Hello Manfred,
Thank you very much for the quick and very informative reply.
I will follow your advice and work on eliminating the turn-on transient. There
are 56 Ohm gate resistors and a small amount of gate to drain capacitance
“negative feedback" (gate and drain wiring is twisted pair). I will look into
implementing your suggestions.
I will build a Pi type LPF to use with this amp as you suggest.
Re. the distortion with center-tap primary. Ah, of course. I knew about the 1
turn primary problem but, duh, failed to have the insight that this would be
the same with 3T. Thanks for the nice explanation. I think I was lucky not to
have blown the FET’s trying the center-tap feed.
The screenshots that I attached to my post email did not appear on the amps
website. I have uploaded them to:
http://qsl.net/v/ve7vv//Temp/Waveform at antenna.png
<http://qsl.net/v/ve7vv//Temp/Waveform%20at%20antenna.png>
http://qsl.net/v/ve7vv//Temp/Waveform at dummy load.png
<http://qsl.net/v/ve7vv//Temp/Waveform%20at%20dummy%20load.png>
73,
Roger
> On May 27, 2018, at 11:58 AM, Manfred Mornhinweg <manfred@ludens.cl> wrote:
>
> Roger,
>
>> 1) When PTT is asserted (with no RF drive), which results in a relay
>> applying the gate bias and doing the antenna change over, there is a
>> turn on transient visible on an oscilloscope. There are two short
>> bursts of oscillation (at max power) each at about 100 kHz. This is
>> over by about 1 msec after which it seems stable.
>> Should I worry about this? If so, can you suggest a cure?
>
> I would worry about every sort of instability.
>
> During bias change conditions, a FET can run through several instability
> conditions, related to the changing internal capacitances. The best I can
> suggest to cure this is to improve the overall stability of the amplifier,
> for example by heavy, broadband gate swamping. Broadband means all the way
> from DC to beyond the highest frequency at which the FET might try to
> oscillate. That's probably ten megahertz or so for the FET you are using.
> Good groundplane construction is important, very short source leads, good
> bypassing, etc. Negative feedback, directly from drain to gate through a
> resistor and a capacitor, is very useful too in aiding stability. Even a weak
> negative feedback helps.
>
>> 2) The output waveform is close to a sinewave (with no LPF) when
>> driving a 50 Ohm dummy load. However, when driving my antenna,
>> especially on 137 kHz, the waveform is distorted. > On 137 kHz the
> > antenna load is 50 Ohm and close to resistive (at 137 kHz) as best as
> > I can measure.
>
> FETs are inherently nonlinear, even more so the switching FETs, and specially
> when operated at low bias. So a distorted waveform is to be expected. If the
> distortion is moderate, you might not immediately see it on a scope, but it's
> still there.
>
> FETs, just like bipolar transistors and pentodes, perform as current sources.
> If you have a small amount of distortion in the combined drain current, and
> apply this to a load resistor, the voltage you see on your scope only has the
> same small distortion, which might be too small to notice. But your antenna,
> despite offering a clean load on the operating frequency, surely has a high
> reactive impedance on the harmonics. So the small harmonic currents cause
> large voltage drops across the antenna, at harmonic frequencies, and they are
> phase-shifted too. And that's why the voltage waveform you see looks more
> distorted with the antenna than with a resistive load, and probably changes
> with frequency (varying phase shifts between fundamental and harmonics).
>
>> I tried inserting a 1:1 balun at the output of the
>> output transformer, which had no effect.
>
> No surprise. A balun shouldn't be frequency-selective.
>
>> A LPF cleans up what comes
>> out of the filter nicely. However, the input to the filter (amp
>> output) still looks distorted in the same way. I am using a T format
>> LPF which I believe is the correct configuration for the amp output
>> stage which is (correct me if I am wrong) voltage feed b/c of the
>> center tapped choke DC connection.
>
> The amplifier would be a voltage source of RF if it runs saturated, or if it
> has heavy negative feedback. In those cases a T filter would be fine. But you
> aren't saturating it. If you don't have heavy fedback there, it's basically a
> current source, and should be better served by a PI-type filter.
>
>> The input drive to the amplifier
>> looks like a nice sinewave in all conditions.
>
> Which means that you don't have negative feedback, or at least no strong
> negative feedback. Due to the nonlinear nature of FETs, heavy negative
> feedback results in distorting the gate voltage waveform, in such a way that
> it largely compensates for the device nonlinearity, resulting in a pretty
> clean drain voltage waveform.
>
>> Do you see a problem operating the amp into a T (inductor input and
>> output) LPF with the distorted output (voltage) waveform? It seems to
>> be running reasonably cool so I don?t see an overheating problem.
>
> It shouldn't cause a big problem, but be sure to measure IMD to ascertain
> this.
>
>> I
>> would prefer to see a clean output from the amplifier and would like
>> to operate with linear modes occasionally so the distorted waveform
>> leads me worry about IMD (I have not run IMD tests).
>
> Many if not most solid state amplifiers work with very dirty internal
> waveforms. That's fine, as long as the output waveform (after the low pass
> filter) is clean enough, both regarding harmonics and IMD. In fact the most
> common amplifiers working inside nearly all our factory-made HF radios work
> essentially with square-wave current waveforms at the drains or collectors,
> and with very "funny" voltage waveforms. After the filter they achieve - just
> barely - adequate IMD and adequate to decent harmonics. This could be vastly
> improved, but hams aren't demanding better quality from the manufacturers,
> and so these go for the cheapest circuits that work "well enough".
>
>> Oscilloscope screenshots attached.
>
> Where can I see them?
>
>> BTW, I tried using a center tap on the output transformer instead of
>> the separate choke feed. The result was a badly distorted waveform
>> into the 50 Ohm dummy load on all bands.
>
> No surprise...
>
>> I don?t understand why the
>> center tap method resulted in such a different output.
>
> That happened because you have 3 turns on that primary. A center tap results
> in what optimists call one and a half turns on each side. The problem is that
> half turns are a physical absurdity. They don't exist! What you really have
> is two separate cores, with independent, uncoupled magnetic paths, and one
> turn through BOTH cores plus one turn through ONLY ONE core, on each side.
> This results in a large uncoupled inductance on each side, resulting in the
> same behavior as if you had separate feed chokes for each FET: It forces the
> total drain current (the sum of both drain currents) to be a pure DC, and
> thus forces the load current to be square wave or something close to it. But
> near the zero crossings your sine drive waveform doesn't turn on the FETs
> enough to conduct that fixed total current. As a result the drain voltage on
> both FETs shows high spikes during the signal zero crossings.
>
> 99% of commercial ham equipment shows the same problem, attenuated to various
> degrees, depending on band, by strong negative feedback, large capacitance
> from drain to ground, and sometimes by avalanche discharge in the FETs!
>
> In your VLF amplifier you can easily avoid this problem by using the separate
> feed choke, or by feeding the amplifier through the transformer's center
> point BUT using an even number of primary turns, and building the transformer
> and the wiring in such a way that the leakage inductance is low enough (which
> means very low). In a 48V 200W VLF amplifier this is really easy, instead in
> 13V 100W amplifiers that have to work through 10m or even 6m it's hard to
> achieve, and with 48V 1.5kW ones it's nearly impossible. The problem is that
> the higher the voltage, the more ferrite cross section or more turns are
> needed. This results in higher leakage inductance. But the higher the
> frequency, and the lower the drain impedance (given by supply voltage and
> output power), the lower is the maximum acceptable leakage inductance. In
> your VLF amplifier the equation solves easily, but in a big amplifier that
> runs at up to 54MHz, and has an effective drain-to-drain impedance of 3 ohm,
> meaning 0.75 ohm drain load impedance on the FET that's conducting, you would
> need Harry Potter's help to get it right.
>
> Manfred
>
> ========================
> Visit my hobby homepage!
> http://ludens.cl
> ========================
> _______________________________________________
> Amps mailing list
> Amps@contesting.com
> http://lists.contesting.com/mailman/listinfo/amps
_______________________________________________
Amps mailing list
Amps@contesting.com
http://lists.contesting.com/mailman/listinfo/amps
|
