40MHz Amplifier.

What gets hot?

Did you simulate it in SPICE and with the proper input and output capacitances of the FET? Is the current intake commensurate with the output power or is it way high?

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

t

Nothing gets hot as such, heatsink get's moderately warm but no more than i'd expect.

I've simulated the input and output matching networks, but not sure how to go about simulating the actual FET, I guess I need to edit a generic FET model to match the parameters of the BLF177.

The current seems high for the output, probably around 1A @50V for 10W into the load, also I've noticed that the input begins to distort as power is increased, is this usual? One thing I haven't considered is the RF choke, I just put a 500nH air cored inductor in, how critical is the actual value, is there a rule how to calculate it?

Cheers,

Jim.

looking in the datasheet the Idq for 28MHz is set for .7A and for

108MHz is set for .1A

could it be that you should pick something in between for 40MHz ?

-Lasse

Do you have a hybrid or directional coupler that lets you see what the input VSWR is at the power level you are using?

Jim

I?ve built a 40MHz amplifier based on a BLF177 RF FET. The circuit is similar to the 108MHz class B example presented in the datasheet, except I have recalculated the input/output matching networks for the lower frequency. The amplifier works to a point, but performance is poor. With Vds=50V Idq=0.1A and 1W input I get about 10W of output power, this is far lower than the 20dB gain I?m looking for. I can?t see what the problem is. I?m pretty sure the matching networks are correct, the values are not too far away from the those given in the

28MHz class AB example also in the datasheet. Can anyone take a shot in the dark as to what the problem may be?

Cheers,

Jim.

That sounds like an input matching issue then. But, further down in your post it looks very different ...

If you have a low-C FET probe check what gets to the gate but this may not be your problem.

This could help:

With FETs usually not. This could be an indication that it oscillates or something. Hard to say from the distance though.

Only 20% efficiency? Ouch, something is wrong here. If you mean the choke to V+ that looks ok. I usually go a bit higher in impedance though, like 10x the load. But make sure it doesn't resonate.

If 40W of dissipation doesn't get toasty you must have one hellacious monster-heatsink there.

One test you can do is move the exitation frequency up and down to see if the efficiency (RF output power versus DC input power) gets better, and whether it peaks somewhere. If so then your output matching network is out of tune for 40MHz.

Just a hint: Such RF amplifiers require super-clean and short ground connections. Don't rely on pigtails, lead-ins and such, those do not work at RF. A few hundred milli-ohms here and there and the whole enchilada goes berserk.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Nope, that is because 0.7A is for class AB operation, for single-sideband transmissions where there must be no crossover distortion or you'd get into serious tussles with the folks using neighboring channels. Or with the Federales, from the FCC.

0.1A is for class B operation, for CW, FM and other modes that do not require linearity. If Jim needs a class B amp then 0.1A is going to be about right.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

is

output

can=92t

The reasoning looks sound, though at 40 MHz either SSB or FM may be used.

is

t

=92t

Well, at the moment the output is straight into the dummy load, but if I ever get it working then it's to be used to generate plasma, not transmitting. Hence I'm not too bothered about output distortion, as long as the output looks sinusoidal on the scope it will suffice. Class B should be ideal, maybe even class C...

The layout should be reasonable, I've used a dremel rotary tool to cut the tracks on one side of a double sided Cu FR4 board. Plenty of via's to the ground plane underneath and the FET is bolted to a BIG Al block. I have a power/SWR meter connected between the amp and load, there is never any reflected power despite chaning the variable caps in the output network, this seems odd.

t is

he

put

n=92t

e e t

how about class-e ?

-Lasse

uit is

the

is

utput

can=92t

are

the

hot

t

be

Class-e is great, I've made a few 100-300W amplifiers using cheap switching mosfets in the low MHz range. At 40MHz i'm not sure what devices you could use, perhaps those IXYS 500V VHF Fets, I imagine driving them would be an issue. Another drawback is the amplitude of the output is related to supply voltage not the input signal, this means I'd need a variable input voltage to get a variable output, more complexity! This isn't an issue in class-b, the output is proportional to the input.

Same for class C. Good old plate modulation. ;-)

If the modulation is audio frequency, just put a class D amp in series. If not even that, heck, a buck converter will suffice.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

They only give examples, probably preferenced by market size. That's one problem with RF transistor datasheets. Very scant information in there, often no SPICE models, and then 2-3 "typical" applications in cookbook format with part numbers and all. Doesn't help guys like Jim or me who often do more unorthodox designs.

For that class C should suffice. But is the power level going to be enough? Must be a small plasma :-)

Is the contact between ground plane and Al block ok? Or riding on "wire vias"?

If the SWR meter is calibrated at 50ohms and you have a good 50ohm dummy load connected you may not see much. If you mess with capacitances or inductances around the dummy load area you should see something. But be careful not to blow the FET, they aren't exactly cheap. That's also a concern with a plasma because that kind of load can show nasty impedance jumps.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

If you're feeding a purely resistive dummy load what would you expect ?

--
Best Regards:
                     Baron.

I've killed a few mrf150's doing that ! The swr can jump to 10:1 or more in an instant. Certainly a lot faster than you can servo the output network tuning. Tubes are far more tolerant in that service.

--
Best Regards:
                     Baron.

[...]

How did you finally tame the MRF150 amp? As you wrote, servoes aren't going to cut it here. All you've got is fast diodes and maybe another RF transistor to throttle things down really fast. Some of the old Motorola transistors had sort of a "graceful" failure mode where some portions of the emitter area would blow but not all of it. I've used one of those in a 2m ham rig for more than 10 years. Instead of the usual 10W it only did 6W but very reliably.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

circuit is

datasheet,

for the

performance is

of output

I can=92t

networks are

in the

a shot

for

?

not

to be

be

if

Gosh. Now why does class e sound so hauntingly like what i worked with previously called class s.

=A0If

If you go this route beware of beats between the bucker's switching frequency and the 40MHz. Syncing the bucker to a subharmonic of the

40MHz may be needed if you don't want to have low frequency modulations on the output.

Why would you worry about the VSWR of a dummy load? If it isn't unity, you've got a real problem.

More better you use the meter at the INPUT of the amplifier.

And please snip for those of us who are on limited bandwidth.

Jim

I have a power/SWR meter connected between the amp and load, there is never any reflected power despite chaning the variable caps in the output network, this seems odd.