superlead with master volume redplates

[Andy Le Blanc]

Would improved inverter balance help?

DC balance at the plates?

[raiken]

Would improved inverter balance help?

DC balance at the plates?

DC balance would do nothing. A balanced phase inverter would be worse. An unbalanced phase inverter tube (unbalanced in the right direction, lower on the 82K side) or a tube with unequal triode sections with about half or one quarter the mu on the 82K side would help.

In fact, it might be interesting to see what a 12DW7 does - it is a dissimilar triode with a gain of 100 on one side and 20 on the other, with the 1,2,3 side being lower, which should be the proper side. Of course, it will lower the gain, especially when running clean, so you may have to goose it a bit more for proper distortion.

I suppose you could also try greatly reducing the 82K or increasing the 100K to deliberately unbalance the PI to see if you can reduce the redplating, just to get a feel for what is going on.

If you do any of this, you also will lose some of your duty-cycle modulation that gives the amp it's second-harmonic character on heavy overdrive...

Randall Aiken

[Roe]

You may want to consider the fact that you are running the poor EL34's way over their max dissipation limits with 4...
What can be done to fix this? Well, if you must run a 3.4K (1.7K) output transformer, you must reduce your plate voltage to no greater than around 400-420V to stay within the ratings. You may get by with a bit more, since the edges of the load line fall under the 2x curve. Alternately, if you want to run 480V, you must change your output transformer to around 4.8K - 6K (2.4K - 3K for 100W) to stay within the ratings. Any time you exceed these rules, you run the risk of redplating.

Randall Aiken

Randall thanks a lot for the explanation! I will do more testing. I will try running a 16ohms load into the 8 ohms secondary.
The amp already works well with 395v on the plates but I very much prefer the sound with 470-480v. All the tubes I've tried have redplated at 480v. So I guess it could be oscillation.

[Andy Le Blanc]

"duty-cycle modulation caused by the long-tail pair phase inverter design"

That's a mouthful...... can you elaborate upon the terms used...

and provide reference's ...

[raiken]

"duty-cycle modulation caused by the long-tail pair phase inverter design"

That's a mouthful...... can you elaborate upon the terms used...

and provide reference's ...

Duty-cycle modulation: A dynamic change in the square-wave duty cycle as the signal level is increased, when the output is heavily overdriven.

Long-tail pair phase inverter: a phase inverter topology that has a single resistor connected as a pseudo-current source from the junction of two tube cathodes, with the outputs taken off the individual tube plates, one in phase with the input signal, and the other out of phase with the input signal. The circuit gets it's name from the "tail" resistor connected to the cathodes.

Following are two simulations of a Marshall 1987 at different drive levels. The red and blue trace are the phase inverter outputs, the green is the output. As you can see, as the drive is increased, the output square wave duty cycle changes. Look at the high time vs. the low time on the green trace, particularly at the right side of the graph.

[img:926:648]http://www.aikenamps.com/1987_1.jpg[/img]

[img:927:648]http://www.aikenamps.com/1987_2.jpg[/img]


And here are the FFT plots showing the increase in second and other even-order harmonic distortion caused by the duty-cycle modulation:

[img:926:648]http://www.aikenamps.com/1987_fft_1.jpg[/img]

[img:927:648]http://www.aikenamps.com/1987_fft_2.jpg[/img]

If you like, I can also post some scope pics of an actual Marshall 50W to show you the effect, but it looks similar to the simulation.

Randall Aiken

[Structo]

Was the test signal 1KHz?

[raiken]

Was the test signal 1KHz?

Yes.

[Andy Le Blanc]

That's a long way to go to get more tone color out of an inverter that reduces
even order distortion of its own nature. And the other solution of increasing
the chosen plate load to keep the excursion of the max dissipation line to a
minimum will increase third order derivatives as a natural characteristic of a
power pentode.... looks like the glass ceiling of that design approach.

Which inverter type may be a better choice for the application?

[raiken]

That's a long way to go to get more tone color out of an inverter that reduces
even order distortion of its own nature. And the other solution of increasing
the chosen plate load to keep the excursion of the max dissipation line to a
minimum will increase third order derivatives as a natural characteristic of a
power pentode.... looks like the glass ceiling of that design approach.

Which inverter type may be a better choice for the application?

The LTP is an excellent choice for the application - it has plenty of gain, nearly matched output impedance drive, and the asymmetrical clipping characteristics promote even-order harmonics because of the duty-cycle modulation, which enhances the overdriven output tone.

The only problem here is that the output stage design of this particular amp runs the tubes on a load line that is perched too far above the max dissipation curve. This is further supported by the fact that it doesn't redplate at idle, only when running at power, so it is not a bias issue.

The choice of phase inverter is only relevant in that the LTP duty cycle change aggravates the dissipation, but if you designed it out in favor of a phase inverter that is more balanced when clipping, you would lose a lot of the character of the tone - a symmetrical duty-cycle square wave doesn't sound as good. In my opinion, it would be better to keep the LTP and properly design the output stage.

Having said that, if you did design a more symmetrical PI, it would reduce the tendency to redplate on only one side, and it would either redplate both, or depending on where you ended up on the operating curve, it may reduce the problem enough to be livable. In this case, something like a split-load PI driving either two equal amplification stages , a differential amp (Williamson), or two cathode followers, would all lead to more symmetrical duty cycles.

You can also increase the symmetry by going to a current source for the "tail" or even simply increasing the value of the 10K tail resistor up to something like a 100K to act as a better "pseudo" current source.

You may even be able to decrease the effect by adding a large resistor, such as a 100K - 470k in series with the PI input coupling cap to keep the PI from distorting as hard, which will increase the symmetry.

To clarify, I wasn't suggesting the tube changes and plate resistor modifications as a "fix" for the problem, rather as a means of testing things to see what is going on and to show the effects of the duty-cycle changes on output tube dissipation.


Randall Aiken

[Andy Le Blanc]

Its always fun when one issue brings another to light.
Designing for distortion, on the edge of stability, is in the complete opposite direction
of nearly all valve design. Your explanation of the inverters duty cycle reminded me
of the that peavey texture/class control I've seen, a post inverter MV on
just one leg of the push-pull circuit. One description might be that it takes the output of fairly well balanced
cathodyne inverter and with asymmetry, mimics the duty cycles of other inverters.

Could you get away with using a greater reflected load to address the high voltage issue,
as Roe has suggested, "I will try running a 16ohms load into the 8 ohms secondary. ",
or is it wiser to use, as you have all ready stated, a different opt?

[Roe]

I did more testing:
1k7 - 395v: no redplating
3k4 - 480v: no redplating
1k7- 480v: redplating after playing on 10 for a few minutes

So it seems Randall is pointing to a problem with the combination of high voltage and low Z (1k7).

Another strange thing: I can measure a litte AC on the tube sockets when running. Is this due to power supply ripple, bad coupling caps or bias circuit?
All sockets seem to be the same. I don't really think the coupling caps and bias circuits can explain this since I use 4 coupling caps and 2 separate bias circuits.

I mistakenly said the kt88s redplated, their bias was of.

[raiken]

Could you get away with using a greater reflected load to address the high voltage issue,
as Roe has suggested, "I will try running a 16ohms load into the 8 ohms secondary. ",
or is it wiser to use, as you have all ready stated, a different opt?

A transformer doesn't have an impedance, it merely reflects the load impedance on the secondary back to the primary. However, it does have a primary inductance, which sets the low frequency -3dB point of the response. If you change the secondary load and reflect back a higher or lower impedance than it was originally designed for, you will get either a higher or lower cutoff frequency, depending on which way you go.

The formula for determining the low frequency response is: f = Z/(2*pi*L), so if you reflect back a higher load, the low frequency cutoff point goes up, because the primary inductance is fixed.

Following is a plot showing the effect of doubling the reflected impedance from 3.4K (1.7K) to 6.8K (3.4K) on the 480V/26mA EL34 setup. As you can see, it shifts you out of the "danger zone" of the 2x max dissipation rating, so it will stop the redplating.

The graph also shows that you can lower the impedance somewhat and still be in the safe area. It turns out that you can go down to 5.2K (2.6K)for 480V, or 4.6K(2.3K) for 450V and always remain in the safe area, if you bias no higher than 26mA (ignoring the effects of any significant voltage sag, which would improve the situation).

In addition, the graph shows that it would be a good idea to lower the screen voltage so the loadline crosses the "knee" of the curves. Lowering the screen voltage causes the curves to shift downward, which increases the linearity of the output. If you don't decrease the screen voltage, the non-linearity of the curves will increase the distortion of the signal, which may or may not be a good thing. Lowering it to somewhere around 350V to 400V would allow the new loadline to intersect the knee again. In addition, this greatly reduces the screen grid dissipation, and you still end up with about the same power output as before.

[img:927:650]http://www.aikenamps.com/loadlines.jpg[/img]

Randall Aiken

[raiken]

On a related topic, you also have to consider screen grid dissipation.

When running tubes at very high voltages, you will likely see the screens glowing brightly in time with the music. This is a sign that they are probably running too hot - you can tell by the brightness/color of the glow. If they are white hot, they won't last very long, and you need to either lower the screen voltage or up the value of the screen grid resistor.

Changing the load impedance will affect the screen dissipation, particularly when you increase the impedance and are running heavy overdrive. This is why attenuators that present a load impedance of twice, or four times the normal load cause fuses to blow and tubes and output transformers to die.

You can measure the screen dissipation by measuring the voltage from screen grid to the cathode (usually ground, unless cathode biased), then measure the current through the screen grid resistor by measuring the voltage drop across it and dividing by the resistor value, then multiply this current measurement by the screen voltage measurement to get the screen grid power dissipation. You will see it radically change from idle to full power, especially if you heavily overdrive the output stage.


Randall Aiken

[Roe]

Randall, yes I'm use to measure voltage drop over the 1k screen resistors, at least if I run a hot bias. the early 100watters with the 1202-84 OT run the screens hard btw. Z=4k and b+ is originally 560v. screen voltage is 559v. this is very hard on the kt66s

[raiken]

Randall, yes I'm use to measure voltage drop over the 1k screen resistors, at least if I run a hot bias. the early 100watters with the 1202-84 OT run the screens hard btw. Z=4k and b+ is originally 560v. screen voltage is 559v. this is very hard on the kt66s

Yes, running KT-66's at a 560V and 4K impedance is unsafe at any speed. You can, however, run them at 560V with a 6K impedance if you bias no more than 27mA to avoid exceeding the dissipation limits anywhere in the operating range. You'll hit right at 30W when it hits 260V peak. You still have to be careful, though, because the absolute max plate voltage is 550V and you are exceeding that.

Randall Aiken