Amps
[Top] [All Lists]

[Amps] Re: IM Distortion

To: <amps@contesting.com>
Subject: [Amps] Re: IM Distortion
From: RFlabnotes at aol.com (RFlabnotes@aol.com)
Date: Wed Mar 5 13:15:18 2003
Ian:
       I totally agree with everything you say, especially the technical 
issues. As I have preached many times, the real problem(s) appear when we 
attempt to reduce complex technical phenomena to one-liners. We may have a 
genuine disagreement here, but it will be predominately semantic, and that we 
can easily fix.

       The immediate problem is that I do not view two (or any multiple) tone 
testing as a means of testing power supplies. Or of testing ANYHTHING other 
than the raw IM of an amplifying device. We are obviously going to have to 
agree on what I mean by that last reference, and to clearly illustrate, let 
me give a clear and common example.

       Suppose I just got a pair of some new, hot power MOSFETS with mfrs. 
house numbers and nothing more than max values of breakdown voltages, and I 
am led to believe by the tooth fairy (that dropped them on my pillow last 
night) that they will get me a KW out on 50MHZ.  So, I go ahead and build up 
a proto amp circuit on a ground plane and make my measurments to see if I 
received Gold or Pyrite from the fairy.

       I realize that not only is it gold, it is platinum. 1200W out at the 
rated VCC and good efficiency. So now on to the final tests, which as you 
know, will include some way of estimating IM. At this point, I am trying to 
establish the most optimistic scenario, knowing fully well that it is 
controlled by factors over which I have NO control (mainly, the inherent 
linearity of those transistors). To do so, I apply a generic test which I can 
always use as a reference. I do my two-tone tests and find them to exceed 
(i.e. satisfy) my requirements by a healthy margin. Incidentally, this is 
precisely the data which any transistor mfr. will give you in his data sheets.

       I realize that the final amplifier will contain a power supply that 
may sag, resulting in transient IM on voice peaks or CW attacks. But at this 
point, I can't worry about that yet. As an engineer, I have found it 
desirable and efficient to work in stages, from simplest to advanced.I 
measure and document the status of my design at every stage, using the most 
basic standards that apply. I try to utilize standards which have as much 
validity as possible, so that if possible they can be accounted for as the 
second and higher order parts of the design start to stack up.

       In short, in the illustrated case, and may similar ones, we 
standardize using the simplest models, hoping that they have enough 
generality to allow intelligent extrapolation when needed. For IM, the 
standard is based on a unit which is essentially the ratio of the amplitude 
of the distortion component of the output to the fundamental component.

       Now, in the present case, I am not disputing the idea of having a 
"larger-scale" standard for IM specs, and this is obviously driving the FCC, 
as well as other factions within the communications industry. Maybe what we 
will end up with is an "effective IM" or "spectrally-weighted IM" spec that 
applies from input terminal to output terminal  of ANY DUT. Meanwhile, there 
are still many places where people are going to restrict what the input and 
output terminals are, and within that range, the IM is well-defined and 
easily measured using muliple CW signals. It is actually the ONLY formal 
definition we have, and unlike you, I think it is a very good predictor of 
performance when the system is expanded. How relavent it is when the system 
gets expanded (e.g. to include power supplies or other devices that can 
result in serious overdriving) is very complex. For example, it could turn 
out to be as simple as this: measure the amp traditionally, specifying some 
subset of the range of IM-order (e.g. 3rd,5th). The add an additional 4db 
(this is a guess, for illustration) for junky power supplies. My intuitive 
"feel" as a design engineer is that IM is predominately controlled by the 
lack of perfect linearity in amplifying devices. In fact, it is somewhat 
amazing and fortuitous that they are as linear as they are, especially 
BipolarTransistors. I think all other factors (in properly operating 
amplifiers) make minor additions to the IM (yes, they can even cause partial 
reductions at certain frequencies). This does NOT include cases where even 
modest overdrive occurs.

       I think I disagree with your prediction about non-correlation between 
results via multi-tone testing, and the other one ( I am not sure I like any 
of the titles we have applied to it so far.) Unless, of course the DUT is 
pathological in its non-amplifying basis (power supply sag, peak clipping, 
eg.) In the case of extreme power supply variation, which is probably more 
prevalent than we want to know, we are just confronting a well-known and 
dreaded monster: TIM (Transient IM). I have been in that one before, and 
don't ever want to go back. There are probably other second-order effects 
arising from non-constant supply voltages, and I don't deny that they exist.

       Here is another interesting question to ponder. Suppose that we agree 
on a system-level definition for IM (that's the title I'll use in this 
paragraph) and write a formal definition, which leads to a properly 
documented measuring procedure. In order to use this (this is where the final 
ultimate reduction of this complex issue to a one-liner is REALLY 
established) we now have to assign a numerical limit - say 50 microzorkons 
per root-hectacycle. So, all of the faithful one-liner believers will very 
carefully monitor their station to produce less than 50 micr.......      of 
splatter. Then one day they discover that the zorkon meter just went to 50.6, 
so they reduce the drive level by just barely touching the drive control.  
Here is the real question: where is that number 50 going to come from? Who is 
going to set it and based on what data?

The most immediate need is what you and I have both agreed upon, and upon 
which we are both speculating oppositely about the likely outcome. How well 
can we hope to get a good correlation between zorkons and multi-tone product  
db's? If we can't, then we will have to switch to zorkons - of course, after 
giving it a respectable name!
Tell GW4(???) that he is not finished yet - we still need him.

73
Eric K8LV
 
<Prev in Thread] Current Thread [Next in Thread>