It looks like the LTP is a 12AY7 (mu = 43), not a 12AX7.
Adding a small negative power supply to feed the cathodes of the LTPs would surely help matters. Perhaps a little 6.3VAC toroid, rectified and smoothed to -5VDC.
Adding a small negative power supply to feed the cathodes of the LTPs would surely help matters. Perhaps a little 6.3VAC toroid, rectified and smoothed to -5VDC.
baudouin0 and rongon,
I am glad to hear from you.
Yes, global negative feedback would be very easy to apply to the stationary grid of the input tube that now simply goes to ground through a 1k resistor.
Compensation should be easy too, because there are only 2 stages, and the output transformer has a good hi frequency square wave response.
The LM334 has a low enough burden voltage, and the 12AY7 has enough bias voltage and enough gain to give plenty of linear drive voltage to the output tubes.
Applying global negative feedback would probably change that, perhaps requiring a negative supply for the LM334.
One thing about this kind of phase splitter is that the 2nd harmonic distortion of each triode is mostly cancelled by the opposite phase of the two cathode currents, that is a form of local negative feedback.
As you later noticed, the input tube is a 12AY7. . . . Sorry for the blurry schematic.
Please remember, my design goal was to make a simple low power 7591 push pull amplifier, and I think I achieved that.
Note: Using a constant current sink that is returned to a negative voltage has been known to damage some tubes that have very fine grid wires.
I had a friend that used a constant current sink from a negative supply to the cathodes of a 12AX7 phase splitter.
Both power up, and hot start events, can drive current from the cathodes to the sensitive grids. The 12AX7 was destroyed over a few power ups and hot starts.
I believe that small signal diode anodes connected to the grids, and diode cathodes connected to the triode cathodes can prevent the triode cathodes from going more than 0.6V negative with respect to the grids, and thus limiting the grid current at power up and hot starts; preventing the grid burnout scenario.
I am glad to hear from you.
Yes, global negative feedback would be very easy to apply to the stationary grid of the input tube that now simply goes to ground through a 1k resistor.
Compensation should be easy too, because there are only 2 stages, and the output transformer has a good hi frequency square wave response.
The LM334 has a low enough burden voltage, and the 12AY7 has enough bias voltage and enough gain to give plenty of linear drive voltage to the output tubes.
Applying global negative feedback would probably change that, perhaps requiring a negative supply for the LM334.
One thing about this kind of phase splitter is that the 2nd harmonic distortion of each triode is mostly cancelled by the opposite phase of the two cathode currents, that is a form of local negative feedback.
As you later noticed, the input tube is a 12AY7. . . . Sorry for the blurry schematic.
Please remember, my design goal was to make a simple low power 7591 push pull amplifier, and I think I achieved that.
Note: Using a constant current sink that is returned to a negative voltage has been known to damage some tubes that have very fine grid wires.
I had a friend that used a constant current sink from a negative supply to the cathodes of a 12AX7 phase splitter.
Both power up, and hot start events, can drive current from the cathodes to the sensitive grids. The 12AX7 was destroyed over a few power ups and hot starts.
I believe that small signal diode anodes connected to the grids, and diode cathodes connected to the triode cathodes can prevent the triode cathodes from going more than 0.6V negative with respect to the grids, and thus limiting the grid current at power up and hot starts; preventing the grid burnout scenario.
Last edited:
That's common for cathode followers but I would not think you would have had a problem with just a few volts maybe only -2-5V or so and ~1ma. Normally its arcing that is the source of failure. Its only the same as raising the grid voltage but you know this.
I think you could get enough current (1ma or so) off the existing heater supply by using the 100R's as the return path to ground.
Applying GNF will not affect the distortion cancellation in the LPT as its in the forward path. You may not need such low plate resistors for the first stage which would increase gain and raise the cathode voltage further. However with GNF the cathode ac voltage will increase and the LM334 needs at least .9v to function correctly. If it runs out you will soon see a AC in balance on the 12ay7 plates.
Nevertheless if the design brief is few components and good distortion that's what you have achieved. I think its a nice simple design.
I think you could get enough current (1ma or so) off the existing heater supply by using the 100R's as the return path to ground.
Applying GNF will not affect the distortion cancellation in the LPT as its in the forward path. You may not need such low plate resistors for the first stage which would increase gain and raise the cathode voltage further. However with GNF the cathode ac voltage will increase and the LM334 needs at least .9v to function correctly. If it runs out you will soon see a AC in balance on the 12ay7 plates.
Nevertheless if the design brief is few components and good distortion that's what you have achieved. I think its a nice simple design.
Last edited:
baudouin0,
I wanted to use a choke input filter, and I already had the power transformer with its rated secondary voltage output.
I wanted to use the 12AY7 triodes somewhere in the middle of their more linear range, and also to have enough plate current to be able to slew the g1 to g2 Miller Effect Capacitance that is caused by Ultra Linear screen voltage swing.
The fixed B+ voltage, plate voltage, plate current, and bias voltage all dictated that I use 47k plate loads.
Increasing the plate loads to more than 47k would have destroyed that part of the design, given the B+ that I had to work with (I would have to change from choke input, to cap input; and that does heat up the power transformer primary more).
It is as simple as that, and as complex as that.
As you noted, using global negative feedback pretty much requires a re-design of the input stage, unless you want to deal with lower overall gain.
And, the LM334 was the perfect thing to use because I did not use global negative feedback, so the dynamic compliance range was within the LM334 limits.
The case of another amplifier, where a 12AX7 had a negative supply for the constant current sink . . . each cathode current was between 1mA to 2mA, and the tube died very early on, when the grids were destroyed, or whatever. I seriously doubt that the grid cathode interface was arcing.
I did not want to experiment with the life of the 12AY7 grids, another reason to stay away from using a negative supply for the LM334. I am sure the 12AY7 grids can take more than a 12AX7, but the LM334 and 12AY7 was higher current than was used in the 12AX7 case.
Building a negative supply for the LM334 would have been as simple as adding a pair of schottky diodes, cap filter, series resistor, and cap filter.
As few parts as that is, it would have been difficult finding room to stuff them all under the chassis.
Tradeoffs, tradeoffs.
I wanted to use a choke input filter, and I already had the power transformer with its rated secondary voltage output.
I wanted to use the 12AY7 triodes somewhere in the middle of their more linear range, and also to have enough plate current to be able to slew the g1 to g2 Miller Effect Capacitance that is caused by Ultra Linear screen voltage swing.
The fixed B+ voltage, plate voltage, plate current, and bias voltage all dictated that I use 47k plate loads.
Increasing the plate loads to more than 47k would have destroyed that part of the design, given the B+ that I had to work with (I would have to change from choke input, to cap input; and that does heat up the power transformer primary more).
It is as simple as that, and as complex as that.
As you noted, using global negative feedback pretty much requires a re-design of the input stage, unless you want to deal with lower overall gain.
And, the LM334 was the perfect thing to use because I did not use global negative feedback, so the dynamic compliance range was within the LM334 limits.
The case of another amplifier, where a 12AX7 had a negative supply for the constant current sink . . . each cathode current was between 1mA to 2mA, and the tube died very early on, when the grids were destroyed, or whatever. I seriously doubt that the grid cathode interface was arcing.
I did not want to experiment with the life of the 12AY7 grids, another reason to stay away from using a negative supply for the LM334. I am sure the 12AY7 grids can take more than a 12AX7, but the LM334 and 12AY7 was higher current than was used in the 12AX7 case.
Building a negative supply for the LM334 would have been as simple as adding a pair of schottky diodes, cap filter, series resistor, and cap filter.
As few parts as that is, it would have been difficult finding room to stuff them all under the chassis.
Tradeoffs, tradeoffs.
Last edited:
@6A3sUMMER thank you for rising the point of the negative supply for CCS.
I use it and I've never had issues, but I always thought it was better not to supply it at the same negative voltage of the source followers that drive the output tubes, but at around -12V.
I use it and I've never had issues, but I always thought it was better not to supply it at the same negative voltage of the source followers that drive the output tubes, but at around -12V.
Certainly I have used a few negative volts on the cathode before warmup on many 12ax7 and had no issue. The datasheet does say max positive grid voltage of 0V. I think as long as the grid current is very small all will be OK, so a AC coupled input stage with say grid leak 100k to ground is not going to cause a problem as only uA could flow which will not damage the internals. I guess you had a circuit which could cause significant grid current to flow and blow the fine wire. It would be interesting to see the circuit that pops the 12ax7.
Triode-strapping the 7591s would be the way to go then. How do you like the sound? I once helped in the building of a PP 7591-triode amp. It used something like a 5687 LTP with a center-tapped plate choke, DC-coupled to the 7591-triodes, with stacked B+ supplies to enable the DC-coupling. The builder and I decided we didn't care that much for the sound of the 7591s in triode, but we were comparing them to 2A3, 6B4G, and the like. I love the idea of the 7591 -- easy to drive, low control grid voltage requirement (usually about -16 to -22V or so for class AB1 operation). The plate dissipation isn't huge, at only 22W (plate+g2).
Looking at the load lines for 7591A triode, I think PP Class A with Zpri = 5k p-p, Vp = 350V, Ip = 60mA (for Pdiss = 21W), Rk would be about 200R per triode (or a shared 100R 5W). Looks promising...
baudouin0,
It was not my circuit that blew out the 12AX7. But I believe the current sink was between 1.5 ma and 2mA per tube (3 to 4 mA total).
As I best can remember, the input signal was to only one grid, the input resistor was high resistance to ground.
But the other grid was connected to a 100 Ohm resistor to ground. So, there may have been 3mA or 4mA all going through the one grid.
That seems like a probable cause of the destruction of the 12AX7.
A lot of this scenario requires that the negative supply voltage for the constant current sink comes up quickly, solid state diodes; and requires a slow starting B+ that uses an indirectly heated tube rectifier, which brings up the B+ after the negative supply is already at full voltage.
rongon,
The 2.5k load line you drew for a 5k plate to plate push pull transformer only applies when both tubes are conducting current (Class A).
When one tube is cut off, like in Class AB, the other tube sees a 1.25k load line. That is a much steeper load line.
That kind of operation requires lots of negative feedback, or there will be large amounts of 3rd harmonic distortion at larger output power (when one tube is at or near cut off).
Using -12V quiescent grid bias, the load line shows 12V positive direction swing to 0V, but only shows 6V negative direction swing to -18V. The rest of the symmetrical swing is from -18V to -24V.
There is very little plate current when the grid is at -18V to -24V (not even shown on the tube curves).
At that time, the other tube will effectively be working into a 1.25k load line all by itself, it will not be working into a 2.5k load line then.
It was not my circuit that blew out the 12AX7. But I believe the current sink was between 1.5 ma and 2mA per tube (3 to 4 mA total).
As I best can remember, the input signal was to only one grid, the input resistor was high resistance to ground.
But the other grid was connected to a 100 Ohm resistor to ground. So, there may have been 3mA or 4mA all going through the one grid.
That seems like a probable cause of the destruction of the 12AX7.
A lot of this scenario requires that the negative supply voltage for the constant current sink comes up quickly, solid state diodes; and requires a slow starting B+ that uses an indirectly heated tube rectifier, which brings up the B+ after the negative supply is already at full voltage.
rongon,
The 2.5k load line you drew for a 5k plate to plate push pull transformer only applies when both tubes are conducting current (Class A).
When one tube is cut off, like in Class AB, the other tube sees a 1.25k load line. That is a much steeper load line.
That kind of operation requires lots of negative feedback, or there will be large amounts of 3rd harmonic distortion at larger output power (when one tube is at or near cut off).
Using -12V quiescent grid bias, the load line shows 12V positive direction swing to 0V, but only shows 6V negative direction swing to -18V. The rest of the symmetrical swing is from -18V to -24V.
There is very little plate current when the grid is at -18V to -24V (not even shown on the tube curves).
At that time, the other tube will effectively be working into a 1.25k load line all by itself, it will not be working into a 2.5k load line then.
Last edited:
Yep I cannot find any data for what actual current would destroy the 12ax7 grid. The circuits I has used would be 1ma CCS, HT and heaters comes up quick and there are 2k2 resistors in series with each grid. I guess being fine wire in a vacuum not much current is needed.
If the datasheet states that the absolute maximum g1-to-cathode voltage is zero, then a well designed circuit for volume production should never go above that value. For DIY is fine though.
Yes that's all I could find and I think that's with valve warm. I assume it is the current that melts the wire. A few ma does not sound very much but its very thin and in a vacuum. Obviously when the circuit is operating normally it won't happen, its as it warms up grid current could flow if either HT is missing or there's not enough emission yet to sustain the CCS. Anyway some grid resistors will limit the current as long as the CCS source voltage is not too negative.
I see a +/-100V window of operation in which the output stage should stay close to class A operation. I think 20mA of plate current is enough to keep the 7591-triode conducting. Granted its rp will increase compared to the other 7591 in the pair at that point, but it will still be conducting.
I don't see any way to run push-pull 7591-triodes in class A with something like a 400V plate-cathode voltage. With an 8k p-p load, the plate-cathode voltage would need to be down around 320V. You won't get much power out that way, but it will be clean.
You're right that a 10k p-p load looks better. Not much power, but better damping for sure.