Optimizing screen modulation level of a plate modulated transmitter

[K8DI]

So I’m building a rig that should end up in the vicinity of the data sheet values for things. What are some ways to set the screen resistance for best performance?

The stage in my case uses 4-65a tubes. It’ll run between two columns on the data sheet, with plates at 730v. I’m starting with a 12k screen resistor because the data sheet value for screen current, and screen voltage, makes that the right value to drop the right amount of voltage at the listed current. It happens that the resistor I have is adjustable.

I know we modulate the screen because if we don’t, it’s ’not linear’. My question is, in what way? What do I measure and optimize? Am I looking for a nice sine wave or triangle wave modulation envelope, or a low thd number on the demodulated audio or a clean signal on a spectrum analyzer? Or something else? 

How do you experts do it?

Ed

[W3SLK]

K8DI said:

So I’m building a rig that should end up in the vicinity of the data sheet values for things. What are some ways to set the screen resistance for best performance?

The stage in my case uses 4-65a tubes. It’ll run between two columns on the data sheet, with plates at 730v. I’m starting with a 12k screen resistor because the data sheet value for screen current, and screen voltage, makes that the right value to drop the right amount of voltage at the listed current. It happens that the resistor I have is adjustable.

I know we modulate the screen because if we don’t, it’s ’not linear’. My question is, in what way? What do I measure and optimize? Am I looking for a nice sine wave or triangle wave modulation envelope, or a low thd number on the demodulated audio or a clean signal on a spectrum analyzer? Or something else?

How do you experts do it?

Ed

Ed, Woody N4MQ posted a youtube video about that here: https://www.youtube.com/watch?v=yXZBU3OPZdc . However, Dave, W2VW has issues with that because there is no real trapezoidal output on the scope. Furthermore, I beleive Don, K4KYV agrees that watching a trapezoid pattern is the way to go. Currently, I run my MK-214D with a audio choke, (installed by the previous owner, WD2AFJ). I may try this method out on the trapezoid pattern to determine linearity.

[w9jsw]

Read chapter D5 of this document for Orr's perspective on screen voltage.

[KD1SH]

   I'm not nearly any sort of expert, but on my little HB "Hail Mary" rig, I found this to work very well:
http://www.amwindow.org/tech/htm/scrnmod.htm
http://www.amwindow.org/tech/htm/scrnmodd.htm
(the second link is to the schematic and calculations)
   I followed his calculations to the letter—or number—and I get a superb triangle waveform, though I'll admit that I haven't yet tried a trapezoid test yet. (wondering how to do that with my digital scope, since there's no direct access to the plates)

  Am I looking for a nice sine wave or triangle wave modulation envelope, or a low thd number on the demodulated audio or a clean signal on a spectrum analyzer?


  I'm thinking that all of those are intrinsically linked, waltzing to the same tune on the ballroom floor; the question is what sort of opera glasses are best for viewing them.

[KD1SH]

  Alright—yes you can do a zap-a-troid on a digital scope, but it's all ugly digi-dots. This is my "Hail Mary" transmitter at roughly 100% mod and 3khz.
  And I did some measurements, in line with this thread, of my modulated screen voltage compared to the modulated B+, with the following results:
With my signal generator's amplitude (at 3khz) set so that the audio signal imposed on the B+ was exactly double the DC value (a theoretical 100% modulation), my screen voltage, measured right at the screen terminals of the 6146, increased by a factor of 1.74, rather than the factor of 2 that one would expect if using the standard single resistor from the screen to the modulated B+. (my screen arrangement, remember, is based on this: http://www.amwindow.org/tech/htm/scrnmodd.htm).
  But even if I was using the more standard arrangement, would we really see the screen voltage increase by a factor of 2, as the modulated B+ does? Since the voltage at the screen would depend on the screen current, we'd be assuming that screen current changes linearly with plate voltage. But does it?

[K8DI]

  Alright—yes you can do a zap-a-troid on a digital scope, but it's all ugly digi-dots.
...

With my signal generator's amplitude (at 3khz) set so that the audio signal imposed on the B+ was exactly double the DC value (a theoretical 100% modulation)

I haven't messed much with trapezoids.. In some part from not definitively knowing where to get the audio from. Split the audio off somewhere before the modulator, After the modulator, or demodulated RF?

One other thing re 100% mod as twice B+
The tube will cut off a bit above zero. So 100% negative occurs before the plate swings from zero to 2x. If the modulator is linear, and we don't exceed -100%, we will never see 2x B+...
So ideally we need the output RF to be twice carrier, requiring a specific amount of nonlinear response to get there....right? 

I realize I'm splitting hairs here. My transmitters will never be that perfect and frankly, I like to finish them and use them so I'll cut corners a little to get to operating ;-)

Ed

[KD1SH]

  I really haven't spent much time with trapezoids, either; I normally rely on the usual sampled RF envelope with sine or triangle inputs to the modulator. I built the RF sampler in the picture for that purpose.
  When I did the trapezoid in my last post, I applied the RF from that sampler to the Y axis, and the audio component (scope set for AC) taken directly off the modulated B+ to the X axis.
  Besides the annoying digi-dots, I have my doubts about that digital scope rendering a truly accurate trapezoid, since when I set my scope to the X/Y mode, those signals are still the scope's digital interpretation and subject to its whims, unlike an analog scope where you're bypassing all the amplifier circuitry and applying signals directly to the deflection plates.
  I chose that 2X B+ setting just because that's the commonly accepted figure—assuming the tube is following square-law—for 100% positive modulation, but on the negative RF peaks, as you say, you're probably never going to zero.
  The learning and experimenting process is great fun, but like you say, there's a time when you've got to get the rig off the bench and on the air. Push the bird out of the nest—it's time to fly.

I haven't messed much with trapezoids.. In some part from not definitively knowing where to get the audio from. Split the audio off somewhere before the modulator, After the modulator, or demodulated RF?

One other thing re 100% mod as twice B+
The tube will cut off a bit above zero. So 100% negative occurs before the plate swings from zero to 2x. If the modulator is linear, and we don't exceed -100%, we will never see 2x B+...
So ideally we need the output RF to be twice carrier, requiring a specific amount of nonlinear response to get there....right? 

I realize I'm splitting hairs here. My transmitters will never be that perfect and frankly, I like to finish them and use them so I'll cut corners a little to get to operating ;-)

Ed

[KD1SH]

  Another zap-a-troid, this one using one of my analog scopes. Not as clean as I'd like, but I suspect that we're seeing some scope artifacts, since this scope—like a great many "modern" scopes—doesn't offer true direct access to the deflection plates; the X and Y inputs are still going through the amplifiers and attenuators and all that; we're just not using the internal time-base.

[KC2ZFA]

the ancients say the analogue scope pattern
above shows “the effect of a too long time
constant in the screen circuit in an amplifier
getting its screen voltage through a dropping
resistor, both plate and screen being modulated.”

[KD1SH]

  Indeed, that might very well be the case. After repeating the test using a couple of my other analog scopes, just to assure myself that I'm not seeing scope artifacts, I will play around a bit with the value of capacitor C2 in the schematic.
  Schematic, text, and formulas courtesy of Dean, WA1KNX, who is unfortunately a silent key. I would have very much enjoyed discussing this approach with him.
http://www.amwindow.org/tech/htm/scrnmod.htm

the ancients say the analogue scope pattern
above shows “the effect of a too long time
constant in the screen circuit in an amplifier
getting its screen voltage through a dropping
resistor, both plate and screen being modulated.”

[KC2ZFA]

C1 and any plate bypass cap will also come into the interplay
of time constants.

[KD1SH]

  Yes. I'm also suspicious of a couple of bypass caps that I recently added, at the jacks on the rear of the chassis for the unmodulated B+ out to the external modulator and the modulated B+ that comes back from the modulator. I was concerned about RF getting into the audio via those unshielded cables, but I might have inadvertently created a problem. Probably better to use RF chokes than bypass caps, in that instance. Time for some more testing.

C1 and any plate bypass cap will also come into the interplay
of time constants.

[KD6VXI]

Having a low ESR for the audio cap is necessary.  Keep that in mind.  Sometimes a back to back electrolytic (negs tied together positives for connection each side of the resistor) makes a good cap.  It's non polarized and modern day caps are really low ESR.

Go too high in value and low frequencies will impart a DC charge on the cap, especially if you use a single polarized electrolytic.  Makes the carrier pump.

That cap is fairly important.  Moreso than most give it attention for.

--Shane
WP2ASS / ex KD6VXI

[WD5JKO]

The Resistor divider off the modulated B+ seems to be causing phase shift over the audio range, 1 Khz and beyond.

When I do this test, I use a 100X AC compensated probe meant to be hooked to a scope. You can make a divider yourself as well.

This divider needs to be AC compensated like a 10X, or 100X scope probe where the big value series resistor needs a tiny capacitance across it, and across the bottom resistor is another Capacitor.

As a starting point, make the top R-C time constant the same as the bottom R-C time constant. The upper Cap needs to be high voltage, or several low voltage caps in series. Getting rid of

the lower cap, if there is one now, should improve things. Remember the scope input usually has a 47pf and 1 meg-ohm input.

Jim
Wd5JKO

[KD1SH]

  I was using a B&K 100X probe at the time, compensated using the scope's square wave calibrator, which is, as I remember, 1khz. Maybe I need to get fussier and adjust the compensation using a higher frequency square wave?
  The ratio C1/C2 was arrived at using Dean's formula, but I neglected to account for the additional bypass capacitor I installed at the feed-through insulator in my copper-clad divider. It's "upstream" of resistor R2—on the unmodulated side of the mod transformer—but it still represents a time constant in conjunction with R2.
  The scope shots I took, with sine and triangle, were done before I added the bypass cap at the jacks for the external modulator, so they might be entering the picture, too.

The Resistor divider off the modulated B+ seems to be causing phase shift over the audio range, 1 Khz and beyond.

When I do this test, I use a 100X AC compensated probe meant to be hooked to a scope. You can make a divider yourself as well.

This divider needs to be AC compensated like a 10X, or 100X scope probe where the big value series resistor needs a tiny capacitance across it, and across the bottom resistor is another Capacitor.

As a starting point, make the top R-C time constant the same as the bottom R-C time constant. The upper Cap needs to be high voltage, or several low voltage caps in series. Getting rid of

the lower cap, if there is one now, should improve things. Remember the scope input usually has a 47pf and 1 meg-ohm input.

Jim
Wd5JKO