Are you sure you want to do this project?
There are a lot of things in this thread that are not meshing in my mind.
You say "100mW into 32R is the design spec", but then worry about 300Ω. If it is not in the spec, you do not care what the heck it does in 300Ω. If you suspect that you will finish the design and then they tell you about 300Ω, get that written into the spec before you go any further.
You are already designing the output transformer, despite a spec that you know is incomplete, but you seem to have an unclear understanding of, not just tubes, but general power amp design.
Every time we talk about power, you talk about feedback. Feedback does not change the power available.
Twice you have used the purple-face in phrases like "Me with tube audio:
" Maybe you would rather be working with chips?
> So the voltage swing on the plate does increase with reduced loading of the secondary. Whither NFB? If the plate has not been overloaded, then why does the voltage swing change at all. The input stage will compare a fraction of the output voltage, see it is incorrect, and do something about it. Or at least that's how it works in power supplies, transistor amplifiers, servomechanisms, etc.
Well, no.
Below overload, a feedback amp will hold constant voltage, true. Take a 100W 8Ω audio power amp, drive the output to 1 Volt, and it will hold it constant into loads from 1MegΩ to 1Ω.
But note that 1V into 8Ω or 1Ω is only 0.125W or 1W, far less than the amp's 100W rating.
The notable thing about Power Amps is that you can be sure they will be run right up to the overload point! Power costs money. In a Power Amp, the Power Rating is the second most important spec (after the price!). In fact the only hard spec you have shown us is "100mW into 32R", the power spec.
You seem to know transistor power amplifers. What is their rating with varying load impedance?
If NFB worked as you describe: a "100W 8Ω" amp would make 200W in 4Ω, 400W in 2Ω, and 800W in 1Ω. In our dreams!
In fact a 100W 8Ω amp typically makes 160W in 4Ω. The voltage sags. At 2Ω the protection circuit will cut-out at very low level, but if you bypass protection and double-up the output transistors you may see 200W in 2Ω, 250 in 1Ω. Where is the "constant voltage" going???
In a transistor amp, a little is lost across the transistors. And the power transformer is a big resistor, which becomes significant when you ask for more power than you paid for. The rails voltages sag at high power. +/-55V at no-load, +/-40V working full-blast in 8Ω, and +/-25V if you try for full power in 1Ω. The power supply has source impedance. NFB can not make a 40V peak when the supply rails are 25V.
Oh sure, you can fix this: use a MUCH bigger power transformer. Of course more power is always more money.
The only thing different in tubes is: we usually run near class A, so you don't get the huge current variation of an AB tranny amp and thus the rail-sag may not be bad. However the best electron-flock in vacuum is a pretty crummy conductor. Tube impedance is almost always a major factor in design.
Stepping back: why are you even planning push-pull in a 100mW amp? Especially with a low voltage tube and a transformer? The minimum heater power for a push-pull amp with phase-splitter will be something like 4 watts. To get 0.1Watts output??? While heater power is cheaper than plate power, this thing is power-inefficent. Also you have at least 16 pins per channel to wire. With a single-ended amp you could do it in 8 pins. SE is not as clean as P-P but it is faddish. And since 0.1W is such a small amount of power compared to what ordinary radio-TV tubes can do, you can aim for 0.5W at 10% THD then rate it at 0.1W at low THD. The simplification of wiring and space might pay for better transformer (which SE needs anyway) and better caps (SE will need fewer signal caps). With cheap modern bulk caps, power supply buzz in SE should not have to be an issue.
Or you can do what many do: use a 18V supply driving an LM377 chip, and stick a tube on top of the chassis for decoration.
There are a lot of things in this thread that are not meshing in my mind.
You say "100mW into 32R is the design spec", but then worry about 300Ω. If it is not in the spec, you do not care what the heck it does in 300Ω. If you suspect that you will finish the design and then they tell you about 300Ω, get that written into the spec before you go any further.
You are already designing the output transformer, despite a spec that you know is incomplete, but you seem to have an unclear understanding of, not just tubes, but general power amp design.
Every time we talk about power, you talk about feedback. Feedback does not change the power available.
Twice you have used the purple-face in phrases like "Me with tube audio:
" Maybe you would rather be working with chips? > So the voltage swing on the plate does increase with reduced loading of the secondary. Whither NFB? If the plate has not been overloaded, then why does the voltage swing change at all. The input stage will compare a fraction of the output voltage, see it is incorrect, and do something about it. Or at least that's how it works in power supplies, transistor amplifiers, servomechanisms, etc.
Well, no.
Below overload, a feedback amp will hold constant voltage, true. Take a 100W 8Ω audio power amp, drive the output to 1 Volt, and it will hold it constant into loads from 1MegΩ to 1Ω.
But note that 1V into 8Ω or 1Ω is only 0.125W or 1W, far less than the amp's 100W rating.
The notable thing about Power Amps is that you can be sure they will be run right up to the overload point! Power costs money. In a Power Amp, the Power Rating is the second most important spec (after the price!). In fact the only hard spec you have shown us is "100mW into 32R", the power spec.
You seem to know transistor power amplifers. What is their rating with varying load impedance?
If NFB worked as you describe: a "100W 8Ω" amp would make 200W in 4Ω, 400W in 2Ω, and 800W in 1Ω. In our dreams!
In fact a 100W 8Ω amp typically makes 160W in 4Ω. The voltage sags. At 2Ω the protection circuit will cut-out at very low level, but if you bypass protection and double-up the output transistors you may see 200W in 2Ω, 250 in 1Ω. Where is the "constant voltage" going???
In a transistor amp, a little is lost across the transistors. And the power transformer is a big resistor, which becomes significant when you ask for more power than you paid for. The rails voltages sag at high power. +/-55V at no-load, +/-40V working full-blast in 8Ω, and +/-25V if you try for full power in 1Ω. The power supply has source impedance. NFB can not make a 40V peak when the supply rails are 25V.
Oh sure, you can fix this: use a MUCH bigger power transformer. Of course more power is always more money.
The only thing different in tubes is: we usually run near class A, so you don't get the huge current variation of an AB tranny amp and thus the rail-sag may not be bad. However the best electron-flock in vacuum is a pretty crummy conductor. Tube impedance is almost always a major factor in design.
Stepping back: why are you even planning push-pull in a 100mW amp? Especially with a low voltage tube and a transformer? The minimum heater power for a push-pull amp with phase-splitter will be something like 4 watts. To get 0.1Watts output??? While heater power is cheaper than plate power, this thing is power-inefficent. Also you have at least 16 pins per channel to wire. With a single-ended amp you could do it in 8 pins. SE is not as clean as P-P but it is faddish. And since 0.1W is such a small amount of power compared to what ordinary radio-TV tubes can do, you can aim for 0.5W at 10% THD then rate it at 0.1W at low THD. The simplification of wiring and space might pay for better transformer (which SE needs anyway) and better caps (SE will need fewer signal caps). With cheap modern bulk caps, power supply buzz in SE should not have to be an issue.
Or you can do what many do: use a 18V supply driving an LM377 chip, and stick a tube on top of the chassis for decoration.
PRR:
Let us approach this situation from an analogous perspective. I specialize in switching power supply design; another engineer might specialize in digital ICs. It is highly unlikely the digital guy will have the know-how to make the deceptively simple push-pull converter. Conversely, it might take the pain of death upon me before I could figure out how to make the basic 4-bit adder/accumulator (a particularly painful final exam question I recall even 12 years later). The difference between my perspective and yours, apparently, is that I don't automatically consider the digital guy an idiot for not knowing how to design a very tricky SMPS circuit.
I am not exactly well-versed in tube audio amplifier design. I own Morgan Jones' "Valve Amplifiers" (1st ed.) and while I've read it many times over - mostly because it's a damn fine read - I am not under the illusion that reading one book on this topic makes me an expert. Not even an amateur, even. Once I apply what is in the book to making a tube amp from scratch, I will have then become an amateur. Build a few dozen more with widly different topologies and I'll be well on the way to being an expert (but not quite). So this little headphone amp project is a relatively unambitious one that will afford me a bit of practical experience in an area of technology that wasn't even mentioned when I went to school. In other words, mostly for fun, but also because I want to see for myself what all the fuss over tubes is about.
I care because I am making this for myself and I own headphones with 32R, 48R and 300R impedance. I specified 100mW into 32R probably because of my familiarity with solid state design - one is most interested in the maximum current required, after all, when designing such.
I keep bringing up feedback, because you keep describing situations that shouldn't happen with feedback operating!
PRR, you are the one that keeps bringing up overload. I never once asked "what will the amp do at overload?" because I already know the answer: the plate voltage will sag and feedback will be powerless - ahem - to stop it. You stated that the plate voltage will rise with lessened load, which implies that, somehow, NFB won't keep it constant. This statement struck me as fundamentally incorrect, but I allowed for the possibility that I might be unaware of some "fundamental aspect" of tube design. I can see, for example, how reducing the load on the secondary would reduce the voltage drop across the primary, but I still fail to see how this causes the signal swing on the plate to rise. In fact, it should decrease slightly, since NFB is no longer calling for more signal swing to offset the IR drop of the transformer.
I described the inverse order, actually, but if the power supply were perfect and the output devices could handle the current, then, yes, that is precisely how a power amplifier would behave. Of course, such is not the case in real-life, and sagging supply voltage will rapidly erode the peak-to-peak voltage swing the output devices can deliver before distorting.
This is rather elementary, and I would think you could tell from context that I had these sorts of things down already.
To eliminate even-order distortion as much as possible and to allow true balanced operation, if so desired.
Was this really necessary, PRR? An obsolete amplifier chip of the same class as the (in)famous LM380?
I was trying to add a bit of humor to the mix, PRR, and, I admit, moreso trying to use some of the quirky Smilies here to good use.
Finally, I spend a good portion of my time - awake and not - designing rather complicated power supplies. When I come here, its with the intent of sharing in the camaraderie of like-minded individuals, but not necessarily like-skilled. I have no interest in getting into ****ing contests with other "contributors" about who is smartest or most knowledgeable - for one thing, I don't care; for another, it's not very becoming, is it?
Before you head to Texas, think about what I've written and how I wrote it.
Regards,
Jeff
Let us approach this situation from an analogous perspective. I specialize in switching power supply design; another engineer might specialize in digital ICs. It is highly unlikely the digital guy will have the know-how to make the deceptively simple push-pull converter. Conversely, it might take the pain of death upon me before I could figure out how to make the basic 4-bit adder/accumulator (a particularly painful final exam question I recall even 12 years later). The difference between my perspective and yours, apparently, is that I don't automatically consider the digital guy an idiot for not knowing how to design a very tricky SMPS circuit.
I am not exactly well-versed in tube audio amplifier design. I own Morgan Jones' "Valve Amplifiers" (1st ed.) and while I've read it many times over - mostly because it's a damn fine read - I am not under the illusion that reading one book on this topic makes me an expert. Not even an amateur, even. Once I apply what is in the book to making a tube amp from scratch, I will have then become an amateur. Build a few dozen more with widly different topologies and I'll be well on the way to being an expert (but not quite). So this little headphone amp project is a relatively unambitious one that will afford me a bit of practical experience in an area of technology that wasn't even mentioned when I went to school. In other words, mostly for fun, but also because I want to see for myself what all the fuss over tubes is about.
PRR said:
I care because I am making this for myself and I own headphones with 32R, 48R and 300R impedance. I specified 100mW into 32R probably because of my familiarity with solid state design - one is most interested in the maximum current required, after all, when designing such.
I keep bringing up feedback, because you keep describing situations that shouldn't happen with feedback operating!
PRR, you are the one that keeps bringing up overload. I never once asked "what will the amp do at overload?" because I already know the answer: the plate voltage will sag and feedback will be powerless - ahem - to stop it. You stated that the plate voltage will rise with lessened load, which implies that, somehow, NFB won't keep it constant. This statement struck me as fundamentally incorrect, but I allowed for the possibility that I might be unaware of some "fundamental aspect" of tube design. I can see, for example, how reducing the load on the secondary would reduce the voltage drop across the primary, but I still fail to see how this causes the signal swing on the plate to rise. In fact, it should decrease slightly, since NFB is no longer calling for more signal swing to offset the IR drop of the transformer.
I described the inverse order, actually, but if the power supply were perfect and the output devices could handle the current, then, yes, that is precisely how a power amplifier would behave. Of course, such is not the case in real-life, and sagging supply voltage will rapidly erode the peak-to-peak voltage swing the output devices can deliver before distorting.
This is rather elementary, and I would think you could tell from context that I had these sorts of things down already.
To eliminate even-order distortion as much as possible and to allow true balanced operation, if so desired.
Was this really necessary, PRR? An obsolete amplifier chip of the same class as the (in)famous LM380?
Twice you have used the purple-face in phrases like "Me with tube audio:
I was trying to add a bit of humor to the mix, PRR, and, I admit, moreso trying to use some of the quirky Smilies here to good use.
Finally, I spend a good portion of my time - awake and not - designing rather complicated power supplies. When I come here, its with the intent of sharing in the camaraderie of like-minded individuals, but not necessarily like-skilled. I have no interest in getting into ****ing contests with other "contributors" about who is smartest or most knowledgeable - for one thing, I don't care; for another, it's not very becoming, is it?
Before you head to Texas, think about what I've written and how I wrote it.
Regards,
Jeff
Hi,
While I can understand why you would like to have a balanced operation of the amp, striving to suppress even order harmonic distortion is like saying I want this thing to sound like transistor.
I am not saying PP amps sound like transistor amps but even order distortion is something I can live with, odd order HD is much more unpleasant sounding.
So, as someone else suggested already you need a little PP power ( if you want to stick to balanced operation throughout) amp that can cope with loads varying from say 32R ( I'd make it 8 Ohms so you can also hook up a speaker) up to 300R.
No big deal and you won't even need to resort to lots of global NFB to obtain good results.
As EC8010 suggested, design for a 4 to 5W PP Class A amp and that would cover all targets in one go.
The endresult may vary on how the PSU is designed and a ton of other details...at least you'd have something to go from.
A PP 6C19Pi would be nice to have for this...a few other tubes will qualify nicely as well.
Cheers,
While I can understand why you would like to have a balanced operation of the amp, striving to suppress even order harmonic distortion is like saying I want this thing to sound like transistor.
I am not saying PP amps sound like transistor amps but even order distortion is something I can live with, odd order HD is much more unpleasant sounding.
So, as someone else suggested already you need a little PP power ( if you want to stick to balanced operation throughout) amp that can cope with loads varying from say 32R ( I'd make it 8 Ohms so you can also hook up a speaker) up to 300R.
No big deal and you won't even need to resort to lots of global NFB to obtain good results.
As EC8010 suggested, design for a 4 to 5W PP Class A amp and that would cover all targets in one go.
The endresult may vary on how the PSU is designed and a ton of other details...at least you'd have something to go from.
A PP 6C19Pi would be nice to have for this...a few other tubes will qualify nicely as well.
Cheers,
I see, especially by your remarks at Headwize, that you are a highly savvy SMPS designer.
I admittedly am NOT: the one SMPS I designed and built ran like an audio amp with low efficiency.
Each field, as you say, has its own "self evident" rules of design. You and I might struggle just to guess the number of gates needed for a basic 4-bit adder/accumulator; the experienced digital dude can glance at the specs of a novel digital problem and get a good estimate of gate-count immediately.
> I don't automatically consider the... guy an idiot
Not an idiot. Clearly you are no idiot, nor even ignorant. But you so very experienced in the SMPS field that you don't seem to see the quite different basic problems of audio amp design.
I am trying to shock you out of that mind-set. I don't have the finger-stamina to teach you tube audio amp design here; anyway you say you have read-up and there is no point me duplicating the words of my betters. So I feel I must "smack you in the head with a 2x4 and get your attention". I think you would be very good at audio design, but you have to stop thinking like a SMPS.
If I succeed, and you say "OH!", then my initial rudeness will be forgotten in the clarity of new insight.
> mostly for fun, but also because I want to see for myself
A side-point: because you are a Professional Designer, your initial post sounded to me like you had been engaged on a commercial project. Which is why I asked "Are you sure you want this project?" and "Use a 377 with a bottle on top", a general plan which IS a successful commercial product (Yes, there are far better choices than 377!) Now that it is clearly a personal fun project, I don't have to care about your professional and company success--- the risk is 100 hours and $100, not tens of thousands and company bankruptcy.
Now, differences between SMPS and Tube Audio:
> deliberately place 30R of resistance in series...? {from here}
No, ....
> solid state design - one is most interested in the maximum current
And this may be one difference. Tubes always have significant output resistance(*) and this is always where you start a triode design. For triode voltage amps an experienced designer may not refer directly to the Rp, but in triode power amps Rp is your first parameter.
(*)(In fact output resistance tends to be the product of Mu and Cathode area: cathodes and their heaters are costly. Cathode-costs dominate most audio power amps, and transformer winding issues are a lot about how low you can make Rp.)
So you "never" add resistance to a tube's output. You have too much already. You want to best-match the unavoidable resistance to the load, and also avoid high transformer winding costs. For full-power work you must also include power supply impedance reflected through the amp topology. (For peaky speech/music, with modern large PS caps, PS impedance tends to be insignificant.)
For a fixed-Z load and a triode with high plate ratings, Rl should be 2X to 10X Rp. 2X is best power, 10X is significantly lower distortion but much less power.
For a triode with tight voltage and power ratings such as 6DJ8, the optimum must be found by playing with the plate curves or a simulator or a breadboard.
For variable-Z work, try an oversize amp with lowest-Z load about 2Rp (or whatever looks best on the curves). Then compute power in the highest-Z load and see if it meets goal. Iterate with different matching (you can load a triode with less than Rp without gross distortion), tapped transfromer, or bigger power until everybody is happy (or the budget bursts).
> I never once asked "what will the amp do at overload?"
Well, maybe I assumed that was your initial question. But maybe we are just coming at it from entirely different perspectives. You asked "For maximum power transfer ... this gets a bit complicated when the load impedance ranges from 32 to 300 ohms.
That sounds like an overload question to me.
Do not take offense if I state the "obvious", because I want to see where we are just-missing each other's thoughts.
Part of it is: a SMPS will typically run at fixed output voltage. And it may see a load Z that varies constantly (motors turn on and off, CPUs shut-down idle parts of the die).
Audio, as you know, is a constantly varying voltage. Also, at least to first approximation, the load Z does not vary within a single listening session. You have 32Ω and 300Ω phones, but you can not change from 32Ω to 300Ω instantly. You have to un-plug and re-plug, take-off and put-on. The change in power available from 32 to 300 cans is not an issue: users are normally advised to turn volume down before plugging phones. This is advisable even if both phones are the same Z, because different phones have VERY different power efficiencies.
We do tend to run NFB and constant-voltage sources in audio. Speakers like it, and Y-cables need lo-Z drive so that level does not change with number of loads. Also NFB reduces device variation and reduces distortion.
However NFB in the amplifier is "local". The whole-sound-system is under another feedback loop: the user's ears and the hand on the Volume knob. If all headphones had similar power sensitivities, it might be elegant to put "constant power" NFB around the amplifier. But the variation in headphone power sensitivity is even greater than the variation in power transfer for the common headphone impedances. And a voltage-gain pot is a lot simpler than a power control.
So a volume knob (somewhere in the system) is necessary: users' needs/tastes vary, and even at the same Z headphones vary. The variation of power coupling from 32 to 300Ω is probably the least issue.
Now as to why I consider overload: users can and do listen at any level from very soft to very loud. Most of the time the overload point is not an issue, EXCEPT in that it influences initial price. However when a user can't get enough loudness, he complains. The designer's defense is that "you got what you paid for". Nearly any listening level is possible, at a price. In low-efficiency loudspeakers, the price may become fairly high if the user has high demands.
And unlike a SMPS, you can't tell users "don't do that!" A SMPS will typically be used in a known system and users understand that it will shut-down if overloaded with low-Z load. The audio case is more like a SMPS user turning the voltage-set up too high, and again SMPS users should know there are limits, but audio users will do it anyway.
I admittedly am NOT: the one SMPS I designed and built ran like an audio amp with low efficiency.
Each field, as you say, has its own "self evident" rules of design. You and I might struggle just to guess the number of gates needed for a basic 4-bit adder/accumulator; the experienced digital dude can glance at the specs of a novel digital problem and get a good estimate of gate-count immediately.
> I don't automatically consider the... guy an idiot
Not an idiot. Clearly you are no idiot, nor even ignorant. But you so very experienced in the SMPS field that you don't seem to see the quite different basic problems of audio amp design.
I am trying to shock you out of that mind-set. I don't have the finger-stamina to teach you tube audio amp design here; anyway you say you have read-up and there is no point me duplicating the words of my betters. So I feel I must "smack you in the head with a 2x4 and get your attention". I think you would be very good at audio design, but you have to stop thinking like a SMPS.
If I succeed, and you say "OH!", then my initial rudeness will be forgotten in the clarity of new insight.
> mostly for fun, but also because I want to see for myself
A side-point: because you are a Professional Designer, your initial post sounded to me like you had been engaged on a commercial project. Which is why I asked "Are you sure you want this project?" and "Use a 377 with a bottle on top", a general plan which IS a successful commercial product (Yes, there are far better choices than 377!) Now that it is clearly a personal fun project, I don't have to care about your professional and company success--- the risk is 100 hours and $100, not tens of thousands and company bankruptcy.
Now, differences between SMPS and Tube Audio:
> deliberately place 30R of resistance in series...? {from here}
No, ....
> solid state design - one is most interested in the maximum current
And this may be one difference. Tubes always have significant output resistance(*) and this is always where you start a triode design. For triode voltage amps an experienced designer may not refer directly to the Rp, but in triode power amps Rp is your first parameter.
(*)(In fact output resistance tends to be the product of Mu and Cathode area: cathodes and their heaters are costly. Cathode-costs dominate most audio power amps, and transformer winding issues are a lot about how low you can make Rp.)
So you "never" add resistance to a tube's output. You have too much already. You want to best-match the unavoidable resistance to the load, and also avoid high transformer winding costs. For full-power work you must also include power supply impedance reflected through the amp topology. (For peaky speech/music, with modern large PS caps, PS impedance tends to be insignificant.)
For a fixed-Z load and a triode with high plate ratings, Rl should be 2X to 10X Rp. 2X is best power, 10X is significantly lower distortion but much less power.
For a triode with tight voltage and power ratings such as 6DJ8, the optimum must be found by playing with the plate curves or a simulator or a breadboard.
For variable-Z work, try an oversize amp with lowest-Z load about 2Rp (or whatever looks best on the curves). Then compute power in the highest-Z load and see if it meets goal. Iterate with different matching (you can load a triode with less than Rp without gross distortion), tapped transfromer, or bigger power until everybody is happy (or the budget bursts).
> I never once asked "what will the amp do at overload?"
Well, maybe I assumed that was your initial question. But maybe we are just coming at it from entirely different perspectives. You asked "For maximum power transfer ... this gets a bit complicated when the load impedance ranges from 32 to 300 ohms.
That sounds like an overload question to me.
Do not take offense if I state the "obvious", because I want to see where we are just-missing each other's thoughts.
Part of it is: a SMPS will typically run at fixed output voltage. And it may see a load Z that varies constantly (motors turn on and off, CPUs shut-down idle parts of the die).
Audio, as you know, is a constantly varying voltage. Also, at least to first approximation, the load Z does not vary within a single listening session. You have 32Ω and 300Ω phones, but you can not change from 32Ω to 300Ω instantly. You have to un-plug and re-plug, take-off and put-on. The change in power available from 32 to 300 cans is not an issue: users are normally advised to turn volume down before plugging phones. This is advisable even if both phones are the same Z, because different phones have VERY different power efficiencies.
We do tend to run NFB and constant-voltage sources in audio. Speakers like it, and Y-cables need lo-Z drive so that level does not change with number of loads. Also NFB reduces device variation and reduces distortion.
However NFB in the amplifier is "local". The whole-sound-system is under another feedback loop: the user's ears and the hand on the Volume knob. If all headphones had similar power sensitivities, it might be elegant to put "constant power" NFB around the amplifier. But the variation in headphone power sensitivity is even greater than the variation in power transfer for the common headphone impedances. And a voltage-gain pot is a lot simpler than a power control.
So a volume knob (somewhere in the system) is necessary: users' needs/tastes vary, and even at the same Z headphones vary. The variation of power coupling from 32 to 300Ω is probably the least issue.
Now as to why I consider overload: users can and do listen at any level from very soft to very loud. Most of the time the overload point is not an issue, EXCEPT in that it influences initial price. However when a user can't get enough loudness, he complains. The designer's defense is that "you got what you paid for". Nearly any listening level is possible, at a price. In low-efficiency loudspeakers, the price may become fairly high if the user has high demands.
And unlike a SMPS, you can't tell users "don't do that!" A SMPS will typically be used in a known system and users understand that it will shut-down if overloaded with low-Z load. The audio case is more like a SMPS user turning the voltage-set up too high, and again SMPS users should know there are limits, but audio users will do it anyway.
So the design of an audio power amp goes:
Pick a power/price ratio that will sell well. (In this case: a power that should satisfy you, and at headphone power levels cost is not real linear with power; price tends to be a different issue in hand-built experimental gear but you would not normally, for instance, pick a quad of 300B for headphones.)
Pick a tube that seems likely to give more than enuff power. If you stick to the usual suspects, the databook is your friend.
But power triodes, and especially very low power triodes, have not been in fashion since 1934. You have to take your best guess then simulate (on curves, in SPICE, or on a breadboard) to estimate the parameters. Your best first guess for load-line is 2Rp, but tubes not meant for Power use may have voltage, power, or current limits which do not allow easy optimization.
Then compare the estimated parameters to the costs of power supply, transformer, and driver.
Here is a rule of thumb: for a transformer coupled triode with R-C driver working from the same bulk supply as the output stage, you want a Mu (Amplification Factor) of about 6. Any lower, and drive voltage becomes so high that you need an extra power supply for the driver. Any higher, and Rp becomes too high for good power at a given supply voltage, also the transformer costs will tend to be high.
In this case, there are no small twin triodes with Mu about 6. However there are some with apparently exceptionally good Gm at modest current, designed for TV tuner preamps. 6BQ7 is classical, 6BK7 is a high-spec 6BQ7, and 6DJ7 is an improved amplifier but with lower plate ratings. They have high Mus and are not great power amps, but at part-Watt power levels we can live with that. And the easy drive is a nice little benefit.
It should be noted that none of these types were expected to run AT their ratings for many hours: in a TV set, if you are getting a good picture the tuner is under AGC and preamp current is less than maximum. I once used 6BK7 as a headphone amp within ratings but very hot and did observe "softening" of power after say 100 hours.
There are also the old-style medium-Mu types like 12AU7, mentioned in another forum. While the 12AU7 has lower Mu than 6DJ8, it has much lower Gm at typical current. The plate resistance is high. To make power, you need a high plate voltage. To deliver that power in "hi-fi", you need an awful lot of turns on the transformer. 12AU7 was often used as pro-gear line amps (a job not unlike headphones), but more for ready availability and known reliability than for best audio performance.
For a hobbyist headphone amp where tube wear-out is no big deal, 6DJ8 and similar look better than 12AU7 or 6BK7. The lower Rp significantly reduces transformer issues, and it is easier to drive. When I picked 6BK7, the BK was $1.38 and the 6DJ7 had not become available (yes, it was a while ago).
6DJ8 at high current runs Rp near 3K. In single-ended we might load it in 6K minimum for best power and distortion over a range of load impedances. If we assume minimum load is 64Ω, the transformer ratio is 10:1. If we run push-pull pure-A, we use two such ratios with a 32Ω load and each tube will still see about 6K load.
Here is what SPICE says:
push-pull 6DJ8
B+ = 120V
Rk = 200 ohms
transformer turns ratio,
one side of primary to whole secondary: 10:1
According to SPICE (Koren model):
16 ohms, 0.17W, 0.9%, (AB)
16 ohms, 0.12W, 0.6%, (A)
32 ohms, 0.38W, 0.9%, (AB)
32 ohms, 0.22W, 0.5%, (A)
100 ohms, 0.19W, 0.3%, (A)
300 ohms, 0.09W, 0.3%, (A)
600 ohms, 0.05W, 0.3%, (A)
Unloaded Gain:
(one grid) : (out) = 1:1.38
32 ohm gain:
(one grid) : (out) = 0.858
Zout = 19 ohms
The distortion figures are rough. "0.3%" is numeric noise; it will drop in higher impedances.
So it has ample power in 32Ω. (More than ample; recall that the user has a Volume control and need not run it at 0.22-0.38 watts.) Power is less than 100mW in 600Ω but the 600Ω phones tend to have better power sensitivity. I assure you that this range of power will be enough for most users. Dozens of mW satisfy many listeners in most phones.
As you know, the transformer is a critical piece. But the design of the tranny has to start with an estimate of source impedance and voltage swing, so we need to look at likely tubes first. We may iterate: 6BL7's low-Z plate makes a nice transformer but a large chassis for a tall octal and big heater. 12AU7 is about the cheapest little power-bottle around but its hi-Z plate makes transformer design messy. But stick with the 6DJ8 and see where it goes....
We start with a core that can probably handle the flux and give us the inductance we need. If you aim for 20Hz, you need a larger core than you would use on a 50/60Hz power transformer of similar VA. Since iron price does not scale much with VA in the range 0.1W to 5W, and we need a LOT of turns and hate hyper-fine wire, pick a large core. 5VA is cheap and not too large for good bass response, though it may wind-up with more capacitance than we like for high-end response, especially under NFB.
There are special irons for audio, and even now funny-stuff I would call "ferrite" (probably not the right word). But for small audio power, conventional power iron will give a workable design. Work it up with ordinary power-transformer iron to get in the ballpark. Then when 99% happy, check what is possible with other irons.
Since it will still be fine wire on the primary, insulation wastes spece. Set aside about 2/3rds of the window area for primary, deduct some fishpaper, and fill it with fine wire. Compute the inductance and the flux. Flux level in audio transformers is a complex topic. For starters: design for much less than half the flux you would allow in a power transformer. The inductance for a triode amplifier should give a bass impedance similar to triode plate impedance. (Plate R in parallel with load Z for fixed-Z loading, but with 32Ω to 300Ω we can't count on the load.)
When you have enough core and winding to give flux and inductance, find the leakage inductance and stray capacitance.
There are estimating tools in Radiotron 4th. Not an easy task. What really happens is you wind a sample with a small number of sections, and measure it in-circuit, then if it won't meet spec you change the number of sections (or change the specs).
Transformer:
turns ratio, one side of primary to whole secondary: 10:1
I still say: mock one up with a 12VA 240VCT/12+12V 50/60Hz power transformer. Because these things barely-work at 168V peak at 50Hz, it may get a little soft at 20Hz even at more like 70V peak. But we rarely need full level at 20Hz. 50/60Hz power transformers are not designed for 20KHz response, but it has to be much better than 60Hz to pass rectifier pulses. What seems to get you first is eddy-current losses: the response droops but not at 6dB/8ve.
Also: try the secondaries in parellel and one alone, other open. See how much difference there is. Sure, the parallel connection will have lower copper loss, but the difference is less than a dB. The parallel connection has lower leakage inductance, but also more stray capacitance, so the frequency of the top-resonance does not change much. Because of core losses, the ringing is less than you might think from a simple lumped-circuit model. I'm not seriously advocating using a power transfromer, but it will be a solid performer 60Hz-5KHz and let you test load impedance and power relationships before you move on to...
Custom Tranny:
For 20-20KHz with 10dB NFB:
Primary inductance, one side: 15H
Primary resistance, one side: 300-1,000 ohms
Leakage inductance referred to primary: under 15mH (K= about 0.999)
stray capacitance referred to secondary: under 0.04uFd
Seat-Of-Pants Interpretation: No good answer exists for heavy NFB with widely varying load with normal audio transformer construction. Response at different loads will peak or dip, phase run wild. Values above attempt to keep phase margin OK out to 70KHz, which with a little fixed compensation in voltage-amp will allow 10dB NBF up to 20KHz. These values probably require multi-section or bifilar windings and even then may not be physically realizable. Or may be easy, if you balance transformer factors well.
Pick a power/price ratio that will sell well. (In this case: a power that should satisfy you, and at headphone power levels cost is not real linear with power; price tends to be a different issue in hand-built experimental gear but you would not normally, for instance, pick a quad of 300B for headphones.)
Pick a tube that seems likely to give more than enuff power. If you stick to the usual suspects, the databook is your friend.
But power triodes, and especially very low power triodes, have not been in fashion since 1934. You have to take your best guess then simulate (on curves, in SPICE, or on a breadboard) to estimate the parameters. Your best first guess for load-line is 2Rp, but tubes not meant for Power use may have voltage, power, or current limits which do not allow easy optimization.
Then compare the estimated parameters to the costs of power supply, transformer, and driver.
Here is a rule of thumb: for a transformer coupled triode with R-C driver working from the same bulk supply as the output stage, you want a Mu (Amplification Factor) of about 6. Any lower, and drive voltage becomes so high that you need an extra power supply for the driver. Any higher, and Rp becomes too high for good power at a given supply voltage, also the transformer costs will tend to be high.
In this case, there are no small twin triodes with Mu about 6. However there are some with apparently exceptionally good Gm at modest current, designed for TV tuner preamps. 6BQ7 is classical, 6BK7 is a high-spec 6BQ7, and 6DJ7 is an improved amplifier but with lower plate ratings. They have high Mus and are not great power amps, but at part-Watt power levels we can live with that. And the easy drive is a nice little benefit.
It should be noted that none of these types were expected to run AT their ratings for many hours: in a TV set, if you are getting a good picture the tuner is under AGC and preamp current is less than maximum. I once used 6BK7 as a headphone amp within ratings but very hot and did observe "softening" of power after say 100 hours.
There are also the old-style medium-Mu types like 12AU7, mentioned in another forum. While the 12AU7 has lower Mu than 6DJ8, it has much lower Gm at typical current. The plate resistance is high. To make power, you need a high plate voltage. To deliver that power in "hi-fi", you need an awful lot of turns on the transformer. 12AU7 was often used as pro-gear line amps (a job not unlike headphones), but more for ready availability and known reliability than for best audio performance.
For a hobbyist headphone amp where tube wear-out is no big deal, 6DJ8 and similar look better than 12AU7 or 6BK7. The lower Rp significantly reduces transformer issues, and it is easier to drive. When I picked 6BK7, the BK was $1.38 and the 6DJ7 had not become available (yes, it was a while ago).
6DJ8 at high current runs Rp near 3K. In single-ended we might load it in 6K minimum for best power and distortion over a range of load impedances. If we assume minimum load is 64Ω, the transformer ratio is 10:1. If we run push-pull pure-A, we use two such ratios with a 32Ω load and each tube will still see about 6K load.
Here is what SPICE says:
push-pull 6DJ8
B+ = 120V
Rk = 200 ohms
transformer turns ratio,
one side of primary to whole secondary: 10:1
According to SPICE (Koren model):
16 ohms, 0.17W, 0.9%, (AB)
16 ohms, 0.12W, 0.6%, (A)
32 ohms, 0.38W, 0.9%, (AB)
32 ohms, 0.22W, 0.5%, (A)
100 ohms, 0.19W, 0.3%, (A)
300 ohms, 0.09W, 0.3%, (A)
600 ohms, 0.05W, 0.3%, (A)
Unloaded Gain:
(one grid) : (out) = 1:1.38
32 ohm gain:
(one grid) : (out) = 0.858
Zout = 19 ohms
The distortion figures are rough. "0.3%" is numeric noise; it will drop in higher impedances.
So it has ample power in 32Ω. (More than ample; recall that the user has a Volume control and need not run it at 0.22-0.38 watts.) Power is less than 100mW in 600Ω but the 600Ω phones tend to have better power sensitivity. I assure you that this range of power will be enough for most users. Dozens of mW satisfy many listeners in most phones.
As you know, the transformer is a critical piece. But the design of the tranny has to start with an estimate of source impedance and voltage swing, so we need to look at likely tubes first. We may iterate: 6BL7's low-Z plate makes a nice transformer but a large chassis for a tall octal and big heater. 12AU7 is about the cheapest little power-bottle around but its hi-Z plate makes transformer design messy. But stick with the 6DJ8 and see where it goes....
We start with a core that can probably handle the flux and give us the inductance we need. If you aim for 20Hz, you need a larger core than you would use on a 50/60Hz power transformer of similar VA. Since iron price does not scale much with VA in the range 0.1W to 5W, and we need a LOT of turns and hate hyper-fine wire, pick a large core. 5VA is cheap and not too large for good bass response, though it may wind-up with more capacitance than we like for high-end response, especially under NFB.
There are special irons for audio, and even now funny-stuff I would call "ferrite" (probably not the right word). But for small audio power, conventional power iron will give a workable design. Work it up with ordinary power-transformer iron to get in the ballpark. Then when 99% happy, check what is possible with other irons.
Since it will still be fine wire on the primary, insulation wastes spece. Set aside about 2/3rds of the window area for primary, deduct some fishpaper, and fill it with fine wire. Compute the inductance and the flux. Flux level in audio transformers is a complex topic. For starters: design for much less than half the flux you would allow in a power transformer. The inductance for a triode amplifier should give a bass impedance similar to triode plate impedance. (Plate R in parallel with load Z for fixed-Z loading, but with 32Ω to 300Ω we can't count on the load.)
When you have enough core and winding to give flux and inductance, find the leakage inductance and stray capacitance.
There are estimating tools in Radiotron 4th. Not an easy task. What really happens is you wind a sample with a small number of sections, and measure it in-circuit, then if it won't meet spec you change the number of sections (or change the specs).
Transformer:
turns ratio, one side of primary to whole secondary: 10:1
I still say: mock one up with a 12VA 240VCT/12+12V 50/60Hz power transformer. Because these things barely-work at 168V peak at 50Hz, it may get a little soft at 20Hz even at more like 70V peak. But we rarely need full level at 20Hz. 50/60Hz power transformers are not designed for 20KHz response, but it has to be much better than 60Hz to pass rectifier pulses. What seems to get you first is eddy-current losses: the response droops but not at 6dB/8ve.
Also: try the secondaries in parellel and one alone, other open. See how much difference there is. Sure, the parallel connection will have lower copper loss, but the difference is less than a dB. The parallel connection has lower leakage inductance, but also more stray capacitance, so the frequency of the top-resonance does not change much. Because of core losses, the ringing is less than you might think from a simple lumped-circuit model. I'm not seriously advocating using a power transfromer, but it will be a solid performer 60Hz-5KHz and let you test load impedance and power relationships before you move on to...
Custom Tranny:
For 20-20KHz with 10dB NFB:
Primary inductance, one side: 15H
Primary resistance, one side: 300-1,000 ohms
Leakage inductance referred to primary: under 15mH (K= about 0.999)
stray capacitance referred to secondary: under 0.04uFd
Seat-Of-Pants Interpretation: No good answer exists for heavy NFB with widely varying load with normal audio transformer construction. Response at different loads will peak or dip, phase run wild. Values above attempt to keep phase margin OK out to 70KHz, which with a little fixed compensation in voltage-amp will allow 10dB NBF up to 20KHz. These values probably require multi-section or bifilar windings and even then may not be physically realizable. Or may be easy, if you balance transformer factors well.
>> why are you even planning push-pull in a 100mW amp?
> To eliminate even-order distortion as much as possible and to allow true balanced operation, if so desired.
OK, no strong argument there.
Especially when my last bottle can-amp was push-pull all the way through. Long-tail 12AX7 driving 6BK7.
At the time, there was not yet a Cult Of The Second Harmonic. In fact most consumer stuff had so much distortion that any good clean amp stood out. Push-pull triode working well below its overload is a nice little thing, even if the 2nd is not hiding the 3rd.
The input and feedback were single-ended: balanced in/out operation is in my opinion mostly for high-noise environments like studios and networks. Anyway the common heaphone plug forces common ground.
Another big argument for push-pull: higher efficiency, is not so much an issue at this power level. Although with the shortage of nice SMALL triodes with low Rp, it isn't a non-issue.
Push-pull sure does reduce the problem of DC flux, which really dominates SE design.
Myself, I'm now leaning toward SET. Maybe for simplicity, maybe just because I already did P-P. I think the "best" designers focused on a single topology and learned all its secrets, but I'll never be one of them.
> To eliminate even-order distortion as much as possible and to allow true balanced operation, if so desired.
OK, no strong argument there.
Especially when my last bottle can-amp was push-pull all the way through. Long-tail 12AX7 driving 6BK7.
At the time, there was not yet a Cult Of The Second Harmonic. In fact most consumer stuff had so much distortion that any good clean amp stood out. Push-pull triode working well below its overload is a nice little thing, even if the 2nd is not hiding the 3rd.
The input and feedback were single-ended: balanced in/out operation is in my opinion mostly for high-noise environments like studios and networks. Anyway the common heaphone plug forces common ground.
Another big argument for push-pull: higher efficiency, is not so much an issue at this power level. Although with the shortage of nice SMALL triodes with low Rp, it isn't a non-issue.
Push-pull sure does reduce the problem of DC flux, which really dominates SE design.
Myself, I'm now leaning toward SET. Maybe for simplicity, maybe just because I already did P-P. I think the "best" designers focused on a single topology and learned all its secrets, but I'll never be one of them.
This was a lot to absorb, hence the delay in my response. I've had to read your posts a couple of times over and ruminate on them a bit.
You suggested the 6BQ7, 6DQ7 and 6DJ7 as good twin triodes to use. The DJ7 caught my eye because, as I understand it, the terminal number in this classification system often indicates the number of elements. The 6DJ8, then, has 1 heater, 2 cathodes, 2 grids, 2 plates and 1 shield. The 'DJ7, one presumes, is lacking the shield (lacking a heater would also bring the total to 7 while preserving symmetry, as well, but then you'd have to fire up the blowtorch to get the thing to work, eh?)
At any rate, I had already bought (2) Sovtek 6SN7GT; (2) Sovtek 6DJ8 and (2) National 12AU7A tubes because they are used extensively in Morgan Jones' book Valve Amplifiers and so I felt comfortable with them. Sure, they may not be the best tubes for the job, but I selected them based on the curves in the back of the book and the extensive discussion spread through ibid.
High Mu is not necessarily a good thing, though, correct? It typically is accompanied by poor linearity, IIRC. High Gm is always a good thing, though, and it is Gm that determines the Rp, correct also?
This is what I came up with, too, so I feel better already.
Hmmm. The total impedance of the primary will be 24k, but each side of the centertap will present 6k to the anodes. But how does this convert the secondary from 64R to 32R?
Which is another reason why multiple secondaries (or at least 2) would be better than one, eh?
Good lord: estimate Ll? Not me. I just short one side and measure the inductance still present on the other.
No surprise, there: it should droop at anything but a nice round number per octave. Transformer losses can only be approximated with curve-fitting to rational exponentials like x^1.34 and such. A real mess.
You've really done a fantastically thorough job of expressing the requirements here. So much so that you've practically taken the pai.. I mean, "fun" out of doing it myself.
I will try some 60Hz iron for the test run - excellent suggestion and one that I hadn't even thought of as I pored over the available transformers at the usual (and unusual) suspects.
Thanks again,
Jeff
You suggested the 6BQ7, 6DQ7 and 6DJ7 as good twin triodes to use. The DJ7 caught my eye because, as I understand it, the terminal number in this classification system often indicates the number of elements. The 6DJ8, then, has 1 heater, 2 cathodes, 2 grids, 2 plates and 1 shield. The 'DJ7, one presumes, is lacking the shield (lacking a heater would also bring the total to 7 while preserving symmetry, as well, but then you'd have to fire up the blowtorch to get the thing to work, eh?)
At any rate, I had already bought (2) Sovtek 6SN7GT; (2) Sovtek 6DJ8 and (2) National 12AU7A tubes because they are used extensively in Morgan Jones' book Valve Amplifiers and so I felt comfortable with them. Sure, they may not be the best tubes for the job, but I selected them based on the curves in the back of the book and the extensive discussion spread through ibid.
High Mu is not necessarily a good thing, though, correct? It typically is accompanied by poor linearity, IIRC. High Gm is always a good thing, though, and it is Gm that determines the Rp, correct also?
This is what I came up with, too, so I feel better already.
Hmmm. The total impedance of the primary will be 24k, but each side of the centertap will present 6k to the anodes. But how does this convert the secondary from 64R to 32R?
Which is another reason why multiple secondaries (or at least 2) would be better than one, eh?
Good lord: estimate Ll? Not me. I just short one side and measure the inductance still present on the other.
No surprise, there: it should droop at anything but a nice round number per octave. Transformer losses can only be approximated with curve-fitting to rational exponentials like x^1.34 and such. A real mess.
You've really done a fantastically thorough job of expressing the requirements here. So much so that you've practically taken the pai.. I mean, "fun" out of doing it myself.
I will try some 60Hz iron for the test run - excellent suggestion and one that I hadn't even thought of as I pored over the available transformers at the usual (and unusual) suspects.
Thanks again,
Jeff
> hence the delay
Took a while to write, no need to rush to comment.
Anyway that was the last under-90°F day in my office this week. I may have to bail out. (It isn't a lot cooler at home, and the dial-up is too slow for comfy forum participation.)
> The DJ7 caught my eye
Apologies, TYPO. My fingers tangled.
> often indicates the number of elements
Yes, but inconsistently. We have all those 6??7 and 12??7, and then they throw a curve-ball with a dual-triode ending in "8". There is no REAL difference, except as Frank says, there aint no 6DJ7.
> you'd have to fire up the blowtorch to get the thing to work, eh?
Seems to me it could work. Drill up through the cathode, drilling-out the filament and ceramic, thread metal tubing through and seal to the glass. Re-suck the vacuum. Get a teeny-tiny blowtorch and run it up through the tubing. Natural gas is a heck of a lot cheaper than electricity.
Might even make sense on monster transmitter tubes, kilowatt heaters, cathodes you could stick your arm through. (Except most of those are true filaments, not heated cathodes.)
> you've practically taken the pai.. I mean, "fun" out of doing it myself.
No. There is plenty of painful fun ahead. But you were off on a wrong path, and would be snared in irrelevant problems before you found your way again. This way you'll get a pretty-good amp, and working with that will reveal many more fun things to work on. Which won't go as expected, hence ample pain.
Anyway, all that calculation is very quick/dirty. Just gets you in the ballpark. I wager that you never saw a dozen-Henry winding in a modern SMPS. And when you apply SMPS techniques and materials, you will find that it is "impossible". Of course it can be done, but if you get to 10% copper loss you done good, while a SMPS winding often aims more like 2%(?) copper loss. So it is a very different world. (WHich is why my one SMPS just sat there and got hot.)
> I had already bought (2) Sovtek 6SN7GT; (2) Sovtek 6DJ8 and (2) National 12AU7A
Fine. The 6DJ8 is an amusing (if hardly ideal) teeny-power tube, and also a fine voltage amp. 6SN7/12AU7 (there is not a huge difference) is rather low gain for a feedback amp but sure it will work as that, or give a nice gain without overall feedback.
For general play, I would have 12AX7 and 12AT7 too. Yes, the 12AX7 in particular is not real linear, but in many cases the signal level is so small it does not matter, and it does pack so much gain that the nonlinearity may be canceled in NFB. 12AT7 is a nice in-between tube. Between 12AT7 and 12AU7 are the early TV-Tuner tubes like 6BQ7. (The 6DJ8 is a late tuner tube and considerably hotter than the classic audio tubes like 12SN7/12AU7 and 12SL7. High Gm, also high heater power.)
To my eye there are two "obvious" topologies: voltage amp driving split-load follower (the most common push-pull amp driver system) and a long-tail differential pair. You don't need more than one gain stage when the power tube grids need only a couple or three volts signal.
Turns out that if you accept single-ended input and can provide a low impedance feedback network, the amp/follower is "better" because it gives twice the forward gain (thus half the error under NFB). In fact the split-load follower is awful hard to beat for linearity at high levels.
The long-tail diff amp looks elegant but has half the gain (remember single-ended input and low impedance feedback network), and of course needs a negative supply for the long-tail return. (Today you could use an FET; they were novel when I built mine and anyway I had the parts to add a -250V 2mA supply.)
There's no reason not to transformer couple the grids, 1925 style. Except the core losses of a good plate to grid tranny is more power than headphones generally need, and a good hi-Z grid transformer isn't cheap. Nor amenable to overall feedback.
> I will try some 60Hz iron
It will make an excellent table-radio amp. Modern power transformers are perhaps better audio iron than those we had in the 1930s, and your 120V supply will not saturate a common 120/240 winding. It should be very listenable, though it may not have amazing specs.
Expect droop below 50Hz and up in the KHz. One nice thing: these transformers are all wound 2-section, so you can quickly estimate what improvement in leakage inductance you could gain with a custom 3-section or 5-section winding. Though, at 120V, with modern varnish, you could probably wind "bi-filar" (actually 21-filar) and get leakage inductance to almost vanish. Capacitance could be high, but if you only face capacitance you have a near-6dB/8ve drop and little trouble with NFB stability.
Took a while to write, no need to rush to comment.
Anyway that was the last under-90°F day in my office this week. I may have to bail out. (It isn't a lot cooler at home, and the dial-up is too slow for comfy forum participation.)
> The DJ7 caught my eye
Apologies, TYPO. My fingers tangled.
> often indicates the number of elements
Yes, but inconsistently. We have all those 6??7 and 12??7, and then they throw a curve-ball with a dual-triode ending in "8". There is no REAL difference, except as Frank says, there aint no 6DJ7.
> you'd have to fire up the blowtorch to get the thing to work, eh?
Seems to me it could work. Drill up through the cathode, drilling-out the filament and ceramic, thread metal tubing through and seal to the glass. Re-suck the vacuum. Get a teeny-tiny blowtorch and run it up through the tubing. Natural gas is a heck of a lot cheaper than electricity.
Might even make sense on monster transmitter tubes, kilowatt heaters, cathodes you could stick your arm through. (Except most of those are true filaments, not heated cathodes.)
> you've practically taken the pai.. I mean, "fun" out of doing it myself.
No. There is plenty of painful fun ahead. But you were off on a wrong path, and would be snared in irrelevant problems before you found your way again. This way you'll get a pretty-good amp, and working with that will reveal many more fun things to work on. Which won't go as expected, hence ample pain.
Anyway, all that calculation is very quick/dirty. Just gets you in the ballpark. I wager that you never saw a dozen-Henry winding in a modern SMPS. And when you apply SMPS techniques and materials, you will find that it is "impossible". Of course it can be done, but if you get to 10% copper loss you done good, while a SMPS winding often aims more like 2%(?) copper loss. So it is a very different world. (WHich is why my one SMPS just sat there and got hot.)
> I had already bought (2) Sovtek 6SN7GT; (2) Sovtek 6DJ8 and (2) National 12AU7A
Fine. The 6DJ8 is an amusing (if hardly ideal) teeny-power tube, and also a fine voltage amp. 6SN7/12AU7 (there is not a huge difference) is rather low gain for a feedback amp but sure it will work as that, or give a nice gain without overall feedback.
For general play, I would have 12AX7 and 12AT7 too. Yes, the 12AX7 in particular is not real linear, but in many cases the signal level is so small it does not matter, and it does pack so much gain that the nonlinearity may be canceled in NFB. 12AT7 is a nice in-between tube. Between 12AT7 and 12AU7 are the early TV-Tuner tubes like 6BQ7. (The 6DJ8 is a late tuner tube and considerably hotter than the classic audio tubes like 12SN7/12AU7 and 12SL7. High Gm, also high heater power.)
To my eye there are two "obvious" topologies: voltage amp driving split-load follower (the most common push-pull amp driver system) and a long-tail differential pair. You don't need more than one gain stage when the power tube grids need only a couple or three volts signal.
Turns out that if you accept single-ended input and can provide a low impedance feedback network, the amp/follower is "better" because it gives twice the forward gain (thus half the error under NFB). In fact the split-load follower is awful hard to beat for linearity at high levels.
The long-tail diff amp looks elegant but has half the gain (remember single-ended input and low impedance feedback network), and of course needs a negative supply for the long-tail return. (Today you could use an FET; they were novel when I built mine and anyway I had the parts to add a -250V 2mA supply.)
There's no reason not to transformer couple the grids, 1925 style. Except the core losses of a good plate to grid tranny is more power than headphones generally need, and a good hi-Z grid transformer isn't cheap. Nor amenable to overall feedback.
> I will try some 60Hz iron
It will make an excellent table-radio amp. Modern power transformers are perhaps better audio iron than those we had in the 1930s, and your 120V supply will not saturate a common 120/240 winding. It should be very listenable, though it may not have amazing specs.
Expect droop below 50Hz and up in the KHz. One nice thing: these transformers are all wound 2-section, so you can quickly estimate what improvement in leakage inductance you could gain with a custom 3-section or 5-section winding. Though, at 120V, with modern varnish, you could probably wind "bi-filar" (actually 21-filar) and get leakage inductance to almost vanish. Capacitance could be high, but if you only face capacitance you have a near-6dB/8ve drop and little trouble with NFB stability.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.