Parallel 2n3904 transistors as VHF PA

Try putting a resistor in the emitter of T2.

This will have the efect of making it more linear as well as dropping the current.
You could well be overdriving the device, and without an emtter resistor there is nothing to restrict the current.

You could start with about 4.7 Ohms and see if it is stable (and cool). Then you could reduce the value until you get optimum performance.

Hope this helps.
BR Harry

Final amplifier design using two 2n3866 transistors. Can anyone help me design a proper matching and filter network for T2?

T2 is significantly cooler when feeding a 75ohm dummy load. It was very hot to touch when feeding a 50ohm dummy load.

https://ibb.co/swfPpGh

This is the new amp circuit with decoupling added to the base biasing supply. The RFC has been replaced with a 680ohm resistor. Gain is 12dB @ 90MHz

https://ibb.co/XkjTx39

Now I need to add one more stage for 1W output.

dare4444 wrote:it was self oscillating when no input was present.

That sounds like improper decoupling!

Proposed new oscillator design based on Franklin topology for another FM TX. It should be more stable.

https://ibb.co/TqyDt1j

Hi Ruud,
I did try biasing the 2n3866 directly from 12V but it was self oscillating when no input was present. Biasing it through 7805 solved this problem.

Yes, I'm measuring the peak voltage and it's a very accurate way of calculating the output power. I modified the amp and now it has 47ohm input impedance and frequency drift is now minimal and TX is back to rock stable. Output power is lower though (150mW). One more 2n3866 should bring it up to 1W level.

I need to work on the output matching and filter network. The series LC matching circuit I'm using is part of a L network. I've eliminated the trimmer cap to the ground as output was a little less so now it looks like a series tuned circuit. Add a trimmer to ground and it's a L matching network.

https://ibb.co/x8yLRxd

I am still having problems with the series L-C output circuit...
Because I'm almost certain that this will result in an impedance mismatch.
On the other hand, I wouldn't start with a very complex filter, without sufficient equipment to measure the behavior of the filter itself.
Just start with a simple Pi-filter to match the output of the transistor to 50 ohms. (And get the highest efficiency!)

By the way: with the diode/capacitor detector circuit to measure the output power, you are in fact measuring the peak voltage.
So, to get the effective power, you should multiply the measured peak voltage with 0.707 and use this value to calculate the effective power in 50 (or 47) ohms.

1n cap across 10ohm emitter resistor removed. Frequency now seems to be stable like before. The 1n cap was giving too much gain. Now there is just a 10ohm resistor in the emitter of 2n3866. It has stabilized the amp. Output power is now 320mW at 89MHz without a low lass filter. Once I get the PCB ready with the Oscillator coil etched on PCB it would make an awesome FMTX project. I'll house the 2n3866 amp in another box separate from the Exciter board to keep the heat of 2n3866 away from the Oscillator stage.

I tried a 3 pole Butterworth filter to match the collector of 2n3866 to 50ohm but output power is 50mW only. I used this tool

https://rf-tools.com/lc-filter/

Cut off is 125-127MHz and I entered 200 to 50ohm. It's not working.. where as with my series L and C I get a good match and efficiency.

How does it work? Does the L cancel out the collector capacitance?

The circuit is drifting as it's laying open. Should I use a 47ohm resistor between the output of FMTX and the base of 2n3866? I tried it and output dropped to 60mW. Output network at T4 was designed for a 50ohm load and the base of 2n3866 is around 3ohm only at 90MHz and 60ma of current. Zin = (800 Ft/ 90MHz) * 26 / current in ma.

I don't want to overload the final 2n3904 stage of FMTX as it will pull the oscillator frequency. So I must use two 2n3866 stages in the amp.. one driver and one PA and connect the output of FMTX via a 47ohm resistor to the amp to keep the oscillator as stable and lightly loaded as possible.

My initial goal was low power and high Stability. Now I want both high power and stability.

Okay I'll try and update you. I've a standing current of 7-8ma with the present design.

No, not as long as the supply voltage is constant.

Wouldn't it make the amp unstable?

I suppose you could omit the 7805 and the 560 Ohms resistor and use a 1500 Ohms resistor from the base to the +12 Volts rail instead.

https://ibb.co/RyVn4BL
https://ibb.co/xm2Lt5V

Tested a 2n3866 class B amp. Output is 390mW at 89MHz into a 47ohm dummy load. It's a perfect amp for my high Stability FM TX.

dare4444 wrote:"If you physically match the diameter against the length. You should be able to get the temperature coefficient to be zero. if the coil is short and fat, the increase in diameter will increase the inductance as the temperature rises, if its long and narrow the inductance will decrease as the coil gets longer."

Hi,
I have never heard that. AFAIK a coil has always a positive temperature coefficient. Capacitors with negative temperature coefficient (NPO) are used for its compensation.

BR from Ivan

Thank you. I've saved the circuit and I'll add it to my FM TX before I publish it in Electronics For You. I have a few TL084. Will check my junk box for TL071.

Here is a much simpler active pre-emphasis filter that uses a single power supply:

R4 and R5 set the output of the OpAmp at 1/2 the supply voltage.
For low frequencies the gain is ~2 x, or 6 dB.  (15K + 470 ohm + 15K) / 15 K.
R3 and C1 have a time constant of about 50 uSec. ( 15 x 3.3 = 49.5 )
R2 restricts the maximum gain for very high frequencies.
For the OpAmp you can use almost any single OpAmp, for example TL071 or the 'good old' UA741.

"If you physically match the diameter against the length. You should be able to get the temperature coefficient to be zero. if the coil is short and fat, the increase in diameter will increase the inductance as the temperature rises, if its long and narrow the inductance will decrease as the coil gets longer."

Is the above statement true? It explains the drift in my FM TX. As the temperature rises the frequency goes up as well. My coil is long 1.5cm in length and only 2mm in width.

Okay I'll incorporate an active pre emphasis filter before I publish my design. I found this circuit online. Does it require a dual power supply of + and - 12V?

https://www.circuitlib.com/index.php/schematics/product/45-active-fm-pre-emphasis/category_pathway-32

Just a note about the pre-emphasis filter you are using:
If you look at the pre-emphasis curve for 50 uS, you see that a frequency of 15 KHz has about 14 dB boost, referred to low frequencies.

https://www.radiomuseum.org/forumdata/users/133/Pre_Empf_Bode_590.png

In your circuit, the maximum HF boost can only be 6 dB:
The 10K/10K form a 6 dB voltage divider. (For low frequencies)
At high frequencies, the 50 nF will (almost) short the 10K series resistor.
So, at high frequencies, the attenuation of the input will be (almost) 0 dB.
This means that the difference between the highest frequency (15 KHz) and low frequencies can only be 6 dB, instead of ~14 dB.

The best way to solve this is with an active pre-emphasis filter, built around an OpAmp.

https://ibb.co/wKqmW9P

New updated schematic. The 2n3904 temp sense has been removed. Output is now taken from collector of the colpitts oscillator instead of emitter and Frequency stability improved further. Total drift in 24 hour period was 7KHz (2Khz upward during daytime at higher room temperature and 5KHz downward during night at lower room temperature).

The board was covered with a plastic bowl but it was not air tight.

Next step is to change the 16V zener with 3 X 5.6V ones.

This design outputs 10mW (88-92MHz).
All transistors are 2n3904. No hard to find RF transistors have been used.

Hi again,
My thoughts are:

1 - You are using the Zener diode reverse-biassed as a varicap, but the voltage to capacitance characteristic is not linear. If you had, say, 6 volts then the Zener/capacitor would be in a shallow part of the slop, which is more linear(ish). But if you are running at 1.2V then the slope gets steeper and less straight-line so it would give rise to distortion. It would not be clipping-type of distortion, but would cause a sinusoidal waveform to be much more compressed on the more positive half cycles.

You may want to try a test tone and watch the recovered audio on an oscilloscope.

2 - "Should this cap be variable?". I believe so as this would be your coarse FM deviation control, if you have not used any audio level limiting/compression before the modulator.

3 - The method you have used for controlling the temperature drift is quite clever, but you may like to try an "amplified diode" so that you still get voltage/temperature correction, but allowing a higher DC voltage.



Best regards from Harry - SM0VPO

Check my 2n3904 temperature compensator.

https://ibb.co/6Hs3WTP

Oscillator is most stable in this arrangement when the voltage to Zener is only 1.1 volts or so. With that low of a voltage, is the zener linear enough for good quality sound? It sounds okay on the receiver.

My TX goes down in frequency when room temperature decreases. So I added a 2n3904 and when room temperature falls then base junction temperature also falls and the conduction voltage increases and it draws less current from the voltage divider so the voltage goes up a bit and vfo frequency rises a little.

This arrangement has reduced drifting to + - 5KHz. I am using cheap ceramic disk capacitors in the oscillator section.

My question is, should I keep this arrangement and supply the Zener with 1.1V for FM modulation purpose but for frequency tuning I replace the 30p capacitor with a variable one? Frequency is around 88MHz with the fixed value. Making it variable should cover the rest of the FM band.

Sounds good? Let me know. I’ll finalise the circuit.

More testing done. Drift is only temperature dependent now. The buffer and amp section is all good. PCB etched coil and NP0 caps should help? How about using 3 x 5.6V zeners in series with 470k balancing resistors between them? These are the only remaining modifications.

Circuit board sitting on a wooden table and covered by a plastic bowl.

11:58pm
91.8980 mhz

12:51am
91.8996 mhz

1:12am
91.9010 mhz

2:23am
91.9008 mhz

2:29am
91.9011 mhz

3:24am
91.9005 mhz

4:10am
91.9012 mhz

4:45am
91.9020 Mhz

Tube light on

5:03am
91.9033 mhz

Tube light off

8:39am
91.9029 mhz

Fan on, doors open. Cold. Time 9:13am
91.8888 mhz

Fan off. Doors closed.
9:22am
91.8964 mhz

9:27am
91.8975 mhz

9:44am
91.9000 mhz

9:53am
91.9014 mhz

10:13am
91.9042 mhz

12:50pm
91.9076 mhz

1:06pm
91.9090 mhz

You said if the FM frequency stays within 15KHz then the receiver would not need retuning. It's either way + or - ? For example for 92.000 MHz the free running LC oscillator should be between 91.985 to 92.015 MHz? Please confirm from your experience. I'm still testing the drift. Your WBFM V6 xtal TX has a 0.17uH pcb etched coil. Incidentally the inductor in my circuit is also 170nH. I'll have the coil dimensions copied and create a PCB. Thanks.

That looks better.
As a matter of interest, I have a thin plastic rod, like a miniature magic wand. At one end I have a ferrite slug from a tuning can. At the other end I have a sawn-off brass screw.
With this "tuning wand" you can apply ferrite or brass to a coil to increase or decrease inductance to check resonance :-)

It is a really good tool to have in the tooolbox.

BR Harry

Output filter plus matching now looks like this. Po = 10mW
https://ibb.co/KrF1pjC

Clean looking sine wave after the filter

https://ibb.co/YQSqMHt

@Ruud

Should I try these values for output matching for 10mW output? It's a Butterworth low pass filter.

https://ibb.co/cJRxvdJ

Below is stability of the TX for 30 minutes after a 10 minute warm up period. It's enclosed in a plastic cover to prevent temperature changes.

3:36am
91.8528 mhz

3:40am
91.8523 mhz

3:47am
91.8528 mhz

3:56am
91.8518 mhz

4:06am
91.8526 mhz

@Ruud
I agree with you Ruud. It looks very strange.

@zsolt
Take a look at my 400mW amplifier and copy the output stage. You should get the correct values without having to draw something for you.
Yes, it is a bad idea to simply parallel 2 transistors. It's like putting two random kids on a see-saw. The fat one will alwars be the lowest.

As Ruud pointed out. Simply putting a small resistor in each individual emitter (1-Ohm?) should be enough to ballance with modern transistors from the same batch. It will also halp to linearise a little more and reduce other rubbish at the output.

Ok, have fun. BR Harry

Yes, there is no filtering at all. I just used an air core coil and trimmer capacitor and I adjusted both for maximum output power into a 47ohm load. Then I took their values using a LCR meter. I lost my USB SDR dongle so I am unable to see the output spectrum but you are right. I calculated the resonant frequency and it was 30MHz something for the output network for the given LC values. It's a crude form of matching but works well.

This circuit is the most stable yet

https://ibb.co/PMmT20p 

The circuit will no doubt work. But I think the biggest improvement could be the matching of the impedance of the output transistor to 50 Ohms.
(Or the impedance you prefer...)
The only thing I see is a series resonant circuit between the collector and the output.
I have the impression that this will not result in an optimal impedance matching.
(The impedance at the collector will probably be far away from 50 ohms.)
Also the value of 165(?) pF in the series resonance output circuit looks at least a bit 'strange' to me, for a frequency of 95 MHz.
Also: a correct impedance matching circuit could work as a band-pass filter, reducing unwanted output signals. (Now there is no filtering at all.)
Maybe Harry can shine his light on this?

I guess it is like riding a tandem bicycle...
It is easy to let the other person do all the work! 

Ruud wrote:In general, it is a bad idea to connect two or more transistors 'hard' in parallel.
Since there are always differences in the Vbe voltage (due to tolerances) and Hfe, there is a chance that one transistor is doing (almost) everything while the other transistor is doing (almost) nothing...
When transistors are used in parallel (as in power supplies for example), you usually see small emitter resistors to equalize the differences.
Even high power HF transistors are made with multiple chips in parallel, with integrated emitter resistors.

  that sounds like  life  at some jobs when one is doing almost everything while the other is doing almost nothing. I wonder what is the real life equivalent of the equaliser resistors then   ? (an angry boss or a big pay check? )

Thank you both Rudd and Harry for your comments. I take notes from Harry and his comments and suggestions are saved in a word file.

I'm in India right now and I do have BF199's in my junk box. In my quest to design a stable FM transmitter I was experimenting with various designs.

This circuit is the most stable yet

https://ibb.co/PMmT20p

It uses four 2n3904 transistors and output power is +10dBm. In my previous experiments I added another 2n3904 in parallel in the final PA stage with an emitter resistor of 47ohm. Power output doubled but I didn't have access to a spectrum analyzer to check the quality of the amplified signal or whether it was generating dirty RF.

You're right about the Ft. The final T4 in the above design is only giving a 4dB gain or so amplifying 4mW from driver stage to 10mW and it needs to be driven hard. I used it as an extra buffer to further isolate the antenna from the oscillator and to boost the output a little over 10 milliwatts at 91MHz.

The low Ft of 2n3904 is noticeable once the frequency goes above 95MHz. Gain falls drastically near 100MHz.

I'll draw it up nicely and write a description and want you to publish on your website. It's better than my previous design of a stable FM TX where I had used 2n3866s.

In general, it is a bad idea to connect two or more transistors 'hard' in parallel.
Since there are always differences in the Vbe voltage (due to tolerances) and Hfe, there is a chance that one transistor is doing (almost) everything while the other transistor is doing (almost) nothing...
When transistors are used in parallel (as in power supplies for example), you usually see small emitter resistors to equalize the differences.
Even high power HF transistors are made with multiple chips in parallel, with integrated emitter resistors.

Hi Joy,
The 2N3904 has an Ft of 300MHz, but in my 400mW project the 2SC3781 has an Ft of 500Mhz.
See https://pe2bz.philpem.me.uk/pdf%20on%20typenumber/0-9/2SC3781.pdf

For a VHF amplifier you really need to have a gain-bandwith product of 5x to 10x the operating frequency to be useable. In my 400mW amplifier I also drove the final stage quite hard due to the low Ft.

The 2SC3781 is designed for up to 30MHz in CB type equipment, or 35Mhz / 40MHz R/C equipment.

With two 2N3904s in parallel you can expect a much lower RF gain, and you would have to drive it even harder. It would be a lot better to choose a device with a much higher Ft, such as the BF199. They are really common and cheap. If I were to design a 400mW VHF amplifier today I would choose something more suitable, such as the 2N4427 or 2N3866. These devices also have a metal can so you can get a bit of heat away from them.

I power amplifier stage should really have a power gain of about 10dB to 20dB, so an Ft of 5x is about the minimum for 100Mhz.

I may have a few BF199 devices kicking around at home. If you want I can take a look and if I find them I can send you a few?

Best regards from harry - SM0VPO

Harry, I was wondering if it's possible to parallel five or six 2n3904 transistors to get 200mW output on the domestic FM band. They have an Ft of 300MHz and base capacitance of 8pf (ON Semiconductor). The PA transistor you used in your 400mW VHF PA has the same Ft of 300MHz. What are your thoughts? I lost my cheap USB SDR Dongle spectrum analyzer. Once I order a new one I'm planning to test a VHF PA by paralleling 2n3904's.

Regards,
Joy