Screen volta6H6ge - JJ 6V6

[nuke]

The best setup for screens is to stick them to a nice stiff power node that is at or below the rated G2 voltage and definitely below the plate voltage. That last bit is important.

That's hard to do in practice in the typical guitar amp power supply, where you get plate supply, then a choke or a resistor to drop it for the screens (not by much usually) then the rest of the supply nodes on a resistor divider.

For grins, look up the Silvertone 1484 schematic. The voltage multiplier power supply in this amp provides a well-regulated screen supply well below the plate voltage. I just restored one of these amps and it sounds pretty darn good, and it operates the power tubes in a conservative manner that should result in long tube life.

The screen grid functions as a "virtual anode", and the plate current through the load becomes largely independent of the plate voltage as control grid voltage becomes more positive. The screen current also changes, with signal, thus the screen supply needs a low source impedance, so the screen voltage stays steady.

The common catastrophic failure in power tubes is melting the screen grid. That happens when the plate dips below the screen supply voltage. If the screen grid becomes the most positive voltage element in the tube, it will conduct a lot of current and the fragile screen wire will melt very quickly. As the screen's voltage gets relatively closer to the plate voltage, screen current increases, the screen wire will get red hot and generally it will deform and degrade the tube, then fail.

There is a difference between power pentodes like the EL34 and EL84, vs the beam-power tubes like the 6L6GC and the 6V6GT. The beam power tubes are constructed with the screen grid wire windings carefully aligned "in the shadow" of the control grid. That's often not the case in power pentodes, like the EL34 and EL84.

In practical terms, the "western" data sheets for 6V6GT all seem to have the same data from lates 30's through 1950's and are pretty tame. The tubes seemed to get better with time and more able to handle high voltages in reality, since they were used also as vertical deflection tubes in television with higher peak voltages.

Some of the sketchy data from JJ and others of modern 6V6GT indicate they can handle an absolute maximum of 450v plate and 400 screen, 14W plate and 2.2 W screen.

That would seem like good advice as an absolute limit if you want long life, design a little more conservatively than the max. If you can setup the power supply for 5 watts of available screen current at a suitable voltage below the plate supply, should be in good shape.

[B Ingram]

... G2 voltage ... definitely below the plate voltage. That last bit is important.

...

The screen grid functions as a "virtual anode", and the plate current through the load becomes largely independent of the plate voltage ...

The common catastrophic failure in power tubes is melting the screen grid. That happens when the plate dips below the screen supply voltage. If the screen grid becomes the most positive voltage element in the tube, it will conduct a lot of current and the fragile screen wire will melt very quickly. ...

Consider an output tube where the screen is connected to a power supply node as you mentioned. Let's imagine the screen voltage stays unchanged all the time.

Half of every signal-cycle, G1 is driven in a positive-going direction, and plate current is increasing. That plate current creates a voltage-drop across the output transformer's primary impedance, and momentarily reduces the plate voltage. Plate voltage is reduced further during each of these half-cycles until the amp is delivering its maximum clean output power (where the plate voltage is momentarily very much lower than screen voltage). The tube screens aren't melting in most amps during this, so what gives?

There are huge gaps between each turn of the screen grid, so electrons accelerated in that direction by the screen voltage will mostly just "miss" and travel onward to the plate.

The only time this becomes an issue is when plate voltage is extremely low (like 75v for the 6L6GC) while screen voltage is quite high (like 400v for our 6L6GC example).

• The attached extract of the 6L6GC data sheet shows that when plate voltage gets below the "knee of the curve" (or what is also called the "diode line" of plate current vs grid voltage), screen current skyrockets as plate current falls away.

• The Red line in the attached image is drawn at 75v on the plate. Screen Current is drawn as a dashed-line for a given G1 voltage (G2 is being held at 400v).

• The Blue Oval area is where plate voltage is finally low enough to cause screen current to spike up sharply.

• The example shows that if the loadline for the output tube "stays above the knee of the curves" that screen current won't be excessive.
  (The dashed lines for "Ec1 = +10" and "Ec1 = +20" are irrelevant, because typical guitar amp circuits cannot push an output tube grid positive of its cathode.)

The point here is that under normal operation (with the plate at/above around 100 volts), we need not be concern about excessive screen current just because the screen voltage is higher than plate voltage.

High Screen Current.png

[nuke]

Yup, I'm aware of the 6L6GC, and one of its design features and improvements over the 6L6GB and earlier versions was much more robust screen performance.

And as I mentioned, the beam-power tubes are generally designed with the screen grid wires wound "in the shadow" of the control grid, quite intentionally. The shadowing construction is less true of power pentodes, like the EL34 and EL84.

If we take the 6L6GC in particular, the maximum screen dissipation is listed as 5-watts. If the screen dissipates more power than that, it will heat excessively and begin to deform.

If you plot the 5-watt (400V x 12.5ma) screen current line on that graph, most of those screen currents are well above the screen dissipation rating. We don't need to reach the sharp spike along the left side of the graph, as we can see the screen current starting to inflect up as the plate voltage decreases. (although I'm not certain about this graph, since the super-imposed Ic2 lines are depicted against Ec1?)

Here's a graph of a very relevant 6L6GC power amplifier from the Tung-Sol data sheet:

6L6GC-amplifier-dissipations.jpg

What's interesting here is the point where power output is increasing, while plate dissipation is decreasing, implying that AB1 operation has shifted toward the class-B area and one tube is approaching cut off and the other is in the high-current region, which means the plate voltage has dropped considerably across the load presented by the output transformer, the screen dissipation sharply increases.

The operating points in this graph are very similar to many guitar amp power stages built on 6L6GC pairs. You can see how close this is to the screen limit.

If we step down to the 6L6GB, the screen is only rated for 2.8 watts, or the 5881 claim of 3 watts at much lower maximum voltages, then you can get into trouble a lot easier.

From practical experience, the most common power tube failure I've seen is screen failure accompanied by burned screen resistor.

I tend to believe the screen suffers abuse, deforms from heat, then is subject to suffering more abuse since it is no longer precisely aligned.

Consider an output tube where the screen is connected to a power supply node as you mentioned. Let's imagine the screen voltage stays unchanged all the time.

Half of every signal-cycle, G1 is driven in a positive-going direction, and plate current is increasing. That plate current creates a voltage-drop across the output transformer's primary impedance, and momentarily reduces the plate voltage. Plate voltage is reduced further during each of these half-cycles until the amp is delivering its maximum clean output power (where the plate voltage is momentarily very much lower than screen voltage). The tube screens aren't melting in most amps during this, so what gives?

There are huge gaps between each turn of the screen grid, so electrons accelerated in that direction by the screen voltage will mostly just "miss" and travel onward to the plate.

The only time this becomes an issue is when plate voltage is extremely low (like 75v for the 6L6GC) while screen voltage is quite high (like 400v for our 6L6GC example).

• The attached extract of the 6L6GC data sheet shows that when plate voltage gets below the "knee of the curve" (or what is also called the "diode line" of plate current vs grid voltage), screen current skyrockets as plate current falls away.

• The Red line in the attached image is drawn at 75v on the plate. Screen Current is drawn as a dashed-line for a given G1 voltage (G2 is being held at 400v).

• The Blue Oval area is where plate voltage is finally low enough to cause screen current to spike up sharply.

• The example shows that if the loadline for the output tube "stays above the knee of the curves" that screen current won't be excessive.
  (The dashed lines for "Ec1 = +10" and "Ec1 = +20" are irrelevant, because typical guitar amp circuits cannot push an output tube grid positive of its cathode.)

The point here is that under normal operation (with the plate at/above around 100 volts), we need not be concern about excessive screen current just because the screen voltage is higher than plate voltage.

[LOUDthud]

Question about the 6L6GC graph posted by B Ingram in the second reply on page 2. The graph says the screen Voltage is 400V. The screen current lines use the scale on the right side of the graph. I'm wondering if the decimal points have disappeared from the numbers. 25mA at 400V would be 10 Watts, well past the screen's dissipation rating. At the top of the graph 100mA at 400V would be 40 Watts. It would make more sense if instead of 25mA it was 2.5mA and dissipation was 1 Watt.

Looking at the RCA data sheet for a 6L6GC push pull class AB1 amp with 450V Plate supply and 400V screen supply, the idle screen current for both tubes is 5.6mA. 400V at 5.6mA would be 2.4 Watts for both tubes.

[martin manning]

Here is how I have it at 400V Va and Vg2, 4k load line for two tubes.
edit: Corrected screen power trace and max to reflect constant 400V.

[B Ingram]

Yup, I'm aware of the 6L6GC, and one of its design features and improvements over the 6L6GB and earlier versions was ...

You're missing a very basic fact: for all pentode & beam power tubes, Plate Voltage is lower than Screen Voltage half the time. Unless we never apply a signal.

Question about the 6L6GC graph posted by B Ingram in the second reply on page 2. The graph says the screen Voltage is 400V. The screen current lines use the scale on the right side of the graph. I'm wondering if the decimal points have disappeared from the numbers. 25mA at 400V would be 10 Watts, well past the screen's dissipation rating. ...

Here is how I have it at 400V Va and Vg2, 4k load line for two tubes.

Martin has the right idea.

Assume a Class AB push-pull pair

• For most of the signal cycle, screen current is <3mA. Some large part of this time, the tube's plate current is cut off.

• The Screen Dissipation is the average dissipation during all this time.

• Pretend Screen Volts is constant (though it would probably sag due to a series screen resistor).

• The actual Screen Dissipation is then proportional to the average screen current, not the instantaneous "blip" of high screen current that happens during the instantaneous blip of peak plate current.

How would you calculate this "average screen current"?

• Lay out a plate load line & decide on an idle operating point (Martin chose 400v plate, 400v screen, -40v G1)

• Note the G1 voltage-change implied by the idle bias: Max Clean Power happens when G1 is driven to 0v, idle is at -40v, so Signal Input for full power will be 40v Peak AC.

• Break up the signal cycle into equal-angle intervals (say, "every 10 degrees") and make a note of the Screen Current that happens at each interval.

• "10-degree intervals" would give us 37 values from 0º to 360º, and we would average those.

Assuming a sine-wave input signal, the first value would be screen current at 400v plate and -40v G1. The second value would be at G1 = [40v Peak x Sin(10º)] - 40v = -33.054v along the plate loadline drawn, and whatever plate voltage results from voltage-drop across the primary impedance at that point. Rinse & repeat 35 more times (or use Spice, if you have models of plate current & screen current at your chosen screen voltage).

[martin manning]

Had to fix my plot above, Vg2 is constant 400V :^) LTSpice version agrees. Unfortunately it won't integrate this type of sweep.

[nuke]

I'm not ignoring the thread, I've been hitting the "too many requests error" nearly every time I visit the forum for the last few days.

I think we're kind of talking past each other.

there's also something odd about that chart with the plate and screen curves super imposed on each other...

[martin manning]

No worries, same here.

Which curve are you referring to? I replaced the original one I posted with the marked up GE data sheet, and I think it is correct now. The LTSpice sim matches the plate and screen currents and dissipations very closely.

[pjd3]

Hey nuke,

Just to mention, I've been receiving the same "too many requests" message as well, and not given access to the website.

Best

Phil D