Date: Sat, 24 Dec 2011 12:21:42 -0500
From: jeff millar <wa1hco@wa1hco.net>
Subject: Re: [Amps] Fwd: Linear Amplifier Tuning---PROPERLY!
Here's more about IMD than you may want to know...
I used to work with very low IMD amplifiers for the cellular industry and spent
a lot of
time studying IMD sources and means of control. I learned the most about the
causes
of IMD from a PhD thesis, and later book, by Joel Vuolevi...
http://herkules.oulu.fi/isbn9514265149/
Most literature describes Intermodulation distortion as coming from harmonics
of
the
fundamental mixing with the main signal. 3rd order IMD means
2 x fundamental +/- 1 x fundamental. (2+1 = 3rd order)
But, the total contribution to the IMD also comes from 3 F +/- 2F, 4F +/- 3F,
etc up to
the 9th harmonic and beyond. All those products stack up on the same frequency
at the output.
## The spacing between tones on a simple 2 x tone will also be the spacing
between
each tone and all IM products. IE: tones are 1000+ 1500 hz. IMD-3 will then
be 500 hz
above the high tone..and 500 hz below the low tone. IMD-5 will be 500 hz
above/below
the IMD-3 products etc.
Each of these products has a phase and the resulting IMD depends on how all the
phases line up. This can make the IMD level very sensitive to the test signal,
power level,
temperature, bias, etc. It also can make the high and low side IMD have
different levels.
## The 2 -tone test is flawed. It's a static test, and all the power
supplies, like bias,
B+, screen V, etc, are all in a steady state. Start talking instead, on SSB,
and now you can
toss in lousy B+ regulation, and ditto with screen + bias regulation. that
just screwed up your
load line right there. Then the big kicker is ALC. Notice how commercial
marine/aircraft SSB
radio's don't even have a mic gain control ! The ALC meter's in most
yaesu's, per the service manual,
are supposed to be aprx 10 db at full scale ALC. I have 4 x yaesus's and
they vary from 5-10db. None
of em appear to be calibrated. Trying to hook up ALC between the xcvr and the
amplifier is another mess.
That's like trying to close the barn door after the horse has bolted. Then you
have yet anothert problem. The
time constants in the ALC in the linear amp will not be the same as the ALC
time constants in the xcvr. A lot
of the older amps don't even have adjustable ALC.
The FCC recognizes this effect and mandates that RF transmitter tests uses a
random modulation to measure IMD, so the testers don't tweak the test to find a
sweet spot in IMD performance where some of the components cancel out.
## agreed. 2 x tone testing is no longer used just for that reason. You could
always tweak the spacing between the 2 x tones...and find a spacing that will
minimize IMD. IF you used say 100 or 200 hz tone spacing, like 1400+ 1600 hz,
then
most of your IMD-2-5-7 will then be IN band IMD. If on the other hand you
used really
wide tone spacing, like say 100 + 3100 hz, then ur IMD-3-5-7 will all be out
of band.
## Rob Sherwood tried another approach..and used white noise into the xcvr's
mic
jack. Then look at the resulting IMD. He then measures the total imd, end to
end BW..at the
-45 db points. The typ Icom is 10 khz wide at the -45 db points when tested
that way. The
new yaesu 5000, in Class A mode is only 3.5 khz wide, also fed with white
noise. Although
the white noise doesn't dynamicly test power supplies, it's an easy test to do,
and results
are always repeatable between same models of xcvr's. Most Xcvr's are 7-8
khz wide,
at the same -45 db points, when fed with white noise. As can be seen, the
yaesu's Class A,
are aprx 1/3 the BW of your typ Icom xcvr.
The level of harmonics and their phases in the output matching network affects
the total resulting IMD. But it's nearly impossible to design for or control
all those effects. Every matching and bias network presents a different
impedance at all
those harmonics.
Most UHF power transistors designed for very low IMD (-50 dB) have an output
match
specification that is _NOT_ the same at the optimum power transfer match.
### Moot point anyway on SSB. This isn't FM. With SSB, your power output
will
vary from 0-max pep output..and everything in between. Your resulting load
Z will
then vary from optimum at full pep output...... to no where near optimum..at
low
levels. You can see this on a scope, just looking at a xmas tree pattern.
So, the optimum IMD setting of the tune capacitor is _probably_ not peak power
output but there's not enough information available to set it...and it would
drift
around anyway.
## whoa. This is not a SS amp, with fixed xfmr ratio's that never seems to be
optimized
for anything. Again , with SSB, the po will vary from 0- 1500w pep. The
best we can do is
to tweak the tune for max out, then slightly over couple the load cap. Plan B
is to adjust the
linear's tune+ load control for max output at say 1700w pep out [apply more
drive to do this],
then back off the drive, so the amp is normally driven to just 1.5 kw pep out.
If it's a real linear amp,
it should be linear from zero to full power output. It would be easy to test
your concept out though,
just look at the SA, while cranking the tune control 1st to one side,CW....
then to the other side,CCW.
The biggest contributor to IMD is the amplifier going into compression on
peaks. This
produces a nonlinear input vs output response and those non-linearities produce
the
harmonic mixing effects that create IMD.
## whoa. Nobody is operating a tube type linear amp anywhere near it's
compression point.
The shape of the compression curve affects the level of IMD. A curve with a
sharp break
has more IMD than a more gradual curve. As a guess, a more heavily load
amplifier
will have a softer compression curve and produce less IMD when driven into
compression.
A heavily loaded amplifier limits on current and a lightly loaded amp limits on
voltage swing.
Intuitively, voltage clipping seems "sharper" than current limiting.
The cellular industry implements predistortion of the input signal to
compensate for
amplifier compression. The amplifier compression curve bends down, so the
predistortion
circuit bends the the gain curve up to compensate.
### The swiss ADAT ham xcvr uses pre-distortion.
The future of amplifiers for _should_ include an IMD monitoring circuit and
predistortion.
This would produce significantly more power with a smaller and more efficient
amplifier.
## it wouldn't be any more eff.... just less imd. But if it's cheap to
implement,
then go for it.
IMD levels should be set to about -40 to -50 dB. This would have to apply to
both
transceivers and amplifiers, because much of the IMD these days comes from the
crummy amplifiers in the transceiver.
Crowded bands would seem completely different if everybody's IMD dropped 10-20
dB.
## agreed, but you don't have to resort to exotic techniques like
pre-distortion to get
good imd these days. this is all old news. Ham gear, TX wise is pure junk,
plane and simple.
You use a 300w PA, say like the old MRF-422 used in the yaesu 767GX.. [ and
also used in the yaesu 1000-D]
The 767GX has -40db pep IMD3. It runs at 100w pep. the 1000-d uses the same
final devices... but at 200w.
Take any 1000-D..and reduce the power to just 100w pep out..and eff drops to
36%. Now throw in your
pre-distortion techniques...and you could probably get the IMD3 from -40db pep
down to -55db pep on a
properly designed radio. Then no Class A required.
## if u want low ind + good eff, you have to back the power way down to almost
a 1/3 of the rated max output.
You also need max device voltage applied for best imd. Then u have to design
the output xfmr around the 100w
power level... for max eff.
## Like other's have said many times before, ham gear on TX should meet the
ITU specs for TX imd, then we would not be
having this discussion. All the marine and Aircraft meet ITU specs for TX imd.
And none of em use pre-distortion techniques nor
Class A blast furnace tricks. 100 w pep out, in Class A would never work on
something like a sailboat anyway, you would kill
the battery in no time at all.
Later... Jim VE7RF
jeff, wa1hco
_______________________________________________
Amps mailing list
Amps@contesting.com
http://lists.contesting.com/mailman/listinfo/amps
|