Hi Tom,
Thanks for your comments. After considerable rumination, I think
I have reconciled your arguments and mine along with some
previosuly fuzzy thinking on my part.
Mixers are always non-linear devices, if they aren't they won't
"mix". Why doesn't the mixer in your receiver create close-in IMD
or cross modulate? Many receivers use broadband block
convertors, some systems even use single diode switches for
mixers. Where's the odd-order IMD problem when processing
multiple signal frequencies?
Yes, I would agree that mixers are non-linear devices - no question
about it. I would also argue that they do produce close-in IMD and
cross modulation. They do this to the degree that the input-to-output
transfer characteristics are non-linear. If I take a mini-circuits SBL-1
and drive the LO port with +7 dBm, apply a moderate signal,
say -30dBm to the RF port, and then look at the IF port with a spectrum
analyzer, I will see an IF signal at around -36 dBm. If I increase the
the input signal to the mixer by 1dB, I would expect to see a 1dB
increase in IF output power. The input/output relationship is reasonably
linear, hence the IMD performance is good. If instead of increasing the
input signal to the RF port by 1dB, I increase the LO drive by 1dB, then
I will see little or no change in the level of the IF output power.
The LO port of the mixer is very non-linear with respect to the IF port.
If I put a multi-tone signal into the LO port, I will have lots of IMD at the
IF port.
Vacuum tubes or transistors in class AB are non-linear when
viewed at a fractional cycle rate, why do they often provide
acceptable IMD performance?
This is a good question and admittedly is where my theoretical insight
starts to falter. I guess this example illustrates the your argument about
fractional versus envelope distortion rather well when I think about it.
A fractionally conducting amplifier can still produce more RF output
power when the drive level is increased implying a somewhat linear
input to output relationship on an envelope basis even though a good
fraction of each RF cycle is being chopped. From a mathematical
point of view, the chopping (fractional distortion) is a even order effect
which would be consistent with minimum IMD impact, since IMD is
an odd order effect. Second order cross modulation would occur,
but wouldn't be relevant since it occurs far from the passband (F2-F1,
F2+F1).
Okay, I have reconciled my uncomfortableness with the language of
your argument. I think we are on the same page.
It's incorrect to think fractional cycle distortion causes performance
shortfalls in the application being discussed.
Yes, I am sure that your are quite correct that IMD produced by a small
perturbation of a sine wave would seem insignificant when set against
the fractional conuduction angle of an RF amplifier. My only argument is
that some level of finite IMD would accompany the fractional RF cycle
distortion because distortion implies non-linearity which in turn implies
some loss of linear independence. Okay, but I see where my thinking
is incorrect. I was trying to argue that any non-linearity was a suffcient
condition for IMD. This is not true, if the non-linearity is purely even
order, then IMD will not occur, only even order cross modulation which
isn't relevant in narrow band RF applications (this is why CATV amps
have to be ultra-linear).
While the part about the core causing distortion at 14 MHz and not
10 MHz is fantasy enough, the most dazzling part of the sideways
thinking is the implied assumption a ferrite core carrying only a tiny
fraction of the applied RF current driving a tube that conducts just
over 1/2 a cycle will somehow be meaningful in distortion
performance of that PA.
I can't really comment on the distortion that Rich measured (no pun
intended). I wasn't there and I don't recall Rich going into much detail
about the test setup or the exact nature of the waveform distortion. My
point was this - if I really do have a core material that is subject to
saturation, and I drive it hard enough with a sine wave, I would expect
to see some harmonic distortion develope in the circuit. If the saturation
is symmetric (flat topping of positive and negative peaks of the RF
waveform), you will get odd harmonics. If there is some asymmetry in the
resultant waveform (perhaps due to DC bias), then you will also get even
harmonics. Now, if I take this same core material that is distortiing my sine
wave and drive it to the same peak RF swing with two sine waves of different
frequencies, I would expect to see some IMD (e.g. 2*F1-F2, 2F2-F1, etc)
or cross mod (F2-F1, F2+F1) products develop. In other words, the minute
some non-linearity is introduced into the circuit, IMD or cross modulation
will appear since the two sine waves are no longer linearly independent.
I would argue that while even order fractional cycle distortion won't produce
IMD, odd order fractional cycle distortion (symmetric crossover
distortion, slew rate limiting, etc) would cause some IMD.
I knew a fellow who wanted to run push-pull tubes in a 75 meter
linear amp to improve the audio. The junk-science about fractional
cycle distortion causing IMD problems fits the same category of
"clueless conclusions".
I must admit I am a little clueless here too. I am probably overly accustomed
to thinking of amplifiers as block boxes with mathematical (albeit somewhat
non-linear) transfer functions (P1dB, 3IP, etc) as opposed to considering the
subtleties of their inner workings. This is what happens when you do to much
glorified system engineering and not enough real circuit design, I guess.
73 de Mike, W4EF................
73, Tom W8JI
w8ji@contesting.com
--
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
--
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
|