The AM Forum

i found this amp design in "Experimental Methods in RF Design": http://i.imgur.com/vrwAY.png

K1 is 12 trifilar turns on a FT50-43
K2 is 12 bifilar turns on a stack of two FT37-50
K3 is a 2:3 transformer but the number of turns for each wire is unlisted

i have a couple questions about this amp:

1. each transformer poses a different challenge for simulation. how do i calculate the inductance values for a center tapped trifilar winding, how do i estimate the inductance caused by double stacking ferrite beads, and what values should i use in the 2:3 winding?

2. LTSpice does not have an IRF511 spice model. Furthermore IRF511 is difficult to find and unreasonably priced where i can find them. IRF510 is readily available and LTSpice includes an IRF510 spice model. would IRF510 be an acceptable substitute in this circuit?

3. what class is this amplifier and will it be able to amplify a low-level modulated AM signal without distortion?

Hi. The 510 is quite fine for CW. I suspect it will do well for AM when biased correctly. The amp pix I posted several weeks ago is a 510 and looking at an 830. They do 50-60 W at 1 W input, push-pull in the HF band.

On the xmfrs. You will need the AL factor from the data sheet. The AL is the inductance per unit turn. Assume the k factor is 0.9. If you stack them, you will get more L per unit turn. Without a measurement, for now assume the AL doubles (this really needs to be checked). Of course the dot convention and the start finish polarity needs to be observed, you have already learned that lesson!

Again on the 2:3 winding, implementing a 4:9 impedance transform, you are going to need the AL factor for that material mix... type 43. I believe all the AL factors are listed... check the AMIDON website (https://www.amidoncorp.com)  or FAIR RITE web site.

As a rule of thumb on these xmfrs, you would like the reactance of the xmfr to be 5x the impedance transformed. This requirement to minimize the insertion loss. So, if the drain push-pull impedance level were 25 ohms, then each xmfr drain side to CT should have a reactance of 125 ohms min. At the frequency of interest, calculate the required L.

73; GL with your new ticket!

oh good. i'm glad i can use IRF510.

here is the circuit with inductance values based on the info at toroids.info: http://i.imgur.com/YjGv6.png

in LTSpice, do i specify the reactance by setting the series resistance of the transformers? i assumed this to be the case and set rser to 125. is that not correct or is there something else wrong with my circuit? i am only getting 100mV Pk output from a 7mhz 7V Pk input.

Danger, this amp as drawn will blow up. The gates do not have a return path to ground so you will develope uneven voltages across the gates. Two options 1. put a little DC bias on C3 and C4 to run in AB mode. 2. tie R3 and R4 directly to ground for class C operation.
Download yourself a copy of Motorola App note AN749 on the design of broad band transformers. Check out CCI web site for a number of motorola app notes on broad band amplifiers.
A guy in EU  built a 1KW linear for 6 meters using 16 IRF511 a few years ago.
Good Luck Fun stuff

The series resistance 125 Ohms is not the right way as it is like putting a resistor in series, but I don't know what the right way is for that.

You beat me to it.

In the simulation, for the bias to each gate, use a dc gen. to ground with the positive side of the gen to a >100 uH inductor and then to each gate. Try varyinging the DC gate biases to between 1.5 and 3.75 volts. I would also put a 10 ohm series resistor in the gate circuit just before the actual gate connection and after the bias node.

In real circuits, the reason for separate bias voltages is that each device will have different gains at different bias voltages. Series gate resistors reduce gate ringing.

Phil - AC0OB

here is the circuit with the bias mods phil and wa1gfz suggested: http://i.imgur.com/ANSER.png

in the real circuit how do you guys suggest i produce the bias voltage? should i use a voltage divider, regulator ic (lm317?), or some other bias method?

is the voltage variable on spice signal sources Peak to Peak, RMS, or something else? i have it set to 7V but it looks like 3.4V Pk-Pk. also, is there a way to display RMS voltage in LTSpice?

i also had to add resistance to the 28V input to L5. without a resistor, ltspice was reporting a loop with L5 and L6. will the series resistance of the real life inductors negate the need to add an additional resistor to the circuit?

It depends on your PS system. For Class A and AB, I would recommend a +,- 5 volt biasing system using 78L05 and 79L05's with a 5k potentiometer such that the wiper of the pot goes to the RF choke inductor. A 0.1 cap is usually place from the wiper to ground for various reasons.

Phil - AC0OB

You don't need to go negative 0 to 5 volts is good. The gate threshold is around 3.5V positive. 5V on the high side of the pot is plenty. Also if you want to run in AB mode put 1 ohm in series whth each source to make the bias more stable over temperature. A resistive divider is fine as long as the PS voltage is stable.

i think i'll go the 78L05 route. can i put two pots in parallel off of one 78L05 or will one not produce enough current for both bias circuits?

Put a diode in series with 5 volt end each pot to protect the system if 1 FET fails drain to gate. This way you won't take out the regulator and other FET.

i'm attaching a schematic that includes all the suggested modifications so far.

with a 7V (Pk?) 7mhz input i get a 25W output. With a 3.5MHz imput i get a 64W output! is that going to be way too much power for the mosfets? will i burn them out?

my goal is a general purpose final amp stage for low level modulated AM signals on all the HF bands.

i want to try to put together a transmitter that uses a diode ring mixer to modulate the carrier, then a push/pull AB amp to get to ~1W and then this IRF510 push/pull amp as the final. will this work or should i be trying to do series modulation of the final amp stage?

The output power will be limited by the size of your heat sink and the current rating of the device. I would think 50 watts is a good power rating for a pair.
I run 14 IRF840s push pull parallel class e on 160 and can hit close to 2 KW PEP output flat out. It has been on the air since 1996.

woah! 14 IRF840s! thats crazy. do you have any photos of your rig?

i just ordered all the parts i needed for this amplifier. hopefully everything will arrive soon and i will be able to assemble it over the weekend.

the only power supply i have that can do 28V is a variable supply with a max current of 750mA. does that mean my max power output would be 21W?

Series mod on the final is much more efficient, and I'd expect to see more Pout.  Class C,D, etc. are MUCH more efficient.  BUT, I'm also enjoying watching this project take effect.  :)

BUT, low level mod has it's bennies, too.  And it's problems.  Don't be expecting 150 percent peaks from a low level stage...

--Shane
KD6VXI

Nope.  That would be your power input.

Output, figure half to three quarters, Class AB to C.


--Shane
KD6VXI

check out QEX Nov. 2005 my 22 FET push pull parallel for 80m article
nicer looking. Same layout though.

well in that case i would be better off with a 12V 3A power supply versus 28V 750mA?

i am going to give low level modulation a shot but i would like to also try series mod. would that look like this?: http://www.eie.polyu.edu.hk/~ensurya/lect_notes/commun_cir/Ch8/Fig8-19.jpg

i tried doing something like that but it doesnt seem to work (circuit image is attached to this post).

Frank that looks a lot like my IRF840 P-P modulator - same basic bias scheme. I get a lot of audio out of a pair.

For the smulation, I would recommend you replace the 78L05 with a DC voltage generator and sweep the voltage from 1.5 to 4.5 volts to observe the effect of bias with output voltage, since I assume you have the 510 model imbedded within the simulation code. 

Phil - AC0OB

Then add 1 ohm source resistors and you will see how to tame the bias.
They also produce some negative feedback

WAIGFZ, the circuit i last posted has the 1ohm source resistors (R9/R10).

DMOD: it actually is using a dc voltage source. i couldn't find a 78L05 model so i just used a 5v dc source with a label placed over the symbol ;D adjusting the 5K resistors adjusts the bias.


the best documentation on high level modulation i could find is this: http://www.eie.polyu.edu.hk/~ensurya/lect_notes/commun_cir/Ch8/Fig8-19.jpg

the arrl handbook doesn't really get into it much.

My LTSPICE modulation circuit from the last image i posted where i have an audio input coupled to the DC power input via a transformer does not seem to be working. i have tried a variety of inductor ratios and audio power levels but no dice.

just for fun i tried using a signal generator as a power source for my LM3866 AB amp (from my other current thread). I set the signal generator in the audio range and cranked up the output voltage as high as it went. it modulated the amplifier decently.

that was a fun little experiment but i cant imagine it being a good idea to use an audio amp to drive a 50W final RF power amp.

in proper 'plate modulation' i am supposed to use the audio source to modulate the power supply right? what kind of transformer do i need and what sort of power levels do i need to run my audio source at?

i ripped a transformer out of a broken radio receiver and tried using it to do plate modulation. i hooked it up every way i could think of but had no success producing modulation. i'm attaching a photo of the transformer to this post.

[EDIT] reformatted text a little bit to clarify the post.

Your circuit still has a problem. There needs to be a bleed resistor between each gate and ground. 10K will be fine. The gate is like a cap and will store a charge based on drive waveform. A bleed resistor will keep it at the right DC voltage.

do you mean like this?

also, can IRF510 really handle 50mA Pk gate current? that would destroy any other transistors i have experimented with.

No, Connect the resistors directly to ground, not in series with caps. DC current does not flow through a cap.

Gate power= CV^2F so the higher the frequency the higher the gate drive power.
The gate draws no DC current. The drive signal just has to charge and discharge the gate C to turn it on and off. At some point the gate can't handle being charged and discharged any faster so it over heats and fails
The gate power is due to ESL and ESR since there is some resistance and inductance in the gate structure.

Frank what should I do. DC flows though my caps  :o

Attach each cap to a line cord and plug it in.
This will release excessive smoke built up in the package
Once excessive smoke is released no current of any type will flow.

WA1GFZ, can i accelerate that process using a blowtorch to gently singe the capacitors?  ;D  ;D  ;D

but seriously. i wasnt thinking when i added those caps but with 10k resistors gate->ground the output waveform looks funky: http://i.imgur.com/QYiyE.png

Looks like crossover distortion. one fet turns off before the next one turns on.
I suspect you need to increase the bias voltage a bit. What is the no signal resting current?
You can add some RF feedback by putting a resistor in series with a cap say 100 ohms and .01 uf.
Connect between gate and drain of each FET. This will reduce the gain a bit
1 ohm source resistors will also require slightly higher bias

with no signal there is 80pA at the gates and 10pA at the source.

ahh i see. correcting the crossover distortion is a matter of getting the bias pots (R3 and R4) set to just the right level.

also those gate shunt resistors really drop my total power output. i go from ~22W to ~14W. how do i get that 50W you mentioned earlier?

Yup you need more bias voltage. Yes the 10K resistors will pull the bias down a tad. They still need to be in the circuit. I'm thinking your drain current should be around 100 ma.

so i went back through the thread and tried to add all the modifications you mentioned. here is the circuit: http://i.imgur.com/yiI47.png

 i had originally put 1ohm series resistors on the drain rather then the source (i mixed up p and n channel). i switched that resistor to source and added gate-drain feedback as you mentioned. 

the good news: i get the same gain regardless of the input signal's frequency. before i was getting more gain with lower frequencies.

the bad news: i am only getting ~1.5W output now. :(

I think the feedback resistors  are too low reducing the stage gain. Try changing the 100 ohm resistors to 470 ohms and see what happens. Gate Z is about 100 ohms and feedback 100 ohms  so that is heavy feedback.
Waveform looks good.

cool! that seemed to do the trick. i think i am about ready to assemble a prototype. how big of heatsinks do you think i am going to need?

i have some of these guys. will they do the trick: http://www.allelectronics.com/make-a-store/item/HS-6030/COPPER-HEATSINK-FOR-TO-220/1.html

until i can get a larger supply, i am going to test the circuit on a 12V 3A regulated supply.

No you will need a lot bigger heat sink. 2 to 3 square inches of surface area for every watt of dissipation.
Enclosed my 300W 3 stage amp (modified MRI amp)
The heat sink has 3 inch fins to dissipate up to 400 W PEP.
This is overkill so I can run class A at 100 watts
The heat sink has 750 square inches of surface area.
I don't think i will need a fan.

You can do a lot with something like this for what you are talking about. These things are all over surplus on the net and at hamfesters. Black anodized aluminum is good. Drill and tap for the TO-220's and use a good insulator and shoulder washer and chickeSH%^.

http://www.ebay.com/itm/Black-Heatsink-6x-1-3-4-x-3-5-/120364128285?pt=LH_DefaultDomain_0&hash=item1c0642d81d#ht_500wt_922

oh wow that is a huge heatsink. considering i am only planning on 50W from two IRF510s, would two of these heatsinks be okay: http://www.allelectronics.com/make-a-store/item/HS-167/T0-220-HEATSINK-W/MOUNTING-LEGS/1.html

(i live near allelectronics so its really easy for me to pickup stuff from their storefront.)

That still looks small you want about 80 square inches of surface area
the formula you want to use is 130 degrees C / Watt / Square inch

So say you have 80 square inches and the amp will be about 50% efficient so 40 watts dissipation That will be 65 degrees C temperature rise making the heat sink 100 degrees C at 100% duty cycle. SSB is around 30% AM linear would be higher.

Here is one example, 4x5 x2 inch sink and sits atop a FAN. 70W out on 20 M and key down forever. This is a bit extreme but replacing the devices dead bug style is a chore. 

this is the biggest heatsink i was able to find: http://i.imgur.com/BYcSC.jpg

its about 3"x2.5"x1" plus the fan. will it dissipate enough heat?

that might work with a fan

hmm. well in my initial experiments i will be running it off a 12V 3A power supply which will limit total power output to a level that this heatsink can handle. i saw a 28V 15A power supply at the store today that i will eventually be using but i didnt have the cash to buy it today.

i started putting the circuit together today. i am taking my time with the assembly to make sure i dont burn anything out when i turn it on. ugly style circuit construction is fast but it sucks when you have to replace a part. :(

here is a photo of the amp in its current state: http://i.imgur.com/MH3gx.jpg

tomorrow i might have time to finish it.

i found another decent sized heatsink today.

http://i.imgur.com/x20e9.jpg
this one is 4.5"x5" with 10 5"x1" fins. i guess that makes the total surface area 72.5"

the smaller one i posted earlier is 2.5"x3" with 26 2.5"x1" fins. oddly enough it also happens to have 72.5" of surface area. however the fins are much more closely spaced.

which one do you guys think i should use?

A 5 X 1 inch fin would be 10 square inches since it has two sides.
The larger deck gives you more room to play if you decide to go to 4 FETs

oh cool. in that case would these heatsinks work without fans at 50W? i dont understand the formula given earlier for heat dissipation.

Basic formula for temperature rise VS area VS power dissipation
Say your area is 1 square inch and dissipation is 1 watt, ther temperature rise in C is 130 degrees C
increase area to 10 square inches and the temperature rise drops to 13 degrees

so to restate what you said: 1 watt dissipated into 1 square inch of surface area will raise the temperature 130 degrees celsius? 

is the watts dissipated the difference between the power into the amplifier and the power out?

would a 50% efficient amplifier with 100watts input need a heatsink that can dissipate 50watts?
would such an amplifier would require 100 square inches of heatsink to reduce the temperature down to 65 degrees celsius?

do i understand all of that correctly?

Yes.  130 * (Watts / Square Inches)
Fins vertical is the most efficient heat radiator
Fins up second best
Fins down useless without a fan.

i've got everything but the bias system setup. should the bias be: "78L05->5kPot->Diode->fet Gate" or should i have an rf choke in there somewhere?

oh and, can i run both bias systems off of one 78L05 or should they have separate regulators?

I wouldn't use a choke but I did put a 10k  resistor in series with the diode in my MRI final. I only used 1 voltage regulator for the 2 drivers and 2 fianls. The diodes provide isolation from shorted FETs and a short to ground protects the bias circuit with the series resistors. (1 to 10k not a critical value)
A choke may have a resonance at some weird frequency that could cause instability

well. i finished assembling the amp but i'm not sure if its working properly. i was trying to test it with a hp3112a function generator. the most power it can put out is 3.5VRMS. is this not enough power to drive the circuit?

i'm running it off a 13.8V 3A power supply.

heres some photos of the output:
http://i.imgur.com/8eCnC.jpg
input at 3.5VRMS 7MHz with the bias all the way down. i turned up the bias slightly and R9/R10 burned out (1ohm source resistors). all the resistors in my circuit are 1/2W. is that not good enough?

EDIT:
i had been using alligator clips but i just replaced them with bnc connectors and now the waveform just looks like it has nasty crossover distortion: http://i.imgur.com/41mQM.jpg

however, when i turn up the bias, the source resistors start smoking.

EDIT2: with no input signal the circuit is drawing 3.3A. is that normal or is there possibly a short somewhere?

EDIT3: i just looked at my spice model again and realized that there is several watts running through the source resistors. i guess i need to use some sort of power resistors for the source->ground shunts?

i did some more tests and now i am even more confused about whats going.

i noticed that the bias voltage would go higher as i raised the resistance on the bias pots. that seems counter intuitive to me. i am using these pots: http://www.bourns.com/data/global/pdfs/3386.pdf

i had the pots connected so that the wipers went to the 78L05 and pin3 going to the 1n914diodes (into the fet gates).
i switched them so that the wipers go to the diodes and pin 1 goes to the 78L05.
i also got rid of the source to ground 1ohm resistors because i only have 1/2W resistors and they burn out in this circuit. the fet Sources are soldered directly to ground.

when i turn on the power supply and have no input signal i see this on the scope: http://i.imgur.com/dCN4K.jpg
this only started happening after i removed the source resistors and switched the pins on the bias pots.

with an input signal and bias turned all the way down: http://i.imgur.com/pqIid.jpg

as i turn up the bias the output voltage gets up to around 7VRMS max and the waveform looks pretty bad. previously it was going up to 20V. (all this is with a 3A 13.8V power supply). sorry i did not get a photo of what this looks like because the the plastic washer on one of the IRF510s melts when i have the bias turned up. this only happens on one of the FETs but i dont want to leave it turned on like this for very long.

adjusting the bias doesnt change the output very much. it stays between 2vrms and 7vrms. the bias pot on the non-melty FET seems to have more impact on the output waveform then the pot on the melty fet.

heres some photos of the circuit board:
http://i.imgur.com/g70D0.jpg
http://i.imgur.com/0G1fb.jpg
http://i.imgur.com/pgevb.jpg

is all this happening because i removed the 1ohm source->ground resistors?

i hope i didn't dump too many problems in this thread all at once. i'm guessing whats going on is related to me running the source directly to ground with no 1ohm resistors.

in LTSpice with a 1W input signal, 5V bias, and a 28V power supply there is 22W running through the source->ground resistors the amp produces between 40W and 80W output (depending on the frequency of the input signal).

14MHz: http://i.imgur.com/04GQE.png
7Mhz: http://i.imgur.com/nA5hq.png
3.5Mhz: http://i.imgur.com/f53PK.png
1.8Mhz: http://i.imgur.com/B3F5n.png

does that sound normal and does this mean i need to use massive power resistors from source->ground? will ceramic tube type resistors work?

i dont plan to run the bias that high but i dont want blown resistors to be one twist of the bias pot away.

also, is there any danger i may have damaged the IRF510s by running 13.8V/3A through them at full 5V bias with the source shunted directly to ground?


edit: in my simulation i made a typo no my frequency input. edited post to acccount for correct values.

I would recommend putting a SPICE voltage node on one or both sources to see what the voltage across the 1 ohmer's might be.

Then you can calculate real circuit value properties for the 1 ohmers.  

Again, this is why I suggested you do a SPICE voltage sweep (in increments of 0.25 V) for the bias voltage from 1.5 to 5 V and then observe how bias affects output.  

Phil - AC0OB

phil i must have missed your earlier comment about setting a voltage sweep. i learned my lesson. one of the FETs was totally blown out. i replaced it with a new one and i put in 5W 1ohm (well actually 10% .87ohm) resistors on the sources.

the amp is behaving much better now but i am still not getting super clean waveforms. actually, above 12MHz or so i get a really nice looking waveform. anything below that and it starts to look distorted.

as an example, here is the waveform at 7MHz (6.3Vrms input into the amp): http://i.imgur.com/S6QUb.jpg
i have the bias adjusted to get maximum output (18Vrms). through bias adjustments alone i cant seem to get the waveform much cleaner then this.

again, my power supply is 13.8V/3A.

do you think i would get a cleaner output if i was able to drive the amp with a higher power input signal?

you should only need 4 to 5 volts on the gates. 18 volts you are in saturaation
Sounds like you are pushing the FETs pretty hard. You can go lower with the sourec reisitors but check simulation. 1/2 ohm seemed ok for me. Power rating is an issue I^2R.
Your bias should be set so there is 100 to 300 ma drain current without drive. If you go higher the signal will be cleaner at the expense of a lot of heat.

18Vrms is voltage across the dummyload. i'll check the drain current and setup the bias accordingly when i get a chance to run tests later today.

in LTSpice with a 12V power source i get a clean 10W sinewave output with ~4V bias.

I hope you are using thermal pads under the FETs not plastic washers. Mylar or Kaptan tape will work in a pinch. You need to transfer heat from the FET to the heat sink plastic won't work.
Source resistors provide negative feedback so help the waverform quality.
If you have a clean signal at 12 MHz and crappy at 7 your transformer inductance is too low.

the washers are on the top side of the fet. i'm using to-220 package fets. there is a mica insulator between the FETs and the heatsink.

which transformer are you talking about? the final output transformer (K3 in my screenshots)? do you think an additional turn on both primary and secondary windings would help?

The FETs should not run that hot.
Yes add a primary turn or two to get the amp to run at a lower frequency then add a secondary turn to increase output power.
You may need to do both input and output transformers.

I agree. I am using a different FET but idle current is 20-50 mA each device. Pin is 1 W and Pout at nearly 70 W after the LPF. Power gain about 18 dB. Supply V at 24-28 V at 3.5-4 A.

I added 2 turns to the primary and one to the secondary of the output transformer  (BN-43-7051) making it 4 windings on the primary and  4 windings on the secondary. this didn't seem to help. in fact now i am getting (with the same signal source and power supply) is 14VRMS output and i get what looks like a little crossover distortion at 12MHz now. i guess i should try adding windings to the input transformer as well? should i also try more windings on the output transformer?

Is this perhaps a Davemade or XForce design?

80 watts of output from 50 watts of output seems a bit distorted to me :)

40 watts = 80 percent efficiency.  80 watts would be > 100 percent.

Methinks you might be getting into voltage dropping causing your dirty output.

Are you modeling this all with a 3A power supply too, or a pretty much unlimited voltage / current source for B+?  That may show you your distortion your getting is due more to b+ rather than any design type issue.

--Shane
KD6VXI

er, sorry for the confusion.

in LTSpice i was testing using a voltage source with no current limit. the 80W output example was with a 28V voltage source. i havent figured out how to setup a current limit, is there a simple way to do that in ltspice?

in real life i am using this power supply: http://amzn.to/uYjqk2

what is 'voltage dropping'?

when you set up the voltage source open window and select source impedance. input a value of your source Z. This will make the output voltage drop when current increases. You can't input a current limit but you can monitor current and just stay below maximum.

i had a busy past few days but i finally got a chance to get back to this project.

i made a new input transformer so now the input has 14 trifilar turns and the output transformer as 4 turns on primary and 4 turns on secondary. here is what the output looks like with the same ~3.5VRMS input: http://i.imgur.com/U9Sz6.jpg


also, without any input signal i see 3.4A being drawn from the power supply. when i turn on the input signal then the current meter drops to 2.8A.  i any idea why that is happening?


WA1GFZ, is the point of introducing voltage drop into the spice model to see if i can get the spice model to produce the same problems as my real life circuit?