From the above statement, I can tell that you have very little experience
with doubly balanced mixers, especially the ones from Mini-Circuits.


You're right. I ran some of their stuff through the bench many years
ago and was disappointed, so I never used it. As for size, Analog
Devices has been making some remarkable stuff in the last few years.

I have worked with some of their newer stuff, and it has been very good. AD
got their act together pretty well, in the RF arena.
The Analog Devices AD831 isn't a bad design; it does have a good IP3, but in
order to reach the NF of a Mini-Circuits SRA-3 however, you need to have a
preamplifier ahead of it. With its 12dB NF, it isn't a bad mixer for HF use
up to 30MHz. I had started a receiver design using the 831, but things got
so busy at work that I shelved that project for awhile.


The
LAVI-XXX series of mixers have IP3s in the +33 to +40dBm range.


You used dB before, which I assumed was carrier attenuation. Still,
I'm not impressed.

I thought the only reference to dB was port to port isolation and SSB
conversion loss.


The only
type of discrete mixer that can even come near this type of performance is
something that uses either a quad JFET ring, a quad CATV bipolar ring, or
a
dual power FET type that uses something like the Siliconix VN66. Your
typical balanced dual JFET mixer, as used in some of the Yaesu and Icom
transceivers will achieve IP3s in the +10 to +15dBm range, which isn't
bad.
This is without having the preamp switched in.
Now, to even be able to measure that type of performance, you need to have
all of your RF sources very clean.


Exactly! That's why I pointed out those numbers are "lab numbers". If
you want to get some realistic numbers you have to test it under
realistic conditions, which isn't that hard. The only drawback is that
the numbers will be relative; i.e, it's a comparison test against
other circuits. But if you do you will find that what I'm saying is
true -- discretes perform much better. And yes, you have to carefully
match the curves. This added labor, along with higher assembly costs
and parts counts, are the primary reasons why discretes are rejected
over mini-bricks; it rarely has anything to do with performance.

I agree on those points. Unless the LO in the actually is actually filtered
to the point where all higher terms are at least -65dBc, that performance
won't be realized. The manufacturers I worked for over the years were quite
happy with -25dBc for the 2nd harmonic of the LO.


It also serves as a buffer to the mixer, which is essential for
reducing mixer IMD. The RF amp is generally a good idea.

The RF amp will not reduce IMD..........it will actually degrade the IMD
performance of the mixer by the amount of gain that the RF amp provides.
It
is very easy to see this if you are making IP3 measurements on a mixer.
Add
10dB of gain ahead of that mixer, and IP3 degrades by 10dB.


I can see that you are locked into a voltage-only mode. Feed your
mixer under test with signals of varying impedance. I think you will
be suprised, if not shocked.

You do make a good point; an unconditionally stable low gain RF amplifier
will satisfy this requirement. I have done the measurements that you
mention, and I have noted some level of disparity between real world
conditions and manufacturers' specs. I know................too many
manufacturers play the "numbers game". As long as they stick to the same
standards, one can use these numbers to initially select a product but the
devices still need to be characterized before those numbers are actually
believed.

Pete