It's characterized for 100 MHz to 2690 MHz, which probably means no internal matching network, so it might be good for fast switching.
Weirdly it has no I_Dmax rating.
$22 at Richardson
Cheers
Phil Hobbs
Interesting, but the specs are terrible. No DC curves, no Idss, no Imax, no capacitances, and a diagram that pretends that it's two amplifiers.
I wonder where the 100 MHz limit comes from.
Biasing is the usual RF nonsense: play with it until it works.
The key is that it is intended for implementation of a Doherty Amplifier, which are widely used in linear RF final amplifier stages because it's far more efficient with spikey data comm waveforms (which look like Gaussian noise) than class A or B.
.
Joe Gwinn
On a sunny day (Wed, 15 Sep 2021 13:16:27 -0400) it happened Joe Gwinn snipped-for-privacy@comcast.net wrote in snipped-for-privacy@4ax.com:
Doherty amps use DC too. They need to be biased.
I'm not a fan of s-params, but they aren't on the data sheet either. Doherty amps are pretty nonlinear, so a Spice model would be nice.
RF people seem to be allergic to Spice.
It is two amplifiers--it's intended for a boosted class-B application, sort of like a Class G. The amps are in fact dual depletion FETs with different characteristics wired common-source. (The pad is the source contact.)
Cheers
Phil Hobbs
Am 15.09.21 um 20:33 schrieb John Larkin:
Power-up/down sequence is described in detail.
Spice means "Simulation Program with Integrated Circuit Emphasis"
When you do boards, then things like transmission lines already require hacks to work at all. Pin diodes do not, since SPICE does not know the concept of carrier lifetime.
On page 10/11 they say they have a s2p file. These are 2 port s-parameters in Touchstone format, probably multiple by operating point.
Input return loss of these devices is probably negative at low frequencies. IE you put 1 mW in and 100 mW come back. That is no fun to stabilize and do something you want at the same time.
We once tried this with early Agilent phemts for use at 144 MHz. When we had that thing stable, noise factor etc was uninteresting. On 432 MHz all it took was some inductive source degeneration.
Maybe they'll have an ADS device kit, but buying Keysight Advanced Design System with the proper options costs an arm and a leg.
Gerhard
They work in the frequency domain, not the time domain, and use different types of simulator to assess the non-linear effects, e.g. Harmonic Balance Simulators
From
LT Spice seems to do a reasonable job with reverse recovery on various diodes, if maybe a bit rectangular about the snap-off. It doesn't have any PINS in its library.
Spice has lossy and lossless tx lines. I use them a lot.
Given a depletion fet, I would have expected Idss to be supplied.
With a conjugate match, probably so, but not in a 50-ohm system surely? The abs max gate current is only 10 mA or something.
A 2N3904 emitter follower will oscillate if you run it hot enough and connect the base to AC ground. (An RC + emitter follower was a popular power-on reset circuit in the early personal computer days, till folks found out how badly it oscillated.)
Well, it occurred to me that a 125-V, 8W FET that had 18 dB gain at 3 GHz and worked down to below 100 MHz might be quite an interesting pulse output stage. ;)
Cheers
Phil Hobbs
One would, wouldn't one. Or at least I_Dmax, but no.
Hey, it's only 21 bucks, cut them some slack. ;)
Cheers
Phil Hobbs
I suppose that the RF boys want to sell billions of parts, and assume they will do a lot of hand-holding for a few big buyers.
They don't want our kind.
Get in touch with your inner grad student. ;)
Cheers
Phil Hobbs
That bit me once, using a 2N2219 as the logic reset on a military project. We didn't know what the heck was going on.
My latest instant-on Colpitts oscillator, with a SAV541, has a ** 499 ohm ** series gate resistor.
The EPC GaN fets are nice and dirt cheap.
Love the AliExpress blob-top construction. ;)
(Yes, I know why you're doing it, but no, I couldn't pass that one up.)
Cheers
Phil Hobbs
The EPC parts need physical protection. We had adventures finding a glob-top that didn't eventually break the BGAs. We had failures that I hoped weren't a circuit design problem; they weren't. The epoxy was cracking them.
The stuff we're using now is a medium-hard-set 1-part thing that cures at 85c.
How did that wind up working?
I see the P500 is on your web site--it's nice looking, and a very decent advance, especially in the trigger jitter department. Factor-of-2 improvements don't come easily once you're down in the 20-ps neighbourhood, especially when vendors pull the rug out by discontinuing all their pHEMT parts. :(
(I still secretly like VFDs though.) ;)
Cheers
Phil Hobbs
Too-hard epoxy is a famous problem in optics and optomechanics as well--at low temperatures it's liable to rip chunks out of lenses, squash glass bead thermistors so they become intermittent, and destroy things like pots and electrolytic caps.
For electronics potting, a bottom layer of silicone helps a lot, but a glob-topped BGA is more like a lens assembly--I can believe that soft epoxy is the ticket. h Cheers
Phil Hobbs
Neg. return loss = S11 means that in the 50 Ohm system there comes more out than gets in.
Connecting the base to AC ground is not enough. It takes an inductance. If you measure into the base/gate of a capacitively loaded follower with a network analyzer, you see a smallish capacitance with a negative resistor in series. It takes some inductor to create a resonance for the oscillation frequency. And the gate stopper resistor may need to be surprisingly large to overcompensate the negative resistance for sure. Bad for noise performance.
(Note to selves: what is the resulting Vnoise of +1KOhm and -1KOhm in series in Spice? Is it ~0 by optimizing the resistors away?)
------ That is my pet problem: A usual low noise amplifier for baseband has a FET with some small source resistor, optional cascode and then an op amp with feedback into the source.
With the feedback closed, the FET is no longer CS because the source follows the gate quite precisely. It is off just by ~loop gain. The FET thinks ist works as a follower. The 0.2 Ohm source resistor gives a wrong impression.
The low drain impedance of the cascode only completes the world as perceived by the input FET: I'm a follower!
But the feedback is late through the op amp. The source voltage is like being capacitively loaded. ---> the Fet generates negative Rin like cap. loaded followers like to do.
If you spice the circuit in AOE3, page 152 and plot re(Vgenerator/Igenerator), you'll see the negative real part of the input impedance over some frequency range. Vgenerator is the voltage source at the input.
I did not find a good solution. Replacing the op amp with a VCVS works nicely but it is hard to solder. A THS4303 3 GHz op amp is fast enough, but has an astronomic 1/f noise that dominates despite the fist stage.
Compensating everything to death works, but I want at least 1 MHz BW. I finally left the loop open. That is surprisingly stable over temp. I'm less than happy because I like to enforce the operating conditions.
Any ideas?
------
Gerhard
BTW: Where is Win? Last time he suddenly disappeared was during the preparation of AOE3. Can we expect a new book? I hope he's well!
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.