I found some "tube rails" for an electronic organ on Ebay. 12 positions with 9 pin sockets (9/10 pin sockets on one) and 3 more empty positions for extra sockets.
12 musical note stickers attached near the sockets. I don't know what brand name organ these were from or what kind of tubes they had in them.
It occurred to me that these could be used for paralleling 9 pin tubes to make a big power tube equivalent, often with unique features. Using plenty of Ferrite beads and grid stoppers of course to prevent parasitic oscillations. Curve tracer matching obviously needed. Probably need to mount them horizontal for cooling. I've got bins of 29GK6, 6HB6, 12HL7, 6197, 6HZ8, 6JH8, maybe some others.
6197 has remarkable curves (pics 3 and 4),
12HL7 has 20,000 gm per tube, 10 Watt and remarkable triode curves too (pics 5 and 6) 12 x 20,000 = 240,000 gm, 12 x 10 = 120 Watts,
29GK6 ( = EL84) at 12 Watts each 12 x 12 = 144 Watts (pic 7),
6HZ8 has a triode driver and 8 Watt pentode,
6HB6 has 20,000 gm per tube, 10 Watt,
6JH8 beam deflection is like an LTP with a CCS tail, all in one tube.
Now, if I could just find another 50 to 100 rails, I could start on the VT-9 AI computer project, that's been on the back burner for decades.
12 musical note stickers attached near the sockets. I don't know what brand name organ these were from or what kind of tubes they had in them.
It occurred to me that these could be used for paralleling 9 pin tubes to make a big power tube equivalent, often with unique features. Using plenty of Ferrite beads and grid stoppers of course to prevent parasitic oscillations. Curve tracer matching obviously needed. Probably need to mount them horizontal for cooling. I've got bins of 29GK6, 6HB6, 12HL7, 6197, 6HZ8, 6JH8, maybe some others.
6197 has remarkable curves (pics 3 and 4),
12HL7 has 20,000 gm per tube, 10 Watt and remarkable triode curves too (pics 5 and 6) 12 x 20,000 = 240,000 gm, 12 x 10 = 120 Watts,
29GK6 ( = EL84) at 12 Watts each 12 x 12 = 144 Watts (pic 7),
6HZ8 has a triode driver and 8 Watt pentode,
6HB6 has 20,000 gm per tube, 10 Watt,
6JH8 beam deflection is like an LTP with a CCS tail, all in one tube.
Now, if I could just find another 50 to 100 rails, I could start on the VT-9 AI computer project, that's been on the back burner for decades.
Last edited:
I too have many boxes of lo power tubes in the 12W region and thought of aisles of 10 or more lined up but and an issue that haunts is the monstrous heater power when compared to perhaps a couple of 50W anode rating tubes capable of punching the same job. The only real advantage is the much easier Lo Z output tranny but with the increased b/w, the emphasis is also driver capability.
Complementary Futterman ?
Bench Baron
Complementary Futterman ?
Bench Baron
How do the internal capacitances come into the equation with paralleled tubes? Aren't they all added together so the appearance to the driver is one tube with very high internal capacitance, so limiting the band width?
I have half a storeroom of ECL82, and boxes of obsolete UL84 having a 45V 100mA heater (aka EL84 main spec). Having for example,10 in each P_P row with total Cg1-k equating to 72pF each half (so here we go), not including Miller capacitance, which is also crippling. Only way with conventional med Z output drivers is B/W curtailment and never stray out of class A.
My solution would be a current dump driver using a complementary mosfet drivers. Not exactly true tube but it would seem a pity if the output stage itself becomes the limiting factor instead of superb output transformers.
Since I bave two output trannies each rated 100W, of 1K A-A imped spec´d -3dB @50Khz and -3dB @15hz as per Williamson config, it would appear a shame not to fully utilize their performance.
Another real bonus with multiple parallel outout pairs is the reduced B+ requirement.
The mass produced ECL82 appeared in 1960´s record /tape players and the UL84 in those dangerous line TV output stages with a baretter as heater ballast. The ECL82 in P-P could just "scrape" 10W output.
I´m listening ! Only my views !
Bench Baron
My solution would be a current dump driver using a complementary mosfet drivers. Not exactly true tube but it would seem a pity if the output stage itself becomes the limiting factor instead of superb output transformers.
Since I bave two output trannies each rated 100W, of 1K A-A imped spec´d -3dB @50Khz and -3dB @15hz as per Williamson config, it would appear a shame not to fully utilize their performance.
Another real bonus with multiple parallel outout pairs is the reduced B+ requirement.
The mass produced ECL82 appeared in 1960´s record /tape players and the UL84 in those dangerous line TV output stages with a baretter as heater ballast. The ECL82 in P-P could just "scrape" 10W output.
I´m listening ! Only my views !
Bench Baron
Please don't confuse UL84 and EL84! UL84 and PL84 are EL86/6CW5's with different heaters. EL84/6BQ5 is quite another beast.
I've never seen any UL84 in a TV set. PL84 was the TV tube, mostly used as AF final tube, but some TV's also featured it as the frame (not line!) deflection amplifier.
Best regards!
I've never seen any UL84 in a TV set. PL84 was the TV tube, mostly used as AF final tube, but some TV's also featured it as the frame (not line!) deflection amplifier.
Best regards!
The internal capacitances and heater powers do add up. But it depends on what tubes you compare as to the sums versus a big tube.. There is also some wiring capacitance to consider. With a ten Watt driver tube too, you can easily drive banks of these.
Big 40 Watt Tube:
6LW6 6.3V 2.65Amps Cin/Cout/Crev: 40 14.5 1.0 pf gm=12,000
4x10 Watt Tubes: (or equiv.)
3x 6GK6 6.3V 2.28Amp Cin/Cout/Crev: 30 21 0.42 pf 3x13.2W=39.6 Watt gm=33,900
4x 6HB6 6.3V 3.04Amp Cin/Cout/Crev: 52 32 0.72 pf 40 Watt gm=80,000
5x 6197 6.3V 3.25A Cin/Cout/Crev 57.5 25 0.625 pf 5x7.5W=37.5 Watt gm=55,000
4x 12HL7 6.3V 2.4A Cin/Cout/Crev 60 24 0.6 pf 40 Watt gm=80,000
Not that bad for comparison.
A 10 Watt driver should drive piles of these.
Reduced plate V (like 300V) versus 1000V+ for 50+Watt tubes is certainly quite attractive. Also the reduced Zout. And super gm.
Big 40 Watt Tube:
6LW6 6.3V 2.65Amps Cin/Cout/Crev: 40 14.5 1.0 pf gm=12,000
4x10 Watt Tubes: (or equiv.)
3x 6GK6 6.3V 2.28Amp Cin/Cout/Crev: 30 21 0.42 pf 3x13.2W=39.6 Watt gm=33,900
4x 6HB6 6.3V 3.04Amp Cin/Cout/Crev: 52 32 0.72 pf 40 Watt gm=80,000
5x 6197 6.3V 3.25A Cin/Cout/Crev 57.5 25 0.625 pf 5x7.5W=37.5 Watt gm=55,000
4x 12HL7 6.3V 2.4A Cin/Cout/Crev 60 24 0.6 pf 40 Watt gm=80,000
Not that bad for comparison.
A 10 Watt driver should drive piles of these.
Reduced plate V (like 300V) versus 1000V+ for 50+Watt tubes is certainly quite attractive. Also the reduced Zout. And super gm.
Last edited:
I have a DIY guitar amp that runs a pair of UL84's in a typical push pull output stage. With about 340 volts for the plate supply, 165 volts for the screen grid into a 3300 ohm load, I get a bit over 20 watts of output. The UL84 (45B5) and two UCC85's (26AQ8) are wired in series across the output of an isolation transformer that also feeds a bridge rectifier for the screen voltage and a full wave doubler for main B+.
I have tried two "massively parallel" (5 tubes per push pull side) output stages in the distant past. The first was more of a joke that a real experiment, but it ended with several melted tubes. I had, and still have probably 500 to 1000 used, pulled 6AQ5 tubes. This is after I tossed all the broken tubes and tubes with badly corroded pins. I simply took 5 sockets and ran wire from tube socket to tube socket for the heater and plate pins. The cathode, screen and control grids got a resistor soldered to the socket pin and a wire from resistor to resistor. No chassis or means of mounting was used. Two of these contraptions were built and they were connected to an OPT, power supply, and an early version of what I now call the universal driver board that used 6SN7's instead of 6CG7's. The contraption worked and provided anywhere from 20 to 50 watts of audio power depending on its mood, the phase of the moon and how angry I got with the voltage knob on the power supply. THD was anywhere between very low and square waves, again depending on whim. Get the idea that this thing was very unstable? That would be an understatement I was convinced that it was due to crummy tubes, so I kept trying more tubes, without success.
Anyone who has ever experimented with the 6AQ5 must know that there is a very fine line between dim red plate and hole in the glass. There is literally a couple mm between the plate and the glass. You don't red plate a 6AQ5, at least not for long. All 6AQ5's are not created equal. Some are well made and fully concentric. These will actually survive at the 10 watt plate dissipation limit. 12 watts brings on the dim glow, 14 makes a crater in the glass that will quickly suck air! Crummy tubes will red plate sometimes at 7 or 8 watts.
The problem I had was that a different tube would red plate every time I turned the creation on. No oscillation was detected with radio, TV or RF spectrum analyzer. After frying lots of 6AQ5's sometimes on purpose, it all (except for the OPT) went into the trash.
About a year later I got a box of 13GB5's from Stan when they were on the dollar menu. Most were Amperex branded Mullard "made in Gt. Brittain" tubes, well made and I could easily squeeze 100 watts from a pair of them. Some were Japanese, also good, but not worthy of operation over 100 watts. There were a few Korean tubes that just sucked. So if one pair made 100 watts, and two pair made almost 200 watts, what should 5 pair make? Yes, I was DUMM enough to try it again. This time I mounted all the sockets on perf board in a pentagon form with a star arrangement for the control grid resistors, and a flying connection above the grid resistors for the cathode resistor / bypass caps. This time it worked better, but I could never make it reliably stable with 5 tubes with anything over about 250 volts. Any 3 tubes (just pulled 2 out) worked fine and made about 280 watts on about 450 volts, as the "80 watt" OPT losses probably sucked up 20 watts because it got pretty warm. 4 Tubes worked but only at a lower voltage which resulted in less power output that three tubes at the edge of the glow zone.
I have seen two pair of 26LW6's or 26HU5's make 500 watts into 1250 ohms on 625 volts. I have not yet tried three pair. Not sure if I want to risk my Plitron OPT's as they do not make them any more.
I have tried two "massively parallel" (5 tubes per push pull side) output stages in the distant past. The first was more of a joke that a real experiment, but it ended with several melted tubes. I had, and still have probably 500 to 1000 used, pulled 6AQ5 tubes. This is after I tossed all the broken tubes and tubes with badly corroded pins. I simply took 5 sockets and ran wire from tube socket to tube socket for the heater and plate pins. The cathode, screen and control grids got a resistor soldered to the socket pin and a wire from resistor to resistor. No chassis or means of mounting was used. Two of these contraptions were built and they were connected to an OPT, power supply, and an early version of what I now call the universal driver board that used 6SN7's instead of 6CG7's. The contraption worked and provided anywhere from 20 to 50 watts of audio power depending on its mood, the phase of the moon and how angry I got with the voltage knob on the power supply. THD was anywhere between very low and square waves, again depending on whim. Get the idea that this thing was very unstable? That would be an understatement I was convinced that it was due to crummy tubes, so I kept trying more tubes, without success.
Anyone who has ever experimented with the 6AQ5 must know that there is a very fine line between dim red plate and hole in the glass. There is literally a couple mm between the plate and the glass. You don't red plate a 6AQ5, at least not for long. All 6AQ5's are not created equal. Some are well made and fully concentric. These will actually survive at the 10 watt plate dissipation limit. 12 watts brings on the dim glow, 14 makes a crater in the glass that will quickly suck air! Crummy tubes will red plate sometimes at 7 or 8 watts.
The problem I had was that a different tube would red plate every time I turned the creation on. No oscillation was detected with radio, TV or RF spectrum analyzer. After frying lots of 6AQ5's sometimes on purpose, it all (except for the OPT) went into the trash.
About a year later I got a box of 13GB5's from Stan when they were on the dollar menu. Most were Amperex branded Mullard "made in Gt. Brittain" tubes, well made and I could easily squeeze 100 watts from a pair of them. Some were Japanese, also good, but not worthy of operation over 100 watts. There were a few Korean tubes that just sucked. So if one pair made 100 watts, and two pair made almost 200 watts, what should 5 pair make? Yes, I was DUMM enough to try it again. This time I mounted all the sockets on perf board in a pentagon form with a star arrangement for the control grid resistors, and a flying connection above the grid resistors for the cathode resistor / bypass caps. This time it worked better, but I could never make it reliably stable with 5 tubes with anything over about 250 volts. Any 3 tubes (just pulled 2 out) worked fine and made about 280 watts on about 450 volts, as the "80 watt" OPT losses probably sucked up 20 watts because it got pretty warm. 4 Tubes worked but only at a lower voltage which resulted in less power output that three tubes at the edge of the glow zone.
I have seen two pair of 26LW6's or 26HU5's make 500 watts into 1250 ohms on 625 volts. I have not yet tried three pair. Not sure if I want to risk my Plitron OPT's as they do not make them any more.
Yeah, bias shift with temp. or ageing is clearly the major uncontrolled concern with this paralleling business.
Does a tube that gets a little hotter start to draw more current?
Auto-biasing would seem like a really good idea for paralleled tubes. Or a bypassed current source type device.
On the input capacitance concern, paralleled tubes are typically going to have higher gm, so less drive voltage swing will be needed. That lightens the load for whatever is driving the tube bank. You can use a lower value plate resistor for the driver to keep freq. response in line.
If the tube bank is a bunch of high gm devices, you'll come out ahead on drive, as long as they aren't oscillating. Some lossy Ferrite beads or resistors in the plate leads will likely be important. And grid stopper resistors.
Does a tube that gets a little hotter start to draw more current?
Auto-biasing would seem like a really good idea for paralleled tubes. Or a bypassed current source type device.
On the input capacitance concern, paralleled tubes are typically going to have higher gm, so less drive voltage swing will be needed. That lightens the load for whatever is driving the tube bank. You can use a lower value plate resistor for the driver to keep freq. response in line.
If the tube bank is a bunch of high gm devices, you'll come out ahead on drive, as long as they aren't oscillating. Some lossy Ferrite beads or resistors in the plate leads will likely be important. And grid stopper resistors.
Last edited:
6AQ5s are out!! 7 pins besides.
Just for fun, 12x 12HL7s paralleled in Triode mode: One 12HL7: mu 25, gm 21000, Triode rp is 1190 Ohms, max DC 50 mA, 10 Watt Pdiss. So 12x would be mu 25, gm 252,000 (or 0.25 Siemens in Mosfet speak), and Zout 99 Ohms, at 120 Watt Pdiss. 600 mA max DC plate current, 1.8 Amp peak. Try and find
-ANY- Triode with specs like that, and near linear perfect Triode curves too.
Just for fun, 12x 12HL7s paralleled in Triode mode: One 12HL7: mu 25, gm 21000, Triode rp is 1190 Ohms, max DC 50 mA, 10 Watt Pdiss. So 12x would be mu 25, gm 252,000 (or 0.25 Siemens in Mosfet speak), and Zout 99 Ohms, at 120 Watt Pdiss. 600 mA max DC plate current, 1.8 Amp peak. Try and find
-ANY- Triode with specs like that, and near linear perfect Triode curves too.
Last edited:
Parallel tube RF oscillations are going to be a lot of fun with that kind of gm. The number of possible modes (each with a slightly different equivalent circuit) boggles the mind. And you need an accurate physical description to model it.
If you add enough losses where it starts affecting the audio band it would probably quiet down. But at some point you start defeating your original purposes.
If you add enough losses where it starts affecting the audio band it would probably quiet down. But at some point you start defeating your original purposes.
I once had an Optimation AC source, with 8 6CA7s in it - and a AC muffin fan directly underneath the bank. Among its output windings was a 16 Ohm, so I tried to use it as a bass amp until I realized it was just too impractical. It had the incandescent bulb arrangement for current limiting, which once triggered, you had to power cycle it to reset... I didnt dare short that, knowing I'd want to sell it one day.
I've had pretty good luck stomping out instabilities on the Tek curve tracer using stopper resistors and lossy Ferrite beads on the 1 foot test leads. But, of course, that was just for one tube. Since I don't need a 120 Watt, 0.25 Siemens Triode, I will likely split the tube rail up into 2 groups of 6 tubes, with the 3 extra socket holes for an LTP driver and splitter to run P-P. Worst case, can drop the whole assembly in dielectric/magnetic Ferrofluid.
I have a 1200 Ohm P-P OT on order for testing the OTL cheat Patent US4614914, with 6528 tubes or various TV Sweeps. So this should work well with that too for various 9 pin tube banks. I just found another one of these "Tube Rails" on Ebay (3x now), so gears are spinning.
(No luck finding 100 of these Rails though. no digital tube computing on the horizon for me, I'd never get out of that hole. Wait, let's see, 6JH8 tubes can do Analog multiplication, an Analog computer! )
The Heathkit H-1 (ANC-1) computer, courtesy of ComputerHistory.org
Tube OP-AMPs with 6JH8 multiplier modules. Someday.....
I have a 1200 Ohm P-P OT on order for testing the OTL cheat Patent US4614914, with 6528 tubes or various TV Sweeps. So this should work well with that too for various 9 pin tube banks. I just found another one of these "Tube Rails" on Ebay (3x now), so gears are spinning.
(No luck finding 100 of these Rails though. no digital tube computing on the horizon for me, I'd never get out of that hole. Wait, let's see, 6JH8 tubes can do Analog multiplication, an Analog computer! )
The Heathkit H-1 (ANC-1) computer, courtesy of ComputerHistory.org
Tube OP-AMPs with 6JH8 multiplier modules. Someday.....
Last edited:
I remember I saw like 20 years ago an Hi-Fi amp with like 12 paralleled 12AX7 as power stage.
Is there anyone who remembers brand and name of the amp?
Is there anyone who remembers brand and name of the amp?
Ooop´s... thanks for mentioning. You´re right about sim to EL86. I´d better recheck the print.
You're welcome! Anyway, I'd need to correct the sequence in my writing, as UL84 was the archetype in this row of valves, while PL84 and EL86 are derivatives 😉 .
Best regards!
I seem to recall a tube Amp with 100 12AX7 tubes, probably a 1 off, quite a while back. Members were complaining about the waste of NOS tubes. There certainly are plenty of OTL Amps around with large #s of tubes in parallel.
Curious, just how does it do that? V1 on G1, V2 on G2 then Plate moves about as V1*V2?
As I recall, two 6JH8 tubes have their cathodes connected together over a CCS tail. Then differential X goes between the grid1s. Then deflectors get D1s connected, and separately the D2s connected. Differential Y then goes between D1 and D2 pairs. The product comes out differentially by summing plate currents from cross pairs.
Tube1 plates: (1+X) (1+Y) = 1+X+Y+XY & (1+X) (1-Y) = 1+X-Y-XY
Tube2 plates: (1-X) (1+Y) = 1-X+Y-XY & (1-X) (1-Y) = 1-X-Y+XY
( 1+X+Y+XY) + (1-X-Y+XY) = 2 +2XY
(1+X-Y-XY) + (1-X+Y-XY) = 2 - 2XY
This scheme was used for SSB by Radio Amateurs and in TV sets for color decoding or something. Some other similar tubes: 6HW8, 6AR8, 6ME8, 7360
Last edited:
I heard at least one of the beam deflection tubes were developed in Europe, so maybe some are still around there.