Ok, let me
think............................................. ...........Actually, I
can't claim all of this knowledge as my own, especially the troublesome 6th
and 9th harmonics of the LO. I have to give Mr. Lu Chen of Mini-Circuits for
that bit of knowledge.
Mr. Paul Vinsand of Mini-Circuits was nice enough to give me that -65dBc
figure when I was running some tests on a mixer that had a +40dBm IP3. Nice
folks.

Pete

"Somebody Somewhere" wrote in message
ups.com...

Pete KE9OA wrote:
IP3 is a figure of merit for dynamic range of a mixer. As already
mentioned,
the higher the number, the better the unit.
Basically, the measurement consists of using three RF generators, with
each
RF port generator running through a 6dB attenuator, a low-pass filter,
another 6dB attenuator, and a combiner. The output of the combiner is
connected to the RF port of the mixer under test. The reasons for the 6dB
attenuators are twofold; first of all, they provide 12dB of isolation for
the RF generators that is added to the 25 or so dB isolation that the
combiner already has. This helps to prevent the generators from "talking"
to
each other, thus preventing IMD to be generated in this portion of the
test
setup. The second reason for these attenuators is to provide a broadband
resistive termination for the low-pass filters, so that they maintain
their
design characteristics.
A third RF generator is connected to a 3dB attenuator, a low-pass filter,
and another 3dB attenuator to the LO port of the mixer. The reason for
the
3dB attenuators is to provide a wideband resistive termination for the
low-pass filter so that it retains its design characteristics.
The low-pass filters are very important in this test setup, since when
the
signals are squared up in the switching function of the mixer under test
harmonics can cause measurement error. Mini-Circuits has a requirement
of
at least -65dBc for all harmonics present in the test setup. The 6th and
9th
harmonics can be especially troublesome when making IP2 and I.F. port
return
loss measurements.
As far as injection levels, the LO generator is set to the level required
to
illuminate this port. For a Level 7 mixer, this would be +7dBm, or 5mW.
With
this type of mixer, the level of each tone at the output of the combiner
that is applied to the RF port of the mixer needs to be at least 10dB
below
the 1dB compression point of the mixer. Since the 1dB compression point
for
a typical Level 7 mixer is abour 0dBm, we would be talking about a
maximum
level of -10dBm for each tone. -20dBm would be a little bit better, just
to
make sure that you are operating within the linear range of the mixer.
Now that we have the proper test setup, we connect all of this to the
mixer,
and we connect the output port, in this case the I.F. port, to a spectrum
analyzer, also making sure that the spectrum analyzer is set up for
maximum
dynamic range so that IMD isn't generated in this portion of the test
setup.
Your test setup needs to have an IMR at least 10db better than the device
you are measuring, in order to minimize any measurement error.
Use at least 50kHz separation between your input tones that are applied
to
the RF port; the reason for this is so that phase noise sidebands from
the
RF generators don't cause measurement error.
Taking a look at the spectrum analyzer, you will see five major tones;
these
are, the LO, which should be suppressed by around 30dB or more, an upper
sideband tone, a lower sideband tone, and upper sideband and lower
sideband
IMD products. These two IMD products are your third order terms.
Next, measure the difference between your upper sideband tone and your
upper
sideband 3rd order term. Do the same for your lower sideband terms. This
difference between the desired sideband and the 3rd order term is
referred
to as your IMR.
The reason for measuring both of these terms is because oftentimes, one
of
these measurements will yield better results.
Choose the worst result when characterizing the mixer.
IP3 is calculated by this method: IP3 = (IMR/2 + Pin), whereby Pin is
defined as the power level at the output of the combiner of only one of
the
input tones that are applied to the RF port of the mixer.
Some of the reviewers make this measurement using a 5kHz offset of the RF
generators that are applied to the RF port of the mixer. This measurement
is
only valid if the noise sidebands of the generators are run through a
very
selective filter, such as a crystal filter. In this case, you need to
make
sure that these levels are well below the level where IMD would be
generated
in the crystal filter itself; also, too much power can shatter a crystal
filter. If I remember correctly, it would be in the ball park of +10dBm.
I remember when I was testing FM communication systems, the test setup
would
be similar, with the exception that the spectrum analyzer wasn't needed
and
the LO generator wasn't needed since the actual receiver's LO would be
used
in this case.
Now, you still need three generators, this time using a 3-port combiner
at
the RF port of the input stage of the receiver.
The two interfering RF port generators would be modulated with a 25%
modulation index. First of all, only the desired signal RF generator
would
be switched on, using a level that would provide a 12dB SINAD. This is
equivalent to using a 10dB S/N+N ratio in an AM system or a 5% BER in a
digital system. Next, the second and third generators are switched on,
increasing their levels until a 3dB degradation is noted. Now, these two
generators need to be set to frequencies so that their difference
frequency
lands on the desired channel where you are making the measurement. Once
again, you need to do this with upper sideband interferors and lower
sideband interferors, since the rejection will not be symmetrical.
The difference between the levels of your desired signal and your
interfering levels is your IMR. IP3 is calculated in the same manner as
mentioned before.
It is no trivial task when making this kind of measurement. The
measurement
itself isn't difficult, but the main thing is to make sure that test
setup
created IMD doesn't cause any measurement error. After you do this a few
times, it becomes relatively easy, that is, until you have to
characterize a
device such as an RF switch that might have an IP3 of +50dBm. This
requires
an IMR in your test setup of at least 110dB. Now, that takes a bit of
work!
I hope this long-winded explanation helps.

Pete



Could you please be more specific? LOL! ; ^ )