Help with high input impedance amp.

I added that tantalum later because I had a capacitive divider with the original 27pf and gate capacitance. I just wanted the same voltage on the input as the gate. If/when I get it to operate the way I want, I will be putting a capacitive divider on it as it will go on a resonant LC circuit with a greatly magnified voltage. Mikek

It is a 0.1uf (104), no signal loss measuring on either side of that cap. Mikek

Idss is 3 mA min on that fet, if it's good, so the gate shouldn't be forward biased.

Who is Dagmar?

There is no LF rolloff.

The real problem is that the fet resents being called a T when it knows it's a Q. It's a status thing.

Yes, I measured the dc voltages posted. Not sure where 16u comes from the only big cap I have is the 22uf tantalum on the input, and as discussed that is not need, I added that to stop the capacitive voltage divider affect of the original 27pf and the gate capacitance. B+ bap is a 104, then I added 180pf right at the drain/collector of the semis. No Change. Mikek

Forward biased, meaning more voltage on the Gate than source? I don't have that. 2.4V gate and 6.5V source.

He was a pretty heavy contributor here for quite awhile, he posted this 5 years ago, I have seen any posts from him for a long time. Mikek

Looking at your board and the Onsemi data sheet, it looks like the input signal is connected to the drain. You might check me on that.

Your scope probe is connected after that final RC, not to node C. That will trash LF gain.

Is that "impact ionization"? Worsens gate current from pA to nA or something? Could be happening here, OP could add a series resistor in drain to drop D-S voltage below 5, can then try bypassing drain to ground or bootstrapping it from emitter too. Tons of solder-smoke fun!

piglet

Wasn't that James Arthur?

piglet

I think that is right. Mikek

Good point. If the pcb is double sided with vast ground plane then the strays to ground might explain all. Perhaps Lamont/Mikek can try peeling off a section of groundplane under the gate nodes and measure if there is an improvement.

piglet

Oh, that's my friend James. He skis in short pants.

I all the idiotic flaming and OT nastiness here drove him away.

The goal here is a high input impedance circuit flat from 50kHz to 30MHz. I'm willing to make another board, should I just remove all the copper and have a positive and ground rail on either side. I can superglue islands for the connection points. I punch and cut islands from Rogers 5880 1/16" teflon PCB and glue it to the pcb. Or I can Dremel another board. Should I use leaded components and space them off the board or just remove all the groundplane and stick with the surface mount resistors? That's all I have, not smd caps or semiconductors. Also, is this a better circuit to try, this one is tried and proven, I think it is out of Linear's databook or some other company.

With the limitation of the parts I have BF256C, J310, 2N3819 2N5485 >> MPSH10, 2N3866, 2n4401, BC549. Mikek

Yeah, that's it.

Worsens gate current from pA to nA or

Microamps!

Could be happening here, OP could add a series resistor in

You can do impedance-matched GHz circuits on dremeled copperclad with surface-mount parts.

At 30 MHz, capacitance to the backside copper won't matter. At a couple GHz, trace impedances might matter some.

It might benefit from vias from top to bottom ground. I like 2-56 screws, which are good ground lugs too.

If the ground plane is causing some of the stray capacitance do I even want a ground plane? Mikek

Yes but OP wants the normally incompatible mix of high impedence and high frequency. Even 0.15pF is going to get Zin down to 30k at 30MHz and even a 0.1 inch by 0.1 inch pad over a ground plane is going to exceed that.

piglet

Just lift the gate node a few mm off the board as a first try. All the other nodes are Lo-Z and can stay as they are.

Could be easy to adapt your existing unit into that other design, major substantive difference is the bootstrapped drain. Try to have the gate up in the air.

piglet