Nah, what I meant was that for triode gain stages, distortion usually falls with increasing RL until around 50 ra is reached. After that there isn't really too much point increasing RL any further. Of course one certainly wouldn't do this for output stages! RL for a SET output stage is usually 3 to 5 times ra.
X*ra
I did know of the 2*ra rule. (MJ mentions it)
I felt kind of puzzled when I found by iterative loadline drawing that a good compromise between noise and power was the 3.7*ra relationship that I found with the 211s. For the 845s and 211s I actually could not get any decent 2*ra loadline in terms of THD compared to >3*ra.
Rada
I did know of the 2*ra rule. (MJ mentions it)
I felt kind of puzzled when I found by iterative loadline drawing that a good compromise between noise and power was the 3.7*ra relationship that I found with the 211s. For the 845s and 211s I actually could not get any decent 2*ra loadline in terms of THD compared to >3*ra.
Rada
> > to get 3V peak-to-peak of grid swing and 100V peak-to-peak plate swing, I had to lower the resistance of the load
That's like tying a boulder to your car to keep it down to 65MPH. Cars have throttles, audio systems have volume controls. You only crank it up as much as you need to. A power amp's input sensitivity can be almost anything, within reason. Some amps need 2V or more to make full output, and may need a little gain in a preamp. Some amps have sensitivity under 1 volt: you just don't turn the volume knob so far.
Yes, volume-pots are imperfect devices. But necessary (a pot, switch, or some type of gain/output control) and not as "bad" as making a tube pull a heavy rock or resistor.
Why are "Low Loads BAD!"? Look at how a triode works. (Pentodes are funny beasts.)
A triode is a feedback device. The effective field at the cathode (where the electrons start) is the sum of the grid and plate voltages, actually the plate voltage divided by a number "Mu", generally between 5 and 50. Mu can be measured with a ruler: to a good approximation it depends only on physical dimensions. Therefore if you work a triode into an infinite load impedance, loadline dead horizontal, the gain is always Mu at any instant: zero distortion.
OTOH, a tube's transconductance Gm is a function of current. And for many tubes, Gm varies about as square-root of current. So if you work a triode into a zero ohm load, dead vertical loadline, gain is different at any instant over an audio cycle, and for large swings runs around 10% before gross clipping.
In real tubes: Mu isn't dead-constant. When Vg gets to Vp/Mu, you would expect current to cut-off; in fact most of the cathode cuts off but now we discover electrons leaking around the ends of a finite cathode-grid structure. And perfect Mu means perfect grid winding, which never happens; and sometimes the winding is "shaped" to give better results at typical currents while allowing atypical operation (or very wide-swing operation) to slack-off. Gm should vary with current, and does; but many of the old-time audio tubes can get into a high-current range where Gm is nearly constant (acting as if there were some dead resistance in the cathode, which may be true).
Still, unless a tube is very bent, you get lower THD and higher voltage gain with high impedance loads, much more than Rp. Of course the current gain and power available tends to zero, and there are no infinite loads, so you have to pick something practical. But reducing load impedance more than necessary just causes strain. "Excess" gain is rarely a problem. You have a volume control. If it has to be set awkwardly low, and there isn't an obvious design change to reduce gain (lo-Mu tubes are sometimes cleaner and damp better), then two resistors as an input pad is only $0.24 and does little to no harm.
Here is a classic curve for triode loads. It is "flawed" because drive voltage is held constant. Actually, the higher-Z loads can take more drive before clipping. So at 5*Rp, both power and distortion could be higher.
Jason may be right that the curve falls out to 50*Rp. In resistor or choke loading, we can rarely manage loads that high: you'd need a current-source and buffer more complicated (and potentially flawed) than the raw amplifier. And the power output would be very low, or require huge heat in the tube to drive the load. Between Rp and 50*Rp, numbers like 2*Rp to 10*Rp tend to be most practical.
That's like tying a boulder to your car to keep it down to 65MPH. Cars have throttles, audio systems have volume controls. You only crank it up as much as you need to. A power amp's input sensitivity can be almost anything, within reason. Some amps need 2V or more to make full output, and may need a little gain in a preamp. Some amps have sensitivity under 1 volt: you just don't turn the volume knob so far.
Yes, volume-pots are imperfect devices. But necessary (a pot, switch, or some type of gain/output control) and not as "bad" as making a tube pull a heavy rock or resistor.
Why are "Low Loads BAD!"? Look at how a triode works. (Pentodes are funny beasts.)
A triode is a feedback device. The effective field at the cathode (where the electrons start) is the sum of the grid and plate voltages, actually the plate voltage divided by a number "Mu", generally between 5 and 50. Mu can be measured with a ruler: to a good approximation it depends only on physical dimensions. Therefore if you work a triode into an infinite load impedance, loadline dead horizontal, the gain is always Mu at any instant: zero distortion.
OTOH, a tube's transconductance Gm is a function of current. And for many tubes, Gm varies about as square-root of current. So if you work a triode into a zero ohm load, dead vertical loadline, gain is different at any instant over an audio cycle, and for large swings runs around 10% before gross clipping.
In real tubes: Mu isn't dead-constant. When Vg gets to Vp/Mu, you would expect current to cut-off; in fact most of the cathode cuts off but now we discover electrons leaking around the ends of a finite cathode-grid structure. And perfect Mu means perfect grid winding, which never happens; and sometimes the winding is "shaped" to give better results at typical currents while allowing atypical operation (or very wide-swing operation) to slack-off. Gm should vary with current, and does; but many of the old-time audio tubes can get into a high-current range where Gm is nearly constant (acting as if there were some dead resistance in the cathode, which may be true).
Still, unless a tube is very bent, you get lower THD and higher voltage gain with high impedance loads, much more than Rp. Of course the current gain and power available tends to zero, and there are no infinite loads, so you have to pick something practical. But reducing load impedance more than necessary just causes strain. "Excess" gain is rarely a problem. You have a volume control. If it has to be set awkwardly low, and there isn't an obvious design change to reduce gain (lo-Mu tubes are sometimes cleaner and damp better), then two resistors as an input pad is only $0.24 and does little to no harm.
Here is a classic curve for triode loads. It is "flawed" because drive voltage is held constant. Actually, the higher-Z loads can take more drive before clipping. So at 5*Rp, both power and distortion could be higher.
An externally hosted image should be here but it was not working when we last tested it.
Jason may be right that the curve falls out to 50*Rp. In resistor or choke loading, we can rarely manage loads that high: you'd need a current-source and buffer more complicated (and potentially flawed) than the raw amplifier. And the power output would be very low, or require huge heat in the tube to drive the load. Between Rp and 50*Rp, numbers like 2*Rp to 10*Rp tend to be most practical.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.