Screen resistor values

[firemedic]

Quick question:
Is 2.2k 5W too much resistance for a screen resistor (6L6/6V6)? All my builds so far have had higher screen voltage than plate & I wanted to try a lower screen voltage for once.

[HotBluePlates]

Try it and listen to the amp. Do you like the result? If so, the 2.2k is fine.

I think I (re-)discovered a wrong way to build an output stage a while back, and suggested it to others as a way to add sag. If you want every last watt from your amp, you want as little impedance between a tube screen and its power supply as possible.

The maximum plate current that the tube can pass is governed by screen voltage. If the screen voltage drops (think 10's of volts, rather than 1-2 volts), then the maximum plate current from the tube is reduced. As an output tube transitions from no output to full output, the screen current rises. If you insert some resistance between the power supply and the screen, then ohm's law tells you that as the screen current rises, the voltage drop across the screen resistor gets bigger, and the resulting voltage at the screen drops.

If the drop in screen voltage is significant, then the power output from the tube (plate voltage swing times plate current swing) will drop.

If you must have zero compression and every last watt from your design, you use no series resistor to the screen. Quite a few old hi-fi amps didn't use series screen resistors, but rather used a single common resistor feeding screens on both sides of the push-pull output. The idea was that the rise in screen current on one side would be offset by the drop in screen current on the other side, and result in a steadier screen voltage. Reality is that the screen voltage is steadier, but the screen current do not change in equal and opposite directions.

But protection against excessive screen dissipation is also important, especially if you whack the snot out of the output tubes to get your distortion sound. Screen current increases sharply when you drive the control grid positive. So you'll see production amps with the smallest screen resistor they can use and still provide some degree of screen protection, in order to make all the output power possible. However, if you can accept some degree of compression and a slight maximum volume loss, you can increase the screen resistor value well over the typical values. Some folks around here routinely use a screen resistor around 2.2-3.3k.

[jjasilli]

I tried 2.2K for my el-84 Traynor Bassmate yba-2.  I thought they made the amp sound horrible.  Still tinkering with it.  (Been getting my workbench in order.  Now my scope is down for no apparent reason.)

[Fresh_Start]

Is it worth considering the value of the filter cap at the screen grid node on the power rail as well as the screen resistor?  My thought here is that a bigger screen filter cap would tend to offset a larger dropping resistor between the plate node and the screen node in terms of voltage drop.  IOW maybe we want the screen node filter cap to be big enough to be able to offset the potential voltage drop across that isolation resistor?

OTOH in a push-pull amp, there are two screen grids drawing current off of the screen filter cap so maybe it doesn't matter in that context. 

What about in a single-ended amp?  Could you create some sag by using a smaller filter cap and/or bigger dropping resistor in the power rail before the screen node?

Just daydreaming...

Chip

[firemedic]

Well I do like some sag so I'll leave the 2.2k resistors in there for now. This will be a heavily modified Pro Reverb/Vibroverb. I'm replacing the normal channel with a single 5879 preamp stage, stock verb channel, so I'll be able to bridge both channels together.
I know I'll need at least 1k screen resistors since I'm including a pentode-triode switch which draws the "triode" screen voltage from the plates.

[HotBluePlates]

Is it worth considering the value of the filter cap at the screen grid node on the power rail as well as the screen resistor?  My thought here is that a bigger screen filter cap would tend to offset a larger dropping resistor between the plate node and the screen node in terms of voltage drop.  IOW maybe we want the screen node filter cap to be big enough to be able to offset the potential voltage drop across that isolation resistor?

What about in a single-ended amp?  Could you create some sag by using a smaller filter cap and/or bigger dropping resistor in the power rail before the screen node?

That was good thinking, but you're thinking about how the resistor and cap interact with respect to ripple voltage. I strongly suspect that the cap would do little to sustain voltage drop in this situation, especially with the cap not being placed right at the screen grid, the way a filter cap would normally feed a power supply node.

Further, the time constant of the resistor and cap would have to be at least seconds long instead of milli- or micro-seconds long to sustain the voltage.

As for SE vs. push-pull, it should still work assuming the output tube screen draws enough current in use to make the effect noticeable. You could always make the resistor very large.

[kagliostro]

Mesa used that trick in the Lonestar

Kagliostro

[PRR]

> Screen current increases sharply when you drive the control grid positive.

No. When plate voltage falls very low.

Hold plate at some voltage near screen, screen current does not go way up even with zero on G1.

Go back. Look at a true Tetrode such as Type 24A. Whichever is gigher, plate of screen, gets "all" the cathode current. In use, this means we never want plate to fall below screen. Since we need a high screen to suck current from the cathode, this limits plate swing.

Now add the Suppressor grid G3. Now plate may fall far below screen before screen current shoots up, and screen current usually won't get as high as max plate current.

I am not sure what the factor is, but it appears to be similar to Mug2. A 6L6 with 250V on G2, plate current equals screen current only when plate falls to 25V.

It IS confusing, because a Well Designed Power Amplifier will usually have the zero-grid peak and the low plate voltage peak about the same point. Hot-rod gitar amps are often not "well designed" in this sense: they over-volt the screens to get adequate peak plate current without taking G1 near zero volts. This gives an overload cushion before grid-blocking. The downside is that when everything is over-volted, tubes intended for sane use will melt.

[HotBluePlates]

> Screen current increases sharply when you drive the control grid positive.

No. When plate voltage falls very low.

Strictly, this is true.

What I was thinking more is that *if* the loadline cuts the G1=0v line below the knee, then screen current goes up. If it cuts that curve above the knee, screen current stays well-controlled. One does also result in plate voltage swinging lower than the other.

This line of thinking does have a problem: it requires the load seen by the tube to be higher than we're probably using. But we can also make the argument that there are a lot of reasons why the load may appear higher than nominal. One reason could be the variability of the impedance reflected from the speaker.

I guess real experimaentation and measurement is called for.

[PRR]

> reasons why the load may appear higher than nominal

It's always higher than nominal.

A "8 ohm" cone loudspeaker is 6 or 7 ohms of dead copper, a LARGE bass resonance, and an inductance. Impedance at bass resonance rises about 5X the copper resistance, falls in midrange, then rises above 1KHz.

In almost all applications, the bass resonance is an important part of speaker response and a fair amount of voltage goes there.

Take 5F6A. Nominal 4KCT load. Take most guitar speakers. Bass resonance is 70Hz-110Hz. Guitar plays down to 82Hz. You may not use the low open string often, but you might. And the resonance extends out to 150Hz-200Hz or higher. (It actually runs into the treble inductance. That's why 7 ohms DCR is never less than 8 ohms across the useful bandwidth.)

At bass resonance the "4K" load is more like 20K. Since the 4K nominal load cuts pretty near the knee, the 20K load is sure to cut way below the knee.

> Quick question: Is 2.2k 5W too much....

Is NOT a "quick question"; or rather, there is no quick answer. The calculations are far too messy.

I will observe that EL34 (higher screen current than the 6V6/6L6) at 400V cathode-bias 6.6K load did not suffer from a 2K common screen resistor. Power was 20V when it maybe could be 22W-23W with zero screen resistance. OTOH, nothing bad happens when grossly over-driven, quite mellow. Mellow is not always good, of course. But for studio use in a jazz school, it sounded fine and was pretty abuse-resistant.