[AMPS] Network analysis of suppressors

[Tom Rauch]

Hi Steve,

> > Can you measure S21, and have you measured it?
> 
> S21 for this network is 46 + j74 at 50 MHz.

We must have attended different engineering schools Steve. I 
thought S21 was the reverse direction gain or loss. I'll dust off the 
books and look again, just to be sure.

> > What is a "nice characteristic"? What was the goal?
 
> The goal was to dissipate about 1 to 2  watts per suppressor at 29 MHz.

I can do that easily with a conventional parallel R/L suppressor. 
How does that tie into the nichrome, the plate choke, and all the 
rest? 

> 572B tubes have significant feedback capacitance and this particular amp
> arced and sparked with the original suppressors (6 turns of bus wire
> concentric with a 2 watt 100 ohm carbon resistor). 

I'd bet it would. Clippertons have two sore spots, one between 20 
and 35 MHz and one in the lower FM BC band.

The lower one is caused by high feedthrough in the parallel 572's, 
the upper caused by a parallel grid resonance in the tube.

The cure for the low one is neutralization, the cure for the upper 
one is a purely resistive low impedance at the anode near the low 
end of the FM BC band. 

I don't understand then what 1.5 GHz has to do with 29 MHz, and 
how the choke and everything else enters this equation.

 You could also tell
> that original suppressors were at one time very hot as the coating on the
> resistors had bubbled.  It was time for some new ones.  The network
> analyzer showed that the new network behaved as an inductor in parallel
> with a resistor all the way past 1 Gigahertz and the impedance was
> consistent with my dissipation goal.  This is what I meant by "nice
> characteristic".

I always quit looking at about 200 MHz, because a 572B has zero 
dB available gain at that frequency. Besides that, anything you do 
OUTSIDE the tube envelope has little effect on the internal 
impedances since the leads in that tube are so long and skinny, 
and since the capacitances are so high. 
 
> > What frequency did you determine was involved in the problem you
> > saw? How did you determine the frequency of the problem?
> >
> > > The construction technique was also critical.  Wrapping the wire
> > > around the body of the resistance added capacitance lowering the
> > > resonance frequency. Winding spacing also changed the resonance.
> >
> > What frequency are you talking about?
> 
> This network behaved as a resistor in parallel with an inductor all the
> way to 1.5 Gig at which point the intrinsic capacitance of the resistor
> and the winding resonating with the inductance.  The frequency that I
> refer to is the self resonance of the circuit.  One of my goals was to
> move this self resonance as high as possible.

Why? Resonance, by itself, has little to do with anything. It is the 
impedance that matters.

A resonance at the correct frequency range with the right 
impedance can actually aid stability.

> > > Differences between the two types of wire were only evident at very
> > > high frequencies where it lowered the Q of the self resonance and very
> > > low frequencies where the network behaved as a pure resistor.
> >
> > 1.5 GHz is ultra high frequency. Why are you concerned about the
> > system at that frequency?
> 
> The network would have probably worked fine even if it resonating at 500
> MHz.  My natural tendency is to optimize.  I was just noting that with
> proper construction, one could build a network that behaved as intended to
> very high frequencies.

If you measure the impedance seen by the anode, it won't be 
anywhere similar to the impedance of the suppressor. You spent a 
lot of time making a flat suppressor when the real goal is a certain 
load at the frequency where feedback is high.  
 
> > What was the impedance these "problems" presented right at the
> > anode of the tubes? How much did you change the anode
> > impedance with you mods, as measured right at the anode?
> >
> > Why is resonance bad? Can't it be good also?
> 
> The anode output impedance at the operating points is about 6K ohms per
> tube.  Even at 29 MHz, the network added no more than 26 + j50 to the
> anode impedance.  However, the small change was enough to operate this amp
> on the 10 meter band.  Something that I couldn't do before.

Sure, you loaded the tank with high losses at 29 MHz. That's a 
band-aid for the real problem you never addressed, the high 
feedthrough capacitance of four parallel 572B's.

Nothing wrong with that, as long as you keep that Clipperton 
loaded heavily so the small ESR you added has enough loss to 
push the thing over on the stable side of the gain curve.
. 
I just want to be sure we don't confuse 30 MHz oscillations with 
oscillations at 100 MHz, and understand what the high loss at 
lower frequencies really did. 

> > What do harmonics have to do with parasitics?
> >
> > If any oscillator has more regeneration than loss, it will oscillate. If
> > it doesn't, it won't. External excitation has NOTHING to do with this,
> > since even random shot or thermal noise in the circuit will start the
> > oscillation.
> >
> > You have an oscillator, or you don't.
> 
> Circuits designed with poles on the Y axis will not intrinsically
> oscillate but will ring for long periods of time if excited at their
> resonance frequency.

So what. Ringing, by definition, never gets higher than the initial 
excitation. 

 We use this type of circuit for high frequency
> (Gigahertz) clock recovery and other related problems.  It is very common
> for these circuits to be designed to ring at a harmonic of the excitation
> frequency.  It works kind of like a poor man's phase locked loop.

Sure, and they never explode.
 
> > Now certainly gain has something to do with a parasitic, but it also
> > requires feedback of the correct phase and the correct impedance at the
> > proper spots are critical to this.
> >
> > What did you do to define the problem? That would be valuable in
> > understanding if the cure was logical...or just an exercise in looking
> > at one narrow parameter that has little to do with the overall problem.
> 
> This type of problem is very commonly encountered in the designs of high
> frequency switching power supplies.  The fundamental of the switch may be
> 1 Megahertz but the ringing can be as high as 100 Megahertz and is the
> result harmonic excitation.  Since there is insufficient load at these
> high frequencies due to component inductance, the ringing manifests as
> radiated RF.  Something the FCC frowns upon.  Careful attention to design
> and the addition of lossy networks reduces this activity to acceptable
> levels.

Resistors make poor harmonic filters, plus this PA is not operated 
even remotely close to the conduction angle of a switching supply. 

Layout and shielding has many-fold more importance than simply 
adding a tiny bit or resistance in series with a circuit area that 
already has an impedance in the hundreds or thousands of ohms.
 
> To be honest, I did not go to the trouble of predicting why this amplifier
> misbehaved at 29 Meg.

I have. It may be old-fashioned, but I like to know what I am fixing 
before I fix it. Especially if I offer it as a cure for others to copy.

Now there's nothing wrong with what you did, but it is always nice 
to know what you need to do before starting on the cure. 

> > I can make any stock Clipperton, or any Clipperton blessed with
> > nichrome, oscillate by misterminating the input and/or output, and
> > adjusting the controls "properly" provided nothing is done to
> > address the HF feedthrough problems.
> 
> Maybe so.  The additional 1 ohm of resistance dissipates less than 1 watt
> at 160 meters where the resistance wire is the dominant characteristic of
> the network.

I'm sorry, I disagree with that. On 160 meters, just like on ten 
meters, the tube impedance and tank impedance dominates the 
system...not the resistance wire. The resistance wire is a tiny little 
change that just barely makes a difference.

If you look at N7WS's web page, you'll find details on what he 
measured. It agrees with what I measured, and what others have 
measured.

Anything you can do with nichrome, I can do with conventional 
components....except make good tasting toast. Nichrome is much 
better for that.

Now there's nothing wrong with nichrome used as a resistor, but 
there's nothing magic either.


73, Tom W8JI
w8ji@contesting.com

--
FAQ on WWW:               http://www.contesting.com/ampsfaq.html
Submissions:              amps@contesting.com
Administrative requests:  amps-REQUEST@contesting.com
Problems:                 owner-amps@contesting.com
Search:                   http://www.contesting.com/km9p/search.htm