Zyg Skrobanski wrote:
>Hi Ian,
>
>>That's the key question, because you have been talking all the time
>>about 3rd order IMD, which is close to and underneath the speech signal
>>itself.
>
>Does a transmitter with poor 3rd order IMD therefore have any noticeable
>distortion in the voice quality due to the speech intermod products within
>the audio bandwidth?
>
Often it doesn't, because the only IMD you'd hear inside the normal
receiver passband would be coming from the low-frequency components of
the speech signal, and would be covered by the higher-frequency
components.
That's why a "hi-fi" report on your speech quality is absolutely useless
as an indicator of IMD.
You have to tune away from the signal in order to see what's really
happening. There is a step-by-step guide on my web site (URL below)
under "No More Splatter!"
>>A 3rd-order IMD figure on its own is no guide to what the amplifier
>>sounds like further out.
>
>If the system/amplifier is IMD'ing in the audio frequencies, then
>intuitively, the problem causing that IMD is probably also causing splatter
>products outside the audio bandwidth too.
>
Probably, but the audio frequencies are the worst possible place to look
for the IMD.
Putting it brutally, the rest of us don't care much about the quality of
your audio - but we care a *lot* about how much more of the band you're
taking up!
>>All these complexities mean that when someone talks about "IMD" without
>>going into very specific detail, it's likely to be incomplete or
>>meaningless. We all do it, but we also need to be aware of it.
>
>That is the point I was trying to get to, what really constitutes a
>meaningful lab test for splatter?
>
GW4FRX made some very useful tests using a borrowed analyser with a
digital peak hold facility. By continuing to sweep across the whole
signal for several minutes, holding the highest signal level found at
each frequency step, the analyser builds up a good statistical picture
of the overall bandwidth that the signal occupies.
The composite spectrum is composed of IMD products from all speech
frequencies, and it generally looks like a ragged triangle with the
higher-order IPs disappearing below the noise. The longer you sweep, the
better the statistics, and the smoother-edged the triangle becomes.
It would not be difficult to codify this into a lab test. We now have
the means to record a standard voice signal, copy it around the world,
and reproduce it with digital perfection in any lab (and it can also
include different types of voices). We then have to specify the analyser
IF and video bandwidths, the sweep span and rate, and - most important -
how long the test must run, in order to build up an accurate statistical
picture.
Not many hams have a spectrum analyser with digital peak hold, but the
same could be done using the old HP-141T mainframe with the storage CRT.
The other difficulty is to establish benchmarks for judging the
"quality" of this composite spectrum.
I just saw Tom's message saying that the FCC has adopted this method.
Can we have more details, please (ideally a URL for the test protocol)?
Then we can start to apply some leverage to ARRL labs and RSGB.
>>Ironically, the most meaningful test of IMD performance for SSB is also
>>the simplest - just tune away from the signal, using a good receiver
>>that isn't being overloaded, and notice how far away you have to go
>>before the splatter dies away into the noise.
>
>Using the (meaningful) receiver method, I have heard terrible, very wide,
>signals transmitted from "high end" transceivers, and excellent, no
>splatter, signals from the so called "cheap" transceivers!
That's the biggest problem of all - IMD depends so much upon the
operator as well as the rig.
--
73 from Ian G3SEK 'In Practice' columnist for RadCom (RSGB)
Editor, 'The VHF/UHF DX Book'
http://www.ifwtech.co.uk/g3sek
|