FET linear RF amp.

[Wendymott]

Hi peeps.....
I have been attempting to get one of these darned FET RF LINEAR amps to work properly. I attach the schematic and 4 scope visuals to highlight my mis understanding...
Firstly its for "Topband" and is intended as a AM RF PA...... the HT/DC to the output pair will be by a series pass transistor assy to AM Modulate.
I can squeeze 5W of RF into 50R via a Low pass filter and matching tuning unit.
However I do not understand the following.. The RF drive to each gate is as the photo's show...... thus both fets are not symetrical. I have checked every component ... both transformers.. re wound both transformers in case one leg insulation was iffy.... changed the feedback caps from SMD ceramic to leaded minature polys..
I set the gate bias to 30 m/A per fet, but I noticed when running that...
1. when powered up.. the RF increased from a miserly 3W to 5W over a period of about 2 minutes.... and if I disconnected the RF input the DC Bias to the output pair had risen to 200 m/A.
The heatsink is common to the driver and the output pair, with a small fan giving it some help.
Am I expecting too much ... why would the waveforms be not symetrical.. is it that the circuit has a lower frequency limit ?
I increased the value of the RF chokes in the bias supply in case the slider decoupling caps were getting too much RF across them, but the reactance of the chokes was 2.5K.... and no real change if I increased to 1mh = 11.3k
The FETS are now IRF530's.
I am now getting wayyyy past my understanding and throw myself at your mercy.

[Terry_VK5TM]

Don't know why I didn't see this before, but - why are the right hand ends of R12 & R13 grounded?

Effectively grounding (for RF) the gates of the IRF530's.

And the reason the bias has increased is there is no temperature compensation in the bias network, have a look at the RD16 circuit we looked at - the two diodes in the common leg of the 5v reg.

[Wendymott]

Hyas Terry........ I did wonder yes everything is grounded....... I did omit the bias temp compensation....as in all previous builds it seemed un necessary...... however.. I didnt pay much attention to measurements then...... I just set the bias and thought that was it..... obviously not... I can rejig the pcb to incorporate the compensation diodes...... but I still cannot get my head around the inequality of the drain waveforms.. or even the drive for that matter.... when everything is equal... as in identical components with a short run to the gates.... I suspect the fets characteristics may have something to do with it... but it always seems to be Fet 2 that is lacking.. I have put it to one side for the moment and concentrate on the PIC control pcb...

[Radio Wrangler]

Firstly, assuming that erroneous ground isn't there, the drive transformer drives the input of one FET against the input of the other. If one is lower capacitance, it will hog the drive voltage swing. If one FET has higher Gm, it will give more gain.

You'd also think that the output transformer, T2 would try to balance the output voltages, see-saw style, but that's not so because of the presence of choke L4.

Balancing transformer action would need the centre tap to be decoupled to ground. L4 floats it and allows common mode stuff on the drains.

This circuit needs things to be very well balanced between the two halves. If it isn't you'll see different waveforms.

T1 primary top end is decoupled, lower end has signal. If the secondary is simply bifilar, one end gets capacitive couple to hot end of primary, the other end of the secondary gets capacitive couple to cold end of primary. So the transformer is implicitly unbalanced. No wonder you get asymmetry?

My trick is to use trifilar. Make the primary have the same number of turns and connections, but to have the borrom end of the new winding go to the cold end of the primary (top in your diagram) so you have what looks like a balanced primary winding pair. THe hot end of the new winding goes nowhere. It looks superfluous it looks silly, but the inverted signal voltage at the hot end of the new winding completes the capacitive coupling to make the transformer output better balanced. It looks like a winding to nowhere, but it is really a winding to the internal inter-winding capacitance of the transformer! Doesn't half fool people.

David

I will now chuck a spanner in the works, Top Band, you could do that with PWM, lots of SMPS chips run at MHz and can be synced to a clock. Just put audio in the control loop.

[G6Tanuki]

I think the 'design error' here is that the centre-tap of the two 22-Ohm resistors should be grounded, and the gates of the output-transistors should be fed from the two 'hot' ends of those resistors (the ends attached to the secondary of the driver-traqnsformer.)

That way the two resistors would form a resisitive centre-tap across the winding, and that will allow some semblance of push-pullness to occur.

It's a bit of a kludge though - much better to make the secondary winding bifilar and ground the centre-tap - thta way you do at least have a chance of somewhat-symmetrical drive.

[Wendymott]

Ok David... I see what you are meaning.....So it seems obvious that all the FET P/A's on the market are way off. I will go back to it.
Tanuki... I did think about centre tapping the secondary and putting the R's in series...but I decided to ask the question first.
Merlin.... I really wanted a traditional TX/RX, Using a SSD to generate the TX drive for AM and SSB......and I.F offset for the RX.
Thanks guys for your inputs

[Terry_VK5TM]

Just looking around at the various push-pull IRF510 amps on the net, the most common circuit is either as per the pic below or with two seperate secondaries.

Then there is the 'current balun' type drive where one winding is in series with the drive to one gate and the other winding is connected between ground and the second gate.

[Wendymott]

Hi Terry..... Thanks for the ideas. It wont be too difficult to implement . Just rebuilding the PIC controller at the moment.... I cannot think why I dismantled the 80M control panel....but it was before the 160M idea formed. Dohhhh

[Terry_VK5TM]

Per merlinmawell's post, you might want to have a look at this - so much simpler https://people.physics.anu.edu.au/~dxt103/160m/

You can replace the oscillator module with the squarewave output of the DDS.

Also, push-pull IRF510/530 amps are a world of pain, a single ended design will give you just as much RF out and are much simpler to set up if you really want to go that way.

[Wendymott]

Hi Terry.... I have had a look ...... I see what you mean..... Sorry Merlin....I Mis understood the concept.... looks like a busy weekend Still.. it is sleeting outside..dark as .... at 10am..... this "rock n roll" Life