Lots of good input there, Andy; I appaeciate it.

The part in question claims it is stable working into and from 50 ohms; I do
put too much emphasis on matching to that somewhat arbitrary value at times.

---Joel

AndyS wrote:
Joel Kolstad wrote:
On a reciever, I'm using a MMIC LNA at VHF

Andy writes:
Joel, all of the advice I've read on this post is good advice, but
I didn't see any reference to the Linville or Stern stability factors.

Devices, ESPECIALLY at VHF, which are matched into source
and load often oscillate. Hell, they USUALLY oscillate.
Hence , S12 is a major factor in these designs....

Most of the time, devices in this range must be mismatched
to some degree to provide stability.

Motorola used to have an excellent App Note on amplifier
matching, and the degree of mis-matching to ensure stability
It was called " Effects of Y parameters" or " Calculating Stability
in... " or some such, and the number was in the 100s or 200s.

If you know the Sparamters ( or Y parameters, for the app note)
you can easily do a Linville on the numbers to see if the LNA
is unconditionally stable for all source and loads.....

It probly won't be.....

Then, you can do a Stern to see what degree of mismatching is
needed to just reach the stability threshold.... If you program
the equations into a computer ( I use BASIC but an EXCEL
spreadsheet will work well also) you can quickly decide what you
want to do. Note that the amplifier configuration is the source
of the parameters and not just the device. For instance, common
base is often unconditionally stable while common emitter more
often than not is never unconditionally stable. The S or Y
parameters for the amp in these configurations show it
when a Linville is applied.

It may be that optimum noise figure can't be reached with
unconditional stability...... so you just do the best you can.......
Personally, I've always found the best noise figure while
diddling the circuit, rather than calculating it..... but that's
just my own experience...


Good luck on your design. And don't get so hung up on this
"maximum power transfer Thevin" stuff, since any amplifier that
is Thevenin matched on it's output could only achieve 50% efficiency,
and most amplifiers do a lot better than that...... for the simple
reason that max power transfer isn't the main goal of an amplifier.....

Andy W4OAH in Eureka, Texas

Note that the free RFSim99 can display stability circles on the Smith
chart display of S11 and S22. They change as you move the cursor
frequency, and redraw quickly so it's easy to check over a frequency
range.

I don't know why the myth persists that you can achieve no more than
50% efficiency if a load is matched to the source impedance. A
Thevenin-equivalent source does not necessarily look anything like the
Thevenin circuit inside; it simply behaves WITH RESPECT TO THE LOAD in
the same way (over the range of loads and outputs for which it behaves
linearly, since the Thevenin model is a linear one). But that's all a
red herring anyway: in an LNA, "efficiency" is probably one of the
LAST things you should worry about. You should worry about noise. You
should worry about stability. And you should worry about distortion.
You should even worry about where you'll put it in the system. All
those would normally come a long ways ahead of concerns about
efficiency.

Cheers,
Tom

Joel Kolstad wrote:
Lots of good input there, Andy; I appaeciate it.

The part in question claims it is stable working into and from 50 ohms; I do
put too much emphasis on matching to that somewhat arbitrary value at times.

---Joel

Put its S-parameters into a two-port instantiation in RFSim99, Joel,
and you can play with matching and stability to your heart's content.
It won't help you directly with noise, but if the data sheet tells you
about how to match the input for lowest noise, that would be at least a
good starting point for the input network. I think you posted that
it's an MMIC; they often don't provide as good a set of info as do
"bare" transistor data sheets. NEC, for example, seem to do a good job
providing data about S-parameters versus frequency and operating
conditions, and about noise figure versus input matching as well.

Cheers,
Tom