Attempt on a 7591 Se design

[bea]

Let me just mention that the setup would also work with the 6L6GC (i am aware of the different pinout...) - the range of the bias pot is sufficient. So if i build it would be relatively easy to migrate to a 6L6GC if the 7591 turned out not to be useful.

Another remark: the 7591 is a real pentode. The 6L6GC is a tetrode, and its abovementioned high voltage cousin has been the 7027A.

[ThermionicScott]

You sure about that? The pictures of JJ 7591s that I can find look like they have beam-forming plates inside, not a suppressor grid.

As far as the old ones went, perhaps they were like EL84s/6BQ5s in that they were made both ways. (I have some GEs at home with beam-forming plates.)

[tubeswell]

If you want a mathematical formula for working out load resistance for output stages, Merlin's page on SE output stage design has information on optimum load resistance for centre-biased class A operation This is applicable no matter what tube you're using because it calculates load resistance based on plate power and plate voltage.

[Mike Sulzer]

You sure about that? The pictures of JJ 7591s that I can find look like they have beam-forming plates inside, not a suppressor grid.

As far as the old ones went, perhaps they were like EL84s/6BQ5s in that they were made both ways. (I have some GEs at home with beam-forming plates.)

This: http://www.mif.pg.gda.pl/homepages/f...127/7/7591.pdf calls it a beam power pentode, whatever that means. I always thought that they were very similar to a 6L6 in construction, but that is just a memory from long ago.

[Mike Sulzer]

If you want a mathematical formula for working out load resistance for output stages, Merlin's page on SE output stage design has information on optimum load resistance for centre-biased class A operation This is applicable no matter what tube you're using because it calculates load resistance based on plate power and plate voltage.

I would think the only way to do get the maximum power is to try load lines on the tube curves, and then derive the impedance from that. "Any tube" covers a lot of possibilities.

[nickb]

Here's my two cents worth:

EDIT: OK It looked like the top cathode resistor was 820 ohm but now I see it's 0.1m and R20. It's just too blurry for my poor old eyes, I guess Those two resistors are just a hangover from simulation, I guess.

The bias arrangement itself presents too low an impedance to the vol pot & driver. I would move the 150K to be in series to the bias pot wiper and connect the other end to the grid and C4.

One you've fixed these things, the loop gain will shoot up and you might find you have stability issues that need to be addressed, assuming it is stable as shown.

What is the thinking behind having 390K anode resistors?

Having C2 and C6 in parallel to form a 3.3nF seems a little wasteful.

[bea]

Here's my two cents worth:

EDIT: OK It looked like the top cathode resistor was 820 ohm but now I see it's 0.1m and R20. It's just too blurry for my poor old eyes, I guess Those two resistors are just a hangover from simulation, I guess.

Yes, i wanted to be able to switch between cathode bias and external bias with relatively few clicks.

The bias arrangement itself presents too low an impedance to the vol pot & driver. I would move the 150K to be in series to the bias pot wiper and connect the other end to the grid and C4.

I had some problem with blocking distortion and increased the grid stopper. But anyway, the preamp is work in progress. I'll probably use the setup from my G-2000.

One you've fixed these things, the loop gain will shoot up and you might find you have stability issues that need to be addressed, assuming it is stable as shown.

I am not sure. There is a (really) slight increase toward 10 kHz which might be suspicious.

What is the thinking behind having 390K anode resistors?

Maximising gain. In the current state, V1 has 390 k. When i'll change the preamp (what i will definitely do), i'll use 220k.

Having C2 and C6 in parallel to form a 3.3nF seems a little wasteful.

In reality, C6 is switchable.

[nickb]

Yes, i wanted to be able to switch between cathode bias and external bias with relatively few clicks.



I had some problem with blocking distortion and increased the grid stopper. But anyway, the preamp is work in progress. I'll probably use the setup from my G-2000.



I am not sure. There is a (really) slight increase toward 10 kHz which might be suspicious.


Maximising gain. In the current state, V1 has 390 k. When i'll change the preamp (what i will definitely do), i'll use 220k.



In reality, C6 is switchable.

A slight increase in gain? C4 is loaded by at most 5.5k (bias network) and the source impedance is 50K-250k so the attenuation is about 10 at best.

Also 390k/3.3k won't give you max gain. For example 100k/1k will give you about 1.5dB more gain as the Gm increases with anode current.

[bea]

A slight increase in gain? C4 is loaded by at most 5.5k (bias network) and the source impedance is 50K-250k so the attenuation is about 10 at best.

I am currently playing a bit with the model. Fixed the bias circuit (the bias was was initially larger - 100k). Now the grid stopper is 15k and a resistor of 220 k providing grid leak through the bias circuit has been added. Now the frequency response looks as expected.

I also played a bit with the anode resistors and the values of the pots. I'll stick with the 390k anode resistors. This will give slightly more gain at V1 which i need if i want to overdrive V2, but change the gain pot to 500k. At V2 i am a bit unsure wether i want more gain or not. Overdriving the power stage is not necessarily intended, and with the 7591 so close to its max dissipation heavy overdrive of the power stage should probably be avoided.

[tubeswell]

I would think the only way to do get the maximum power is to try load lines on the tube curves, and then derive the impedance from that. "Any tube" covers a lot of possibilities.

Well yes - the tube needs to be biased correctly to ensure the operating point corresponds with the plate voltage and the target dissipation at the screen voltage you are working with. Nevertheless I did state 'centre-bias' 'Class A operation' as assumptions.

[Gingertube]

I played with some 7591 when I did an Ampeg Echo Twin restoration not so long ago.
Did an A B on the AVO MK3 Tube tester betwen some old RCA and new JJ. The new JJ had half the gm of the old RCA. That is, they were not much like a "real" 7591. They did bias at the same voltage thou'.
I made myself a note to try ElectroHarmonix istead of the JJ next time I needed 7591. May not be any better but I don't recommend the JJ.
Cheers,
Ian

[bea]

If you want a mathematical formula for working out load resistance for output stages, Merlin's page on SE output stage design has information on optimum load resistance for centre-biased class A operation This is applicable no matter what tube you're using because it calculates load resistance based on plate power and plate voltage.

The formula is the one i used in my estimate. It appears to me not as power maximum but as being derived from the distortion minimum condition. As the internal resistance of the tube is large compared to the load resistor, the available power is determined by the anode voltage, the current determined by the load impedance and the minimum usable voltage due to the "knee". To first order You will see the same output power from any tube provided it can handle the current and the voltages.

[jazbo8]

It appears to me not as power maximum but as being derived from the distortion minimum condition.

No, that is just not the case, the quiescent variables Ep0, Ia0, and Pmax have little to do with the harmonic distortion per se. To calculate the distortion, as others have already mentioned, you need to do load line and/or transient analysis.

[bea]

Well, here the latest simulation. The IMO logical next step - large signal behavior.

This time i'll show the result for the 6L6, which i did in parallel to the experiments with the 7591. This time i did not repeat the small signal analysis but ran a sine of 1000 Hz through the circuit, the amplitude large enough to reach saturation. I adjusted the bias for max undistorted and symmetric output.

As You see, the max clean amplitude at the output is quite exactly +-10 V, i.e. 7 V eff which means 6.1 W at 8 Ohms. For the 7591 i found 5.5 W. In order to find this, i needed to slightly reduce the cathode current of the output tube to 60 mA.


EDIT: at least to a large degree that's the preamp.

[nickb]

Well, here the latest simulation. The IMO logical next step - large signal behavior.

This time i'll show the result for the 6L6, which i did in parallel to the experiments with the 7591. This time i did not repeat the small signal analysis but ran a sine of 1000 Hz through the circuit, the amplitude large enough to reach saturation. I adjusted the bias for max undistorted and symmetric output.

As You see, the max clean amplitude at the output is quite exactly +-10 V, i.e. 7 V eff which means 6.1 W at 8 Ohms. For the 7591 i found 5.5 W. In order to find this, i needed to slightly reduce the cathode current of the output tube to 60 mA.


EDIT: at least to a large degree that's the preamp.

Let me throw a couple of simulation tools/ ideas your way.

First, until you get to some level of refinement, keep it simple. this means lose as many non-essential components as possible and simulate one section at a time. The schematic below will give you the idea.

Second, which is handy for seeing the linearity visually, is rather than to step the driving voltage source, is to use an amplitude modulator fed by a linear ramp. You can then see at a glance where the linearity starts to fail off.

Schematic:
Ramp Input Sch with Power & THD 7591 SE Power ramp out sch.pdf

Waveform:
Waveform Ramp Input 7591 SE Power - ramp out.pdf



Lastly, you can estimate the power out and THD when stepping the input using '.four' and '.meas' commands. Note the TRANS statement parameters are crucial to get meaningful results.

Schematic:
Step Input Sch 7591 SE Power step.pdf

Results (in err/log file):
Step Input THD and Power out log 7591 SE Power.log.txt


One more thing, to report the power dissipated in anything when doing a trans, put your cursor over the component and hold the ALT key. On the newly added waveform, hover over the name, hold CTRL and click. The power will be in the pop-up dialog.