here to go to our main page on S-parameters
here to go to our page on mismatch loss, loss factor etc.
to go to our page on VSWR
here to learn some basic network theory
Using manufacturers' S-parameters,
you can make all manner of useful plots, which will help you evaluate
candidate parts, as well as the give you some insight into how accurate
the data really is. The file is called "S-Parameter_Utilites_101.xls",
and its latest version is located in our download
area. Go get it, behold its sheer awesomeness, and use this
page as your guide to working its magic!
Update for June 2007! We
updated our S-parameter file example using new data downloaded from
Hittite's web site for the HMC467LP3 switchable attenuator in its
0 dB state (thanks Mike V. for pointing out the new data!) If you
visited this Microwaves101 page before, you'll remember that the
original data on this switchable attenuator had two flaws, first,
that the S12 measurement was flakey, and second, the part did not
work down to DC as advertised (our guess is that they measured it
in a fixture that had blocking caps). Looks like they corrected
both problems, so we redid the analysis. You'd think by now they
might open the Hittite wallet and sponsor this page!
Before we go further on this
topic, let's give kudos to both Hittite and TriQuint for offering
downloadable S-parameters of their MMICs on their respective web
sites. We'd like to think we shamed TriQuint into this (on this
very page) but Texans never admit it when they latch on to other
people's ideas, nor do they pick up the cue even as the rest of
the world rolls their eyes...
Eleven plots are available in
the latest version of the spreadsheet. If you have any suggestions
to make the analysis more versatile, send
us an email!
(with and without smoothing!)
phase (with and without reference plane extensions!)
and maximum available gain
for "good" S-parameter data
Below we give you some things
to consider when you use our S-parameter utility spreadsheet to
evaluate manufacturer's S-parameters. For more info, see our pages
on basic network theory and K-factor.
- For passive reciprocal parts,
S12 must equal S21. Example of passive non-reciprocal parts are
circulators and isolators.
- For passive parts, S21 and
S12 can't be greater than 0 dB.
- In the intended band of operation,
mismatch loss should be just a few tenths of a dB for a passive
part, maybe a little higher (0.5 dB) for an amplifier.
- For a passive part, K-factor
cannot be less than 1.0, by definition.
- For an amplifier, K-factor
should be greater than 1 at all frequencies, by design,
even outside of the band, or you'll have stability problems using
Example plots: Hittite HMC467LP3,
0 dB state
We downloaded the S-parameters
for this passive reciprocal device from the Hittite web site, which
is very friendly, and provides S-parameters in "S2P" format.
This is a nine column text file with some specific formatting in
the header which you can ignore for now. The nine columns that you
Every once in a while you'll
find data that is not in this order, it's your job to correct it!
To use our spreadsheet, you first
need to read the S-parameter data into Excel in a separate file
(choose "all files (*.*)" as the file type, using the
"delimited" option, with "spaces" as the delimiter).
Then copy nine columns of data from the Hittite file into our spreadsheet
that you just downloaded, on the "Enter_data" sheet, starting
in the first frequency box. That was easy! Save it with a file name
that indicates the data that you entered. You'll need to do three
more things before you even look at the plots. There are two pull
down boxes, one says whether you are in dB
or magnitude or real/imaginary format, and one tells the spreadsheet
what the frequency units are, make sure you have these correct.
You also need to specify the system characteristic impedance (almost
always 50 ohms, but once in a while it will be 75 ohms). You should
copy some of the header info onto the "Enter_data" sheet
on the blank rows near the top of the sheet so you will remember
what data you were looking at, then save it under a new filename.
Now all you need to do is go to the plots, and adjust the axes and/or
make other formatting changes to pretty them up, and add as much
detail to the titles as you want. Voila!
The example we chose is a digital
attenuator, Hittite HMC467LP3, measured in its low-loss state. This
is a passive circuit, with bandwidth advertised from DC to 6 GHz.
Let's look at some of the plots:
This is what the S-parameter
magnitude looks like if it is not converted to decibels. Boring!
Because this is a passive reciprocal circuit, the magnitude of S21
and S12 should be exactly the same. Here's the original data on
the HMC467LP, which we pointed out was problematic. The S12 data
has some ugly ripples, and the magnitudes of S12 and S21 go to zero
at 0 GHz, indicating that the part doesn't work at DC as advertised.
If you ever measure a passive 2-port device where S12 and S21 aren't
equal in magnitude, you've messed up the measurement and need to
Now then, here's
the new data. The data is looking good, and passes our passive reciprocal
sniff test (S21=S12)!
From here down we
discarded the old data.
One thing we'd like
to pass on, if you ever try to put Excel plots onto a web site,
you'll need to create jpg or other images of them. One thing you
DON'T want to do is to scale an image of an Excel plot (we know
this from experience) because the text and lines will soon become
unreadable. Here's what we recommend in order to obtain a crisp
five-inch wide image.
1. Change the plot type from
portrait to landscape
2. Change the right side margin to 2.5 inches
3. Change the lower margin to 6 inches
4. Select all of the text and change it to 12 pitch.
5. Create an image using Paintshop Pro or other software
Now go back and
compare the two S-parameter magnitude plots (above). The upper one
was scaled, the lower one was done as suggested. Which one is easier
to read? Yet another helpful hint for the unwashed masses provided
as a service of Microwaves101...
magnitude in dB
Here are the S-parameters, in
decibels. In our spreadsheet we put S21 on the primary Y-axis and
S12, S11 and S22 on the secondary Y-axis. This is helpful for an
amplifier, but in the case of this passive device, we suggest that
you move S12 onto the same axis as S21 as we did here.
In order to convert
S-parameters to impedances, you must specify Z0. Usually it's 50
ohms, sometimes 75 ohms.
to get from S-parameters to impedances is more complicated than,
for example, VSWR. This is one of the reasons the Smith
Chart was invented, you could enter coordinates either way and
the graph would solve the equations for you. Here's one form of
the equations, sent by an alert engineer named Steve:
He sure like brackets! Here's
the input and output impedance, with real and imaginary parts plotted
separately. Ideally the real part is 50 ohms, and the imaginary
This is just the
above plot, with the impedances normalized to Z0.
Here we've converted S11 and
S22 to VSWR. The part is well matched, under 1.5:1 over the entire
DC to 6 GHz band. Again, input and output VSWR should be equal for
a passive reciprocal part, here they're slightly different, but
no measurement is perfect, and measurement errors are magnified
when you look at them in terms of voltage standing waves. Here
Click here to learn more about VSWR.
Hey, this part is pretty-well
matched! You lose less than 0.2 dB due to reflection.
The concept of loss factor is
discussed on this page. This is a
measurement of how much energy is either radiated or goes up in
heat, due to resistive or dielectric losses. In this case, no more
than 15percent of the input power is "wasted". Loss factor
should be equal in forward and reverse directions for a passive
reciprocal part such as this.
Efficiency factor is also discussed
further here. This measurement answers
the question, "if I could perfectly match this part, how much
loss would it have?"
delay is a measurement of how long it takes electrical energy
to go through the part. The bumpiness of the raw data (blue trace)
is in our opinion because the data points are too close together
for the angle measurement to be accurately resolved between frequency
points. They might have improved this by applying more averaging
to the measurement to reduce noise. At least the bumpiness tells
us that they didn't smooth the S-parameter
You'd get a more realistic look
at group delay if you smoothed
the group delay. Which is something you can actually do using
our spreadsheet! In the purple trace below we've added an "aperture"
of five frequency points along the curve to smooth the data. From
this trace we approximate the group delay at 0.045 nanoseconds,
at least at the band center.
Using our 1 foot equals one nanosecond
rule of thumb, let's calculate the
0.045 foot length of this part in smaller units:
0.045 foot x 12 inches/foot
x 1000 mils/inch=540 mils
long or about 1.4 centimeters.
phase and reference plane extensions
This is described in more detail
on a separate page. Here's the
unwrapped phase of S21:
Now let's move the reference
planes 0.7 cm into device on both sides (1.4 cm total, see group
delay calculation above). The phase is much flatter over frequency.
It's no coincidence that moving
reference planes 0.7 cm on each side (1.4 cm total) to achieve a nearly
flat phase, we've made the part have a near-zero electrical length.
and maximum available gain
Because this part is purely passive,
its K-factor should be greater than 1 by definition (unconditionally
A recent addition to the spreadsheet
is the ability to make Smith chart plots
of the input and output reflection coefficients.
Does the Hittite
HMC467 data pass the sniff tests? Indeed it does for the remeasured