The AM Forum

hi. i am trying to put together a little AB amp design i found in "Experimental Methods in RF Design." it is designed to give about +30dB gain to a 0dBm signal.

here is the schematic and my results in LTSpice:
http://i.imgur.com/ObGfH.png

in the simulation it seems to work except for the fact that the waveform gets clipped. when i put the circuit together on a breadboard i get the same results with a 0dBm input signal but if i drop the input to -5dBm or lower the 2n3866 starts self oscillating at 19MHz. this also occurs with no input signal.

questions:

1. is it possible this is due to some sort of unwanted capacitance or inductance caused by me using a breadboard? should i try assembling the circuit 'ugly style' on a piece of copper?

2. how can i get rid of the clipping which is occurring in real life and in LTSpice. according to Experimental Methods In RF Design, this circuit is appropriate for producing 30dB gain on a 0dBm signal.

thanks,
solomon

Try putting a small cap across the collector to base feedback resistor. This will roll off the HF gain. You can do this in the first stage also.
you are really pushing a 2N3866 for over 1/2 watt output.

Kudos for Building and experimenting!!  Good Luck!

WA1GFZ: a cap across the collector-base resistor of the 2n3866 eliminated the self-oscillation. why does this fix the self-oscillation?

oh and thanks for the kind words Ed. :)

How about a photo of the breadboard?  :)

The capacitor provides negative feedback which reduces the gain of the stage as you go up in frequency.  So its giving more negative feedback due to the Xc decreasing as frequency increases. 

Have fun!

Joe, W3GMS 

A ferrite bead on the base lead and sufficient C to bypass will also help. Ive run a 3866 at 1W on 20M using a small push on heat sink. A 2N5109 is more fun to play with.

Carl

joe, thanks for the explanation.

kb4qaa, here is an image of the breadboard: http://i.imgur.com/m7lUA.jpg

km1h, what is the advantage of a 2n5109 over a 2n3866? the only differences i see are that 2n5109 is rated for 1w of power dissipation whereas the 2n3866 is rated at 5w and the 2n5109 seems to have a higher minimum hfe. anything else? would i be able to get a higher gain with a 2n5109?

carl (km1h), would a ferrite bead on the base potentially eliminate the feedback without limiting gain at higher frequency (unlike a feedback capacitor)?

edit: as a correction to what i said in my first post: according to Experimental Methods In RF Design, this circuit should give 24-27dB of gain (not 30dB).

Circuits with VHF components are very likely to "take off" with that type breadboard. You would be better off using ugly construction. The long leads within the breadboard and the long wire leads just ask for feedback problems. For this type of circuit you need good low impedance grounds and power supplies.

That circuit is long used and trouble free when done with the proper layout. I have used it many times with great sucess.

Pat
N4LTA

okay. in that case i will put it together on a piece of copper and see what happens.

Just get yourself a piece of copper clad non-etched PC board.  Use that as your ground place and for all your circuit common connections.  You can do voltage rails with a piece of 12 ga wire since its stiff enough run horizontally maybe 3/8" above the copper ground plane.  Then you can put your bypass caps right from the voltage rail that you have created to the ground plane.  The caps will keep the voltage buss in place.  As you use that technique soon you will become real comfortable with it.  I sometimes use small strips of PC board material as the voltage buss.  Try to analyze your circuits current loops and that will give you insight into how to lay it out. 

I have some of the same breadboard material that you show and that's not the best when dealing with RF especially at high frequencies. 

Have fun....

Joe, W3GMS       

Wow, I'm surprised that it worked so well without a ground plane under it.
Joe's method will work a lot better.

In our prior 7 page post, dealing with the mixer,  I mentioned to you NOT TO BUILD RF circuits with your technique. I posted pictures and I thought that error was behind us. I know it is tempting to use your push in wire to BB circuits. However, you really need to understand that parasitic elements, unwanted coupling; a REAL problem to get a handle on in RF circuits and  GROUND IS NOT necessarily GROUND.

Good Luck.

The first obvious conclusion is overdrive in one of the stages. Look at the output of the 3904 to determine which stage is doing the dirty deed.

Ah, Yes..............................DC ground is not necessarily RF ground. Sometimes we find out the hard way :o  ;D

alan, i know you made it really clear in the last thread that perfboard is not ground. it is just so tempting to get off the wagon and test circuits on perfboard.   ;D

i'm gonna put the circuit together on copper right now. i'll post pics and a performance report when it is done.

You will always have clipping in an AB stage without a tank circuit flywheel action.
The only way you can run a clean linear in class AB is if you go push pull. This is the reason all HF transceivers run push pull AB.

There is a technique that may work nicely for breadboarding this sort of thing, I think it is called "Manhattan" style. A search on the internet will show examples.

You take a PCB copper side up, then glue (epoxy or crazyglue) pieces of cut out PCB material wherever you want junctions of parts... I use a 1/4" metal punch to make the junction points, punch out a number of them. You could cut or shear strips and then cut squares, etc... it works nicely for proto and fast builds...

                    _-_-bear

it took me a long time but i assembled the circuit "ugly style." i think i prefer drawing the circuit with sharpie and etching. it takes more steps but i can solder much faster that way.

here is the circuit board: http://i.imgur.com/ABJZT.jpg
sorry about the bad photo.

the circuit still self oscillates if i dont have a >150mV RMS input signal.

i am getting +26dB gain from a 0dBm signal but it is severally clipped. http://i.imgur.com/9Yp0S.jpg

i assembled the circuit without a feedback capacitor on the 2n3866. i am going to try adding one right now.

i was hoping to use this circuit to amplify a low level modulated signal to drive a push/pull IRF510 AB amp stage. maybe i should use some other circuit topology to do this?

[EDIT] with the feedback capacitor the circuit does not self oscillate but the waveform is still severely clipped

I'd be interested to see what would happen if you built a pi for the output network with a q=3 to 5.  You might even need more, depending on how much flywheel you need.

As Frank said...  You're going to end up with some clipping because it's single ended.


--Shane
KD6VXI

would a pi-network actually cause the amplifier to work more linearly or would it just clean up the clipped signal? i'm having a hard time understanding this conceptually.

also, for calculating the pi-network, what should i use as the input impedance?

down load a copy of AN749 tells you how to determine output Z. VCC X VCC / 2 X power out.
Yes a pi network will improve IMD (clipping) The network acts like a flywheel on a motor.

great! i'll read AN749 today. thanks for the tip. Z seems to be a really hard concept to get a handle on.

You can also throw in NFB via a .01 and app 68 to 150 ohm resistor from collector to base of the transistor.  These are the values I use from 15 meters up...  You may need to fudge the values of the R and C to get the necessary feedback to stop any oscillations.

It will lower gain, but it really helps to stop oscillations and other such nastiness.  Some also use judicious amounts of ferrite beads on the xister leads, depending on if it's HF or LF oscillations.


--Shane
KD6VXI

I've been watching this one also. I wuz wonderin when someone was going to suggest wrapping a little feedback around it for neutralization ;D  ;D

I built a small (5W) stripline amp some years back for my 220 repeater. I needed an IPA to go from the 300mW exciter to drive the 100w PA. It has been many years back and I dont remember what the transistor was, but IIRC it was an NPN bipolar device, a small "windmill".

This thing would take off so bad that it would quickly fry the transistor. I pulled my hair out trying to stop it. I finally ended up bypassing the Vcc line to the collector with a combination of a .01 ceramic disk, a 100uF electrolytic, and a 5400 series diode to the ground plane of the pc board. that combination tamed it right down and it lived happily ever after and is still running 15 years later.

each stage has feedback My suggestion just increased the FB at higher frequencies to tame it. Good power supply bypass is always a good idea

The 3866 is rated at 1W out in Class C while the 5109 designed as a Class A CATV amp wil double that in Class C and have about 1.5dB more gain. Ive also found it easier to stabilize even tho its Gbw is higher



That depends upon what you mean by higher frequency. If you use the minimum ferrite to eliminate a VHF parasitic then it should have no loss effect at HF. One tiny FB43-101 may be all it takes. They also work on the other leads.

Carl

WA1GFZ, i read through AN749 but i am still unclear about how to determine the input Z circuit. i'm not really sure what i am reading in this article. it seems to be tables of impedance values for various transformer setups. how does that help me know what impedance i need to match?

Simple, the load is 50 ohms on the output.
So if you have a 1:2 transformer the output stage needs to drive 12.5 ohms
Push pull each side drives 6.25 ohms.
Now if you have a 1:3 the output stage needs to drive 5.55 ohms or each side of the push pull 2.7 ohms.
A 1:4 transformer loads the final stage to 3.125 ohms so each side drives 1.56 ohms.
Look at your output stage and select the transformer that matches the power
rating of the output device.
Typical 12 volt 100 watt rig runs 1:4 or 1:5 output transformer.
A pair of 48 volt devices will make 300 watts with a 1:2 output transformer.
Go to CCI site and download a bunch of old motorola applications to get ideas. Helge wrote the bible on this stuff back in the 70s.

i arbitrarily chose a 1:2 ratio and used this calculator to generate the pi-network: http://www.qsl.net/wa2whv/radiocalcs.shtml

i entered:
input Z = 12.5
output Z = 50
Q = 5
freq = 7MHz

and it gave me the part list:
C1 = 2752pF
L1 = .41uH
C2 =1585pF

i added the pi-network to the amp output and it produces a clean sine wave. however it the amp has signifigantly less gain. without the pi-network it was producing about +26dB gain and with the pi-network it produces +13dB @ 7MHz. the gain is very frequency dependent with it producing +20dB at 6.4MHz.

would a different impedance ratio in the pi-network produce greater gain?

[EDITED]replaced all mW and mV measurements with dBm and +dB gain to clarify this post.

attached images are the circuit set to ~6.4MHz and 7MHz.

yes, Your pi network is transforming the impedance
As long as the output device can handle the extra load.

messing around in LTSpice i tried a bunch of Z ratios. i highest peak voltage i was able to get was 4.1V with 25ohm inputZ and 50ohm outputZ ratio. the lower i set the input Z, the lower my peak voltage. for example 5.55ohm input Z yields 3.5V peak output.

does this make sense and seem right?

is the basic process here to test out different Z ratios until you get the circuit to peak?

edit: is it possible to have a pi-network that can be adjusted to work across a variety of frequency bands?

Solid state rigs have a fixed low pass filter for each band. You may have to run class A to get a clean signal through a fixed filter. The output transformer ratio would be 2:1 broadband. this will force 12.5 ohms on the collector and 50 ohms to load/filter. This is a very common configuration for the 3866.
Class AB will produce some distortion when the waveform clips.