[Top] [All Lists]

Re: [Amps] SPE 1.3K-FA and FTDX-5KMP / IMD at rated PEP

Subject: Re: [Amps] SPE 1.3K-FA and FTDX-5KMP / IMD at rated PEP
From: Manfred Mornhinweg <>
Date: Wed, 18 Nov 2015 15:15:16 +0000
List-post: <">>
Joe, and all,

Even linearity correction will not work if the devices are being driven
into saturation - the TV transmitter designers know that very well.
Linearity correction works as long as there is some headroom left in
the Pin/Pout curve but once you've lost the ability to increase drive
to compensate for the compression you've reached the limit of

Of course you are right in this. The highest power an RF amp can deliver is when the FETs are driven 50% of the time fully on, 50% fully off. At that point, no change in bias can extract additional power, nor additional gain. Also, with perfect components, at that point we have 100% efficiency. With real world components, maybe 95%, if the circuit is well optimized.

But my point isn't that. Getting into full saturation puts is too far into the range of the gain curve where it becomes too low. I will explain more in detail what I mean to do, with linearization through bias modulation:

Plain simple linear amplifiers work in class AB, with a fixed bias. Without negative feedback, their gain is a little irregular in the very low power area, around the cross-over zone, and this is strongly affected by the amount of idling current. Then the gain rises a little, reaches a peak at some power level which might be roughly in the middle or two thirds of full unclipped amplitude. From there on it again falls slightly, and as we start driving the amplifier into clipping, it falls more strongly. At full saturation, the gain curve falls so strongly that additional drive results in no additional output.

When we add negative feedback to this amplifier, the overall gain becomes smaller, but also more even, in roughly the same proportion.

Common class AB amplifiers use negative feedback, maybe about 10dB of it, and are set up in such a way that they produce full nominal output at about 1dB gain compression, that is, pretty much where noticeable flat-topping begins. This results in the typical IMD specs of -28dB or so for the 3rd IMD.

Okay. All this is plain common knowledge.

Also it's reasonably common knowledge that the gain of a FET amplifier operating with sine wave drive can be controlled by varying the bias. At any given drive level, changing the bias voltage can vary the drive from its peak value (in class A, with the FET running a DC close to its maximum transconductance value), down to zero, when the bias is set low enough to keep the FET cut off through the whole RF cycle. Never mind about the distortion of the RF waveform that results - the low pass filter fully fixes it.

And let's add as a third point that the open-loop power gain (no negative feedback) of modern LDMOSFETs, used at HF, is very high, ranging from 25 to 40 and more dB, while a ham amplifier designed to put out 1500W from 75W drive requires just 13dB gain. So we have at least 12dB of excess gain to put to good use.

What use could that be? Well, I love the idea of improving the efficiency. Simple class AB amplifiers, used in variable power modes such as SSB, are awfully inefficient. While their efficiency at peak power might vary between 45 and 70%, depending on the implementation, their average efficiency over the whole range of amplitudes is far lower. An amplifier delivering 1500W at 60% efficiency will only have 30% efficiency at 375W, which is roughly the average output during SSB operation. So, let's improve that low efficiency!

What we shall do, then, is this:

- See at what level of compression (flat topping) the amplifier is down to a net gain of 13dB. Set the loading (output transformer turns ratio, and power supply voltage) so that this happens at 1500W output, or slightly above for peace of mind.

- Make a biasing circuit that controls the bias in such a way that the gain of the amplifier is kept precisely at this 13dB level, at all output levels lower than 1500W. This means that over most of the range the FETs will work in class C, with slightly varying conduction angles, and at very low power they will again work in class B and then AB.

Since the gain varies with frequency, temperature, individual devices, etc, the way to implement such a bias is through feedback. One envelope detector at the input, one at the output, and an error amplifier that controls the bias so that the output remains precisely proportional to the input. This is quite simple to do, and only uses low cost parts.

The improvement in efficiency comes from two things: One is that at full output the amplifier runs moderately saturated, allowing at least 80% efficiency. The other is that at most power levels it runs in class C, at zero idling current, and significantly higher efficiency than it would in class AB.

Seen in a voltage and current scenario: The bias-modulated amplifier has a slightly higher load impedance on the FETs (lower transformer turns ratio), so that a larger percentage of the power supply's voltage is fed to the RF output, at all power levels. This gives an efficiency improvement. And then, over most of the power range it operates in class C, meaning that the FETs conduct current only during the part of the RF cycle when the drain voltage is lowest, which further improves efficiency.

The result is that fewer/smaller FETs are required, smaller heatsinks, smaller fans, smaller power supply, lower cost, lower weight, while still producing excellent amplitude linearity, in fact much better than that of a common class AB amplifier.

The problems are two:

- Driving the FETs so deeply into saturation worsens the problem of phase modulation from dynamically varying internal FET capacitances. With high capacitance FETs, this would require predistortion in the radio, which is a show stopper in most ham applications. But with the latest crop of modern, low capacitance UHF LDMOSFETs, the capacitances are so low that the resulting phase modulation might no longer be a serious concern at HF.

- The bias modulation circuit can correct the output amplitude only to the extent that it can correctly detect the amplitude of the driving signal. If the amplifier exhibits a non-linearly varying load impedance to the driver, this would degrade the signal quality. Again, modern LDMOSFETs with their extremely high gain and low capacitances should minimize this problem.

Rounding it up, this scheme would have been quite problematic with the VMOSFETs of the 1980's, like the MRF150, but is much more attractive with modern
LDMOSFETs like the BLF578, thanks to their lower capacitances and higher gain.

If I weren't so lazy, I would built such an amp and try it! As things are, I'm old and lazy, and can only combine theory with experiments I did many years ago, at lower power levels.


Visit my hobby homepage!
Amps mailing list

<Prev in Thread] Current Thread [Next in Thread>