EMRFD Message Archive 81

Message Date From Subject
81 2006-08-08 09:26:52 kilocycles Advice needed on Converters for PC-based FFT SA
I need some advice on enhancing the coverage of my 16 MHz FFT
scope/spectrum analyzer (Velleman PCS64SI). I'm committed to building
the test setup shown in Fig 7.34, page 7.20 of EMRFD. I've made the
PC Board set for the test sources of fig 7.32 and the Feedback amps of
Fig 7.35. I have a homebrew hybrid recombiner, and the correct parts
for a 10-position step attenuator.

My idea is to use a switch-selectable set of crystal-controlled
converters for the SA, with the SA input being the IF. Cheap $.40
crystals for 16, 24, 32, and 48 MHz should give me ranges of 0 - 16
MHz, 8 - 24 MHz, 16 - 32 MHz, and 32 - 48 MHz. My concern is flat
frequency response and reduction of out-of-band signals. For example,
if I view the fundamental of 10 MHz on the SA, I'd like the harmonic
levels of 20 MHz and 30 MHz to correlate with the carrier. I'd also
not want to have artifacts from mixing in the converters show up in
the output.

If I could build an oscillator that can give the 4th overtone of one
of the 16 MHz surplus crystals, I could also have the range of 48 - 64
MHz. That wouild be good, since I could then see the third harmonic
of up to 15 meters.

One idea I have is to use a low-gain RF amp using a J310 in common
gate as an RF amp as in figure 6.32 on all converter bands with
front-end bandpass filters for the 16 MHz bandwidth of each band, if
they can practically be constructed. In reading the text regarding
the design, I'm not sure what the effect of a wideband tunded circuit
would have on circuit performance. At he converter outputs I would
use wideband bandpass filters, or high pass filters. It appears that
the SA display would have to be read in reverse with the converters in
play, since the mixing scheme is IF=RF-LO, hence the reason for
including a 0 - 16 MHz converter for consistency in reading the display.

From what I've read from the topics of the past few days, I should be
able to construct feedback amplifiers for the converter outputs with
resistive pads to balance the converter gains and accomodate to some
extent the gain difference within each band as frequency increases.

Does this appear to be a sound approach? It would seem to be a better
solution than to use a signal generator as a tuneable LO, although
that would look to be simpler.

Thanks,
Ted KX4OM
85 2006-08-13 20:53:47 Wes Hayward Re: Advice needed on Converters for PC-based FFT SA
Hi Ted and gang,

This is a rough one. Spectrum analysis is more demanding that a
simple receiver, for you are trying to get measured data out of it,
beyond some mere sound. A filter or amplifier that might have a
slight gain slope can really get in the way.

It is possible to build a converter that you put ahead of an
existing analyzer that will provide some very useful data.
However, it must be used with care, always keeping the sins of the
converter in mind.

One thing that you always want to do is to avoid a local oscillator
for the converter that is within the tuning range of the input.
So your choice of crystal controlled oscillators may be
problematic. But you can do some interesting things with cheap
crystals. Virtually any of the sub-one-buck computer clock
crystals will oscillate not only on their fundamental, but on odd
overtones. (Crystal oscillators don't function on even overtones,
or multiples. It has to do with the boundary conditions for
oscillation.) So if you have a 10 MHz crystal, you can make a 30
MHz oscillator out of it. This will have output at 30 and the
harmonics of 30 (60, 90, etc), but will have no output at 10.
Don't try to make such an oscillator with a logic gate – you really
need a tuned circuit to guarantee that the circuit will function at
the desired frequency and not elsewhere. Take a look in Chapter 4
at the crystal controlled oscillators.

The first place I ever used this was in a 15 meter transmitter that
used 40 meter crystals. It ran 1 watt input and was, hence, named
a "Milligallon for 15." It was in QST back in the late 1960s.

My main spectrum analyzer for home use is still my homebrew job,
which tuned from almost zero up to 70 MHz. I wanted something that
would let me look at harmonics of my 6M rig as well as 2M rigs and
the like. I built a box with the following:
(1) TUF-1 mixer. (2.) Minicircuits VCO that tuned from about 150
to 290 MHz or so, and (3) a pad of about 3.5 dB. The box had a
pot in it that controlled the frequency of the VCO with a knob that
was just calibrated with the oscillator frequency. The VCO
output was attenuated to +7 dBm and applied to the LO port of the
TUF-1. The 3.5 dB pad was in line with the input. When added to
the 6.5 dB loss of the TUF-1, the system presented a net loss of
about 10 dB, allowing me to use the existing analyzer calibration
with a simple mental adjustment.

Let's say I wanted to look at the 2M band. I would just set the LO
to 130 MHz, which was about the lowest that VCO would go. A 144
MHz signal would then show up at 14 MHz on my baseband analyzer.
But there is a problem – a signal at 130-14=116 MHz would also
produce an output at 14. It is always easy to identify such an
image though. Just grab the knob on the VCO tuning and increase the
frequency. If you start with a 144 MHz input and move the
converter LO to 131, the output will decrease to 13. But a signal
at 116 MHz will come out at 15. The "real" signal and the "image"
will move in the opposite directions.

You can get rid of these problems with preselector filters.
However, such filters will add loss and will probably not be as flat
as you would like. Moreover, unless they are terribly
sophisticated, they won't be sharp enough to completely get rid of
the problems while still preserving the flatness and accuracy you
want. You are better off to live with some spurs with a system
that is intrinsically flat, such as a ring mixer, and then apply
some careful thought to guide you in your measurements.

There is generally no need for any gain. You can get pretty good
measurements without it and encounter a lot of problems if you
include gain. That's a good rule for a receiver, and it applies
many more times when doing spectrum analysis.

Rather than doing too much planning, grab a diode ring and a signal
generator or two and do some experiments. If you don't have extra
signal generators (they are like power supplies – you can always use
more of `em) you can build a simple generator like the one in Fig
7.27. Build it for just the upper band with just one variable
cap. It is still a super useful tool. Once you do some
experiments, I'll bet that you will have a lot better feeling for
where you want to go.

73, Wes, w7zoi