bjt push/pull amp experiment

[ssbothwell KJ6RSG]

hi guys. i was messing around in LTSpice today trying to mix different concepts from various amps i have built in the past and ended up with this BJT push/pull amplifier.

in spice it seems to produce a great looking waveform and has high gain.

is this a good design for a BJT linear?

before i put this thing together on copper, are there any modifcations any of you would suggest?

will the 2n3866 be a suitable transistor or am i going to burn them out? Ie is -34mA, Ic is 34mA, and Ib is 133uA.

thanks for any comments and happy holidays.

EDIT: OOOPS, in the attached photos i had 2N3904 transistors but in actuality i had been planning to use 2N3866

EDIT2: bleh, i just realized this amp doesnt work the way i thought it did. the output is dc biased and not producing all the gain i thought it was. oh well.

[WA1GFZ]

Happy Holidays.
A pair of 2N3866 will generate a 1/2 watt easily. Might put a couple top hat heat sinks on the devices. C2 &C3 could be higher vlaues. Also there are many circuits out there for 3866s and 5109s. Download a copy of the cubic R3030 manual for ideas. Also there are other ways to configure the feedback to make the gan flat over a wide range.

[ssbothwell KJ6RSG]

oh cool. i increased C2&C3 to 0.1uF and added an output transformer and i am getting a nice 1W output. is it safe to run 2n3866s that high? i was thinking this might make a good driver for my IRF510 amp.

is this the manual you are referring to? http://www.premium-rx.org/ref%20mans/cubicr3030/cubicr3030.pdf

[WA1GFZ]

Yes, That is the manual. Look at the schematic of the first IF module. It has a push pull 2N5109 stage as for the LO amplifier that easily makes 1/2 watt.
I think you could get 1 watt from a pair of 3866s with heat sinks. 34 ma of bias isn't all that high. I've run 60 ma through 5109s with 3 fin top hat heat sinks in RX projects.

[ssbothwell KJ6RSG]

very cool. i'll have to dive into that manual.

i just added 50ohm series resistance to the 0dBm 3.5mhz input signal and the output dropped to 7mW. any idea why that happened?

edit:

oh woops. i noticed i had some problems with the wiring on K2. i corrected that and i seem to be getting 1W output with 330mA on Ic. i guess the maximum collector current is 400mA, do you think i am getting too close to it?

[WA1GFZ]

I've never run more than 65ma per device. They must be running very hot. You will need good heat sinks on the 3866s to dissipate that kind of power.
Try reducing the dc voltage on the bases. I bet you can fry spit on the 3866 cases.

[ssbothwell KJ6RSG]

sorry i wasn't clear. i am just doing spice tests at this point. i havent moved on to actually building the circuit and maybe the 2n3866 is not the right transistor for what i want. i am trying to build a decent driver for my 35w irf510 amp.

i tried building one according to a design frits showed me in the 'AM Challenge' thread that uses a vn-10 and a 2222/2907 pair to drive an IRF510 but have been having a some trouble with it. i keep frying the VN-10s and have to 'ride' the bias pot to prevent them from freaking out.

maybe you could recommend another direction to go for a 1W driver?

[WA1GFZ]

You could always use another pair of IRF510s at lower power as a driver.
Use a 1:1 or even step down output transformer to reduce the output power.
1 watt from a pair of 3866s is pushing it a bit.

[ssbothwell KJ6RSG]

oh i had just assumed irf510s need a lot of drive to turn on. i guess i'll try rewinding the output transformer on my 35w amp and see if i can use it to go 0dBm -> 30dBm.

if that works i can just keep it as the driver and build a new 35w final. my current one is on too small a heatsink anyways.

[WA1GFZ]

If you run class AB then the FETs are on and linear. AS a driver with a 1:1 transformer you may need to increase the feedback resistor values to get more gain. You might consider a bigger FET for the final or 4 IRF510s in push pull parallel.

[ssbothwell KJ6RSG]

maybe i should use my salvaged mrf148a FETs as the final?

i still have to check if they are operational (the unit i ripped them from was left outside for an unknown number of years at the scrapyard), but that would be pretty exciting.

[ssbothwell KJ6RSG]

i tried inputting a 100mW signal from my HP3312A into the IRF510 amp and got between 5 and 10 watts output depending on how i set the bias. however, the FETs kept burning out after a couple minutes. i went through 4 or 5 of them before i decided to take a break for te night.

i had assumed that because i was operating at a lower power level i would be way less likely to burn out the FETs, however this is clearly not the case. i just crunched the numbers and 1W->35W is 15.4dB gain whereas 0.1W->10W is 20dB gain. is this increase in gain the likely reason i burnt out the transistors?

i'm guessing that the difference in gain is from the 28V/2.5A power supply i am using not being able to produce 20dB of gain with a 1W input but being able to with a 100mW input. 

i am going to remove a turn or two from the output transformer's secondary to  drop the gain down to about 15dB. if i can get it stable, it might make a good final for a QRP rig.


i am finding tons of uses for this push/pull design. now i am imaging building a 5W QRP rig in addition to a 60W rig with the my MRF18A transistors as a final. i'll have to find some more heatsinks so i build a couple more IRF510 push/pull amps.

[WA1GFZ]

You can run less bias current when you run lower power. It is unlikely running lower power blew the FETs. They could be oscillating or the bias is set too high. Put a 50 ohm load on the input and monitor the terminated output with a scope to see if it is taking off. Try setting the bias at around 200ma or even lower.
Lower gain at high power is due to FET losses and maybe a little transformer loss.

[ssbothwell KJ6RSG]

i was thinking that at lower power the amp is able to produce more gain because at higher power it runs up against the current limit of the power supply. its also entirely possible that i wasnt being careful with my bias settings. i wasnt measuring current last night when i was messing with it.

today i had the bias set at 70mA  with 15.5VRMS out for about an hour and it seemed stable.

i just turned the bias up to 140mA and the output 16.9VRMS. it seems stable but the FETs are 80C. is that too hot?

[WA1GFZ]

80C is too hot for me. I get concerned when I go above 50. RDS on losses increase when the FET junction gets hot.

[ssbothwell KJ6RSG]

woah i just added a low pass filter to the output of the amplifier and it is dropped my output by ~3dB.

i set the FETs bias to 70mA each (140mA total) and went straight into my dummyload. i read 22VRMS across the load and 1.4A drawn from the power supply. i then inserted a 5 element butterworth filter with a 50ohm characteristic impedance and the output dropped to 10VRMS and only 400mA was drawn from the power supply.

i'm guessing the impedance of the low pass filter is interacting with the output transformer's secondary's impedance and changing the load on the FETs? if thats the problem, how do i get around it?

do i need a matching network between the amp's output transformer and the lowpass filter?

i built my LPF according to this calculator: http://www.pronine.ca/butlf1.htm

[WA1GFZ]

Sounds like you designed the filter for the operating frequency. Most of these programs design the filter at the 3dB point of the attenuation curve. So you need to change your design frequency of the filter higher. A push pull amp will have good second harmonic attenuation so the filter can be  designed for a flat attenuation  with a 3dB point above the band on interest.
Simulate the filter and you will see where it needs to be set. you do not need impedance matching since both sides are at 50 ohms.

[ssbothwell KJ6RSG]

i get that the low pass filter could be causing the attenuation i set the cutoff at my desired frequency (which i did) but why does it cause the amp to draw less current?

i assumed the amp would put put out the same amount of power and that the signal and the loss would occur when the signal runs through the LPF. i guess imagining the signal like a flow of water is not the proper analogy in this situation?


how high above the desired operating frequencies should i set the LPF?

[WA1GFZ]

Running RF into a filter along the attenuation slope will rellect a different Z back at the source so I'm not surprised your power went down.
You want to set the corner frequency so the slope starts just above the highest operating frequency so there is no extra attenuation. Remember a push pull amp will attenuate the second harmonic a lot so the corner can be be just below the second harmonic. Simulate the filter to see where the response starts to increase loss.

[ssbothwell KJ6RSG]

i have been house sitting this past week and havent had a chance to work on any projects.

i just stopped by my apt today i rebuild the LPF and retest the amp. i bumped the cutoff point up closer to the first harmonic.

i have the filter setup for the 75m band and set the cutoff based on my available silver mica caps. the cutoff is 5.3kc.

i first tested the filter using a signal generator tuned to 75m and set to 3Vrms (into a 50ohm load). going through the filter this signal looks fine. at the first harmonic the Vrms is 528mV and by the second harmonic there is no detectable signal.

i then used the signal generator to drive the push/pull amp with a 75m signal and filtered the output with the LPF. i dont have a spectrum analyzer so i cant easily check harmonics but what i did do was add some modulation to the signal generator and then use my grundig G3 receiver to listen at the fundamental and the first harmonic.

the first harmonic came through loud and clear. in fact, it sounded clearer then the fundamental frequency.

is this something to be concerned with? is there some other way to check the harmonic suppression using just an oscilloscope?

[WA1GFZ]

not easily with a scope. A step attenuator and receiver would work. Set the step attenuator so the signal just breaks the noise floor at the fundamental. Then tune to the second harmonic and adjust the attenuation to hear the second harmonic.
The change in attenuation number is your harmonic attenaution assuming the RX has the same noise floor on both frequencies. A push pull usually has very good even harmonic attenuation so make sure to check the third.

[ssbothwell KJ6RSG]

i dont have a step attenuator but i can adjust the amount of drive into the amplifier from the signal generator. i'll try doing that when i get a chance tomorrow. thanks.