[Amps] RE: 4CX250B Screen Supplies

[Ian White, G3SEK]

Will Matney wrote:
>
>"I'd like to make a modest proposal here: how about not feeding a 
>"dirty" supply to *either* of the grids?".......... Ian G3SEK
>
>Ian, that's exactly the point I'm trying to get across.

Agreed, we're mostly dickering about different ways of saying the same 
thing.

>The only other thing I mentioned was that voice modulation creates more 
>IMD than does the two tone test, and by what I've read, we are all in 
>agreement it does.
>

Again, agreed.

But taking up Peter's point, even a tone spacing as narrow as 30Hz don't 
come anywhere close to a realistic test for an amplifier that is 
supposed to be used with human voice modulation.

The nasty feature of real voice signals - ordinary spoken words - is 
that they also contain a *very* strong pulsing VLF component at a 
syllable/word rate of only a few Hz.

This VLF component drags down the power supply, and makes the amp very 
vulnerable to the large 300-3000Hz components that are often happening 
at the same time. No two-tone test can simulate this, and even 
three-tone or pulsed two-tone tests are only tinkering with the problem.

If you want to measure the SSB IMD performance, then the  *only* truly 
valid test signal is... the human voice.

The question is, how to develop a  reproducible voice test?

Well, a standard voice signal is very easy to generate, in these days of 
digital audio (I'll come back to the question of *whose* standard 
voice).

For a transceiver review, the reviewer would play this file into the MIC 
input, but otherwise would strictly follow the manufacturer's 
instructions about MIC/PROC settings, ALC readings etc. Then blast away 
into the analyser and see what you get.

The proposed test is based on ideas from GW4FRX, but it has been 
floating around for a long time.

The test uses a spectrum analyser in peak-hold mode. You play the 
standard voice signal continuously in a loop, set the scan width to 
maybe +/-20kHz, and let the peak IMD spectrum build up. Peak-hold mode 
is recording the *worst* that the transceiver produces, on any frequency 
within the scan width.

The analyser is only looking at one spot frequency at any given instant, 
but if you keep looping the audio file, and continue to scan, it will 
eventually catch everything. What you see is a spectrum that starts out 
looking very peaky, but gradually the gaps fill in to create a roughly 
triangular heap of IMD. When nothing new is being added to the spectrum, 
the test is complete.

This is surprisingly easy to do. Unlike a two-tone test, you don't need 
extremely slow scans and very narrow analyser bandwidths. I don't have a 
modern peak-holding analyser, but my old HP-141T storage tube display 
will grab the peaks just as well (just set the persistence to max).

One advantage of this test is that you're not limited to using a single 
person's voice for the standard signal. The same standard .wav file  can 
include several different kinds of voices (for example, male and female, 
speaking in high-pitched languages and low-pitched languages). The only 
difference with a longer file is that you have to leave the file looping 
for longer. With digital audio and the Internet, exactly the same 
standard test file can be available all around the world.

Testing an amplifier is obviously more complicated than testing a 
stand-alone transceiver, because you have to account for IMD in the 
input signal. There seem to be several possible ways of doing this:
* Build a special driver with IMD so low that it can be completely 
ignored (a very tall order);
* Use a "good-enough" driver and display its input IMD spectrum 
underneath the amplifier output spectrum;
* Subtract the last two traces digitally, and display only the 
amplifier's added IMD;
* Subtract the last two signals by analog methods as suggested by Marv.

A formal test protocol would require some additional details, like 
specifying standard scan speeds and filter/video bandwidths, but that's 
the basic idea.

A peak IMD spectrum is a tough test, because it records the very worst 
that the transceiver or amplifier can do. On the other hand, it's also a 
*fair* test: that steaming heap of accumulated IMD is a pretty accurate 
picture of its effect on other band users!

The disadvantage of this test is that you don't get a single easy number 
out the end - but you do get a very good picture of the way the IMD 
spreads out across the band, which is a key feature that's missing from 
typical narrow 2-tone tests. Each test casts its own different light on 
transceiver/amplifier IMD performance, so I think they should always be 
used and published together.



-- 
73 from Ian G3SEK         'In Practice' columnist for RadCom (RSGB)
http://www.ifwtech.co.uk/g3sek