Thank you very much Manfred.
Your notes have wonderful technical information. I have learned a lot from this
correspondence. You have provided an excellent design tutorial!
I will lower the series gate resistors (I chose the 56 Ohm b/c I had used that
on my LF/MF mod of the 1W SoftRock RXTX, w/o looking into what might be the
correct value for the big mosfets).
Thanks in particular for clarifying the gate resistance issues, series and
load. I will improve my gate circuit in accord with your guidelines.
Thanks also for the suggestions for alternative mosfets. I will check out their
datasheets.
I will let you know which changes lead to improved performance.
73,
Roger
> On May 28, 2018, at 10:05 AM, Manfred Mornhinweg <manfred@ludens.cl> wrote:
>
> Roger,
>
>> I will try lower value gate resistors. That will also have the
>> advantage that I will be able to drive to higher output on 160 I
>> think. (My 1W drive is not sufficient on 160.)
>
> Ops... We have a misunderstanding there. I was talking about gate LOAD
> resistors - the ones that connect from the gates to RF ground. Now I realize
> that you are talking about resistors connected in series with the gates.
>
> With such big FETs I would use resistors of roughly 10-20 ohm from each gate
> to ground (through a larguish bypass capacitor), making sure that the bias
> supply has an impedance not over a few hundred ohm and that it is coupled to
> the gates in such a way that the bias supply's internal resistance loads the
> gates fom DC up to beyond the frequency where the reactance of the mentioned
> bypass caps has become irrelevant. This gives you gate loading at a few
> hundred ohm from DC to some low frequency, and of 10 to 20 ohm from there on
> and into the VHF range.
>
> The gate series resistance for such a big FET is typically just 1 to 3 ohm.
> 50 ohm gate series resistance is what I use for 6 watt FETs (like the RD06)
> in the HF range. At your low frequency the resistance can be higher than at
> HF, but what you have, for your FETs having huge capacitances, is far too
> high.
>
>> I can substitute braid for the wire I used for the gate and drain
>> leads for lower inductance.
>
> That would be an improvement. Anyway I wonder how much effect it's causing,
> given that you are running this at relatively high drain impedance and very
> low frequencies. Still, just to be on the safe side, I would still use RF
> construction techniques, with wiring lengths approaching zero as much as
> possible.
>
>> Yes, as far as I could tell, the 5023 is just a 5022 with 0.23 Ohm
>> Rds(on) instead of 0.22.
>
> OK. I have a bag full of APT5020, which I got for free. Given their extremely
> high capacitances, I'm using them in DC applications only! Definitely not at
> RF. Even in switching applications that need to run at 50kHz or so, I prefer
> buying more modern, better suited FETs, than using those 5020.
>
>> I will try adding the negative feedback if other changes do not
>> eliminate the transient.
>
> OK. It also aids linearity. As soon as you find you have any spare gain, you
> should use it up in negative feedback, instead of an attenuator.
>
>> The APT502X series is obsolete and hard to find. Can you suggest a
>> possibly better replacement that can work on 48V and provide 200W? I
>> found the relatively low transconductance and high power dissipation
>> ratings of the APT units attractive.
>
> Have a look at the IXFQ20N50P3, AOK42S60L, IXFQ30N60X, and IXFH26N50P3. They
> are all current and available at Digikey and other distributors. Each of them
> replaces the APT5023 with considerable advantages, either in most or in all
> areas. Compare the output and reverse capacitances of these to the APT5023's
> ones...
>
> Any of these will provide far more gain.
>
> All four are rated at 500-600V, which is really a lot of overkill for a 48V
> supply. But in the optimal range, say, 150-200V, there isn't much to choose
> from. Those lower voltage, high power FETs have extremely high current and
> low RDSon ratings, which brings along extreme transconductance and very high
> capacitances. So it's probably better to stay with the 500-600V FETs. These
> should also be more resistant to hotspotting in linear operation.
>
> > The power output increases
>> linearly with input up to the 1dB compression point at close to 200W.
>
> In fact with 48V and a 3:6 turns output trafo you should be getting closer to
> 300W at the -1dB point.
>
>> I have not used any compensation capacitance across the output
>> transformer primary. The measured Z of the transformer was better on
>> 160 with 500pF across it, but I was concerned the 500V silver micas
>> (two 249 pF in parallel) might not survive. Do you think C across
>> either the primary or secondary would be beneficial for either the
>> transient oscillation or the harmonic distortion?
>
> Probably not.
>
> A properly compensated transformer has a pretty flat response up to a certain
> frequency, then falls off rapidly. Without compensation it will start falling
> off at a lower frequency, but less abruptly. I would think that when
> operating with 2MHz as the top frequency, compensation won't be needed. But
> it really depends on the details of transformer construction. If you find
> that your transformer works well on the VLF bands but struggles on 160m,
> consider adding properly calculated compensation capacitors. The measurable
> symptom of a transformer that's not able to cope with the frequency is that
> the secondary voltage no longer is in the correct ratio to the primary
> voltage (bad coupling factor). In that case one would add the correct amount
> of compensation capacitance ON BOTH WINDINGS, thus absorbing the unwanted
> leakage inductance, that causes the poor coupling, into a PI-type low pass
> section, whose cutoff frequency hopefully ends up above the highest frequency
> of operation.
That's all there is about compensating a transformer. It's not directly
related to linearity, although the frequency characteristics of the transformer
do affect the harmonic structure of the signal.
>
> Since compensation capacitors will shift existing resonances, stability can
> change. But it could change for the better or for the worse. Typically for
> the worse, though, given the fact that adding reactance tendsa to raise the Q
> of those resonances.
>
> Manfred
>
> ========================
> Visit my hobby homepage!
> http://ludens.cl
> ========================
_______________________________________________
Amps mailing list
Amps@contesting.com
http://lists.contesting.com/mailman/listinfo/amps
|