Ah man, had a chance to look it over. It's pretty much the article from aX, but without the schematic. I do like the new color picture of his cased Octode though.
I can scan in the actual schematic to jpg and post it up here if anyone wants to see it as it is not available online at aX.com
Some time ago after reading the ax article, I messed around with it, but couldn't
get it to work very well. Was like the error amp triode wanted or needed to be of
substantially higher Mu than the feedback triode? If same vs same to compliment
out non-linearities, that seemed to be a configuration that loses more drive signal
swing than I got back. And the triode's shared cathode voltage presents another
headroom limitation for drive signal, maybe a longer tail to a negative rail would
have helped?
Maybe I'll mess with it again this weekend, see if I hit the same wall?
get it to work very well. Was like the error amp triode wanted or needed to be of
substantially higher Mu than the feedback triode? If same vs same to compliment
out non-linearities, that seemed to be a configuration that loses more drive signal
swing than I got back. And the triode's shared cathode voltage presents another
headroom limitation for drive signal, maybe a longer tail to a negative rail would
have helped?
Maybe I'll mess with it again this weekend, see if I hit the same wall?
Last edited:
I read the article and the looked at the wrong schematic that was published and forgot about it. The correction was never posted on their web site but it did show up in a later iussue. The circuit shows some original thinking, and I decided like you to add it to my projects list. My list is probably longer than yours.
Partition noise is white, not 1/f. 1/f noise comes from the cathode (probably) via poorly understood mechanisms. A pentode made to behave like a triode due to feedback would still have partition noise because the partition of the electron stream is still taking place. The feedback could reduce the noise, just as it reduces other distortions.
The only way to get rid of partition noise is to merge the two current streams back together, as in normal triode connection. Or I suppose you could take a signal from g2 and feed it back antiphase, but this would be difficult because it would be negative feedback for partition noise but positive feedback for shot noise and the signal.
The only way to get rid of partition noise is to merge the two current streams back together, as in normal triode connection. Or I suppose you could take a signal from g2 and feed it back antiphase, but this would be difficult because it would be negative feedback for partition noise but positive feedback for shot noise and the signal.
Anyway, higher amplification factor and higher load resistance obviously lead to igher level of noises (and signals) no matter what originated them initially. I used pentodes in tape head amps, mic preamps (I don't mention first stage of the power amp - it's obvious) and every time signal / noise ratio satisfied by requirements completely. I even used EF86 in the 1'st stage of a guitar amp, however with lowered filament voltage, and mounted on a coustic damping pad, but it worked as expected.
6AU8?
Triode section for input buffer and plate feedback
With Pentode section for error amplifier and driver
I can think of no reason an error amplifier should be
held back by a Mu of its own. Already in a Mu loop.
Make big Power Pentode slave to a triode is OK.
But AB crossing is gonna be ugly if we cancel out
all the curve at the bottom of that triode. Thus I
see no advantage same vs same triode balance.
Might be even better the error amp were sand...
Triode section for input buffer and plate feedback
With Pentode section for error amplifier and driver
I can think of no reason an error amplifier should be
held back by a Mu of its own. Already in a Mu loop.
Make big Power Pentode slave to a triode is OK.
But AB crossing is gonna be ugly if we cancel out
all the curve at the bottom of that triode. Thus I
see no advantage same vs same triode balance.
Might be even better the error amp were sand...
Last edited:
Tempting direct couple to power pentode's screen instead.
Except triode would have to swing screen currents without
messing up its own linearity... Input triode has to swing the
same AC cathode currents as the driver. Folded cascode...
Except triode would have to swing screen currents without
messing up its own linearity... Input triode has to swing the
same AC cathode currents as the driver. Folded cascode...
Last edited:
Screen current folds back in correct phase to parallel w triode and pentode.
Triode operates at constant current, more of less. (set by R10 at 33V)...
Approximately 50ma per collector idle, if balanced. Screen idles B+ -110V.
I'm not sure this is a realistic operating point yet? Screen should probably
idle lower, a lot closer to cutoff, since thinking this is 1/2 of an AB amp...
Then again, maybe I want screen idle at 1/2 B+ and lower G1 to set AB
crossing bias?
If Pentode G1 wants to be pulled lower for proper bias?
Like input triode, we got 33V easily available to do so.
Not complete sim yet, not by a long shot...
Discuss or disgust?
Triode operates at constant current, more of less. (set by R10 at 33V)...
Approximately 50ma per collector idle, if balanced. Screen idles B+ -110V.
I'm not sure this is a realistic operating point yet? Screen should probably
idle lower, a lot closer to cutoff, since thinking this is 1/2 of an AB amp...
Then again, maybe I want screen idle at 1/2 B+ and lower G1 to set AB
crossing bias?
If Pentode G1 wants to be pulled lower for proper bias?
Like input triode, we got 33V easily available to do so.
Not complete sim yet, not by a long shot...
Discuss or disgust?
Attachments
Last edited:
For 10W maybe a pair of 6BM8 (tP) Mu70+7W? Bootstrap screen to opposite plate?
I dunno what to use for sand right yet... This whole idea might be whack...
I dunno what to use for sand right yet... This whole idea might be whack...
Octoid Lieber
Re: Ken,
For making a P-P stage, I would get rid of both Octode triodes at the input (use pentodes). The same diffl. driver can be used for both output tubes then. The Schade feedbacks come back to the driver screens. And the output tube grids get driven by the diffl. driver plates. This scheme has been around for a while. The pentode driver plates can then either track the screens or operate at a fixed multiplier of the screen voltage swing (depending on "Mu" and feedback atten.). In either case, the screens should look much more constant Z (intercept a constant fraction of cathode current). The plates can be pulled up by CCS's for high loop gain. Closed loop gain is set by the g2/g1 Mu. The best triode is a pentode.
Re: Ken,
For making a P-P stage, I would get rid of both Octode triodes at the input (use pentodes). The same diffl. driver can be used for both output tubes then. The Schade feedbacks come back to the driver screens. And the output tube grids get driven by the diffl. driver plates. This scheme has been around for a while. The pentode driver plates can then either track the screens or operate at a fixed multiplier of the screen voltage swing (depending on "Mu" and feedback atten.). In either case, the screens should look much more constant Z (intercept a constant fraction of cathode current). The plates can be pulled up by CCS's for high loop gain. Closed loop gain is set by the g2/g1 Mu. The best triode is a pentode.
Last edited:
Fine with me, but you will need to know:
Its never any more linear than 1st buffer / plate feedback triode,
and in original circuit, that 1st triode is not operating anywhere
near constant current. Whatever AC currents flow in the driver
must flow in the other half of the UTurn. Plus whatever AC might
leak past the impedance of the short tail. You won't be operating
on the best loadline for Mu..
Linearity claims made for original circuit fall apart in simulation.
Maybe it sounds OK in real life, lots of messed up things do.
There ain't nothing wrong with the basic idea, it can be fixed...
You are going to burn a twin triode for each output Pentode,
and this count still hasn't even included for a phase splitter.
If you pull Octode error amp triode to the opposing Pentode's
plate. And use just one twin triode LTP to drive both Pentodes.
Then you basically got Pete Millet's red board with plate to
plate Schade feedback. This scheme works best (more like a
triode as result) if LTP is pentodes too rather than triodes, for
lengthy reasons already discussed elsewhere.
Edit: I just now see Smoking beat me to that observation...
Regardless if you chose the original circuit or not, first thing
needs doing is a negative supply, so you can lengthen tail
impedance between the cathode coupled pair.
Its never any more linear than 1st buffer / plate feedback triode,
and in original circuit, that 1st triode is not operating anywhere
near constant current. Whatever AC currents flow in the driver
must flow in the other half of the UTurn. Plus whatever AC might
leak past the impedance of the short tail. You won't be operating
on the best loadline for Mu..
Linearity claims made for original circuit fall apart in simulation.
Maybe it sounds OK in real life, lots of messed up things do.
There ain't nothing wrong with the basic idea, it can be fixed...
You are going to burn a twin triode for each output Pentode,
and this count still hasn't even included for a phase splitter.
If you pull Octode error amp triode to the opposing Pentode's
plate. And use just one twin triode LTP to drive both Pentodes.
Then you basically got Pete Millet's red board with plate to
plate Schade feedback. This scheme works best (more like a
triode as result) if LTP is pentodes too rather than triodes, for
lengthy reasons already discussed elsewhere.
Edit: I just now see Smoking beat me to that observation...
Regardless if you chose the original circuit or not, first thing
needs doing is a negative supply, so you can lengthen tail
impedance between the cathode coupled pair.
Last edited:
The Octode is playing one triode gm off versus the other triode gm in an LTP, giving a P-P characteristic as you mentioned earlier (the non constant triode currents). It might be possible to tune out the 3rd harmonic with a specific tail resistor to A-, if it still has enough gain to drive the output grid.
The "Octoid" I mentioned above differs from Pete's design by using crossed plate feedbacks to the driver screens, instead of to the driver plates. The driver plates can then be pulled up with a CCS on each side to maximize loop gain and linearize the internal driver "Mu"s. A resistor tail tuned to null the 3rd harmonic would be a good idea then. Also, the "Octoid" gets some current gain (3 to 20 X depending on tube chosen) from using the screen grid for feedback instead of the triode plate in the Octode. As you said, the same AC feedback current goes around the U turn in the Octode.
The "Octoid" I mentioned above differs from Pete's design by using crossed plate feedbacks to the driver screens, instead of to the driver plates. The driver plates can then be pulled up with a CCS on each side to maximize loop gain and linearize the internal driver "Mu"s. A resistor tail tuned to null the 3rd harmonic would be a good idea then. Also, the "Octoid" gets some current gain (3 to 20 X depending on tube chosen) from using the screen grid for feedback instead of the triode plate in the Octode. As you said, the same AC feedback current goes around the U turn in the Octode.
Last edited:
What would I ever want to do that to one half of an AB amp?
We need controlled non-linearity for smooth crossing. We are
correcting away square law of beam power tetrode to replace
with curve of what? Same vs same drive is more problem than
solution here. OK maybe for SE amp.
We need controlled non-linearity for smooth crossing. We are
correcting away square law of beam power tetrode to replace
with curve of what? Same vs same drive is more problem than
solution here. OK maybe for SE amp.
Last edited:
With the dual feedbacks for P-P, its controlling the final result, not the individual tube behavior. The pentode drivers have probably got a 100X local loop gain (using CCS loads) to make those output tubes bend any which way they want them to. You would still want to bias the outputs so that they had optimum crossover (isolated behavior) so the drivers would have less work to do. I don't see where getting the drivers/outputs to inherently give isolated smooth crossover will really work unless there is a larger global feeback loop to straighten the whole thing out. Class AB just has too big a variation of combined gm across the signal range to leave the corrections only to the small crossover region. Only fixing the small crossover region would leave a lot of 3rd harmonic. (although thats certainly going to sound better than leaving in the high harmonics) Hawksford error correction works quite well to eliminate all crossover, local, by using a huge local loop gain (it actually uses a positive feedback ghost gain stage) and then the usual outer global fdbk staightens out the whole thing.
Last edited:
Ok, so I openly admit I'm a complete noob WRT to tube designs and amp design in general.
But what I am gathering is that assuming the original design, built per the actual schematics that were published in the addendum in the 01/10 issue, sounds good or great, changing the tubes without a lot of consideration for resistor values and such could cause the revised amp to sound like crap. Correct?
I love 6922s, triode strapped 6V6s, and the 6L6s, so I may just bite the bullet and have SumR wind me the correct 8200:6R 40W OPTs and just build the amp as designed.
Then I can try to digest all of this great info from the discussion and mod the design to work better at a smaller 10W output that is better suited to my 60R impedance headphones since they can handle 2-3W of power... Heck at that point I might as well try implementing the feedback scheme from Decware's Taboo Lucid mode too. 🙂
But what I am gathering is that assuming the original design, built per the actual schematics that were published in the addendum in the 01/10 issue, sounds good or great, changing the tubes without a lot of consideration for resistor values and such could cause the revised amp to sound like crap. Correct?
I love 6922s, triode strapped 6V6s, and the 6L6s, so I may just bite the bullet and have SumR wind me the correct 8200:6R 40W OPTs and just build the amp as designed.
Then I can try to digest all of this great info from the discussion and mod the design to work better at a smaller 10W output that is better suited to my 60R impedance headphones since they can handle 2-3W of power... Heck at that point I might as well try implementing the feedback scheme from Decware's Taboo Lucid mode too. 🙂
Ach, but now with CCS loads you've thrown sand back in.
And you mean gyrator loads anyways, else they fight the
tail CCS.
And you mean gyrator loads anyways, else they fight the
tail CCS.
BMF>changing the tubes without a lot of consideration for
resistor values and such could cause the revised amp to
sound like crap. Correct?
Build the original proven circuit verbatim, else hire an engineer.
I have no faith the original circuit works exactly upon theory of
operation as described in AX, else would be duplicatable in sim.
I find operating points finniky and linearity not so good. Maybe
there is some sweet point where crud cancels, I havn't found it.
I only own the issue with the main article, not the latter with a
final schematic. My guesswork may be at fault in the simulation
not working as promised? But if it were so simple to get right, I
should have stumbled upon a configuration that works like the
article by now. Its not a circuit topology that magically makes
everything come together right if it isn't.
resistor values and such could cause the revised amp to
sound like crap. Correct?
Build the original proven circuit verbatim, else hire an engineer.
I have no faith the original circuit works exactly upon theory of
operation as described in AX, else would be duplicatable in sim.
I find operating points finniky and linearity not so good. Maybe
there is some sweet point where crud cancels, I havn't found it.
I only own the issue with the main article, not the latter with a
final schematic. My guesswork may be at fault in the simulation
not working as promised? But if it were so simple to get right, I
should have stumbled upon a configuration that works like the
article by now. Its not a circuit topology that magically makes
everything come together right if it isn't.
Last edited:
"Ach, but now with CCS loads you've thrown sand back in.
And you mean gyrator loads anyways, else they fight the
tail CCS. "
No tail CCS used when using CCS plates loads (Octoid case). No need to, since the feedbacks and drives are already symmetrical in P-P. So go for 3rd harmonic cancellation in the tail instead.
"changing the tubes without a lot of consideration for
resistor values and such could cause the revised amp to
sound like crap. Correct?"
Hard to tell how much tweeking took place for that particular design (triode selection, tail resistance). I'm surprised Ken couldn't get the Sim to work well. Some things to keep in mind when changing it: The feedback current has no gain anywhere, so the triodes had better be able to handle sufficient current to drive the output tube grid cleanly (operate at sufficient tail current). Higher Mu (particularly for the 2nd triode stage driving the output grid) should give better linearity, but only up until its output Z gets too high and can't drive the pentode grid capacitance or bias resistor. So an easy to drive pentode should help (and low rp triode). The tail resistance would, I think, want to be well above the 1/gm cathode resistance of the triodes. The 6922 seems well chosen for high gm and low rp (for driving purposes, but the Mu could be higher for linearity purposes) The negative tail supply Ken suggested could help a lot. There also might be some advantage in choosing the driver triodes for linear ramping gm with current, rather than constant Mu, since the LTP will work more linearly that way (using a CCS tail anyway). A 12AT7 ,or better, a 6LQ8 or even 6S4A triode comes to mind. That would be an experiment. Non dual triodes would allow matching them, maybe quite important. And some grid bias adjustment range on the second triode's grounded grid would be quite helpful.
And you mean gyrator loads anyways, else they fight the
tail CCS. "
No tail CCS used when using CCS plates loads (Octoid case). No need to, since the feedbacks and drives are already symmetrical in P-P. So go for 3rd harmonic cancellation in the tail instead.
"changing the tubes without a lot of consideration for
resistor values and such could cause the revised amp to
sound like crap. Correct?"
Hard to tell how much tweeking took place for that particular design (triode selection, tail resistance). I'm surprised Ken couldn't get the Sim to work well. Some things to keep in mind when changing it: The feedback current has no gain anywhere, so the triodes had better be able to handle sufficient current to drive the output tube grid cleanly (operate at sufficient tail current). Higher Mu (particularly for the 2nd triode stage driving the output grid) should give better linearity, but only up until its output Z gets too high and can't drive the pentode grid capacitance or bias resistor. So an easy to drive pentode should help (and low rp triode). The tail resistance would, I think, want to be well above the 1/gm cathode resistance of the triodes. The 6922 seems well chosen for high gm and low rp (for driving purposes, but the Mu could be higher for linearity purposes) The negative tail supply Ken suggested could help a lot. There also might be some advantage in choosing the driver triodes for linear ramping gm with current, rather than constant Mu, since the LTP will work more linearly that way (using a CCS tail anyway). A 12AT7 ,or better, a 6LQ8 or even 6S4A triode comes to mind. That would be an experiment. Non dual triodes would allow matching them, maybe quite important. And some grid bias adjustment range on the second triode's grounded grid would be quite helpful.
Last edited: