more complete:
List of knee currents for g2 = 150 V, g1 = 0 V, Vp = 60 to 80 V
<tube> <Watts> <mA knee@150Vg2> <gm> <Mu> <maxDCmA> <registered by> <date>
6LF6 40W 1144mA@150V 15K@125mA Mu3 500mADC Amperex 1968
6KG6/EL509 34W 1135mA@150V 13K@150mA Mu3.2 500mADC Amperex 1965
6MC6 33W 1130mA@150V 14K@125mA Mu4 400mADC RCA (6LX6 clone) 1972
13E1 90W(absmax) 1120mA@150V 35K@500mA Mu4.5 800mADC AEI 1961
6MH6 38.3W 1100mA@150V 14K@125mA Mu4 500mADC GE (up-rated 6LX6,6KD6,26HU5) 1972
6MB6 38W 1085mA@150V 14K@110mA Mu3.5 400mADC Sylvania 1971
6LR6 30W 1085mA@150V 16K@140mA Mu3.5 375mADC Sylvania 1968
6LX6/6KD6/26HU5 33W 1080mA@150V 14K@125mA Mu4 400mADC GE 1969/1965/1969
6LW6 40W 1050mA@150V 12K@125mA Mu3.7 400mADC GE 1971
6KN6 30W 1050mA@150V 16K@100mA Mu4.5 400mADC Sylvania 1965 (later versions are 6KD6)
6LZ6 30W 940mA@150V 11K@140mA Mu3 350mADC RCA 1971
6LB6/A 30W/35W 825mA@150V 13.4K@105mA Mu4 315mADC GE 1967
6JE6C/6JS6C 30W 789mA@150V 10.5K@130mA Mu3 350mADC Sylvania 68/69
6JE6 24W 762mA@150V 9.6K@115mA Mu3 315mADC RCA 1962
6JS6/6HF5 28W 749mA@150V 11.5K@130mA Mu3 315mADC GE 1964/1963
6MJ6 30W 740mA@150V 11K@100mA Mu3.6 350mADC RCA 1973
6LG6 28W 740mA@150V 11.5K@90mA Mu3.6 315mADC GE 1967
6LQ6 30W 715mA@150V 7.5K@95mA Mu3 350mADC RCA 1967
6ME6 30W 700mA@150V 9.6@130mA Mu3.5 350mADC RCA 1971
6DQ5 24W 690mA@150V 10.5K@110mA Mu3.3 315mADC RCA 1957
6JF6/6JG6 17W 660mA@150V 10K@80mA Mu4.1 275mADC RCA 1965/1964
6KM6 20W 630mA@150V 9.5K@80mA Mu4 275mADC RCA 1965
6HD5/6HJ5 24W 630mA@150V 10K@80mA Mu4.2 280mADC Raytheon 1962/1963
6JR6/6JU6 17W 600mA@150V 7K@45mA Mu4.7 275mADC RCA 1968/1966
6JZ6/21HB5A 18W 560mA@150V 9K@46mA Mu4.8 230mADC GE 1966/1964
12HE7 10-15W 540mA@150V 8.8K@60mA Mu4.2 200mADC GE (15W if damper disabled) 1964
6CL5 25W 514mA@150V 6.5K@90mA Mu3 240mADC Sylvania 1955
6GB5/29KQ6/EL500 17W 500mA@150V 13K@100mA Mu5.1 275mADC Amperex 1961/Matsushita 1959/Philips 1961?
6KV6/A 20-28W 488/610mA@150V 6K@40mA Mu4 275mADC RCA (re-rated 6KM6?) 1967/1969
6HB5/6GY5/21JV6/6KE6/16KA6 18W 475mA@150V 9.1K@50mA Mu4.7 230mADC GE/GE/GE/Ray/Tung 1962/1962/1965/1965/1964
6EX6 22W 460mA@150V 7.7K@67mA Mu4.2 220mADC Raytheon (up-rated 6CD6) 1959
6CB5/A 23W 440mA@150V 8.8K@90mA Mu3.8 240mADC RCA 1954/1956
6CD6/GA 15/20W 422mA@150V 7.7K@75mA Mu3.9 200mADC RCA/GE 1949/1954
6GT5/6GJ5/6JT6/6JB6/6GW6 17.5W 380mA@150V 7.1K@70mA Mu4.4 175mADC RCA 1961/1961/1964/1962/1961
6GE5 17.5W 350mA@150V 7.3K@65mA Mu4.4 175mADC GE 1961
6GF5 9W 345mA@150V 4.7K@34mA Mu4.2 160mADC GE 1961
6JM6/6JN6/6FW5/6GC6 17.5W 340mA@150V 7.3K@70mA Mu4.4 175mADC GE 1964/1964/1960/1960
6DQ6B/6GV5 17.5W/18W 330mA@150V 7.3K@65mA Mu4.4 175mADC GE 1959/1962
6DQ6/A 18W 280mA@150V 6.6K@55mA Mu4.1 120/155mADC CBS/RCA 1955/1956
6JA5/10JA5 19W 276mA@150V 10.3K@95mA Mu5.5 110mADC GE 1971
6LU8/6LR8//6MY8 14//16W 265mA@150V 9.3K@56mA Mu6.5 75mADC Sylvania 1964/1964//1970(Toshiba)
6AV5/GA///6BQ6/GA 11W 255ma@150V 5.9K@57mA Mu4.3 110mADC CBS/GE 1949/1955 /// CBS/Syl 1949/1953
KT120 60W 221ma@150V 190mADC
KT90 50W 220mA@150V
6Y6G/GT/GA 12.5W 200mA@150V Ray 1937/KenRad 1939/Syl 1954
6550A 35W/42W 190mA@150V 11K@140mA 190mADC
6W6GT 10W 185mA@150V 8K@46mA Mu6.2 65mADC CBS 1939
KT88 35W 170mA@150V 175mADC
6CA7/EL34 25W 107mA@150V 11K@100mA Mu10.5 150mADC Philips 1952
6JC5 19W 80mA@150V 4.1K@43mA Mu7 75mADC Sylvania 1971
6L6/G/GA/GB/GC 30W 77mA@150V 4.7K@40mA Mu8 110mADC RCA 1936/Ray 1936/Syl 1943/Syl 1954/GE 1958
6HB6 10W 70mA@150V 20K@40mA Mu33 60mADC Raytheon 1961
6GK6 13.2W 65mA@150V 11.3K@48mA Mu19 65mADC CBS 1959
6BQ5/EL84 12W 65mA@150V 11.3K@48mA Mu19.5 65mADC Rogers 1956
6V6G/GT 14W 45mA@150V 4.1K@45mA Mu9.8 40mADC KenRad 1936/CBS 1939
List of knee currents for g2 = 150 V, g1 = 0 V, Vp = 60 to 80 V
<tube> <Watts> <mA knee@150Vg2> <gm> <Mu> <maxDCmA> <registered by> <date>
6LF6 40W 1144mA@150V 15K@125mA Mu3 500mADC Amperex 1968
6KG6/EL509 34W 1135mA@150V 13K@150mA Mu3.2 500mADC Amperex 1965
6MC6 33W 1130mA@150V 14K@125mA Mu4 400mADC RCA (6LX6 clone) 1972
13E1 90W(absmax) 1120mA@150V 35K@500mA Mu4.5 800mADC AEI 1961
6MH6 38.3W 1100mA@150V 14K@125mA Mu4 500mADC GE (up-rated 6LX6,6KD6,26HU5) 1972
6MB6 38W 1085mA@150V 14K@110mA Mu3.5 400mADC Sylvania 1971
6LR6 30W 1085mA@150V 16K@140mA Mu3.5 375mADC Sylvania 1968
6LX6/6KD6/26HU5 33W 1080mA@150V 14K@125mA Mu4 400mADC GE 1969/1965/1969
6LW6 40W 1050mA@150V 12K@125mA Mu3.7 400mADC GE 1971
6KN6 30W 1050mA@150V 16K@100mA Mu4.5 400mADC Sylvania 1965 (later versions are 6KD6)
6LZ6 30W 940mA@150V 11K@140mA Mu3 350mADC RCA 1971
6LB6/A 30W/35W 825mA@150V 13.4K@105mA Mu4 315mADC GE 1967
6JE6C/6JS6C 30W 789mA@150V 10.5K@130mA Mu3 350mADC Sylvania 68/69
6JE6 24W 762mA@150V 9.6K@115mA Mu3 315mADC RCA 1962
6JS6/6HF5 28W 749mA@150V 11.5K@130mA Mu3 315mADC GE 1964/1963
6MJ6 30W 740mA@150V 11K@100mA Mu3.6 350mADC RCA 1973
6LG6 28W 740mA@150V 11.5K@90mA Mu3.6 315mADC GE 1967
6LQ6 30W 715mA@150V 7.5K@95mA Mu3 350mADC RCA 1967
6ME6 30W 700mA@150V 9.6@130mA Mu3.5 350mADC RCA 1971
6DQ5 24W 690mA@150V 10.5K@110mA Mu3.3 315mADC RCA 1957
6JF6/6JG6 17W 660mA@150V 10K@80mA Mu4.1 275mADC RCA 1965/1964
6KM6 20W 630mA@150V 9.5K@80mA Mu4 275mADC RCA 1965
6HD5/6HJ5 24W 630mA@150V 10K@80mA Mu4.2 280mADC Raytheon 1962/1963
6JR6/6JU6 17W 600mA@150V 7K@45mA Mu4.7 275mADC RCA 1968/1966
6JZ6/21HB5A 18W 560mA@150V 9K@46mA Mu4.8 230mADC GE 1966/1964
12HE7 10-15W 540mA@150V 8.8K@60mA Mu4.2 200mADC GE (15W if damper disabled) 1964
6CL5 25W 514mA@150V 6.5K@90mA Mu3 240mADC Sylvania 1955
6GB5/29KQ6/EL500 17W 500mA@150V 13K@100mA Mu5.1 275mADC Amperex 1961/Matsushita 1959/Philips 1961?
6KV6/A 20-28W 488/610mA@150V 6K@40mA Mu4 275mADC RCA (re-rated 6KM6?) 1967/1969
6HB5/6GY5/21JV6/6KE6/16KA6 18W 475mA@150V 9.1K@50mA Mu4.7 230mADC GE/GE/GE/Ray/Tung 1962/1962/1965/1965/1964
6EX6 22W 460mA@150V 7.7K@67mA Mu4.2 220mADC Raytheon (up-rated 6CD6) 1959
6CB5/A 23W 440mA@150V 8.8K@90mA Mu3.8 240mADC RCA 1954/1956
6CD6/GA 15/20W 422mA@150V 7.7K@75mA Mu3.9 200mADC RCA/GE 1949/1954
6GT5/6GJ5/6JT6/6JB6/6GW6 17.5W 380mA@150V 7.1K@70mA Mu4.4 175mADC RCA 1961/1961/1964/1962/1961
6GE5 17.5W 350mA@150V 7.3K@65mA Mu4.4 175mADC GE 1961
6GF5 9W 345mA@150V 4.7K@34mA Mu4.2 160mADC GE 1961
6JM6/6JN6/6FW5/6GC6 17.5W 340mA@150V 7.3K@70mA Mu4.4 175mADC GE 1964/1964/1960/1960
6DQ6B/6GV5 17.5W/18W 330mA@150V 7.3K@65mA Mu4.4 175mADC GE 1959/1962
6DQ6/A 18W 280mA@150V 6.6K@55mA Mu4.1 120/155mADC CBS/RCA 1955/1956
6JA5/10JA5 19W 276mA@150V 10.3K@95mA Mu5.5 110mADC GE 1971
6LU8/6LR8//6MY8 14//16W 265mA@150V 9.3K@56mA Mu6.5 75mADC Sylvania 1964/1964//1970(Toshiba)
6AV5/GA///6BQ6/GA 11W 255ma@150V 5.9K@57mA Mu4.3 110mADC CBS/GE 1949/1955 /// CBS/Syl 1949/1953
KT120 60W 221ma@150V 190mADC
KT90 50W 220mA@150V
6Y6G/GT/GA 12.5W 200mA@150V Ray 1937/KenRad 1939/Syl 1954
6550A 35W/42W 190mA@150V 11K@140mA 190mADC
6W6GT 10W 185mA@150V 8K@46mA Mu6.2 65mADC CBS 1939
KT88 35W 170mA@150V 175mADC
6CA7/EL34 25W 107mA@150V 11K@100mA Mu10.5 150mADC Philips 1952
6JC5 19W 80mA@150V 4.1K@43mA Mu7 75mADC Sylvania 1971
6L6/G/GA/GB/GC 30W 77mA@150V 4.7K@40mA Mu8 110mADC RCA 1936/Ray 1936/Syl 1943/Syl 1954/GE 1958
6HB6 10W 70mA@150V 20K@40mA Mu33 60mADC Raytheon 1961
6GK6 13.2W 65mA@150V 11.3K@48mA Mu19 65mADC CBS 1959
6BQ5/EL84 12W 65mA@150V 11.3K@48mA Mu19.5 65mADC Rogers 1956
6V6G/GT 14W 45mA@150V 4.1K@45mA Mu9.8 40mADC KenRad 1936/CBS 1939
Last edited:
Thanks...i have some pieces also with of this tube and I want to put them in use...with that loadline i can now see the nice operating conditions.
@smoking amps
that's a great data and will be useful to us..
Well, I haven't designed any screen drive output stages yet, but I think the procedure is much the same as for grid1 drive. You just plot a load line on the screen grid curve set, instead of the grid1 curve set. Usually one wants the load line to cross just a bit above the knee of the curve for the screen voltage being used, so that overload cannot push it into heavy screen current. That will be even more important here, so stay above the knee.
Since the screen grid curves are so linear versus Vg2, you can choose most any load line for a P-P stage. A steeper load line will give more 2nd harmonic, which will cancel out in P-P, but will produce more 3rd and 5th harmonics instead, if it is not in class A. For class AB you likely want a lighter load. And a lighter load will require less screen voltage to operate, so will be safer for the screen grid (less peak Vg2 for less peak cathode current).
Class A automatically allows the loading on each tube to be higher Z (2X primary Z) since both tubes are operating together always.
As George (Tubelab) has mentioned, the major issue with screen grid drive is overheating the grid2. So it's best to use a lighter load (more B+, less current, lower Vg2 peak). The 18GB5 has a higher internal Mu (around 5.1) than the 40KD6 (4.0), so will require more peak screen voltage. And the other issue is significant grid2 current is required from the driver stage.
One way to avoid the grid2 meltdown scene is to split the drive "50/50" effectively with grid1, by scaling the g2 drive signal by a 1/Mu R divider, which then drives g1 in addition (internal g2/g1 Mu). This will only require half the peak AC grid2 drive compared to pure screen grid drive. However, if g1 is biased in negative territory still, then the peak g2 DC voltage will still be the same as before. Just the peak to peak g2 AC signal is halved then. So some experimenting is required here to see if both grids can be driven positive from 0 volts bias. Driving grid1 from zero volts bias will of course require grid current also.
As an example, using an output tube with an internal Mu of 4, and split scaled g2/g1 drives, this will only require 2X the usual grid1 only drive (AC), from the driver stage, to operate the g2. (and half the usual g1 AC drive on g1 now)
The output Z of screen grid drive is high, just like a pentode, so some local feedback of some type is in order.
The power law (I = kV^n) for screen grids is in the vicinity of 1.1 to 1.3, versus 2.0 to 2.7 for grid 1 drive. So g2 drive is more linear (obvious from the plate curve sets). One interesting aspect of grid2 drive, is that it can also be looked at as a current drive, with a Beta current gain factor, since grid2 tends to intercept a constant fraction of cathode current (until plate V gets down around Vg2). The Beta drops off however as Vp drops below Vg2, causing signal compression. While voltage drive of grid2 can be somewhat expansive instead (in class aB), if it's power law is up around 1.3. So a tradeoff between the two may be possible by putting some resistance in series with the grid2 drive to find a happy medium that is linear. Again, some trial and error required there. Same thing for grid1 in positive territory. (note: in class A, only power law above 2.0 can be expansive)
As a footnote, let me just mention an alternative for linear conventional drive of grid1 only may be possible using differential current feedback of output cathode currents back to an LTP driver stage. This local N Fdbk scheme attempts to flatten the total output stage gm (both tubes taken together, constant gm goal). This is where my interest has drifted lately for linearized class AB. This approach wants as much gm and local loop gain as possible to "flatten", so g1 drive is the choice there.
Another note:
For power law (I = kV^n), gm is KV^(n-1). So n=2.0 gives gm = kV, a linear ramp. In class B, you get two opposing ramps forming a V for gm total, in class aB you get some overlap to form a W, and in class A you get full overlap to form a constant gm sum. Constant gm sum makes a linear amplifier (before any feedback even). A power law above 2.0 makes for gm curves that curve upward individually, and below 2.0 makes for gm curves that bow over downward individually. One can play around with the different gm curves' P-P overlap by biasing, but nothing ever comes out flat for the sum then.
http://frank.pocnet.net/sheets/123/6/6KD6.pdf
http://frank.pocnet.net/sheets/030/e/EL500.pdf
Since the screen grid curves are so linear versus Vg2, you can choose most any load line for a P-P stage. A steeper load line will give more 2nd harmonic, which will cancel out in P-P, but will produce more 3rd and 5th harmonics instead, if it is not in class A. For class AB you likely want a lighter load. And a lighter load will require less screen voltage to operate, so will be safer for the screen grid (less peak Vg2 for less peak cathode current).
Class A automatically allows the loading on each tube to be higher Z (2X primary Z) since both tubes are operating together always.
As George (Tubelab) has mentioned, the major issue with screen grid drive is overheating the grid2. So it's best to use a lighter load (more B+, less current, lower Vg2 peak). The 18GB5 has a higher internal Mu (around 5.1) than the 40KD6 (4.0), so will require more peak screen voltage. And the other issue is significant grid2 current is required from the driver stage.
One way to avoid the grid2 meltdown scene is to split the drive "50/50" effectively with grid1, by scaling the g2 drive signal by a 1/Mu R divider, which then drives g1 in addition (internal g2/g1 Mu). This will only require half the peak AC grid2 drive compared to pure screen grid drive. However, if g1 is biased in negative territory still, then the peak g2 DC voltage will still be the same as before. Just the peak to peak g2 AC signal is halved then. So some experimenting is required here to see if both grids can be driven positive from 0 volts bias. Driving grid1 from zero volts bias will of course require grid current also.
As an example, using an output tube with an internal Mu of 4, and split scaled g2/g1 drives, this will only require 2X the usual grid1 only drive (AC), from the driver stage, to operate the g2. (and half the usual g1 AC drive on g1 now)
The output Z of screen grid drive is high, just like a pentode, so some local feedback of some type is in order.
The power law (I = kV^n) for screen grids is in the vicinity of 1.1 to 1.3, versus 2.0 to 2.7 for grid 1 drive. So g2 drive is more linear (obvious from the plate curve sets). One interesting aspect of grid2 drive, is that it can also be looked at as a current drive, with a Beta current gain factor, since grid2 tends to intercept a constant fraction of cathode current (until plate V gets down around Vg2). The Beta drops off however as Vp drops below Vg2, causing signal compression. While voltage drive of grid2 can be somewhat expansive instead (in class aB), if it's power law is up around 1.3. So a tradeoff between the two may be possible by putting some resistance in series with the grid2 drive to find a happy medium that is linear. Again, some trial and error required there. Same thing for grid1 in positive territory. (note: in class A, only power law above 2.0 can be expansive)
As a footnote, let me just mention an alternative for linear conventional drive of grid1 only may be possible using differential current feedback of output cathode currents back to an LTP driver stage. This local N Fdbk scheme attempts to flatten the total output stage gm (both tubes taken together, constant gm goal). This is where my interest has drifted lately for linearized class AB. This approach wants as much gm and local loop gain as possible to "flatten", so g1 drive is the choice there.
Another note:
For power law (I = kV^n), gm is KV^(n-1). So n=2.0 gives gm = kV, a linear ramp. In class B, you get two opposing ramps forming a V for gm total, in class aB you get some overlap to form a W, and in class A you get full overlap to form a constant gm sum. Constant gm sum makes a linear amplifier (before any feedback even). A power law above 2.0 makes for gm curves that curve upward individually, and below 2.0 makes for gm curves that bow over downward individually. One can play around with the different gm curves' P-P overlap by biasing, but nothing ever comes out flat for the sum then.
http://frank.pocnet.net/sheets/123/6/6KD6.pdf
http://frank.pocnet.net/sheets/030/e/EL500.pdf
Last edited:
I prefer to use real Zero bias tubes, they can handle 1000's of volts have high plate dissipation, so an 811A at 600 volts you can use lots of mills. the 805 can easily take 800 volts at over 100ma.
A TV tube driven on the screen just makes a zero bias triode. The best TV tubes now cost more than a zero bias triode and you don't have to worry about anything melting.
Phil
A TV tube driven on the screen just makes a zero bias triode. The best TV tubes now cost more than a zero bias triode and you don't have to worry about anything melting.
Phil
look at the EA230, a lower powered version...
you can idle at 3 to 5ma, accroding to SY......
Some curves for 6HJ5/6HD5. This tube has a near square law response for grid 1 (Ip = k Vg1 ^2).
6LQ6, and 300B also have this property.
6HJ5/6HD5 Grid 2 power law is 1.3 which is fairly typical.
Most sweep tubes have g1 exponent up around 2.2 to 2.7
6JC5 grid 1 has a 3.0 power law!
6LQ6, and 300B also have this property.
6HJ5/6HD5 Grid 2 power law is 1.3 which is fairly typical.
Most sweep tubes have g1 exponent up around 2.2 to 2.7
6JC5 grid 1 has a 3.0 power law!
Last edited:
For reference, some sweep tube g1 -mid current range average- exponents (2.0 is ideal for class A )
(and the 2nd number is the g2 exponent, which is fairly stable versus current):
[ Ip = k1(-Vo+Vg1)^exp1 + k2(Vg2)^exp2 ]
21HB5A 2.3 1.4
6LQ6 2.05 1.05
6LG6 2.88 1.28
6GE5,6JN6,6GV5 2.5 1.4
6MC6 2.2 1.2
6DQ5 2.01 1.14
6LB6 2.26 1.33
6LR6 2.6 1.3
6LX6 2.8 1.2
6GT5 2.5 -- (g2 probably 1.4, no data avail.)
6LF6 2.1 1.37
6CB5 2.2 1.32
6LW6 2.8 1.4
6HJ5 2.33 1.3
6JC5 3.09 1.33
6L6GC 2.15 1.28
EL34 2.4 1.5
6V6 2.1 1.3
6GB5 2.25 1.47
10JA5 2.27 1.36
38HE7 2.63 1.23
13E1 2.06 1.36
300B 2.0 1.5
Well, that 13E1 is a lot better tube than 2x 6LW6's in parallel after all.
But the beam tetrode winner is the lowly 26DQ5 for only $3
However, the exponent for variation of Mu with current (versus Vg1 actually) in Triode mode is (g1 Exp)-(g2 Exp), and the 300B wins there.
Note:
Any tube with a g1 Exponent above 2.0 can have it reduced down to 2.0 using a low value tail resistor below the P-P class A stage cathodes (the WE harmonic equalizer or Kiebert equalizer), so no need to get obsessed with 2.0 exponent tubes for class A. Variation of the g1 Exponent over the operating current range is an issue however, and this requires detailed info., best actually measured for an actual tube type (rather than the coarse datasheet curves info, where this stuff came from).
(and the 2nd number is the g2 exponent, which is fairly stable versus current):
[ Ip = k1(-Vo+Vg1)^exp1 + k2(Vg2)^exp2 ]
21HB5A 2.3 1.4
6LQ6 2.05 1.05
6LG6 2.88 1.28
6GE5,6JN6,6GV5 2.5 1.4
6MC6 2.2 1.2
6DQ5 2.01 1.14
6LB6 2.26 1.33
6LR6 2.6 1.3
6LX6 2.8 1.2
6GT5 2.5 -- (g2 probably 1.4, no data avail.)
6LF6 2.1 1.37
6CB5 2.2 1.32
6LW6 2.8 1.4
6HJ5 2.33 1.3
6JC5 3.09 1.33
6L6GC 2.15 1.28
EL34 2.4 1.5
6V6 2.1 1.3
6GB5 2.25 1.47
10JA5 2.27 1.36
38HE7 2.63 1.23
13E1 2.06 1.36
300B 2.0 1.5
Well, that 13E1 is a lot better tube than 2x 6LW6's in parallel after all.
But the beam tetrode winner is the lowly 26DQ5 for only $3
However, the exponent for variation of Mu with current (versus Vg1 actually) in Triode mode is (g1 Exp)-(g2 Exp), and the 300B wins there.
Note:
Any tube with a g1 Exponent above 2.0 can have it reduced down to 2.0 using a low value tail resistor below the P-P class A stage cathodes (the WE harmonic equalizer or Kiebert equalizer), so no need to get obsessed with 2.0 exponent tubes for class A. Variation of the g1 Exponent over the operating current range is an issue however, and this requires detailed info., best actually measured for an actual tube type (rather than the coarse datasheet curves info, where this stuff came from).
Last edited:
To clarify some about these grid exponents:
The gm versus Vg curve is the 1st derivative of the plate current law:
Ip = k1(-Vo+Vg1)^exp1 + k2(Vg2)^exp2
gm = deltaIp/deltaVg = dIp/dVg1
for an exponent = 2.0, gm = 2 k1 Vg1 or 2k1 (Vg1) to the 1.0 power
So a 2.0 grid1 exponent (or square law) will give a 1.0 (or linear) ramp for gm versus Vg1. A linear (1.0 power) gm ramp summed with its P-P flipped and overlapped gm ramp, in class A, will sum to a constant total gm for the stage. This gives zero distortion. Hence the big significance of 2.0 for the grid1 exponent in class A.
Only a few output tube datasheets give gm graphs. The E55L datasheet (linked below) gives a graph of gm (or S) versus Vg1 on pages 5 and 7. The S shaped curve shows that the grid 1 power law changes with current (well, with Vg1 as shown). Where the graph is straightly ramping up, the Ip power law is 2.0 (and gm law 1.0). But at low current the Ip power law is above 2.0 where the curve is bending up. And at high current, the Ip power law is below 2.0 where the curve is bending over. So one really needs to see the full gm graph to see if it will approximately sum to a constant gm when flipped and overlapped in class A mode. (Mosfets tend to be 2.0 power law over a central current range.)
The E55L datasheet also gives a plot of gm (or S) versus plate current on page 9. Notice the difference in shape, same tube. This is the form of graph needed when using a CCS tail in class A, since the currents will be complementary there. You might notice that the flipped gm curves here do not add up to a constant gm total, but rather a centrally humped bump. A CCS tail is actually not low distortion for tubes, but rather has a hump in the middle of the gm sum curve, indicating compressive 3rd harmonic distortion.
(bipolar transistors have gm proportional to current, so they do sum to a constant gm sum over a CCS)
The original page 5, gm versus Vg1, curves apply for grounded cathode mode P-P (or over a small tail resistor for WE harmonic cancellation). Even though they can add up to constant gm sum there for linear gain, that does not mean they operate at constant current sum. There is actually a common mode current variation due to the changing gm's which does not get past the OT.
The 300B datasheet also has a gm versus Vg1 graph on page 5, so is nicely instructive also. The straight ramping section in the middle of the gm S curve indicates an avg. 2.0 power law for the plate current equation above. The nicely symmetrical S shaped gm curve sums with its flipped around version in class A to give near constant gm sum. The triode plate feedbacks then further improve on any residual non-linearity. The old saw of linearizing before applying feedback applied in spades here.
http://frank.pocnet.net/sheets/009/e/E55L.pdf
http://frank.pocnet.net/sheets/084/3/300B.pdf
I was long mystified by the gm curves shown on some datasheets, mostly small signal tubes. I am rather pleased now to finally understand the powerful information conveyed by them. Too bad most audio output and sweep tubes don't give this essential information on their datasheets.
This can be remedied however by a device that measures gm or gain. One just uses a triangle or sine wave generator (sweep signal), at low frequency, covering peak to peak the full signal input range. Summed with that input is a low level 1 KHz or so sine wave. At the output of the circuitry or stage under test, one just filters or synchronously detects the 1 KHz signal to display its amplitude vertically on a scope. The low frequency input sweep is applied to the scope horizontal to give an X-Y display.
A flat line then indicates constant gm or gain versus input signal sweep. High gain can be applied to the vertical channel to see small variations of gain. This is essentially the "wingspread plots" given in the D. Self and B. Cordell books. (The Audio Precision analyzers offer this capability.) One might notice the similarity in function to how a gm tube tester works, just more dynamic here. This is so much more powerful that FFT testing because you can SEE where the gain is drooping or peaking, and you can watch the scan change while changing parameters in the circuit. Tweak until you get a flat line. The FFT analyzer now is just for final quantitative test.
The gm versus Vg curve is the 1st derivative of the plate current law:
Ip = k1(-Vo+Vg1)^exp1 + k2(Vg2)^exp2
gm = deltaIp/deltaVg = dIp/dVg1
for an exponent = 2.0, gm = 2 k1 Vg1 or 2k1 (Vg1) to the 1.0 power
So a 2.0 grid1 exponent (or square law) will give a 1.0 (or linear) ramp for gm versus Vg1. A linear (1.0 power) gm ramp summed with its P-P flipped and overlapped gm ramp, in class A, will sum to a constant total gm for the stage. This gives zero distortion. Hence the big significance of 2.0 for the grid1 exponent in class A.
Only a few output tube datasheets give gm graphs. The E55L datasheet (linked below) gives a graph of gm (or S) versus Vg1 on pages 5 and 7. The S shaped curve shows that the grid 1 power law changes with current (well, with Vg1 as shown). Where the graph is straightly ramping up, the Ip power law is 2.0 (and gm law 1.0). But at low current the Ip power law is above 2.0 where the curve is bending up. And at high current, the Ip power law is below 2.0 where the curve is bending over. So one really needs to see the full gm graph to see if it will approximately sum to a constant gm when flipped and overlapped in class A mode. (Mosfets tend to be 2.0 power law over a central current range.)
The E55L datasheet also gives a plot of gm (or S) versus plate current on page 9. Notice the difference in shape, same tube. This is the form of graph needed when using a CCS tail in class A, since the currents will be complementary there. You might notice that the flipped gm curves here do not add up to a constant gm total, but rather a centrally humped bump. A CCS tail is actually not low distortion for tubes, but rather has a hump in the middle of the gm sum curve, indicating compressive 3rd harmonic distortion.
(bipolar transistors have gm proportional to current, so they do sum to a constant gm sum over a CCS)
The original page 5, gm versus Vg1, curves apply for grounded cathode mode P-P (or over a small tail resistor for WE harmonic cancellation). Even though they can add up to constant gm sum there for linear gain, that does not mean they operate at constant current sum. There is actually a common mode current variation due to the changing gm's which does not get past the OT.
The 300B datasheet also has a gm versus Vg1 graph on page 5, so is nicely instructive also. The straight ramping section in the middle of the gm S curve indicates an avg. 2.0 power law for the plate current equation above. The nicely symmetrical S shaped gm curve sums with its flipped around version in class A to give near constant gm sum. The triode plate feedbacks then further improve on any residual non-linearity. The old saw of linearizing before applying feedback applied in spades here.
http://frank.pocnet.net/sheets/009/e/E55L.pdf
http://frank.pocnet.net/sheets/084/3/300B.pdf
I was long mystified by the gm curves shown on some datasheets, mostly small signal tubes. I am rather pleased now to finally understand the powerful information conveyed by them. Too bad most audio output and sweep tubes don't give this essential information on their datasheets.
This can be remedied however by a device that measures gm or gain. One just uses a triangle or sine wave generator (sweep signal), at low frequency, covering peak to peak the full signal input range. Summed with that input is a low level 1 KHz or so sine wave. At the output of the circuitry or stage under test, one just filters or synchronously detects the 1 KHz signal to display its amplitude vertically on a scope. The low frequency input sweep is applied to the scope horizontal to give an X-Y display.
A flat line then indicates constant gm or gain versus input signal sweep. High gain can be applied to the vertical channel to see small variations of gain. This is essentially the "wingspread plots" given in the D. Self and B. Cordell books. (The Audio Precision analyzers offer this capability.) One might notice the similarity in function to how a gm tube tester works, just more dynamic here. This is so much more powerful that FFT testing because you can SEE where the gain is drooping or peaking, and you can watch the scan change while changing parameters in the circuit. Tweak until you get a flat line. The FFT analyzer now is just for final quantitative test.
Last edited:
A simple sound card can also perform gain "wingspread plots" with the proper software. (easier to do than an FFT) Works the same way as above, a slow sweep signal across the full input signal range with some low level HF (say 1 KHz) added to it. Then the output of the circuit is monitored (typically with some type of attenuator) by the sound card and filtered for just the 1 KHz component. Amplitude of the 1 KHz component versus the sweep signal gets plotted.
Maybe some existing sound card analyzer software already offers this gain plot capability. Certainly should be far easier to program up than an FFT analyzer.
Anyone aware of such a "wingspread gain" feature offered on any of the (hopefully cheap) sound card analyzer software products out there?
Also, a digitally controlled curve tracer setup should be able to provide a gm plot for a tube, with some simple calculation using the curve tracer data.
Maybe some existing sound card analyzer software already offers this gain plot capability. Certainly should be far easier to program up than an FFT analyzer.
Anyone aware of such a "wingspread gain" feature offered on any of the (hopefully cheap) sound card analyzer software products out there?
Also, a digitally controlled curve tracer setup should be able to provide a gm plot for a tube, with some simple calculation using the curve tracer data.
Last edited:
Fixed a couple of typos in the grid exponent table and added 6BQ5/EL84 and then ordered the list by linearity (most linear at top)
For reference, some sweep tube g1 -mid current range average- exponents (2.0 is ideal for class A )
(and the 2nd number is the g2 exponent, which is fairly stable versus current):
Ip = k1(-Vo+Vg1)^exp1 + k2(Vg2)^exp2
gm = delta Ip/ delta Vg
and solving gm versus Vg:
gm1 = k1(exp1)(-Vo+Vg1)^(exp1-1)
gm2 = k2(exp2)(Vg2)^(exp2-1)
or solving for gm variation versus Ip:
gm1 = Ip^[(exp1-1)/(exp1)]
gm2 = Ip^[(exp2-1)/(exp2)]
Mu = gm1/gm2
Mu variation versus grid voltage: (-Vo+Vg1)^(exp1-exp2)
Mu variation versus plate current: Ip^[(exp1-1)/(exp1) - (exp2-1)/(exp2)]
300B example [2.0 1.0]:
gm1 versus Vg1 = 2 k1 (-Vo+Vg1)^(2-1) = 2 k1(-Vo+Vg1) [so gm1 varies linearly with Vg1]
gm1 variation versus Ip: Ip^[(2-1)/2] = Ip^0.5 [and so gm1 varies as square root of Ip]
gm2 (the plate, 1/Rp) variation versus Ip: Ip^[(1.5-1)/1.5] = Ip^0.3333
Mu variation versus Ip: Ip^[0.5-0.3333] = Ip^(.1666) [very slow rate of change for Mu]
300B 2.0 1.5
6DQ5 2.01 1.14
6LQ6 2.05 1.05
13E1 2.06 1.36
6V6 2.1 1.3
6LF6 2.1 1.37
6L6GC 2.15 1.28
6BQ5 2.18 1.2
6CB5 2.2 1.32
6MC6 2.2 1.2
6GB5 2.25 1.47
6LB6 2.26 1.33
10JA5 2.27 1.36
6GE5,6JN6,6GV5 2.3 1.4
21HB5A 2.3 1.4
6HJ5 2.33 1.3
EL34 2.4 1.5
6GT5 2.5 -- (g2 probably 1.4, no data avail.)
6LR6 2.6 1.3
6LX6 2.62 1.2
38HE7 2.63 1.23
6LW6 2.8 1.4
6LG6 2.88 1.28
6JC5 3.09 1.33
For reference, some sweep tube g1 -mid current range average- exponents (2.0 is ideal for class A )
(and the 2nd number is the g2 exponent, which is fairly stable versus current):
Ip = k1(-Vo+Vg1)^exp1 + k2(Vg2)^exp2
gm = delta Ip/ delta Vg
and solving gm versus Vg:
gm1 = k1(exp1)(-Vo+Vg1)^(exp1-1)
gm2 = k2(exp2)(Vg2)^(exp2-1)
or solving for gm variation versus Ip:
gm1 = Ip^[(exp1-1)/(exp1)]
gm2 = Ip^[(exp2-1)/(exp2)]
Mu = gm1/gm2
Mu variation versus grid voltage: (-Vo+Vg1)^(exp1-exp2)
Mu variation versus plate current: Ip^[(exp1-1)/(exp1) - (exp2-1)/(exp2)]
300B example [2.0 1.0]:
gm1 versus Vg1 = 2 k1 (-Vo+Vg1)^(2-1) = 2 k1(-Vo+Vg1) [so gm1 varies linearly with Vg1]
gm1 variation versus Ip: Ip^[(2-1)/2] = Ip^0.5 [and so gm1 varies as square root of Ip]
gm2 (the plate, 1/Rp) variation versus Ip: Ip^[(1.5-1)/1.5] = Ip^0.3333
Mu variation versus Ip: Ip^[0.5-0.3333] = Ip^(.1666) [very slow rate of change for Mu]
300B 2.0 1.5
6DQ5 2.01 1.14
6LQ6 2.05 1.05
13E1 2.06 1.36
6V6 2.1 1.3
6LF6 2.1 1.37
6L6GC 2.15 1.28
6BQ5 2.18 1.2
6CB5 2.2 1.32
6MC6 2.2 1.2
6GB5 2.25 1.47
6LB6 2.26 1.33
10JA5 2.27 1.36
6GE5,6JN6,6GV5 2.3 1.4
21HB5A 2.3 1.4
6HJ5 2.33 1.3
EL34 2.4 1.5
6GT5 2.5 -- (g2 probably 1.4, no data avail.)
6LR6 2.6 1.3
6LX6 2.62 1.2
38HE7 2.63 1.23
6LW6 2.8 1.4
6LG6 2.88 1.28
6JC5 3.09 1.33
Last edited:
Where from could we obtain this beautiful tube? I've been seeking for it in the WWW for long, long years, but never could find any.
Best regards!
There was one on Ebay last year in April I think, for around $500. I though it was some rip-off back then, but maybe that was the last one..........
If you just want the sound (Mona Lisa's smile but sorry, no Mona) then try 3 X 26DQ5s in parallel. Each are rated 24 Watts and cost $3 each over at ESRC. They use a nice convenient octal socket too. If you don't like plate caps, then there's 6HJ5 for $5.
If you just want the sound (Mona Lisa's smile but sorry, no Mona) then try 3 X 26DQ5s in parallel. Each are rated 24 Watts and cost $3 each over at ESRC. They use a nice convenient octal socket too. If you don't like plate caps, then there's 6HJ5 for $5.
Last edited:
With SPICE, gm curve can be plotted easily, which ones you would like to see?