Parallel transistors in HF PA

Hello, I have found this simple HF PA made out of many small paralleled transistors instead of a power one.
http://lh4.ggpht.com/_IO6H5riFKbs/TX...002%5B5%5D.gif

Have you seen such designs, would they work and what are the pros and cons of paralleling small transistors instead of using a big one?

You did not say the frequency but I think your many transistors will have much more wiring capacitance than one big transistor and the capacitance will ruin the high frequency gain.

I think the main disadvantage is the maximum power dissipation of each transistor. Suppose its 200mW then ten of them would give you 2W ! not very much compared to a proper high power transistor?
Frank

Efficiency, linearity, and reliability of the circuit get worse paralleling transistors, instead using only one transistor chosen to meet the specifications.
To get a compromise in performance there is an empirical limit how many transistors you can place in parallel for an amplifier, and usually this limit is 2 for BJTs and no more than 4 for MOSFETs.

The design posted is for a 7mhz amplifier. I have built this amplifier with pn2222 transistors, but without the LPF. I did not say it did not worked but the waveform had severe distortion without the LPF. I do not know how I can bias this for class-A operation, just to see the outptut power. Also it would be desirable to work on other frequencies to 25MHz or so.

Reading on your posts I see the disadvantages of the paralleling. However there is an advantage I see when class-A is used. The size of the amplifier is much bigger because of the huge heatsink needed when one power transistor is used. If many smaller transistors are used the heat is divided in each transistor so even the plastic pn2222's can do the job. I know class-A is barely used in RF PAs, unless they are used for UHF TV.

To bias the amplifier into class A, put a resistor in series with the choke that is earthing the bases ,after the input capacitor. feed some current into the top end of the choke, say .6- .8mA, which if you get it from your +12 V means a 18 - 13K resistor. This should provide a little forward bias. You could use a 10K and a 10 K pot in series and set the pot to give a combined PA current of say, 100mA. This gives a total dissipation of .1 X 12 = 1.2W or 120 mW per transistor, depending on its Hfe and the emitter resistor.
Frank

I will try this Frank,
Thanks a lot!

Paralleling the transistors is a very common technique in RFIC Power Amplifiers because the transistors are-almost- identical on the silicon.But it's pretty difficult to catch the same matching with discrete transistors due to higher manufacturing tolerances.It may work if it is well matched but if the "co-operation" can not be maintaned, it will be worse than a single equivalent.

I know in push pull transistors must be matched and the emitter tesistors can help if these are not closely matched. But I did not know that the same applies in single ended amplifiers.
Is it true even if the amplifier operates in class-A? I thought it wouldn't matter then.

I read somewhere that a power Mosfet has thousands of tiny ones in parallel inside.
An RF amplifier frequently has its output tuned then its transistor can be efficiently pulsed with class-C but its output still has a sinewave with low harmonics.

The thing about power mosfets you mentioned is very interesting.

No, I refer to class-A broadband amplifiers without output filters to rebuild the sinewave. Is the matching required in that case?

A class-A transistor produces severe distortion (many harmonics) when its output level is fairly high (but not clipping) and its voltage gain is high.
Usually a lot of negative feedback is added to reduce the gain and reduce the distortion.

I never worked with radio transmitters but I remember learning that when an antenna impedance matches the output impedance of the transmitter then the output radiated power is maximum. If the impedances do not match then reflections cause the output radiated power to be less and the heating of the output transistor to be more.

Matching is always necessary.You should do "Power Matching" for high powers because conjugate matching is slightly different than power matching.
How to do that ? Load Pull technique but it's almost impossible for radio amateurs due to its cost.
But you can still find something to do this.If you can find a sliding tuner for that band of interest, you may find the optimum load impdance.Then you should design a matching circuit that matches this impedance to antenna impedance.

I think you refer to impedance matching. I do not refer to impedance matching, but to transistors matching, when they are used in parallel in a class-A RF amplifier (on HF), like the circuit in my first post. This circuit is not a class-A one but when this is biased to class-A using some base resistors as mentioned earlier.



Yes, I have seen that experimentally as well, my measurements show that if class-A is used, it is better to drive the transistor deep into class-A (more hot) and keep the input signal as low as possible and so the output. It is only then when harmonics are kept low.
All this is no LPF is used at the output, i.e in broadband designs.

Transistor matching is another concern.If you're able, find the same lot numbered transistors as close as possible..

parasitic capacitance would kill the gain at high frequencies.

there is paper on the net where transistors put in parallel using wilkinson power dividers and combiners. Although it is narrowband paper pointed out it is more realistic than transistors in parallel.

Google for wilkinson power amplifier or power divider/combiner power amplifier. For example: http://www.microwaves.dsl.pipex.com/24ghz/24gpa.htm

Thank you for the links. Amplifiers using wilkinson power splitters and combiners are a good option and immune to impedance mismatching problems as far as concern vswr.
However I refer to HF frequencies (1-30MHz) where most transistors would do the job.

I have tested the amplifier posted on #1, but biased in class-A, using a pair of base resistors and it worked ok at 10MHz, I have no idea if this would wirk satisfactorily at 25-30MHz though using pn2222

The datasheet for the PN2222 shows that its minimum current gain at 100MHz is only 2.5 which is not much better than a piece of wire. At 25-30MHz it also has poor performance because it is not an RF transistor.