Liste des Groupes | Revenir à se design |
Am 05.04.24 um 17:55 schrieb John Larkin:On Fri, 5 Apr 2024 11:38:43 -0400, Phil Hobbs
Try putting BLF03VK600 beads in source and drain. Besides being rated at
3 GHz instead of 100 MHz, it has really nice low Q everywhere.
My cascoded lab amp proto is using three SAV-331+'s in parallel like
that, running about 2.5 mA each, and shows no sign of trouble.
Cheers
Phil Hobbs
We tried various fixes, no joy. The circuit is complex, lots of diodes
and comparators and varicaps and stuff, so there are too many resonant
stubs.
We prototyped the new sorta-Colpitts circuit, using the BUF602 as the
gain element, and it's great. Having a closed-loop near-perfect
follower is nice.
Do you have slowish feedback into the source? From the FETs POV
that makes it look like a capacitively loaded follower. It translates
directly into a negative real part of the input impedance.
<
https://www.flickr.com/photos/137684711@N07/34701106245/in/datetaken/lightbox/
>
It is the input impedance of 2*IF3602 and the negative real part
can get REALLY large, out of bead-land.
In the Smith diagram, S11 is decoded at the marker positions.
Where the trajectory gets out of the circle through 0 and inf,
there comes more energy back from the DUT than the VNA sends to it.
Cannot happen with passive DUTs.
It is a really hard problem and even in AOE3 is a bad example.
I got an array of 16* CPH-3910 stable with feedback via a
3 GHz CFB amplifier. But CFB's 1/f noise easily dwarfed the
noise of the 16 FETs even after 40 dB of gain.
I tried driving the feedback with 2 * BFQ19S BJTs as a follower.
It seems it kinda works in simulation. Generates a lot of heat.
Not yet built.
Les messages affichés proviennent d'usenet.