I have traditionally tried to stick with current production tubes in my designs, just for the sake of availability, but I will be picking up some NOS tubes to experimant with, and I was looking for suggestions. I want to do a "just because I can" massive gain stage.
I usually try to use dual triodes for cascode connection for the sake of keeping wiring short and footprint small, but I want to just have merry fun with insane voltage gains, noise figures aside, so two single triodes would be more fun.
I was thinking of course of a high Gm device for the lower device, like a 5842, 6C45, or even a EC8020 if I get really spunky.
For the top tube though, I am not sure what would be best. I assume that a lower mu tube with a low Ri that will take massive plate voltages would be best from the standpoint of output impedance and gain. Would something like a 12B4A work well? Or maybe 6S4A. OR Do I want a higher gain top device with a lower Ri, like a 6er5 or 6ha6?
Suggestions?
I usually try to use dual triodes for cascode connection for the sake of keeping wiring short and footprint small, but I want to just have merry fun with insane voltage gains, noise figures aside, so two single triodes would be more fun.
I was thinking of course of a high Gm device for the lower device, like a 5842, 6C45, or even a EC8020 if I get really spunky.
For the top tube though, I am not sure what would be best. I assume that a lower mu tube with a low Ri that will take massive plate voltages would be best from the standpoint of output impedance and gain. Would something like a 12B4A work well? Or maybe 6S4A. OR Do I want a higher gain top device with a lower Ri, like a 6er5 or 6ha6?
Suggestions?
> merry fun with insane voltage gains
Huge voltage gain is usually a folly. But it may amuse you to prove it for yourself.
I must point out that huge stage gain is the province of Transistors, was their holy grail for years, and led them into some horrid-sounding designs.
> I was thinking of course of a high Gm device for the lower device
Seems almost right. Isn't so. Gain is about Rl*Gm, so Gm is good. But Gm is some function of current. 8417 (pentode or strapped-triode) can give Gm of 20,000uMho, but at current like 100mA. Assuming 300V drop in Rl, Rl must be like 3K. Gain (assuming perfect cascoding) is about 60. Not a lot. Taking a more reasonable current, we have several tubes that will do 5,000uMho at 10mA. Assuming 300V drop in Rl, Rl must be like 30K. Gain is about 150. Yawn. Let's take dull old 12AX7, Gm = 1,600uMho at 1.2mA. Rl is about 250K. Gain is about 400.
In fact (and this was proven in WWII Radar studies) if Voltage Gain is your only goal, you want to run the LEAST plate current you can get away with. You also want a tube with high Gm for that plate current, which may or may not be one of the Usual Suspects we turn to for high Gm at common currents.
Keep following this train, and you will realize that you can have high voltage gain at low frequency, but the high impedance and low operating current leads to crummy performance at high frequency. And for audio, that can mean we need a mA of current and not much more than 100K impedance at the output node.
> For the top tube though, I am not sure what would be best. I assume that a lower mu tube with a low Ri that will take massive plate voltages would be best from the standpoint of output impedance and gain.
For the reasons above, incredibly high nodal impedances give incredibly bad high-frequency performance. The Transistor boys ran into this. Their salvation(?) was integrated circuits, which give a significant reduction in parasitic capacitance (short leads) and such cheap transistors that they can buffer the heck out of all outputs (and inputs).
> The top triode is a transimpedance amp right? So what virtues do I want to find
Transimp..what? Oh, right, yeah, except hollow-heads would just say "grounded grid". You want ALL cathode current to flow to the plate; any neg-grid triode is perfect enough for that. You want the whomping big output swing to NOT appear on the cathode, where it will repress the lower triode's gain. And the number for that is Amplification Factor.
But while Mu must not be low, going high may not be better. Take two 12AX7. Gm of the bottom tube is 1/625Ω, bottom tube's plate resistance is Mu times higher or 62K, top tube's plate resistance is essentially 100 times higher yet or 6,200K or 6Meg. A reasonable plate resistor would be 250K. (250K||6M)/625= gain of 384. An unreasonable plate resistor might be 6Meg which at 1.2mA would drop 7,000 volts(!). Gain looks like 5,000 (half of Mu*Mu). Zout is 3Megs. Taking a small stray wiring capacitance of 30pFd, the -3dB point is 1769Hz, not very broadband. If we reduce Rl (and B+) we reduce gain but raise the -3dB point. For a good broadband audio amp we might want to get near 60K node impedance. Now the low frequency gain is 100, which is only a little better than one 12AX7 can do alone. (Note that the same two sections of 12AX7 in cascAde will give gain of 2,000, with audio bandwidth, and no heroic voltages.)
So if you go heroic voltages with reasonable current, your node impedance is so high that stray capacitance kills the top of the audio band. And if you run less-insane voltages to give node impedance that allows full audio bandwidth, total gain is nowhere near Mu*Mu so the Mu of the upper tube hardly matters.
Note that while the plate supply voltage may be huge, the voltage across the tube will normally be "normal", a couple hundred volts. Raising the plate voltage does not raise the gain. We need high plate supply voltage to cover the huge resistor needed to meet the high potential plate resistance, not to drop in the tubes.
The sand-state boys have PNP current limiters that can work as active loads. We don't have P-type vacuum tubes so we can't copy that trick. We can float a vacuum tube with cathode resistor and constant voltage supply, but the parasitic resistance and capacitance in an extra tube (tubes: we'll probably have to cascode this side too), voltage reference, and heater leakage, usually make this a lot of work for very little or no gain.
So about the only requirement of the top tube is that it pass the bottom tube's current while staying negative-grid.
So: pick a reasonable output node impedance. Pick a semi-reasonable supply voltage. Figure the current. Find the tube with best Gm at that current. Estimate the voltage gain. The product Mu*Mu should be higher than this gain. Since you have already picked a trial lower-tube, this gives a minimum Mu for the upper tube. Find a tube with higher Mu which will pass the current.
A good bet, for reasons I am too tired to go into, are the TV tuner tubes. 6BQ7 is canonical, 6DJ8 is better and very "in". Don't work them at maximum current for maximum Gm; figure your bandwidth-limited impedances and scale for that current. Because Gm reduces as square-root of current, LF gain will actually increase as current is reduced and all else scaled to keep the voltages the same.
Note that VERY small output-input coupling will clobber your gain (unless the input is driven very low-Z). Layout must minimize capacitance, but also trace leakage in sockets may knock-down theoretical gain.
6BK4 has Mu (not cascoded!) of 2,000, plate rating 27,000-60,000, Gm around 1mA of about 500uMho. Thinking about that will keep you busy, but probably will not produce a good audio amp.
Huge voltage gain is usually a folly. But it may amuse you to prove it for yourself.
I must point out that huge stage gain is the province of Transistors, was their holy grail for years, and led them into some horrid-sounding designs.
> I was thinking of course of a high Gm device for the lower device
Seems almost right. Isn't so. Gain is about Rl*Gm, so Gm is good. But Gm is some function of current. 8417 (pentode or strapped-triode) can give Gm of 20,000uMho, but at current like 100mA. Assuming 300V drop in Rl, Rl must be like 3K. Gain (assuming perfect cascoding) is about 60. Not a lot. Taking a more reasonable current, we have several tubes that will do 5,000uMho at 10mA. Assuming 300V drop in Rl, Rl must be like 30K. Gain is about 150. Yawn. Let's take dull old 12AX7, Gm = 1,600uMho at 1.2mA. Rl is about 250K. Gain is about 400.
In fact (and this was proven in WWII Radar studies) if Voltage Gain is your only goal, you want to run the LEAST plate current you can get away with. You also want a tube with high Gm for that plate current, which may or may not be one of the Usual Suspects we turn to for high Gm at common currents.
Keep following this train, and you will realize that you can have high voltage gain at low frequency, but the high impedance and low operating current leads to crummy performance at high frequency. And for audio, that can mean we need a mA of current and not much more than 100K impedance at the output node.
> For the top tube though, I am not sure what would be best. I assume that a lower mu tube with a low Ri that will take massive plate voltages would be best from the standpoint of output impedance and gain.
For the reasons above, incredibly high nodal impedances give incredibly bad high-frequency performance. The Transistor boys ran into this. Their salvation(?) was integrated circuits, which give a significant reduction in parasitic capacitance (short leads) and such cheap transistors that they can buffer the heck out of all outputs (and inputs).
> The top triode is a transimpedance amp right? So what virtues do I want to find
Transimp..what? Oh, right, yeah, except hollow-heads would just say "grounded grid". You want ALL cathode current to flow to the plate; any neg-grid triode is perfect enough for that. You want the whomping big output swing to NOT appear on the cathode, where it will repress the lower triode's gain. And the number for that is Amplification Factor.
But while Mu must not be low, going high may not be better. Take two 12AX7. Gm of the bottom tube is 1/625Ω, bottom tube's plate resistance is Mu times higher or 62K, top tube's plate resistance is essentially 100 times higher yet or 6,200K or 6Meg. A reasonable plate resistor would be 250K. (250K||6M)/625= gain of 384. An unreasonable plate resistor might be 6Meg which at 1.2mA would drop 7,000 volts(!). Gain looks like 5,000 (half of Mu*Mu). Zout is 3Megs. Taking a small stray wiring capacitance of 30pFd, the -3dB point is 1769Hz, not very broadband. If we reduce Rl (and B+) we reduce gain but raise the -3dB point. For a good broadband audio amp we might want to get near 60K node impedance. Now the low frequency gain is 100, which is only a little better than one 12AX7 can do alone. (Note that the same two sections of 12AX7 in cascAde will give gain of 2,000, with audio bandwidth, and no heroic voltages.)
So if you go heroic voltages with reasonable current, your node impedance is so high that stray capacitance kills the top of the audio band. And if you run less-insane voltages to give node impedance that allows full audio bandwidth, total gain is nowhere near Mu*Mu so the Mu of the upper tube hardly matters.
Note that while the plate supply voltage may be huge, the voltage across the tube will normally be "normal", a couple hundred volts. Raising the plate voltage does not raise the gain. We need high plate supply voltage to cover the huge resistor needed to meet the high potential plate resistance, not to drop in the tubes.
The sand-state boys have PNP current limiters that can work as active loads. We don't have P-type vacuum tubes so we can't copy that trick. We can float a vacuum tube with cathode resistor and constant voltage supply, but the parasitic resistance and capacitance in an extra tube (tubes: we'll probably have to cascode this side too), voltage reference, and heater leakage, usually make this a lot of work for very little or no gain.
So about the only requirement of the top tube is that it pass the bottom tube's current while staying negative-grid.
So: pick a reasonable output node impedance. Pick a semi-reasonable supply voltage. Figure the current. Find the tube with best Gm at that current. Estimate the voltage gain. The product Mu*Mu should be higher than this gain. Since you have already picked a trial lower-tube, this gives a minimum Mu for the upper tube. Find a tube with higher Mu which will pass the current.
A good bet, for reasons I am too tired to go into, are the TV tuner tubes. 6BQ7 is canonical, 6DJ8 is better and very "in". Don't work them at maximum current for maximum Gm; figure your bandwidth-limited impedances and scale for that current. Because Gm reduces as square-root of current, LF gain will actually increase as current is reduced and all else scaled to keep the voltages the same.
Note that VERY small output-input coupling will clobber your gain (unless the input is driven very low-Z). Layout must minimize capacitance, but also trace leakage in sockets may knock-down theoretical gain.
6BK4 has Mu (not cascoded!) of 2,000, plate rating 27,000-60,000, Gm around 1mA of about 500uMho. Thinking about that will keep you busy, but probably will not produce a good audio amp.
While it's difficult to add to PRR's fine info, don't overlook commonly available (and cheap as dirt) TV tubes as voltage amplifiers. The 6GH8A with the pentode in triode mode works fine in a single-bottle mu-stage with plenty of gain. Characteristics are suitable for a line driver (a wee noisy for a phono or microphone).
Dissimilar triodes, like vertical sweep tubes (compactron or octal), can offer great performance and tend to be very linear. Your "top and bottom" tubes in one!
Dissimilar triodes, like vertical sweep tubes (compactron or octal), can offer great performance and tend to be very linear. Your "top and bottom" tubes in one!
Now THAT"S what I'm talking about
That's what I was wanting to hear! How far the stage can be pushed. Thanx for the great info.
So I'm going to make some rules of thumb for ideal tube charactaristics:
Bottom tube= high Gm at lowest possible a-k current... time to find some tube charts and make up some GM to Ip-k ratios!
So if I had mad cash to throw around, an EC8020 might fit the bill. I have heard of them developing 50+mA/V at currents around 20mA. I guess current to bias voltage is an important issue too.
Top Tube= Low internal resistoance and mu slightly higher than the bottom tube, and that will take the current that the bottom etube requires, and the heater to cathode voltage that results from the bottom tube's plate.
So maybe if there were a 8020 in the bottom, then with a mu around 55, then something as simple as a common 6hm5/6ha5 would work , with a mu of 70 and a Ri of 5K.
ROCK ON! Can't wait to get some circuits built and start blow-in stuff up.
That's what I was wanting to hear! How far the stage can be pushed. Thanx for the great info.
So I'm going to make some rules of thumb for ideal tube charactaristics:
Bottom tube= high Gm at lowest possible a-k current... time to find some tube charts and make up some GM to Ip-k ratios!
So if I had mad cash to throw around, an EC8020 might fit the bill. I have heard of them developing 50+mA/V at currents around 20mA. I guess current to bias voltage is an important issue too.
Top Tube= Low internal resistoance and mu slightly higher than the bottom tube, and that will take the current that the bottom etube requires, and the heater to cathode voltage that results from the bottom tube's plate.
So maybe if there were a 8020 in the bottom, then with a mu around 55, then something as simple as a common 6hm5/6ha5 would work , with a mu of 70 and a Ri of 5K.
ROCK ON! Can't wait to get some circuits built and start blow-in stuff up.
Geek said:
I am a little fuzzy on both voltage stage pentode design and hu-stage design. I went through the TubeCAD "SRPP Deconstructed" article, but gleaned very little practical info. I get lost in the complexity of the problems they try to solve on that site sometimes.
Do you have any prectical info on pentode and triode cascode/totem/mu stage design? Or maybe a a few schematics that are not fraught with solid state CCS's and voltage regulators.
PRR said:
I just caught that 6BK4 reference. HOLY CRAP! I would never have thought about using one of those monsters as a voltage amp. Cool idea though. Aren't thet like 30,000 volt regulators or something? Maybe I'll make an audio stage with one and then use it to regulate the voltage I need to get the 1.21 gigawatts of dissapation from my flex capacitor... I never could get my amps up to 88 miles per hour though, even when I am cursing and hurling them around the room. Maybe if I could though, I could ride it back to 1955 and pick up some sweet deals on old tubes!!!
Easy question ... maybe
OK. I have been going crazy trying to figure out what a measure of the top tube's merit is.
This is what I KNow:
The top tube is converting the changing current from the bottom tube into voltage. That makes it a transimpedance device. It is operating as a grounded grid amplifier, with the varying current going in through the cathode and out through the plate..
Mu = ratio of the plate's effectiveness over the grid's effectiveness in controlling current flow from cathode to plate
Gm=ability to vary current conduction in response to a change in its voltage
plate's gm= 1/Ri
grid'sGm=mu/Ri
Cathode's Gm=(mu+1)/Ri
This is what I need to know PLEASE: How do I know what tube will do this better? What is a cathode's ability to change current into voltage called!!! Conductance? Do I just use ohms law to see what the resulting gain will be at the plate using the modified Ri of the top tube (ra'+ra*(u+1))? What then is the effect of the grounded grid being in the way? How do I calculate a tube's cathode's ability to convert current to voltage?
PRR said:
OK. I have been going crazy trying to figure out what a measure of the top tube's merit is.
This is what I KNow:
The top tube is converting the changing current from the bottom tube into voltage. That makes it a transimpedance device. It is operating as a grounded grid amplifier, with the varying current going in through the cathode and out through the plate..
Mu = ratio of the plate's effectiveness over the grid's effectiveness in controlling current flow from cathode to plate
Gm=ability to vary current conduction in response to a change in its voltage
plate's gm= 1/Ri
grid'sGm=mu/Ri
Cathode's Gm=(mu+1)/Ri
This is what I need to know PLEASE: How do I know what tube will do this better? What is a cathode's ability to change current into voltage called!!! Conductance? Do I just use ohms law to see what the resulting gain will be at the plate using the modified Ri of the top tube (ra'+ra*(u+1))? What then is the effect of the grounded grid being in the way? How do I calculate a tube's cathode's ability to convert current to voltage?
aletheian said:
Sure! I Spice them, no hand math (real bad at it anyway). Then breadboard them, tweak and finalize them on chassis.
I have examples of practical Loftin-White cathode followers (gain <1):
(my everyday used headphone amp)
http://geek.scorpiorising.ca/contrib/Geek/Concerto_Headphone.png
(as the linedriver to my used-daily stereo control centre)
http://geek.scorpiorising.ca/contrib/Geek/tubeCtrlCtr.png
(in the linedriver for my recording mic preamp)
http://geek.scorpiorising.ca/contrib/Geek/TubeMicPreamp-withoutCompress.png
(generic)
http://geek.scorpiorising.ca/contrib/Geek/12AU7-White-CF.png
And mu-stage (gain>1):
(circuit used in my AT24/2600tuner)
http://geek.scorpiorising.ca/contrib/Geek/6N1P_srpp.png
(generic)
http://geek.scorpiorising.ca/contrib/Geek/12AU7-SRPP.png
Hope it helps!
I thought about the CCS route. I kicked around the idea of a follower too. Both would help actualize the max gain potential.
Any examples?
Most stuff I find with cascodes are small gain circuits. I have ben looking for TV and radio cascode circuits that have HUGE gains, but schematics are scarce.
And I still don't know what the ideal parameters for the top tube are. But I found lots of old literature on grounded gris amps, so I'll be pouring over that stuff over the next few days. I am in the dark with transimpedance stuff... well, outside the opamp world that is
Any examples?
Most stuff I find with cascodes are small gain circuits. I have ben looking for TV and radio cascode circuits that have HUGE gains, but schematics are scarce.
And I still don't know what the ideal parameters for the top tube are. But I found lots of old literature on grounded gris amps, so I'll be pouring over that stuff over the next few days. I am in the dark with transimpedance stuff... well, outside the opamp world that is
Geek said:
B-b-b-but... i like math.
Seriously though. Spice lies to me. I have tried every ******** program I could find and they all lie. I bought Electronics workbench years ago when I was doing the engineering school thing, but I could never get it to work right with tubes... actuallym it worked fine with the 3!!!!!!!!!! ******** tube models that it came with, but it would not let me make new ones as in it would revert all the parameters to zero when I tried to save the library. Tried the multisim demo... same thinf.
Eventually, I got s text based spice to work on my Mac, but text based is SUCH a pain in the ar-se that I'd rather do the math by hand.
What proggies do you use?
Geek said:Circuit Maker Student - it's free and the DC modelling is *deadly* accurate.
AC, well... meh.
For AC power output measurements, divide the result by 5
That's why I DC model the bias, then finish-up on the breadboard.
I'm actually running that right now.. I dusted it off, and am trying to get it going.
It modeled resistance coupled triode stages pretty accurately, but with signal pentodes... it all went to h#LL. I can't even get it close to reality. I mean, a resistance coupled 5879-, with 250V on it and a plate resistor of 470K should have a BIT more gain than 5.73V!!! Like 20x that much.
Cascodes too... no dice. I got millivolts out with circuits that have measured 150X and up gains on my bench. And I modelled an SRPP... pfffft! It just freaked out.
Can't get my netlists to import right either... grrr.
For a top tube in a Cascode try a pentode... 1/gm looking into cathode...
As for Orcad .... You would need to write your own modeling for a tube... The standard 1st order equations are not going to get you any usable real life results... The model is is only accurate as you make it...Common available tube models are not always to good...
I prefer to derive high order polynomials based on real life measured curves of a "boogie" valve..... ALthough RCA, GE manuals have curves....they are typically the cream of the crop for marketing purposes back in the day... Most NOS still fall close to the manual curves..Telefunken would usually surpass thier data sheets!!!!
When stacking tubes in ORCAD....you will get netlist error due to this node not having ground return....So, the fix is to place a 1 GIG resistor from this center of the tubes to ground...Sort of consider this like an "air" resistor..but it makes the simulation converge and run...
As for Orcad .... You would need to write your own modeling for a tube... The standard 1st order equations are not going to get you any usable real life results... The model is is only accurate as you make it...Common available tube models are not always to good...
I prefer to derive high order polynomials based on real life measured curves of a "boogie" valve..... ALthough RCA, GE manuals have curves....they are typically the cream of the crop for marketing purposes back in the day... Most NOS still fall close to the manual curves..Telefunken would usually surpass thier data sheets!!!!
When stacking tubes in ORCAD....you will get netlist error due to this node not having ground return....So, the fix is to place a 1 GIG resistor from this center of the tubes to ground...Sort of consider this like an "air" resistor..but it makes the simulation converge and run...
cerrem said:
I'll be darned, it works. That's going to get confusing...I'll be building circuits with random resistors to ground all over the place. HA
So in Circuitmaker, how do I make another tube? They have tetrodes and triodes in there and diodes/pentodes for "schematic purposes only" but I can only seem to edit the existing ones, not make new ones. So I could lose the 6L6 model by pasting a KT** model over it... Do I have to do that out-of-program? Or can I just import them somehow.
aletheian said:
Pentodes, I dunno how to do
Triodes, I make a copy of a triode in triode.sub and place it at the end of the file and rename the triode to my target triode number.
Setup a DC curve tracer with appropriate voltages for the triode and just start changing parameters from within CM Student until the curves match the one in the datasheet.
It's been so long since I've done it, I can't tell you which param changes what on the curves
Geek said:
Pentodes, I dunno how to do
Triodes, I make a copy of a triode in triode.sub and place it at the end of the file and rename the triode to my target triode number.
Setup a DC curve tracer with appropriate voltages for the triode and just start changing parameters from within CM Student until the curves match the one in the datasheet.
It's been so long since I've done it, I can't tell you which param changes what on the curves
Cool man, I'll give that a try.
Hey, as a side note, do you run the geek zone? I saw your signature down at the bottom there. Great website.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.