I don't think it will oscillate as long as the feedback components are adjusted for good low distortion performance. The values of C2 and C2 can be adjusted as long as the frequency response of the amp is reasonably flat from about 20 kHz to 20 kHz. Also that the feedback is as fast it can be in the important part of that bandwidth. I think using the Triode for preamp is better than using a Pentode as input.
The LE-EL34 AMP 1 circuit needs more improvements as regards component values as you are suggesting I think. Personally I don't think the 12AU7W is a very linear looking triode. Something like the 5670 is more linear assuming a higher Anode to Cathode DC Bias current is used:
Anode Current vs Anode Voltage for 12AUW (ECC82) Triode:
Anode Current vs Anode Voltage for 5670 Triode:
Thank you all for the "forward-thinking" off this project. Don't worry about the trannies these are in solid and good condition. I tested the powertr's with total load of near 140W each (2x 6SN7 filament=1.2Ax6.35 or 8W + 3 lightbulbs of 40W/230V @ 700V or 120W and a GZ32 @ 5V winding or 9.5W) for half a day: 40°C. What do you think of Tubecad EL34 PP https://tubecad.com/2010/03/21/push-pull power amp example for ps-4.png ?
That looks like a very good circuit, local feedback using Cathode resistor, it's better than feedback from the transformer all the way back to the input. Also push pull the two EL34 in opposite phase.
The 6SN7 valve (I am assuming they use 2 x 6SN7 Dual Triode and 2 x EL34 Pentode valves) the 6SN7 looks OK:
The 1M resistors between 1st and 2nd stage are very high, I think 250K would be better and the 0.1uF capacitor there should be higher, more like 2,2uF for good Bass response. Both 1st stage and 2nd stage have negative feedback the 1st stage could do with more feedback (make the 20K Anode resistor lower say 15K).
The third and fourth Triodes give the opposite phase output for the EL34 Pentodes.
The circuit needs more work and carefully testing for good frequency response and low gain. This idea is used in many circuits and was called "phase splitter" in the past.
What do you think @6A3sUMMER ?
1. Post # 39 Schematic uses a Step-and-Repeat Software Library.
It shows that the Suppressor Grid is connected to the EL34 Cathode inside of the tube.
No it is not connected that way, you have to wire a jumper from Pin 1 Suppressor Grid, to Pin 8 Cathode.
By the way, the 6CA7 Beam Power tube has that same requirement.
They are plug and play equivalents, EL34 Pentode, and 6CA7 Beam Power.
12AU7 not linear? . . . it depends on how you use them.
I used a JJ ECC82 (essentially a 12AU7 Plug-and-Play); an LM334 current source to the cathodes, and matched plate loads for the phase splitter.
The outputs were push pull KT77. No negative feedback, except for the Triode Wired KT77s.
It is a special low power amplifier.
At 1 Watt, the 2nd harmonic distortion was -57dBc, and the 3rd harmonic distortion was -57dBc (0.14% 2nd, and 0.14% 3rd. 0.2% THD).
The 12AU7 in single ended is less linear (not as good as my CCS LTP phase splitter).
So build it as a phase splitter, and then only take one phase (crazy, right?). It does have only 1/2 gain, because of the CCS and cathode to cathode action.
Tricks of the trade.
Post # 42 schematic . . .
That is the first time in decades that I have seen the EL34 Suppressor Grid tied to the plate. The SE EL34 Randal amp out of San Francisco was the only other time I have seen that.
They used 100 Ohms screen to plate, but 1k Ohms from Suppressor to plate. Is it possible that either the Suppressor could draw too much current, or is there a possibility of the Suppressor can short when the plate swings maximum positive with signal?
The 3 major low frequency poles are 0.47uF, 0.47uF, and the output transformer's primary inductance.
There is no global negative feedback NOW, but if it is added Later, Caution . . .
If the 3 poles line up, it is the "Williamson Trifecta Oscillator effect".
And two out of 3 Do line up, there are dual 0.47uF and dual 470k; and the next coupling is also dual 0.47uF and dual 470k.
There are lots of stages, input gain, Concertina with gain less than 1, and the next state to make up for the low gain concertina.
All of these can be made to work, if they are implemented properly.
Which one is better, a lot depends on two things:
The output transformers
The loudspeakers impedance versus frequency, efficiency, etc.
It shows that the Suppressor Grid is connected to the EL34 Cathode inside of the tube.
No it is not connected that way, you have to wire a jumper from Pin 1 Suppressor Grid, to Pin 8 Cathode.
By the way, the 6CA7 Beam Power tube has that same requirement.
They are plug and play equivalents, EL34 Pentode, and 6CA7 Beam Power.
12AU7 not linear? . . . it depends on how you use them.
I used a JJ ECC82 (essentially a 12AU7 Plug-and-Play); an LM334 current source to the cathodes, and matched plate loads for the phase splitter.
The outputs were push pull KT77. No negative feedback, except for the Triode Wired KT77s.
It is a special low power amplifier.
At 1 Watt, the 2nd harmonic distortion was -57dBc, and the 3rd harmonic distortion was -57dBc (0.14% 2nd, and 0.14% 3rd. 0.2% THD).
The 12AU7 in single ended is less linear (not as good as my CCS LTP phase splitter).
So build it as a phase splitter, and then only take one phase (crazy, right?). It does have only 1/2 gain, because of the CCS and cathode to cathode action.
Tricks of the trade.
Post # 42 schematic . . .
That is the first time in decades that I have seen the EL34 Suppressor Grid tied to the plate. The SE EL34 Randal amp out of San Francisco was the only other time I have seen that.
They used 100 Ohms screen to plate, but 1k Ohms from Suppressor to plate. Is it possible that either the Suppressor could draw too much current, or is there a possibility of the Suppressor can short when the plate swings maximum positive with signal?
The 3 major low frequency poles are 0.47uF, 0.47uF, and the output transformer's primary inductance.
There is no global negative feedback NOW, but if it is added Later, Caution . . .
If the 3 poles line up, it is the "Williamson Trifecta Oscillator effect".
And two out of 3 Do line up, there are dual 0.47uF and dual 470k; and the next coupling is also dual 0.47uF and dual 470k.
There are lots of stages, input gain, Concertina with gain less than 1, and the next state to make up for the low gain concertina.
All of these can be made to work, if they are implemented properly.
Which one is better, a lot depends on two things:
The output transformers
The loudspeakers impedance versus frequency, efficiency, etc.
Last edited:
richgwilliams,
You said:
"That looks like a very good circuit, local feedback using Cathode resistor"
Which circuit are you talking about?
Post #, please?
You said:
"That looks like a very good circuit, local feedback using Cathode resistor"
Which circuit are you talking about?
Post #, please?
It looks to be happy with higher Ip more like 20mA whereas the 5670 looks good at 5 to 10 mA.
According to EL34 data sheet, tying that G2 to Plate makes the EL34 into a Triode, and they give a characteristic curves for EL34 as a Triode.
richgwilliams,
Post # 50,
It is not the screen grid (g2) that is in question; and triode wiring is also not in question.
The issue is, tying the Suppressor Grid (g3) to the plate.
Post # 49,
The only cathodes that have cathode feedback are the input tube's 470 Ohm resistor, and the driver cathodes 730 Ohm resistors.
They improve linearity, true.
Let's use the data sheet rp of 7700 Ohms.
The un-bypassed Rk also increase the plate resistance, rp, by a factor of u x Rk. Lets use the lower u of 17 (you can also use the u of 20).
17 x 470 = 7990 Ohms, we have doubled the rp. About 15k rp is driving Rl of 20k, so the gain is reduced.
17 x 730 = 12,400 Ohms, we have about increased rp to about 2.6 times; that rp drives Rl of 20k, so the gain is reduced.
The bias voltage times the new lower gain might not have enough signal swing to drive the output grids to full output; that has to be calculated.
You might argue that the concertina splitter also has linear outputs from negative feedback, due to degeneration (well that is true).
But look at it this way, barring nuclear and similar events, electrons do not normally just disappear.
If the concertina tube does not have gas or contaminants, then there is no grid current for the maximum signal before the grid conducts to the cathode.
So, the signal current at the cathode is exactly equal to the plate signal current. Intrinsic equal signal currents, and intrinsically linear.
Post # 50,
It is not the screen grid (g2) that is in question; and triode wiring is also not in question.
The issue is, tying the Suppressor Grid (g3) to the plate.
Post # 49,
The only cathodes that have cathode feedback are the input tube's 470 Ohm resistor, and the driver cathodes 730 Ohm resistors.
They improve linearity, true.
Let's use the data sheet rp of 7700 Ohms.
The un-bypassed Rk also increase the plate resistance, rp, by a factor of u x Rk. Lets use the lower u of 17 (you can also use the u of 20).
17 x 470 = 7990 Ohms, we have doubled the rp. About 15k rp is driving Rl of 20k, so the gain is reduced.
17 x 730 = 12,400 Ohms, we have about increased rp to about 2.6 times; that rp drives Rl of 20k, so the gain is reduced.
The bias voltage times the new lower gain might not have enough signal swing to drive the output grids to full output; that has to be calculated.
You might argue that the concertina splitter also has linear outputs from negative feedback, due to degeneration (well that is true).
But look at it this way, barring nuclear and similar events, electrons do not normally just disappear.
If the concertina tube does not have gas or contaminants, then there is no grid current for the maximum signal before the grid conducts to the cathode.
So, the signal current at the cathode is exactly equal to the plate signal current. Intrinsic equal signal currents, and intrinsically linear.
Yes that is very true and follows from Ohms law.
My problem is that I don't know enough about how Pentodes work. It is of course possible that they just made an error drafting the schematic.
Ideally when @decramer decides which of the many proposed schematic circuits he wants to use he should probably try to use LTspice to finalise some of these component values you refer to. Also LTspice with give an idea of THD in dB.
As regards using a separate feedback winding or taking feedback from the speaker winding. There are two ways of thinking about this:
If you take feedback from the speaker winding you risk picking up interference from long runs of speaker wiring (that's why you see filter components in the feedback) also the speaker may force a response onto the feedback that's delayed or otherwise coloured and therefore problematic. Also the impedance of the speaker (especially if its wrong) will scale the feedback.
If you take feedback from a separate winding well it has taken into account properties of the output transformer but may not be swinging along with the speaker.
I think feedback from a separate winding is a better option.
"# 45 I see a FB winding" Altec Lansing is well known and is a well respected manufacturer.
richgwilliams,
Your comments of Post # 52 (from an earlier schematic):
I think the connection of the EL34 screen grid g2, and the suppressor grid, g3 to the plate, was not a schematic error.
The idea is to connect All elements that are beyond (outside) of the control grid, g1, in an attempt to make that Pentode act as close to a Triode as possible.
The 20 Watt Mullard schematic in Post # 55:
The 82k cathode resistor in the cathode coupled phase splitter is not a very good CCS.
The cathode impedances each are 1/Gm + (180k/u). 1/Gm = 625 Ohms; u = 100 180k/100 = 1800 Ohms
Each cathode impedance is 625 + 1800 = 2425 Ohms
In parallel, the two cathodes look like 1212 Ohms.
1212 / 82,000 = 0.0147.
Well, that is not too bad, the balance is degraded by 1.5% versus a really good CCS.
A pair of 1% plate load resistors on their own, could give up to a 2% error.
Have Fun!
Your comments of Post # 52 (from an earlier schematic):
I think the connection of the EL34 screen grid g2, and the suppressor grid, g3 to the plate, was not a schematic error.
The idea is to connect All elements that are beyond (outside) of the control grid, g1, in an attempt to make that Pentode act as close to a Triode as possible.
The 20 Watt Mullard schematic in Post # 55:
The 82k cathode resistor in the cathode coupled phase splitter is not a very good CCS.
The cathode impedances each are 1/Gm + (180k/u). 1/Gm = 625 Ohms; u = 100 180k/100 = 1800 Ohms
Each cathode impedance is 625 + 1800 = 2425 Ohms
In parallel, the two cathodes look like 1212 Ohms.
1212 / 82,000 = 0.0147.
Well, that is not too bad, the balance is degraded by 1.5% versus a really good CCS.
A pair of 1% plate load resistors on their own, could give up to a 2% error.
Have Fun!
This is a hard circuit for me to understand. As a general comment I don't like the very high resistor values they use. Living in a fairly cold, fairly wet country dampness/condensation has a big effect on resistors 1Meg and above. I suppose the heat from the tubes dries them out but still think high value resistors are best avoided (620K is my maximum).
I think they are using V2 as a long tail pair to split the phase, I have seen this before with a quite high negative B- supply.
Remember this thing:
At the moment I'm trying to find a way to use a current source (above the plate of a Triode) and still have a way to build in local (at the tube) negative feedback. If I put in a Cathode resistor the constant current drops the same voltage across the Cathode resistor and there is no negative feedback and maximum gain always. Any ideas?
richgwilliams,
1. Are you referring to the Mullard schematic in Post # 55?
If you are worried about a high resistance resistor for an LTP, such as 82k, consider that a real CCS could be affected even more, especially if the CCS is on a PCB that picks up moisture.
R10, 1Meg Ohm is even more suseptible to moisture, on or off the PCB.
What % humidity do you have in your home?
Perhaps some other readers have some ideas for de-humidifying your listening room.
2. For a triode with a CCS in the plate, the current in the cathode is constant.
Example:
Suppose the CCS current is set to 5mA; and the cathodes self bias resistor is 1k Ohm. Then the cathode sits at + 5V, the grid to cathode bias is -5V.
You can either use a bypass capacitor across the 1k resistor, or you can leave the bypass capacitor out.
Leaving the bypass capacitor out can reduce the stage gain, depending on all the circuit parameters.
When we apply the negative feedback in the next step below, leaving the bypass cap out can reduce the amount of negative feedback, again depending on the rest of the circuit parameters.
Now, if you want to apply negative feedback to that cathode, insert a 100 Ohm resistor between the bottom of the self bias resistor and ground.
The plate current and the cathode current is still 5mA; but the new cathode voltage has increased to 5.1V.
That means the plate voltage increases by a small amount.
Connect a resistor from the junction of the 1k and 100 Ohm cathode resistors. Suppose it is a 10k resistor coming from the output transformer secondary. Done. Global negative feedback accomplished.
If you do not use a bypass capacitor across the cathode self bias resistor, you can use a series R and C from the negative feedback pick-off point, such as the output transformer secondary. That is the other alternative of how to get negative feedback fed to the cathode.
Understand these two methods, and now you have the basis for the next topology:
. . .
Want to take the feedback from somewhere other than the output transformer secondary, just change the 10k value if needed, and connect to the point you select that will provide negative feedback from (that point has to have the correct phase, the wrong phase makes it positive feedback). Again, if there is an issue of the negative feedback resistor changing the DC bias of the cathode, you can use a series RC to apply the negative feedback.
Careful, a series RC will allow the negative feedback to go away at very low frequencies; but that is similar to DC coupling from the output transformer, because that also causes the negative feedback to go away at very low frequencies.
For an LTP or CCS cathode phase splitter, you can Not apply negative feedback to the cathodes (that would become Both negative feedback And positive feedback; negative to one cathode and positive to the other cathode).
Some designers want a CCS in the cathode, and a CCS in the plate; but because they would both have to be exactly equal under all conditions of the current setting, temperature, B+ changes, warm or hot tube, etc, this does not work.
The solution would be to have Un-equal current settings for two CCSs, and then add a high Ohms resistor across the CCS that is set for less current.
I hope that helps.
1. Are you referring to the Mullard schematic in Post # 55?
If you are worried about a high resistance resistor for an LTP, such as 82k, consider that a real CCS could be affected even more, especially if the CCS is on a PCB that picks up moisture.
R10, 1Meg Ohm is even more suseptible to moisture, on or off the PCB.
What % humidity do you have in your home?
Perhaps some other readers have some ideas for de-humidifying your listening room.
2. For a triode with a CCS in the plate, the current in the cathode is constant.
Example:
Suppose the CCS current is set to 5mA; and the cathodes self bias resistor is 1k Ohm. Then the cathode sits at + 5V, the grid to cathode bias is -5V.
You can either use a bypass capacitor across the 1k resistor, or you can leave the bypass capacitor out.
Leaving the bypass capacitor out can reduce the stage gain, depending on all the circuit parameters.
When we apply the negative feedback in the next step below, leaving the bypass cap out can reduce the amount of negative feedback, again depending on the rest of the circuit parameters.
Now, if you want to apply negative feedback to that cathode, insert a 100 Ohm resistor between the bottom of the self bias resistor and ground.
The plate current and the cathode current is still 5mA; but the new cathode voltage has increased to 5.1V.
That means the plate voltage increases by a small amount.
Connect a resistor from the junction of the 1k and 100 Ohm cathode resistors. Suppose it is a 10k resistor coming from the output transformer secondary. Done. Global negative feedback accomplished.
If you do not use a bypass capacitor across the cathode self bias resistor, you can use a series R and C from the negative feedback pick-off point, such as the output transformer secondary. That is the other alternative of how to get negative feedback fed to the cathode.
Understand these two methods, and now you have the basis for the next topology:
. . .
Want to take the feedback from somewhere other than the output transformer secondary, just change the 10k value if needed, and connect to the point you select that will provide negative feedback from (that point has to have the correct phase, the wrong phase makes it positive feedback). Again, if there is an issue of the negative feedback resistor changing the DC bias of the cathode, you can use a series RC to apply the negative feedback.
Careful, a series RC will allow the negative feedback to go away at very low frequencies; but that is similar to DC coupling from the output transformer, because that also causes the negative feedback to go away at very low frequencies.
For an LTP or CCS cathode phase splitter, you can Not apply negative feedback to the cathodes (that would become Both negative feedback And positive feedback; negative to one cathode and positive to the other cathode).
Some designers want a CCS in the cathode, and a CCS in the plate; but because they would both have to be exactly equal under all conditions of the current setting, temperature, B+ changes, warm or hot tube, etc, this does not work.
The solution would be to have Un-equal current settings for two CCSs, and then add a high Ohms resistor across the CCS that is set for less current.
I hope that helps.
Last edited:
Yes hard to understand because I'm not good with Pentodes. The comment about high value resistors applies more to solid state circuits, high value resistors cause problems if there is moisture around. Take your point about humidity inside a house. Good point about the CCS since by definition they are high impedance.
Yes the problem is that any value resistor between Cathode and Ground provides no feedback. Consider this circuit:
Initially I assumed that the voltage drop across R3 would provide negative feedback and reduce the gain between Input and Output. However there is no negative feedback, the 1 Volt peak input becomes a 30 Volt peak output x30 gain rather than some reduced gain. The reason is obvious now - the constant current across R3 by Ohms law must put a constant voltage across R3 so no negative feedback - none!
I have found one way around this. This very complicated circuit as follows:
The constant current still flows through R3 but the small current through R1 (that is proportional to the output voltage) also must flow through R3. So now there is negative feedback due to R3 and the gain between Input and Output is now x 3. For the 1 Volt peak input the output is about 3 Volts. But this is complicated - needs two bias voltages.
-