Tom said in respect of noise testing:
>I'd love to see that paper! I was trying to follow your description
>but I couldn't, so the concept escapes me.
The original paper was in Marconi Instrumentation, sometime in 'the
early 70's. Whether or not there's a library that has them, I don't
know. Marconi Instrumentation was the technical newsletter of Marconi
Instruments.
The basic idea is that if you have a telephone channel and you load it
up with white noise, the IMP's will appear as noise. If you put a notch
in the middle of this band input of noise, IMD will cause products to
appear and 'fill in' the notch when you look at the output. The noise
approximates fairly well to the effects of multiple transmissions,
especially where one sideband has two telephone channels, and the other
has 16 teleprinter channels.
> Is it something us poor Hams can use at home?
Provided you have a spectrum analyser, then I guess it's not too hard to
build something. I use noise loading without the notch, pushing up
towards PEP, to simulate to some extent the normal signal. Obviously, it
doesn't have the components at the syllabic rate in the strength that
the 3 tone test does, but it gives, I believe, a good idea of how wide
the total transmission is. You can cheat with it as well. In any
multi-tone system with equal power per tone, the PEP is n*2 [n squared]
times the power of each tone, where n is the number of tones. With real
signals, the actual PEP required in the amplifier to provide an
acceptable level of spectral regrowth due to IMD does not rise with
signal numbner once you get above a certain number of signals. With 16
signals, it has been shown that the PEP requirement is 8 to 10dB below
that theoretically required [PEP of 256 times the power of each tone] -
this is because with real signals, the randomness is such that all the
signals being in phase is rare, so the number of times the amplifier
actually limits and causes spectral regrowth is similarly rare. In the
case of data transmission using OFDM techniques, you can actually code
the signal so that the 16 combinations producing the worst case are
never transmitted. This doesn't apply with multiple pure tones, of
course. You can therefore, with noise loading, push the PEP such that
there are noise spikes that are limiting, but because of the amplitude
distribution of noise in time, the spectral regrowth so caused is small
enough that it makes things look better than they are.
I hope that's clear.
Nevertheless, I believe that it's a useful test to see how wide the
signal through the system really is - even pushing the linear hard, you
can see by comparison tests that the limit can be the exciter.
You could make a 'poor man's tester' by using noise with a notch (at say
1KHz) to drive the tx, and a matching filter at audio with a monitor
receiver - say a direct conversion job, and measure the relative levels.
But I think an oscilloscope (if you don't have a spectrum analyser) is
easier - Pappenfus et. al even give the equations for working out the
IMD products from the transfer characteristic. Or a direct conversion
receiver (on 80 metres, you can dig out that old BC221 or LM14!) and a
suitable FFT program for the PC should give you spec. an. capabilities
of a sort
73
G3RZP
--
FAQ on WWW: http://www.contesting.com/ampfaq.html
Submissions: amps@contesting.com
Administrative requests: amps-REQUEST@contesting.com
Problems: owner-amps@contesting.com
Search: http://www.contesting.com/km9p/search.htm
|