From: Brett gazdzinski <[EMAIL PROTECTED]>
> I've had good results with the "three diode" circuit too. Earliest
> reference to the scheme I can find is in QST, October 1956
> using 866 rectifiers.
> Covered again in ER #3, July 1989, this time with solid state
> diodes.
I did not say I had good luck with the circuit, although it looks
like it works on the scope, and on the mod monitor, it results
in a very wide signal. Not much point in running it if you go
50Kc wide...
I tried the 3-diode neg cycle loading circuit, aka "ultramodulation" years
ago. My experience was that I never got a signal report saying the audio
sounded louder or that it was more readable than straight audio, but nearly
all reports said the signal was "fuzzy" or had more distortion.
Think about how the circuit works. What you are really doing is introducing
a certain amount of controlled carrier, and introducing even-harmonic
distortion to the audio. The cortrolled carrier comes from the increase in
average DC plate voltage produced by the rectified audio. The even harmonic
distortion comes from the asymmetrical clipping of the audio waveform, in
which everything below the baseline is attenuated, but everything above
baseline is unattenuated. The abrupt change in slope at the zero point
produces many higher order harmonics. These are relatively low in
amplitude, but if the signal is strong they will be heard.
When you apply modulation, the plate current rises, and the plate glows
brighter, impressive but that doesn't mean the signal is any stronger or has
any more talk power. That additional rf output is almost entirely carrier
power due to the upward carrier shift, and spurious sideband products.
Besides, audio power is too expensive to generate, and then rectify and use
to heat up a large wirewound power resistor.
A method that I found to be much more effective was symmetrical high level
speech clipping. I used a variac on the modulator power supply, and
installed a low pass pi-section audio filter between the modulation
transformer and rf final. I would deliberately drive the modulator tubes to
saturation (way past flat-topping region) and adjusted the modulator plate
voltage with the variac until the negative modulation peaks were just a
little under 100%. To avoid a mushy sound and problems with canting or
tilting of the clipped waveforms, I cut back substantially the low frequency
response below about 200~.
The signal wasn't broadcast quality or even hi-fi, but sounded at least as
good as a stock DX-100 or Viking I or II, but the audio was much louder and
carried a punch that would often rival or exceed that of a SSB signal
running about the same power and signal strength.
But there was one serious flaw. The circuit ate modulation transformers. I
must have gone through half a dozen kilowatt size and larger modulation
transformers in less than six months. I even blew one that was rated at
2000 watts of audio, that I purchased brand new out of box! It was intended
as a replacement modulation transformer for a 4 kw ship-to-shore
radiotelephone transmitter.
I finally figured out what was happening to all the modulation transformers.
The low pass filter, near the knee of the curve at the cutoff frequency,
presents the modulation transformer with a highly reactive load. Modulation
transformers are designed to work into purely resistive loads. One strong
audio frequency component right at the critical frequency and zorch! smoke!
I abandoned the high level clipping and haven't blown a modulation
transformer since.
A better way to achieve the high positive peaks, which produces no audible
distortion, is to take advantage of the natural asymmetry of the human
voice. It requires a good microphone, a low distortion audio chain from mic
preamp to mod xfmr, a flat frequency response at least one octave above and
below the highest and lowest audio frequencies you intend to transmit, and a
modulator capable of substantially higher than normal peak power output.
The ultramodulation scheme requires the high peak power output capability as
well. By properly phasing the polarity of the audio input to the modulator,
most voices will produce from 150% to as much as 200% positive peaks with no
rectifiers in the circuit at all.
Of course, the extreme positive peaks themselves produce little "talk power"
and add little loudness to the signal, even though they look impressive on
the oscilloscope screen. The gain in loudness comes from being able to
modulate closer to 100% in the negative direction before flat-topping in the
positive direction, so that the overall result is a cleaner, louder signal
with less distortion. And you don't need to attenuate the low frequencies
quite so much because phase shift in the rounded unclipped audio waveform
has less detrimental effect.
Don K4KYV
_______________________________________________________________
This message was typed using the DVORAK keyboard layout. Try it - you'll
like it.
http://www.mwbrooks.com/dvorak/
http://gigliwood.com/abcd/
