Read other posts in the “Elephant in the Test Room” series here.
Recap On this Thread
The room: MIL-STD RF immunity testing – 1-18GHz 200v/m
The elephant: everyone in the room is aware that a significant fraction of the ‘calibrated’ 200v/m test field is actually created at the wrong test frequency.
The culprit: notoriously high ‘start of band’ harmonics produced by all octave band TWT amplifiers.
The consequence: equipments purportedly demonstrated to be resilient to 200v/m have in many instances been exposed to only around half this field strength, the rest of the test field created by an unintended frequency.
Welcome back readers, I hope 2014 is turning out well for you all.
I have spent more time on Elephant #1 than I intended, but I really want to plant the seed of the linearization of EMC amplifiers before leaving it. However, I will start introducing Elephant #2 “Lack of Harmony in Harmonic Limits” at the next posting regardless of where we are with Elephant #1. Also, I am a great believer in EMC engineers/technicians being conversant with the most common threat out there – cell phone signals, so in parallel with prodding elephants, I shall do a 101 on 3G and 4G wireless telecom signals (again, starting in the next post).
So where were we last year? – ah yes, Elephant #1, we had a fix for the system harmonics up to 7.5GHz, and were checking the power margin available if we were to place a filter to remove the start of band harmonics emitted by the 7.5-18GHz amplifier.
A back of the envelope calculation showed there was insufficient power if we go with the guaranteed minimum power available. If 250W is truly all that is available at the start of the band, then something fundamental needs to change (higher power amplifier $$$, or higher gain antenna $, or….).
However there is one saving grace. As opposed to the situation where you are buying an amplifier, and do not yet know its performance (risky to assume the available power due to manufacturing spread in the amplifier manufacturing process) you already own the amplifier and can establish the available power yourself.
The picture shows one way of doing this. Note we want to establish the power in the fundamental peak only, the intention is to remove the harmonic to the right will by filtering.
So if we are lucky enough to confirm we have sufficient power overhead, how much will the filter (diplexer) cost? As ever, it depends. If the design already exists I would hazard a guess at $2k-$3k. If a design does not exist, then non-recurring engineering charges will apply, and this could add $5-$7k, making a price of $7k-$10k. So it is worth doing a thorough search of all vendors to see if they already have a product, or at least one close enough that only needs tweaking.
The Linearization of EMC Amplifiers
Linearization is common enough in wireless telecoms, where the push for the technology came from economic necessity. Strangely it has not touched EMC amplifiers, perhaps because the number of amplifiers is relatively small and the market is too small to chase.
In my view, EMC amplifiers are ripe for the technology since compared to the very complex wireless telecom modulation schemes:
1. The signal to be linearized occurs at one spot frequency at a time
2. The EMC modulation schemes are the simplest there are, AND
3. The PC running the EMC software is idle for 99.9% of the time. This means a very powerful tool is available for any linearization scheme we choose to deploy
There are several linearization schemes, some of which I will mention here. Note: I use passive to mean the distortion element behaves the same irrespective of the amplifier output, and active to mean feedback is involved in the pre-distortion.
Passive Pre-Distortion Concept
A non-linear element displaying gain-expansion where the amplifier has gain-compression distorts the input signal such that the resultant output is a straight line.
With active pre-distortion the input signal is compared to the output signal and any difference is used to control an attenuator. This is actually only feasible with an AM modulated waveform, where the AM envelopes are compared and the attenuator is used to ensure the output envelope follows the input envelope.
This is best explained by following the signal through the concept diagram. The input signal in this case is a two tone signal.
The input signal  splits at the input, some going to the input of the amplifier, the rest going to a phase inverter. The amplifier amplifies the input signal as intended  but also introduces unwanted signals (red arrows). The sampled wanted and unwanted signal s  and the inverted input signal  are summed leaving only the distortion products . These are inverted , amplified  and injected into the output. This causes cancellation of the unwanted output signals .
We will tackle the linearization of TWT amplifiers first, and at the next posting I will be proposing a form of active pre-distortion using the PC as the ‘brains’ of the error amplifier. But a different form of linearization could also be used to great advantage in lower frequency solid-state amplifiers since this allows ‘forbidden’ compressed power to be accessed. We will cover this in a later post.
To be continued……
In response to the comment on the previous post, “The Final Piece in the Conventional Approach,” that stated the calculation for required power, P = d^2 x E^2 / 30G watts, is not correct, Tom says,
“It is true that at a plane the power density is E^2/Zo (equivalent to power density = P/4Pixr^2 for a point source of power emanating spherically at a distance r from the point). But we are looking for the power required from an amplifier to create a certain field level at a certain distance using an antenna of certain gain G. I will elaborate on this in the next post. By the way, does anyone know where the magic number 30 comes from in the equation E = sqrt (30xPxG)/d^2 ? Years ago I looked for and couldn’t find the explanation in any textbook, and always thought it was a practical ‘fudge’ factor to make the v/m level come out right (or nearly right). Maybe the explanation is out there somewhere but I could not find it. Eventually, in the absence of any reference, I worked it out myself. Anyone know where it comes from before I tell you at the next post? You have a week.”
The calculation for required power, P = d^2 x E^2 / 30G watts, is not correct. The power of 200 V/m is [(V/m)^2/(free space impedance)], 200^2/120/pi = 106 W. Power density in MKS units is normalized to power per unit area, watts per meter^2. This only works if the amplifier really produces its power at free space impedance; the point of your discussion is that amplifiers don't always produce the power that the manufacturers specifications tell a person that they do at the load of interest, whatever that might be. Thank you for the article. Manufacturers are quite aware of power rating competition and will publish what makes their product look good, so the upshot of your article is to test the amplifier for power output for the load of operation.
Hi Alan, thanks for the comment. Feedback is always welcome. You are right and you are wrong. It is true that at a plane the power density is E^2/Zo (equivalent to power density = P/4Pixr^2 for a point source of power emanating spherically at a distance r from the point). But we are looking for the power required from an amplifier to create a certain field level at a certain distance using an antenna of certain gain G. I will elaborate on this in the next post. By the way, does anyone know where the magic number 30 comes from in the equation E = sqrt (30xPxG)/d^2 ? Years ago I looked for and couldn't find the explanation in any textbook, and always thought it was a practical 'fudge' factor to make the v/m level come out right (or nearly right). Maybe the explanation is out there somewhere but I could not find it. Eventually, in the absence of any reference, I worked it out myself. Anyone know where it comes from before I tell you at the next post? You have a week.
Thanks again Alan