At 11:27 AM 2000-02-15 -0800, Steve Thompson wrote:
>On one amp I did a comparison between long hand measurements and the network
>analyser on amplitude sweep. I got good agreement.
You have to be very careful when using any kind of broadband sweep
generator to measure gain compression, as many of the older ones do not
have good harmonic suppression. Some later models utilize PIN diode LPF
switching to reduce harmonics from the generator; but when the test results
are critical to the project, one should always check the generator for
harmonic energy. For example, the good old standby YIG oscillators used in
the 8350-series sweeper microwave plugins typically show as much as -20 to
-25 dBc harmonic energy. A network analyzer with a tuned receiver will
reject these responses; but at -20 dBc harmonic energy into the
Device-Under-Test, gain compression can easily appear soft.
It never hurts to be cautious and certain.
>I've not noticed that FETs generate more odd order harmonic than other
>devices.
See the application notes in the older Motorola RF Transistor handbooks for
examples of higher 3rd and 5th order distortion by FETs than bipolars. But
an easier way to see the difference is with multicarrier tone generators.
However, the higher-voltage RF power FETs do perform much better compared
to devices designed 10 years ago for operation at 12 or 28 volts. I was
thinking in terms of some FET IC amplifiers which I tested several years
ago to replace bipolar hybrid amplifiers.
>Agreed - although for the measurement in question, absolute amplitude
>accuracy isn't essential; amplitude linearity at any given frequency is
>needed, and typically a spectrum analyser isn't up to that job either.
If you're trying to measure the harmonic energy and calculate the
third-order intercept point, then the analyzer has to be calibrated across
3 times the bandwidth. The amplitude flatness specification for most
spectrum analyzers is also on the order of +/- 1 dB or so.
As you point out, another not-widely-recognized potential got-cha is
amplitude non-linearity, sometimes of the CRT itself. The manuals for most
analog oscilloscopes (the Tek 465/475 series come to mind) and some
spectrum analyzers give such a specification which is again on the order of
+/- 1 dB or so (something like 10% for the oscilloscopes; been too many
years since I've had one of those manuals in hand). In fact, even
late-model digital spectrum analyzers can fall prey to this characteristic.
The HP 8590C Cable TV Analyzer, which is actually nothing less than a 10
MHz to 1.8 GHz digitized spectrum analyzer with 75-ohm input impedance and
uP programs to ease CATV measurements, has a CATV measurements applications
book written by an HP engineer. He repeatedly cautions the user to always
ensure that the measured signal is placed near the top of the graticule.
The bottom line is simply to check and check again your test equipment
before drawing any rock-solid conclusions about a Device-Under-Test. At the
least, you gain a warm, fuzzy feeling of confidence in what you're using
and seeing; and in the best case, you may note something not-quite-right
before your customer does.
73, Steve K0XP
--
FAQ on WWW: http://www.contesting.com/ampsfaq.html
Submissions: amps@contesting.com
Administrative requests: amps-REQUEST@contesting.com
Problems: owner-amps@contesting.com
Search: http://www.contesting.com/km9p/search.htm
|