I concluded the following general rules for myself:
- The higher is the idle current, the lower is the THD
- The higher is the idle current, the closer is the g1 voltage to zero, i.e. the lower is the input voltage swing (grid current limit)
- The higher is the idle current, the higher is the power dissipation
- The higher is the anode voltage, the lower is the THD
- The higher is the anode voltage, the higher is the power dissipation
- If higher input voltage swing is needed, apply a negative feedback to the unbypassed cathode resistor (that sets g1 voltage hence bias current anyway)
- With low power gain stages I set anode voltage = 2/3 x supply voltage, and idle current = 0.8 x (max. permitted power dissipation) / anode voltage
- Always keep within max. permitted operating conditions (anode voltage, idle current, g1 to GND resistor, max. power dissipation)
- The higher is the idle current, the lower is the THD
- The higher is the idle current, the closer is the g1 voltage to zero, i.e. the lower is the input voltage swing (grid current limit)
- The higher is the idle current, the higher is the power dissipation
- The higher is the anode voltage, the lower is the THD
- The higher is the anode voltage, the higher is the power dissipation
- If higher input voltage swing is needed, apply a negative feedback to the unbypassed cathode resistor (that sets g1 voltage hence bias current anyway)
- With low power gain stages I set anode voltage = 2/3 x supply voltage, and idle current = 0.8 x (max. permitted power dissipation) / anode voltage
- Always keep within max. permitted operating conditions (anode voltage, idle current, g1 to GND resistor, max. power dissipation)
Some tube types draw g1 current even when g1 is at about -0.5V, or even -1V versus the cathode voltage.
This is caused by the cathode dimensions and shape, grid dimensions and shape, spacing from grid to cathode, the space charge electrons from the cathode, and spacing from grid to plate.
That small g1 current sometimes loads the previous stage, causing the previous stage to distort.
I am not talking about when g1 is +0.5V or + 1V higher than the cathode.
That cause of grid current is more obvious and more well known.
The g1 to cathode bias voltage that you pick for your circuit, may cause the positive alternations of the signal voltage to draw grid current, even though g1 is still negative with respect to the cathode voltage.
One example:
A 12AX7 / ECC83 is operated at -1/2 Volt grid bias.
The quiescent plate voltage is 40V.
The grid is very close to the cathode, but the grid further away from the plate.
The low plate voltage is not high enough to pull the space charge electrons through the very tightly packed grid wires.
Grid current is drawn, even before signal is applied.
Depending on the manufacturer of your 12AX7 / ECC83, Your Mileage May Vary.
Do spice models of the 12AX7 / ECC83 compensate for this effect?
Just asking.
This is caused by the cathode dimensions and shape, grid dimensions and shape, spacing from grid to cathode, the space charge electrons from the cathode, and spacing from grid to plate.
That small g1 current sometimes loads the previous stage, causing the previous stage to distort.
I am not talking about when g1 is +0.5V or + 1V higher than the cathode.
That cause of grid current is more obvious and more well known.
The g1 to cathode bias voltage that you pick for your circuit, may cause the positive alternations of the signal voltage to draw grid current, even though g1 is still negative with respect to the cathode voltage.
One example:
A 12AX7 / ECC83 is operated at -1/2 Volt grid bias.
The quiescent plate voltage is 40V.
The grid is very close to the cathode, but the grid further away from the plate.
The low plate voltage is not high enough to pull the space charge electrons through the very tightly packed grid wires.
Grid current is drawn, even before signal is applied.
Depending on the manufacturer of your 12AX7 / ECC83, Your Mileage May Vary.
Do spice models of the 12AX7 / ECC83 compensate for this effect?
Just asking.
Last edited:
These discussions, without us all looking at the same set of plate curves, are difficult to the point of useless, but some generalizations are true with a capital T. 6A3sUMMER sets a minimum value for (ordinary, Class A1, no drive current drama) peak positive drive voltage. Idle ("bias") voltage needs to be that amount minus peak signal voltage including some margin. Easy enough.
Triodes (of all kinds) are most linear at the highest idling current that doesn't melt them, and at the lightest loading possible with your(!) design B+ and next stage drive requirements.
So, if you care about (driver, or any other triode stage) linearity you want to provide it with as high a B+ as you can afford (per artosalo), and therefor as large a plate load resistor as possible. For low mu output stages this may call for larger driver B+ voltages than output valves' voltages. Shocking! because it wasn't done in the 1950's or whatever.
We perhaps too easily fall into a trap of continuing design elements from Golden Age audio design. That's not always a bad start, but recognizing strengths and limitations and choosing within the buffet - well, that's where we are today.
All good fortune,
Chris
Triodes (of all kinds) are most linear at the highest idling current that doesn't melt them, and at the lightest loading possible with your(!) design B+ and next stage drive requirements.
So, if you care about (driver, or any other triode stage) linearity you want to provide it with as high a B+ as you can afford (per artosalo), and therefor as large a plate load resistor as possible. For low mu output stages this may call for larger driver B+ voltages than output valves' voltages. Shocking! because it wasn't done in the 1950's or whatever.
We perhaps too easily fall into a trap of continuing design elements from Golden Age audio design. That's not always a bad start, but recognizing strengths and limitations and choosing within the buffet - well, that's where we are today.
All good fortune,
Chris
Last edited:
Have a play with a piece of software like the trioda.com or dmitry's spice parameter programme for curve painting. See Model Paint Tools: Trace Tube Parameters over Plate Curves, Interactively
This will lead to conclude that the flattest loadline, with operating point furthest into top left corner of the graph will, in theory, give the cleanest output subject to your not drawing grid current. The latter isn't obvious from software simulations as few models incorporate parameters for that.
This is also why you consistently see topics and posts which refer to source followers and gyrators, pentodes in triode mode and directly heated triodes ("DHT"). The source followers cater for the grid current drawn by a tube but not supplied by the previous stage. The gyrator type circuits flatten the loadline by mimicking a high B+ voltage with infinitely high load resistor. The DHT's generally present the most linear plate curves and the triode run pentodes have plate curves whose linearity matches them.
kind regards
Marek
This will lead to conclude that the flattest loadline, with operating point furthest into top left corner of the graph will, in theory, give the cleanest output subject to your not drawing grid current. The latter isn't obvious from software simulations as few models incorporate parameters for that.
This is also why you consistently see topics and posts which refer to source followers and gyrators, pentodes in triode mode and directly heated triodes ("DHT"). The source followers cater for the grid current drawn by a tube but not supplied by the previous stage. The gyrator type circuits flatten the loadline by mimicking a high B+ voltage with infinitely high load resistor. The DHT's generally present the most linear plate curves and the triode run pentodes have plate curves whose linearity matches them.
kind regards
Marek
Yes that would work, provided the clipping is not enormous.
Higher output impedance, more Miller capacitance. i.e. you get less bandwidth