I've been meaning to get a web page with this information up for some time (the JoeList post was made in 2000 
). I'll be very interested in people's take on the technique.
http://www.t-linespeakers.org/tubes/triode-trick.html
dave
). I'll be very interested in people's take on the technique.http://www.t-linespeakers.org/tubes/triode-trick.html
dave
How about this? In an output stage the maximum rated screen voltage sets the anode voltage limit of a triode-connected pentode. Dropping it a constant voltage permits a higher B+ and higher maximum output power while (possibly?) retaining triode-like AC operation. What I don't understand is the reasoning behind only dropping it ~30 volts. Of course, this completely ignores the question of the effect on the plate curves.
The maximum screen voltage is routinely exceeded in triode operation of pentodes. Fritz Langford-Smith (of RDH fame) did it when adapting the Williamson design for 807s and found no appreciable decrease in valve life. Telefunken specifies 250V max for the LS50, but then allows 800V under triode operation (or so I think - the datasheet was in German). It doesn't seem internal arcing to or from g2 is particularly likely - especially considering how the grids and their voltage levels are arranged. I suppose internal arcing around the base could be possible, though.
I think the greater risk is exceeding the maximum power dissipation of the screen - so the screen current must be monitored (this is easily done by measuring the voltage across the screen stopper resistor).
Nonetheless, exceeding maximum ratings still makes me nervous...
I think the greater risk is exceeding the maximum power dissipation of the screen - so the screen current must be monitored (this is easily done by measuring the voltage across the screen stopper resistor).
Nonetheless, exceeding maximum ratings still makes me nervous...
rdf said:
This is possible, for instance 6W6, 807 and more significantly 6146 have drastically different screen and plate ratings (or heck, any sweep tube), but there is a flaw: if screen follows plate in parallel at AC, take for example Vp = 600V, Vs = 300V: if plate drops only 300V, screen is already at zero and the tube is past cutoff in this assumption. This is greater than 300V saturation voltage! I figure, if reduced DC screen voltage is the goal, the AC signal must be reduced in respective ratio. (This could be done with a voltage divider to ground feeding a cathode follower supplied by the B+ rail.
But again, in the examples I cited, all have triode voltage limits far above the screen voltage alone; 6W6 is in fact rated for vertical output which requires withstanding a peak above 1kV, cutoff. (6V6 too, although it has no problem with screen voltage.)
The reality is that, at any given point in triode mode, the actual pentode-style operating point is above the high-Is saturation knee. Because screen falls with plate voltage, it is never greater; on the other hand, screen voltage is high, but the plate is high as well.
But anyway, I'm trailing on with long winded, tired replies....
Tim
The other thing not noted is the screen stopper resistor doesn't drop any appreciable voltage at all -- 20mA peak across a 220 ohm resistor is only 4.4V, maybe all of 3% of peak (negative-going, saturation) voltage. Nay, the purpose is to attenuate parasitic oscillations as an RC circuit against mainly the screen-grid (i.e. miller) capacitance.
And again, as I noted, tet/pentodes typically find the saturation knee maybe 10 to 40% below plate voltage, so actually increasing screen voltage (or more accurately, multiplying plate's voltage by the 1.1 to 1.4 factor) would improve performance, assuming screen stays within limits of course.
Tim
And again, as I noted, tet/pentodes typically find the saturation knee maybe 10 to 40% below plate voltage, so actually increasing screen voltage (or more accurately, multiplying plate's voltage by the 1.1 to 1.4 factor) would improve performance, assuming screen stays within limits of course.
Tim
Dave,
The main reason for having a resistor would have to be to prevent parasitic oscillation (some pentodes/beam tetrodes apparently will oscillate if connected as triodes without this resistor). The reason being, even a whopping 50mA of screen current will only develop a 5V drop across a 100W resistor.
EDIT: Oh, I see Tim posted this already. I really should read before posting...
By the way - out of interest's sake, the reason why 6.2V Zeners are preferable is because high voltage Zeners actually use the avalanche effect, while low voltage Zeners use the true Zener action. At 6.2V both mechanisms are present, and they apparently combine to give low temperature coefficient, minimum slope resistance and lowest noise.
Possibly the secret part is something like a TL431? Have a look at the datasheet.
P.S. The Zeners will probably drop only around 3.5V in your schematic (they seem to be forward-biased).
The main reason for having a resistor would have to be to prevent parasitic oscillation (some pentodes/beam tetrodes apparently will oscillate if connected as triodes without this resistor). The reason being, even a whopping 50mA of screen current will only develop a 5V drop across a 100W resistor.
EDIT: Oh, I see Tim posted this already. I really should read before posting...
By the way - out of interest's sake, the reason why 6.2V Zeners are preferable is because high voltage Zeners actually use the avalanche effect, while low voltage Zeners use the true Zener action. At 6.2V both mechanisms are present, and they apparently combine to give low temperature coefficient, minimum slope resistance and lowest noise.
Possibly the secret part is something like a TL431? Have a look at the datasheet.
P.S. The Zeners will probably drop only around 3.5V in your schematic (they seem to be forward-biased).
For the heck of it the scenario was run through LTSpice using an EF86 triode-connected as per Eric Barbour's Svetlana application note: B+ = 250 vdc, unbypassed Rk = 1.5k and Rp = 68k cap coupled to a 500k load. A voltage source replaced the short from plate to screen and its valued varied between 0 (triode) and 200 (!) VDC with an input signal which resulted in 2 V p-p output at 400 Hz in pure triode mode.
Surprisingly little happens. Barbour claims < 0.1% distortion for this circuit, a value the EF86 model used can't match. With the stock circuit and a plate-to-screen voltage Vps = 0 VDC, THD = 1.239444%. With Vps = 200 VDC it sims as 1.197771%. The output drops from 2 to ~ 1.83 V p-p, possible explaining the miniscule reduction in distortion. The biggest change is cathode current, plummetting from 1790 uA to 448 uA. Output impedance increases slightly from ~34k to ~37k.
That's it. If more is happening it's a level simulation or this particular tube model can't capture.
Surprisingly little happens. Barbour claims < 0.1% distortion for this circuit, a value the EF86 model used can't match. With the stock circuit and a plate-to-screen voltage Vps = 0 VDC, THD = 1.239444%. With Vps = 200 VDC it sims as 1.197771%. The output drops from 2 to ~ 1.83 V p-p, possible explaining the miniscule reduction in distortion. The biggest change is cathode current, plummetting from 1790 uA to 448 uA. Output impedance increases slightly from ~34k to ~37k.
That's it. If more is happening it's a level simulation or this particular tube model can't capture.
I gave an EL84 a quick shot under the following conditions:
- primary current = ~48.5 ma
- B+ = 275 VDC
- plate-cathode voltage = ~250 VDC
- cathode resistor = 150 ohms w/ bypass
- OPT = Hammond 1628e approx.
- Vin = 5 volts p-p for about 2.5 volts p-p across the secondary into 8 ohms
The driver is modelled as a perfect voltage source. Triode-connected the THD = 0.413187%, plate current = ~44.45 ma and screen current = ~4.05 ma. Separating the plate and screen by 30 VDC drops the primary current to ~42.45 ma, the plate current to ~38.95 ma, screen current to ~3.55 ma and raises THD to 0.486032%.
The problem here is that a drop in plate + screen current means a drop in current through the cathode resistor and therefore less grid voltage. A triode-connected EL84 models with precious little to spare and runs out of input headroom long before plate current swing. I can't think of any way to simultaneously raise grid voltage, reduce cathode current and apply a plate-screen voltage into a net benefit. Raising the B+ doesn't help, current goes up and grid voltage is regained, but so does idle power and we run smack into the EL84's 12 watt limit.
- primary current = ~48.5 ma
- B+ = 275 VDC
- plate-cathode voltage = ~250 VDC
- cathode resistor = 150 ohms w/ bypass
- OPT = Hammond 1628e approx.
- Vin = 5 volts p-p for about 2.5 volts p-p across the secondary into 8 ohms
The driver is modelled as a perfect voltage source. Triode-connected the THD = 0.413187%, plate current = ~44.45 ma and screen current = ~4.05 ma. Separating the plate and screen by 30 VDC drops the primary current to ~42.45 ma, the plate current to ~38.95 ma, screen current to ~3.55 ma and raises THD to 0.486032%.
The problem here is that a drop in plate + screen current means a drop in current through the cathode resistor and therefore less grid voltage. A triode-connected EL84 models with precious little to spare and runs out of input headroom long before plate current swing. I can't think of any way to simultaneously raise grid voltage, reduce cathode current and apply a plate-screen voltage into a net benefit. Raising the B+ doesn't help, current goes up and grid voltage is regained, but so does idle power and we run smack into the EL84's 12 watt limit.
This muy correct. I'm not sure if the so-called maximum screen voltage rating for multi-grid tube applies to triode connection. I've personally never seen any figures showing current drawn by the screen during operation in triode mode. Anyone know any links?
Also muy correct. I suppose if one is ambitious enough, he could plot the load lines of various pentodes and bpa's with different values for the screen voltages, although it sounds like a full time job to me. It might be easier to make a setup with a separate adjustable REGULATED power supply providing screen potential and diddling from there, measuring distortion as Es varies.
John
It's possible but I have no idea in which direction to look in the EL84 case. The effect of Bill Pearl's circuit is to reduce plate current. Examined purely at the simulation level, this moves a cathode biased EL84 in the wrong direction for maximum output power while appearing to increase distortion at the same time. Maybe it's a technique better suited to push-pull AB than single ended class A?
SY said:
43% DC as in U-L ? I believe that is the point when the internal impedance of the tube suddenly drops and stays low - in other words that point is the optimum ratio (xx%) of the screen load impedance to the plate load impedance - which for example the 807s and KT88s - as measured by H&K and GEC - is 43%.
I hope I am getting this right ..
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.