The AM Forum

hi guys. i have been really busy with school lately and havent had much time to work on my radio projects. my last thread on my push/pull irf510 project is kinda old and i figured i would just start a fresh one in the hopes of getting some help finishing the project.

i'm attaching the current schematic and photos of the output on my oscilloscope.

i am running it off a junky Pyle brand benchtop 13.8V 3A regulated power supply and have 1W RF input. with the bias maximsed i get a 6.3W (17.8VRMS) output but the waveform looks really strange. if i drop the bias down signifigantly i can get a cleaner waveform but the power drops down to 1.8W (9.7VRMS) which is a piss poor amount of gain.

any idea why the output is so distorted? is it normal for such an amp's output to be like this? do i simply need to add a low pass filter or something to clean it up?

apparently one of the images isnt attached properly. i'm attaching it to this post and adding a bonus photo of the board layout.

Why is R8 in the circuit? You have 1 ohm in series with 1/2 the amp.
Output transformer primary inductance looks pretty high.

oh sorry, i forgot to mention R8. it is only in the LTSpice model. its there because spice was reporting the following error without the resistor:

"Inductor L5 is in a loop involving inductor L6 and other voltage source and/or inductors making an over-defined circuit matrix. You will need to correct or add some series resistance."

i assume it has to do with the spice model not including the actual series resistance of the inductors.

what range would you have the output inductance set at? i got these values from 'Experimental Methods in RF Design' where they used 2:3 turns on primary:secondary of a BN-43-7051.

Hi ,
I've never done shunt DC feed in simulation. You will need to balance the circuit and put a second resistor in series with L4. 1 ohm is pretty high maybe you could try a lower value.
Look at each drain with a scope and see if the high frequency ripple is on one devce or both. Remember a linear amp will generate harmonics. Also the length and termination position of your scope probe ground lead cah give some interesting results.
OK on the transformer keep an eye on input current. You may need more turns if the current looks high. I don't think two turns will give you 40uh though.

oh man. i am so dumb. without realizing it, i had been using a wirewound resistor as my load. i replaced my load with one made from proper power resistors and now i am getting a pretty clean 14VRMS output. this image is with the bias turned up almost to the max: http://i.imgur.com/NLm5y.jpg

i dont have an image of it but i was also able to get the output up to 18.2VRMS with a more or less identical waveform.

is this the amount of power output i should expect running off a 13.8V/3A power supply?
13.8V * 3I = 41.4W but what kind of efficiency should i hope to expect from a push/pull design like this?

my output seems kind of low.
 
as far as the output transformer 2 turns on a BN-43-7051 core is 40uH according to toroids.info http://toroids.info/BN-43-7051.php

I stand corrected on the core. I have a bunch of those cores I didn't know you could get such a high inductance from 2 turns. I never looked them up.
Your amplifier is making good power for class A but efficiency low for class AB
Try reducing the bias to see if power comes up. Also you could add a turn to the output transformer secondary. Class AB will degrade the waveform and require filtering on the output.
Glad you found the load problem.

adjusting the bias doesn't get me past ~18VRMS. how exactly do i know when the circuit is operating in AB mode?

hmm. if i am going to have to filter the output anyways, would it make more sense to modify it to operate in class-c rather then try to get it working in AB?
scratch that last sentence, i have a lot to learn from this circuit before i move on to something totally different.

wa1gfz, maybe you can explain something to me. my bias is coming from a 78L05 being run into two 5K pots (one for each fet) that then go through 1n914 diodes into the FET gates.

with no input signal and the bias pots set to 5Kohms, the amp is drawing 2.45A and there is 3.7V across the each bias pot. if i drop the bias pots resistance to 0 then the amp starts drawing 3.4A and i see 0V across the bias pots. why is this?

i would think that with less resistance there would be greater voltage across the pots? whats going on here?

also, is the high idle current indicating that the circuit is running in Class A?

and just to clarify some of my previous. when i talked about high and low bias, i was talking about adjusting the amount of bias resistance to a high or low amount.

Hi Sol,
The typical 12 VDC 100 watt final runs idle bias for a drain/collector current of about 300ma. So you are set quite high at 2.45A.
Now if you want to increase your output voltage just add a turn to the output transformer secondary. So the 24 volt primary gets stepped up by a factor of two. You might want to try even an additional turn.
Again the typical 12 volt 100 watt fianal the output transformer has a ratio of 1:4 or even 1:5 turns.
Every time you add a secondary turn you reflect a lower impedance back at the FETs. At some point the FETs will not be able to drive the transformer  and will get hot. The source resistors may need to get smaller to keep the dissipation down.
You will need to put a couple trim pots between the 7805 and D1,D2 so you can reduce the gate bias down to reduce drain currents.

ahh i see. why does having a step-up transformer on the output of the transistor force the transistor to produce more gain? intuitively i would think the transformer would just change the ratio of voltage and current (increase voltage and drop current) while maintaining the same amount of power.

in the real life circuit i have 5K trim pots between the 78L05 and the diodes. i know its not listed in the LTSpice model but i wasn't sure how to draw a  potentiometer symbol so i skipped it in the model. with 5K of resistance the idle current is 2.45A.

should i keep adding resistance to the bias until the idle current is at 300mA?


edit: i accidentally shorted one of the IRF510s and burnt it out. i bought another one from radioshack. i wonder if there was something faulty going on with the transistor that burnt out because the circuit is no longer drawing 2.45A when idle. with no input it is now drawing 0A and it jumps to 740mA when i input a 1W sinewave. also, the heatsink is not getting nearly as hot as it was before. it is now cool to the touch even when the circuit is outputting 5W.

the wave form also now has a little crossover distortion. i can only get rid of it by dropping the bias voltage to zero or near zero.

oh woah! i switched from my 12V/3A powersupply to a 28V/2.5A one that i bought today and suddenly the circuit is putting out 24W!

34% efficiency is pretty good but i should be able to get 50% with an AB amp, right?

edit: is it possible to blow out the transistors if there are too many windings on the output transformer's secondary coil?

Very cool sounds like you had a defective FET.
Adding more turns to the transformer secondary puts a heavier load on the primary. Look at, as the primary needs to drive each turn to the push pull primary voltage. It takes the same energy to drive each turn so when you add more turns the primary has more work to do. At some point the RDS on of the FETs will limit current. I'm not sure where this point is using the IRF510. My 160 meter rig uses IRF810s. I limit each FET to 25 watts carrier and hit well over 100% positive peak on AM. I will be interested in your final results because I have a heat sink all machined for a 22 FET 11N90 linear with 7000 square inches of heat sink. Simulation shows I can get 1800 watts at 80 volts VCC. At 130 volts 5500 watts.

holy cow are you talking 1800W in class A? isn't that gonna be a massive energy hog? why use such a giant linear amp? is it somehow cheaper then modulating a class C or E final?

i tried bumping up the turns on the output xfrmr secondary and at first it produced more gain but then i think one of the FETs was damaged. the output voltage dropped significantly and now the circuit is drawing 1.5A when idle (like it was doing before). i'll have to go back to the ratshack for a replacement.

i guess i cant put too many turns on that secondary.

no 1800 watts class AB or D.
Take a look at your drains when the stage is operating to see if there is any overshoot voltages generated that will kill the FET. Maybe you might want to use some higher power/voltage devices now that the circuit is functional.

oh okay. how do you modulate an 1800W amplifier?

i'm about to go buy a replacement irf510, i'll check the drain voltage later tonight.

do you think i will be able to get 40-50W out of an IRF510 pair or do i need a higher voltage transistor for that power level?

I adjusted the amount of turns on the output xfrmr's secondary. its at 6 turns now. when i first turned on the amp it was producing 42VRMS but it slowly is dropping down in voltage. its been a couple minutes and its down ti 29.4VRMS.

edit: it seems to have stabilized at 27VRMS and drain->ground voltage is 28V.

Hi Sol,
Very cool, I  figured you would need a 2:6 turns ratio. Now you need to find why the output voltage is dropping off. It could be the FET die or the source resistor heating.
Check your drive voltage to see if it is constant. I'm not sure how much power could be extracted from the 510. An option is to add another pair of FETs in parallel to get more current if they are heating up. Layout becomes critical when FETs are in parallel.
1800 watt amp is a linear so not modulated. Check my article in Nov. 2005 QEX.
(Now WA1QIX jumps in and busts my butt "when you going to modulate that thing and get on the air")

its exciting to be so close to getting this thing working. it would be awesome if i could get this amp running stable at 35W

i checked to see if the source resistance is changing as the resistors heat up. i am using .88ohm 5W sand resistors. when cold they are .8 to .9 ohm and when they heat up they increase to 1.1 to 1.2 ohm. going back to the spice model i see that changing the source resistors from .8 to 1.2 ohms does make a noticable difference in output power. perhaps this is my problem?

i repositioned the source resistors so that they could be pressed flat against the heatsink. this seems to slow down the drop in output power but its still happening. maybe i just need higher watt resistors.

are sand resistors a bad choice? i have some 10W 0.7ohm sand resistors and i have 1ohm 5W metal oxide film resistors. which ones would you recommend i use?

I think sand resistors are wire wound. Yes when they increase in resistance the gain of the stage drops. Metal film resistors will work  and will be less inductive. An option would be to put two in parallel and see how the stage operates. This will increase the voltage swing across the FETs and reduce feedback.
There is a fair amount of voltage developed across these 1 ohm resistors so  you could try going lower or adding 2 more FETs. The linear I simulated has .5 ohms in the sources and has room to go to .33 ohm using 1 ohm 2 watt surface mounted resistors in parallel.
I think you are very close to having a 35 watt output.

well, yesterday ended up being a bad day for me and electronics. someone spilled liquid on my laptop and i accidentally shorted the source to drain on my amp project burning out my last two irf510s. i really hope my computer isn't dead.

i ordered a bunch more irf510s on ebay but they wont be here for a week or two (they shipped from china). i'll stop by a radioshack and see if i can find some more locally.

i had no idea sand resistors are wire wound! i have been using them for all kinds of rf related stuff. they are much easier to find then metal film resistors. when i get new FETs i'll try using 1ohm 5W metal film resistors in parallel and see how the circuit operates.

There are a number of FETs you can use. IRF510, IRF610, IRF710, IRF810 For higher voltages  gate C goes up a bit at each jump
Then there is the IRF820, 830 and 840 that go higher in current.
Again Gate C increases at each step.

Bummer on the lap top. Been nailed by liquids myself. Guess it is your turn and we all get a turn

no dice on the ratshack i went to on my way home. i'll report back again once i find some new fets and test them out.

I know how you feel. Back in the early 80s I found 4 TO3 IRF250s in old breadboards at work. they sold for about $80 back then. I built an amp with them and blew it up on 80 meters a couple weeks later. It took a year to scrape up more parts then the price really dropped

wow you can get irf250s for 1$ each on ebay now. its so crazy how different the electronics market was 30 years ago.

i rebuilt the circuit board to make it more mechanically stable. i think i like 'manhattan' style more then 'ugly' style. the transistors you see in this photo are actually burnt out (i just put them in the photo as models). i need to stop at another radioshack today and try to find new ones.

i think at some point i would like to try making a printed circuit board for this amp. i have a vision of a modular plug and play amplifier where you can add and remove push/pull pairs with ease.

looks good. PC boards are pretty easy for amplifiers. just use masking tape and an XACTO then cut out tape where you want copper to go away.

hey i found some IRF630s and IRF740s. would either of those be suitable?

YES

great :)

i installed the irf630s but some weird thing are happening with the bias circuit.

1. whenever i power on the circuit the solder melts off the anode end of one of the bias circuit diodes.  ???
2. when i check the voltage across the bias pots, the voltage is 0v for almost the full rotation of the pot and then jumps to 5V when i have rotated completely. when i power off the circuit and check the resistance of the pots it goes smoothly from 0 to 5k across the full rotation of the pot.
3. this is likely related to #2 but adjusting the bias does not change the output waveform at all until suddenly it cuts out when i have rotated the pot completely to the 5v bias position.

any ideas?

sounds like a wiring error or one of the FETs is leaky drain to gate. It could have taken an eSD hit.
Try running the stage without bias and measure the gate voltage. If there is DC voltage on a gate the fET is dameged. fc

the weird desoldering of the bias diode stopped happening. maybe something was sitting on the board causing a short that i did not notice.

i didnt measure any voltage between gate and ground when the bias is disconnected. the IRF630s seem to be working fine, its just that adjusting the bias only a small portion of the bias pot's rotation seems to impact the bias voltage and i cant seem to get rid of the crossover distortion for the life of me.

i wonder if i am not properly driving the IRF630s? i didnt make any changes to the circuit to accommodate the switch 630s.

i wish i had some 510s with which to verify the circuit. --edit: i just checked my mail and my order of IRF510s arrived!

i installed IRF510s finally :)

here is the output: http://i.imgur.com/5yLW7.jpg

it still has really nasty crossover distortion which i cant get rid of. bias adjustments are behaving weird as i described earlier.

is it possible that using a 7805 rather then a 78L05 the source of my bias problems? when i rebuilt the circuit i chose 7805 because 78L05 is rated at 20V input and i am using a 28V power supply.

other then the bias and crossover related issues this new board layout seems good. using 2x 1ohm 5w power resistors in parallel for source resistance seems to improve the stability but it is still very slowly dropping in voltage. maybe a fan will help with this.

well it turned out that all the latest problem was the same as all other problems: i didn't notice something obvious. when i rebuilt the board i forgot to include the R1,R2,C1, and C2. oops.

now the amp is working great! when i first turn it on it produces 47VRMS and over ten minutes it dropped down to 38VRMS. i wonder if this is actually the amp putting out less power or my dummy load overheating and having its resistance change. i am using an oil filled load that is rated for 20W so i am going way over its specs here.

i am going to consider this project 90% complete. i would like to test it with a higher rated load and see if this power dropping issue is real or not.

wa1gfz, do you have any final comments, suggestions, or tests i should run?

Go back to the higher voltage 78XX regulator of reduce the input voltage.
I've seen them take a real beating but the one thing that takes them out is high input voltage.
Your output voltage would go up if the load resistance was increasing.
I suspect the FETs or source resistors are warming up.
I would think 50 watts RMS is a good place to be with 2 TO220 devices. I do know IRF840s last forever at 100W PEP/FET in class E. Do the FETs seem hot after 10 minutes?
You could try going lower with the source resistor or add another pair of FETs.
If you add more FETs you could add a turn or 2 to the output  transformer secondary. That core should easily do 100 watts.

i'm using 7805 (35V Max input).

i just remembered i have a temperature probe!
after ten minutes the output is down to 37.5VRMS and the the fets are 140C and the source resistors are 100C! i measured one of the heatsink fins and it is 60C.

i guess i need to put a fan on the transistors?

edit: i aimed an 80mm fan at the fets and the voltage is slowly creeping up. its at 41VRMS right now. i am going to get a giant fan to mount to the bottom of the heatsink and maybe i second one as an intake to the enclosure i eventually put this amp inside.

edit2: with one 80mm fan aimed at the fets and one 80mm fan aimed at the heatsink fins the fets are at 55C and output is 39-40VRMS.

140 degrees C is way too hot for a case temperature. The junction is hanging by a thread. You need a better heat sink. I monitor case temperature of one FET in the 160 meter rig and they never get above about 50 degrees c on a hot day without a FAN.
As the junction gets hot the RDS on must be going up. 

but when you are using your 160m rig it is not left on peak output for extended periods of time. i would imagine that when i modulate the power supply it wont always be reaching peak output so it will have more of a chance to cool down, right?

with fans the transistors are stable at 50-55C. is even that too high a temperature?

Fan is ok if you don't mind the noise.
My 14 FETs are running about 300 watts carrier and about 1500 watts peak

so this brings up something i have been wondering about. when i am driving my amp with an unmodulated VFO and the amp is putting out ~35W, that is my peak output right?

when i setup a modulator circuit i want to set it so that it reaches 35W at 100% modulation, right?

It is your peak power for that drive level and VCC voltage you are running.
You can low level modulate it if you take the output of the VFO and use a double balanced mixer as a modulator. You can't run more than 0 dBM into most mixers. VFO drives LO port and output is the RF port. Audio goes in the IF port with a slight DC offset to set the carrier level. All three ports want to see 50 ohms. you will need a little gain after the DBM to make up for circuit losses. go to mini-circuits site and look at their app notes.

cool, i assumed it was peak power but i wanted to make sure.

if you low level modulate a signal and then input it into a class AB amp, wont you get distortion? i thought that AB is not completely linear.

i was planning to modulate the power supply ( is that called series modulation?), but i already have a working diode ring mixer from a previous thread. i guess i should try it out.

You could run class B if you series modulate the power supply. It will take more drive though. Class AB will work as a linear amplifier if biased properly and has some RF feedbach which you have.

oh interesting. how can i tell if the amp operating linearly? is there a way i can verify it precisely using my scope?

also, in the SS rig challenge thread, W1FVB recommended a modulation circuit. i am attaching it here. do you think it would be suitable for my amp?

That circuit will not go up to 30 volts output peak. Maybe 28 volts. I would simulate it and check the voltage swing. I like using a PNP to drive an NPN. Also need a big heat sink to dissiapate the heat. I don't like to run much more than 3 amps through a 2N3055.
I'm not sure if there will be distortion due to the AC terminated op amp - input but see the need to have it float.

doesnt it only need to reach 28V? thats the voltage of the power supply i am currently using.

yes but it will have a little extra loss

why is that?

VBE of both transistors and the maximum voltage from the op amp.
Not a big deal at this power level but it will limit the maximum output