Tweaking an impedance matching network for different class of PA operation

[vu2tve]

Hi,

I'm new to RF PA design, but am trying to figure out a design for a class-A MRF101AN Amplifier. I contacted NXP and am quoting below, their reply when I asked about tweaking their existing reference design board (designed for class-AB) to operate in class A.

Total power consumption of a class-A amplifier does not depend on the signal level. This amplifier dissipates maximum heat power when RF signal is off (RF power is zero). The heat power equals to RF power at maximum signal (50% efficiency).

Suppose a heat sink can dissipate 100W. Use thermal resistance value from MRF101AN datasheet. Junction-to-case temperature rise at 100W power dissipation will be dT=100*1.1['C/W]=110'C. To keep maximum junction temperature, the heat sink temperature must be less than Tj-dT=175['C]-110['C]=65'C. It looks reasonable.

Note! In class-A mode, the maximum RF power is 1/2 of maximum heat power dissipation -> maximum output power must be less than 50W to avoid overheat.

Example. If the heat sink temperature is 45'C then maximum dissipated power is (175-45)/1.1=118W -> maximum Rf power 118/2 = 59W.

NXP MRF101AN 88-108MHz RF board was designed for class-AB operation. It can't be used in class-A mode. The circuit need to be redesigned for operation in class-A mode.

The reason is that class-A mode requires different load impedance (larger value). If you simply increase quiescent current of the original circuit, the maximum power of the amplifier will exceed the limit.

The load line resistance is given approximately by: RL = VCC^2/(2*Pout)

In Class-AB mode, RL = 50^2/(2*100) = 12.5 Ohm
In Class-A mode, RL = 50^2/(2*50) = 25 Ohm

Actual load impedance value includes some reactance related to parasitic capacitance and inductance of the transistor. The estimated RL value for Class-AB mode is close to the load impedance values provided in the device datasheet.

So, from this, am I to understand that with a tweaked output impedance matching network, the board could work in class-A? I experimented a little in SimSmith to tweaked the matching network to match essentially double the impedance than what is provided in the data sheet (went to 26.24 + j10.42 FROM 13.12 + j5.21)

I don't have access to a proprietary tool like ADS to accurately model the device characteristics. Is there anything grossly wrong with this approach?

My next step is going to be trying this new matching network and setting Idq of 2A.

[rfclown]

I assume that you are wanting Class A based on the assumption that it will be more linear than Class AB. You will find that for an RF power amplifier Class A doesn't buy you anything except heat. The source and load impedances which the manufacture provides are usually based on load pull measurements. They will show you where the part works best. Years ago when I was designing RF amplifiers (mostly 800-900 MHz) and adjusting source and load tuners and gate (or base) biases to find optimal efficiency and linearity, I was surprised to find that Class AB gave as good a linearity as Class A. I assumed Class A would be way more linear, and I was wrong. But it does run way hotter.

(edit)... I see that you have two threads on this design question.
(edit)... Three threads... So you want to put muliple carriers in your PA. You'll need linearization.

[vu2tve]

I tested the a modified MRF101AN_START board to work in linear region with 4 carriers, and am seeing pretty good linearity. I don't see 2nd or 3rd order harmonics in the produced signal. That said, I havent pushed the amp to its limits.

Yeah, linearity is important, and because in multicarrier signals PAPR (Peak-to-Avg-Power-Ratio) is high, maintaining linearity in low signal amplitude range would be especially important, I'm guessing.

You're right that a class-AB push-pull amplifier might be configured just right to offer good linearity, and am willing to consider as the next iteration in the design process. People apply Digital Pre-Distortion techniques to improve signal linearity in class-AB, but for an initial stab, a heat generating class-A seems like an easier bet with my current knowledge (although I could be wrong when it comes to thermal management).

I don't have any experience designing a class-AB push pull at VHF frequencies, using transformers and transmission lines for matching, but am willing to give it a go. Could you point to any good resources or reference designs?

Thank you!

[rfclown]

Helge Granberg wrote great application notes for Motorola on PA design. I have a book by him, "Radio Frequency Transistors, Principles and Practical Applications." Jerry Sevick has written good stuff on transmission line transformers. I have three of his publications: Transmission Line Transformers; Transmission Line Transformers Handbook; and Understanding, Building, and Using Baluns and Ununs.

It would also be educational to buy something and see what they did. I just Googled "88-108 Pallet Amplifiers" and saw this:
https://indo-rf.com/product/rf-power-amplifier-fm-88-108-mhz-pallet-only/
I'm NOT recommending this company (have no idea who they are), it's just an example. The point is that in my optinion $78 isn't much to spend to learn something. You could see how it works as is, then crank up the quiensient current and see if any linearity improvement. I did the same when I was new to HF (1.5 - 30 MHz). I had worked VHF-UHF for 20+ years, then found myself at an HF company. I bought a few PAs on my own dime as part of my education. Wrapping coaxes around ferrites was not how I had done matching circuits for 900 MHz amplifiers.

[vu2tve]

Thanks for the many links and pointers. I'll follow up..

I also followed up on some NXP ref designs for push-pull amps, and even within their own documentation, the ref. design for 88-108MHz generates significant enough harmonics (-20dB 3rd order harmonic)

Case in point: https://www.nxp.com/products/radio-frequency/rf-industrial-scientific-and-medical/1-600-mhz-broadcast-and-ism/1-8-600-mhz-150-w-cw-50-v-wideband-rf-power-ldmos-transistors:MRFE6VP5150N

I do think I would learn more and try a class-AB design as a 2nd iteration because it brings so much improvement in efficiency.