
  • 4
    \$\begingroup\$ "since the transistor is a FET, isn't the gate current zero?" - At DC, yes, but definitely not at 100MHz. \$\endgroup\$
    – Linkyyy
    Dec 22, 2022 at 9:39
  • 4
    \$\begingroup\$ I'm super curious: 100 MHz @ 1nV -- what's it for? \$\endgroup\$
    – jonathanjo
    Dec 22, 2022 at 9:52

1 Answer 1


Before, take a look to this and this.

Perhaps a "distributed" amplifier should do it or minicircuits RAM-8A.
For reference, an example with "bad" BJT transistors.

enter image description here

It is perhaps what you don't expect, but here is a circuit that could be tested (100 kHz -> 20 MHz).
Testing in the nV range is really a challenge, so good luck ...

enter image description here

Tried also something like analog "spectrum analyzer".
1 nV input -> something like 8 uV output.

enter image description here

And here, the behavior with square waves (1 uV) and recalculated filters. Not bad ...

enter image description here

See this for comparison with the input stages an "old" spectrum analyzer, page 9-2.

enter image description here

  • \$\begingroup\$ Many thanks for you answer. I will take a look at that BJT for sure. However I have the problem of the input impedance since the signal is very low at 1 nV (or even lower). The AC analysis seems great but the transient is terrible... \$\endgroup\$
    – mabeco
    Dec 23, 2022 at 6:55
  • 1
    \$\begingroup\$ TRansient depends on some components and "bandwidth" as a "filter". So define exactly the specs of "signal". \$\endgroup\$
    – Antonio51
    Dec 23, 2022 at 16:27
  • \$\begingroup\$ I stand corrected. I just simulated your circuit and Vo follows Vi (assuming no noise, of course). Sorry for doubting you. I incorrectly thought that ib would always be zero in FET and non zero in BJT but that is only true at DC, which is far from 10 MHz. I don't exactly the spec of signal (I'm kind of new to this). I want it coming from an antenna and then passing through a mixer so I guess the input impedance for this circuit is ~200 Ohm and @1 nV peak to peak. I would be happy with a bandwidth from 1MHz to ~80 MHz. But as minimum up to 40 MHz. \$\endgroup\$
    – mabeco
    Dec 23, 2022 at 20:38
  • 1
    \$\begingroup\$ No worries. If your signal comes from an antenna and would pass through a mixer, then it is the "classic" schematic of the input stages of a "spectrum analyzer". And "generally" the amplifiers are after the multipliers. There can be as many as 3 multipliers and amplifiers. This is the principle of an HP spectrum analyzer I used ... many years ago. You should add these pieces of information to your question. \$\endgroup\$
    – Antonio51
    Dec 24, 2022 at 9:30
  • 1
    \$\begingroup\$ Hope that help. If you need more information ... \$\endgroup\$
    – Antonio51
    Dec 24, 2022 at 18:41

Your Answer

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

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