switching between UL and Pentode

[crisisat22]

I recently built a 100W that uses the 4 KT88 power section of a sunn model t. The OT is a hammond with UL taps. I set it up originally with the UL taps. Sounds great.
then I installed a 2p2t that switches the screen supply of both banks of tubes to the first node in the PS after the plate supply. when it is in Pentode mode, the amp sounds amazing for a few seconds and then blows a fuse as the volume is brought up. Also I should point out that it appears as if one of the plate resistors (50 ohm, 5W) somehow arcs to the chassis when this happens.

I am sure the switch is set up correctly, and have tried it hardwired with no switch only to get the same result.
Va is around 540. Vg2 is a few shades less. Both are under the rated max's for the tubes.
dissipation at idle is 28W.

Any suggestions would be much appreciated- The amp functions perfectly as an UL amp, and normally I would not mess with it, but the tone I got for those brief moments in pentode mode is definitely worth some investigation!!

thanks-

[Merlinb]

This probably has something to do with screen arcing. It always does with KT88s. They are not the same beast are your common EL34 / 6L6 types, they need a MUCH lower screen voltage in pentode mode, or a really low load impedance.

It may help if you stick some really big screen-grid stoppers on during pentode mode, like 3.3k or maybe even more since you're operating them way off the map (although the valves may be damaged already).

[loudthud]

I built an amp like this for bass and always liked it better in Pentode mode.
1) You shouldn't switch modes unless you put the amp in standby mode. Between modes the tubes will switch off momentarily and create a spike from the inductance of the OT.

2) Can you change the lead dress to move the 50 ohm resistors farther from the chassis?

3) Connect a 270 ohm 10W resistor across the speaker terminals. This will reduce or eliminate any spikes.

[crisisat22]

yeah, the amp was in SB when switched.
I will think a bit more about the screen voltages and see what I can get away with there.

When I hit this wall, I started looking for other ways to open the amp up. Replacing the 33k feedback resistor with a 330k! did the trick. It took away some of the unwanted tightness, and made it sound a bit less UL IMO.

thanks!

[defaced]

Would you mind posting what you find out after you work with the screen voltages? I'm heading down a similar road and wouldn't mind learning/collaborating. I'll be using 560ish B+, similar screen, and a Hammond 1650T OT.

Merlin, what would be on the map for the screen voltages for KT-88s in pentode mode?

[Merlinb]

Merlin, what would be on the map for the screen voltages for KT-88s in pentode mode?

With what load impedance?

[Steve Conner]

Are you sure merlinb? I was always under the impression that KT88s could handle more screen voltage than 6550s, and certainly EL34s. After all, they're rated for UL operation up to quite high voltages.

Still I don't think running them off 550V with no screen resistors is a good idea! I'd guess 1k or maybe 1.5k for the screen resistors (per tube) - 3.3k seems too much. I'd start with something like 1.5k and adjust it by trial and error until I got the same output power I was getting in UL.

[Merlinb]

Are you sure merlinb? I was always under the impression that KT88s could handle more screen voltage than 6550s, and certainly EL34s.

You're right, their basic screen voltage maximum IS higher than most, but the problem isn't voltage alone, it's screen dissipation and the mode of operation (see below).

After all, they're rated for UL operation up to quite high voltages.

Here's the rub: in UL, the average screen current does not increase under working conditions (well, not very much). So as long as you aren't overpowering them at quiescence, they won't be overpowered the rest of the time either. No problems there then, you can use high voltages.

BUT: When in regular pentode mode the peak screen current can become HUGE as the grid is driven towards 0Vgk. The higher the screen voltage, relative to the load impedance, the bigger those screen current pulses are. And in a guitar amp we may be overdriving the KT88 continuously, so the screen dissipation goes off the chart, and that's what kills KT88s. You either need to use a low load, a low screen voltage, or some screen-current limiting (big stoppers).

Just becuse we are used to seeing Va and Vg2 similar in most amps, doesn't mean all tetrodes are alike, you have to consider the working conditions. And good tone doesn't come from just arbitrarily using high voltage. Voltage does not 'make a tube sound good', it just makes it work. It's the dynamic conditions that lead to that killer tone, and the voltages needed to get you there aren't necessarily always high (or power scaling would never sound any good, would it!)

If you're lucky you'll get a bottle that takes the strain, which leads to people thinking, " if it worked once, it must be ok", when in fact they just got lucky. This in turn leads to attitudes like "KT88s made by X are bad, they always die early", when in fact, they are not SUPPOSED to take that kind of abuse. It's the amp design which is bad.

[Steve Conner]

Yes, but all of what you said about dynamic screen dissipation applies to every pentode or beam tetrode. More so to true pentodes like the EL34, since they draw more screen current, both static and dynamic. I believe this was the reason why the early Marshalls burnt through EL34s so quickly: Jim Marshall used too small screen resistors, causing the screen grids to melt.

So what is it about KT88s in particular?

[defaced]

With what load impedance?

4 x KT-88, 1.9K primary impedance load, pentode class AB1.

[Merlinb]

Yes, but all of what you said about dynamic screen dissipation applies to every pentode or beam tetrode. More so to true pentodes like the EL34, since they draw more screen current, both static and dynamic. I believe this was the reason why the early Marshalls burnt through EL34s so quickly: Jim Marshall used too small screen resistors, causing the screen grids to melt.

So what is it about KT88s in particular?

KT88s have a lot more screen mu than the other bottles. It's kinda hard to explain without drawing it out graphically, but here goes:

Compare the anode characteristics of the EL34 and KT88 at the same screen voltage. For example, see the graph of the EL34 at Vg2 = 250V on page 6, and look at the KT88 250V screen curve on page 6. We're now comparing the 0V grid curves.

EL34: http://tdsl.duncanamps.com/link.php?target=00259703
KT88: http://tdsl.duncanamps.com/link.php?target=00259D50

The EL34 oV curve flatten off around 270mA. But the equivalent KT88 curve is 'higher up', that is, it flatten off around 370mA!

Now suppose you add a load line, the same value for both valves. In the EL34 it might pass through the knee of the curves. In the KT88 it will pass way BELOW the knee. The further below the knee you are, the bigger the screen current pulses. So at Vg2 = 250V, into the same load, we would expect the peak screen pules in the KT88 to be about 370 - 270 = 100mA bigger than the EL34!! What's more, because it's a tetrode, the KT88 screen is rated for only 6W, versus 8W in the EL34. So it's screen is already more fragile!

If you wanted the loadline to pass though the knee of the KT88's characteristics, you'd need to lower the screen voltage, or use a lower load. That's it in a nut shell. Did any of that make sense?

[Merlinb]

4 x KT-88, 1.9K primary impedance load, pentode class AB1.

It varies a bit with anode voltage, but assuming you have conventinal 1k screen stoppers and you want an amp that responds in broadly the same way as 99% of other pentode/tetrode guitar amps by having the loadline pass below the knee, but not TOO far below:

Va = 200V to 300V, Vg2 = 200V

Va = 300V to 400V, Vg2 = 250V

Va = 400V to 500V, Vg2 = 300V

(And I'm being generous there. Personally I would err on the side of 250V even with 450V on the anode...)

[Steve Conner]

Yes, it makes sense.

Working from the graph on the KT88 datasheet page 6, I'd choose a lowest anode voltage of 75V, since this is just on the knee. That corresponds to an Ig2 of 75mA peak at 300Vg2. So with 550V supply we need (550-300)/0.075 = 3.3k screen resistors.

And we get 400mA peak plate current, hence the peak power is 0.4*(550-75) = 190W, so we get 95W RMS from a pair of tubes, or 190 from a quartet, which is reasonable. The load impedance should be 4*(550-75)/0.4 = 4.75k for a pair of tubes or 2.4k for a quartet.

So I take back what I said, 3.3k screen resistors were about right Since the OPT is 1.9k and the supply rail is only 540, and would sag further under load, we might want to go a little lower to make our 200 clean watts.

[defaced]

Beautiful, thank you.

[crisisat22]

Beautiful, thank you.

Agreed!
Thanks guys for the wonderful explanation!