IMS2012 Cost-Effective Millimeter-Wave Signal Analysis Approaches | Keysight Technologies
0 Comments


Hello, everyone. I will do a presentation
on Agilent’s millimeter-wave options and, basically, the commercial state of the art
for what is available. The paper was written by Ben Zarlingo, who is a quite prolific micro
app writer, and I was recruited to take his place. My name is Bruce Erickson. I am one
of the hardware design architects at Agilent, and I primarily work with spectrum analysis.
Millimeter-waves, of course, is not so different than a lower-frequency measurement system,
but there is quite a few extra challenges. I will talk a little bit about spectrum analysis
options, the connector challenges — coax versus waveguide and so forth — and some
choices along those lines. I will make mention of some signal processing methods to get additional
sensitivity. So millimeter-waves has extra challenges for many reasons. Hopefully, when
you are putting together a test system you will not have all of these challenges; hopefully,
only some of them. Typically the cost of a system becomes more expensive. Millimeter-wave
test system, it is frequently expected that will cost more. It is not necessarily the
case, and I will describe some of that. I have mentioned the additional difficulty.
Of course, I think most people appreciate that there is more loss — connection losses,
impedance match. A lot of people who work at lower frequencies do not realize that if
you get a 1 meter coax cable at 70 GHz, that could have 5 to 10 dB of loss. So if you specified
a very accurate sensitive mixer for 70 GHz, and then you add a 1 meter cable, maybe that
is foolish. Or maybe your requirements are not very stringent. I mentioned mismatch can
be a problem, but waveguide is quite nice. Waveguide is very forgiving, has excellent
match, so although coax cable gets quite challenging and expensive, for banded systems waveguide
solves the problem nicely. So generally with a millimeter-wave system you would not expect
to have the same specifications as you might at 5 GHz. However, that is changing, I will
show some products approach low-frequency performance at millimeter waves. So these
are definitely changing. I do not know if my pointer is working here, but, in particular,
the displayed average noise level has been improved very much in recent products. I know
if you look around the floor and you stop by the booths, and you look at some of the
external mixing solutions and some of the millimeter wave spectrum analyzers, you will
see that the displayed average noise level can actually be very low. In recent years
there has been a lot of consideration of measurement accuracy regarding mismatch errors of connectors,
and historically the philosophy has either been a worse-case computation or an RSS computation.
It has become more fashionable now to use the relay distribution, and if you do that
you will find that gives you a better representation, in most cases, and you will have a smaller
error budget. You see, Agilent does have an application note on the topic. That is an
interesting subject for all frequencies. So if you are putting together a millimeter-wave
system, there are many choices. One of them is where should the first mixer be? Well,
it should be first. Right? What we mean here is physically. So I mentioned the oppressive
losses of coax cables. So you want to keep any coax cabling as short as possible and,
of course, that is a benefit with waveguide as well. So an external mixer, of course,
would have the advantage that it could be installed right at the device under test.
Yes? Versus an instrument that had internal coverage. Let’s say we build a 0 to 500
GHz spectrum analyzer in a big box. You would probably interconnect with a cable. Yeah?
So an external mixer solves the cable link problem. The oscilloscopes are very popular.
They are excellent tools for looking at wide bandwidth modulated signals, and you will
find that most spectrum analyzers have an IF output where an oscilloscope can be integrated
to look at these wider bandwidth signals. So frequently you will see a hybrid system
with an external mixer, a spectrum analyzer and an oscilloscope. Very common. There are
2 products that have been recently updated in the Agilent portfolio. This year the new
product is the N9010A EXA. The frequency coverage has been extended to 32 and 44 GHz. The interesting
thing about this product is, it is really a very low-cost product that supports external
mixing. Most external mixing support options exist in the higher performance products that
are very much more expensive. So this is really a very low-cost way of getting into millimeter-waves.
It also works with a new product called the M1970 smart harmonic mixer. This is a USB
device, it is a USB external mixer. It will self-recognize its plug and play. Very easy
to use. Pre-amplifiers have become easier, and most of the 40 to 50 GHz spectrum analyzers
now have available internal pre-amplifiers. I mentioned that the sensitivity is improving
at millimeter-waves. The internal pre-amplifiers are very beneficial. That was something that
was generally not available previously. If you are working with wideband signals, some
of you may realize that many spectrum analyzers have a tracking filter to reject the image,
frequently using YIG technology. That is a nice pre-selector that works well if you are
looking at narrow-band signals. However, if you had a wideband modulated signal greater
than, perhaps, 50 MHz, the pre-selector would introduce many flatness, phase and amplitude
artifacts, and it would be very difficult to achieve a good EBM measurement through
a YIG filter. So you will see many products have introduced a concept called the YIG bypass.
What that will do is bypass the YIG filter so you can get the entire IF bandwidth, and
so in the case of Agilent products, if you bypass the YIG filter, you can have 160 MHz
bandwidth internally, and if you want more than that, you use the IF output option, where
you can use an oscilloscope or external equipment to go up to, perhaps, 500 MHz bandwidth. One
nice thing about using a millimeter-wave spectrum analyzer as opposed to constructing a test
system from individual pieces is all of the features that have been developed for other
spectrum analysis applications. Such as the software used to analyze pulses and demodulate
vector signal analysis to look at INQ, digital demodulation, phase noise analysis, noise
figure, these are all available options that will automatically be available in the millimeter-wave
range, and these would all be challenges to create on your own. So it is very nice to
either buy a millimeter-wave instrument or extend the frequency range, and then you automatically
have this long list of available capabilities. One thing that has been introduced in a number
of products is called external source control. Some people may remember long ago there were
things called tracking generators. They were an internal source. We have found that many
people would like modulation and better capabilities than a simple tracking generator, and so most
people would prefer to use an external source that has very much more capability. It could
be modulated; it could be used for other applications. We are able to synchronize sweeps and step
through the range and get the equivalent behavior as a tracking generator with external sources
now through the entire frequency range. On an EXA this could be done to 44 GHz, on a
PXA it could be done to 50 GHz. It is like you have a 50 GHz tracking generator, which
is quite nice. I mentioned one choice to be made. I think anyone building a system for
millimeter-waves is faced with this choice: Should they buy an instrument that covers
50 GHz or should they use an external mixer and possibly not have a cable linked to the
device, and they are both very attractive options. It depends on the specifics of your
case. The nice thing is there is a variety of products available to support any of those
configurations. I think most people probably realize an external mixer supplements the
internal mixer. So you see here is a picture of a PXA interfaced with a legacy 11970 mixer.
Notice the RF input port in the corner is not used at all. The only interface with the
instrument is the IF and LO, which are diplexed in one cable. I mentioned the new M1970 smart
harmonic mixers, they have very good performance. You will see the sensitivity is quite good.
The phase noise is quite good, and, really, the performance is not so different than what
the base box would be. There is not so much difference between buying a 50 GHz spectrum
analyzer and buying an external mixer with this new product. It does, however, not have
the YIG filter, so if you are looking for spurious responses at very low level, that
is something to consider. Again, I mentioned phase noise in all the applications are available,
and I mentioned the coax versus waveguide trade-off. Really, we have to support both,
and solutions for both are available. The last thing I want to mention is the product
has internal noise floor subtraction that automatically can happen in the background.
So unlike manually trying to subtract noise, that feature is supported internally very
easily. Thank you. Any questions? Thank you.

Leave a Reply

Your email address will not be published. Required fields are marked *