What do you propose as been the sonic advantage to this. It definately requires a pentode as the driver to get the best performance from Partial Feedback. Most people don't use one because they either have a spare high impedance triode - or don't want the extra complexity. Alex Kitics designs are all compromised by his reluctance to use a pentode driver.
can you explain further?
Shoog
Great looking circuit! I have been working on something similar, using 6W6's for the input and an 6SL7 driver... the next step is using a pentode (6SJ7??) instead of the 6SL7.
Just a note on the LM317. Following the recommendations by Broskie I eventually used the LM337 in a similar application. Vin (and tab) on this one should be grounded: I just bolted 4 LM337's on a heatsink and grounded the heatsink: works perfectly and doesn't need insulators.
http://www.glass-ware.com/tubecircuits/Tube_Auto_Biasing_3.html
Just a note on the LM317. Following the recommendations by Broskie I eventually used the LM337 in a similar application. Vin (and tab) on this one should be grounded: I just bolted 4 LM337's on a heatsink and grounded the heatsink: works perfectly and doesn't need insulators.
http://www.glass-ware.com/tubecircuits/Tube_Auto_Biasing_3.html
Hey Shoog,
Seems you missed the difference between the two
! In the triode case a 15k feedback resistor is present.The solution I propose gives you a defined Zin of +15k. Without the 15k resistor, Zin will be very low as we can look at it as a virtual earth point.
So we does not require a pentode driver. In "my" case you only need a LowRi-triode capable of driving the load defined by the input feedback resistor.
I totally agree that Kitics solution is not the best and results in gross distortion. Triodes can not drive into very low impedances while pentodes should.
Where does the term "partial feedback" come from? Shouldn´t "inverse feedback" as Schade calls it be more appropriate?
Except for the Schade-paper, this could also be interesting to read about inverse feedback:
http://www.aikenamps.com/FeedbackAmp.htm
Last edited:
Thanks for explaining that. I now see the difference - though I am not certain I would hear the difference
I just added a bypass cap to the cathode caps, might have smoothed things out a tad.
Good pointer to the LM337, though I understood that they performed slightly worse at high frequencies. I would guess that the difference would not be noticeable.
Shoog
I just added a bypass cap to the cathode caps, might have smoothed things out a tad.
Good pointer to the LM337, though I understood that they performed slightly worse at high frequencies. I would guess that the difference would not be noticeable.
Shoog
I´d like to add about Kitics solution:
A triode with an unbypassed cathode resistor gets characteristics looking a little pentodish! Not enough, though.
At the same time a pentode with with inverse feedback gets triodelike characteristics.
The BBC-paper about feedback at Milletts site is great!
http://www.pmillett.com/Books/bbc_feedback.pdf
A triode with an unbypassed cathode resistor gets characteristics looking a little pentodish! Not enough, though
.At the same time a pentode with with inverse feedback gets triodelike characteristics.
The BBC-paper about feedback at Milletts site is great!
http://www.pmillett.com/Books/bbc_feedback.pdf
Last edited:
Thanks for the BBC paper, looks interesting and will read it later.
Did a side by side comparison with my Tabor clone with a friend today. Result was that it didn't match the Tabor (I would be surprised if it did since the Tabor is more complex by a large margin). The self splitter was more forward and generally zingier which tended to be a bit fatiguing over time. So to my mind I believe there is something a bit wrong. First thing to check is OT polarity. If its not that then some form of high order harmonic distortion is creeping in which maybe intrinsic to the circuit. Could also be the coupling cap is to small for the low impedance node of the PL84 grid. Might even be power supply ringing. Any suggestions beyond the output transformer would be appreciated.
Shoog
Did a side by side comparison with my Tabor clone with a friend today. Result was that it didn't match the Tabor (I would be surprised if it did since the Tabor is more complex by a large margin). The self splitter was more forward and generally zingier which tended to be a bit fatiguing over time. So to my mind I believe there is something a bit wrong. First thing to check is OT polarity. If its not that then some form of high order harmonic distortion is creeping in which maybe intrinsic to the circuit. Could also be the coupling cap is to small for the low impedance node of the PL84 grid. Might even be power supply ringing. Any suggestions beyond the output transformer would be appreciated.
Shoog
Complexity should not be necessary for good sound
In my experiments with self-splitting stages, I found no intrinsic high order
distortion nor any mechanism for it. It produces even order distortion
intrinsically because there is no inherent cancellation.
Can you measure it? basic frequency response and look for ringing or instability.
Cheers,
Michael
The 120 Ohm cathode resistor for the 6AU6 is less than 1/gm of the 6AU6, so it is still exhibiting some pentode related 3/2 power distortion. Not really convenient to increase it though, without messing with the 6AU6 biasing. (just raising the cathode resistor will reduce the 6AU6 current and further increase 1/gm, so not helpful) Some version of "trioding" of the input stage as Revintage suggested could fix this.
Some interesting variants would be to try connecting the Schade feedback to just the 6AU6 screen grid instead. This makes for an odd situation though where the 6AU6 internal "triode" (ie, screen grid) is running the sound show and the left PL84 is running in anti-triode mode to the 6AU6 effectively. (and the right PL84 would be anti-anti, or effectively directly emulating the 6AU6 "triode".)
Leaping even further, an AC Schaded feedback network from C1,C2,R5 junction back to the 6AU6 screen might remove just odd harmonics generated in the output stage. But this would not triode-ify the 6AU6 pentode. So either a higher value cathode resistor for the 6AU6 would be needed as well, or another Schade network enclosing just the 6AU6.
If the OT xfmr secondary has 0,4,16 Ohm taps, it could be inserted in at the junction of C1, C2, R5 to give CFB for the PL84s and still preserve the anit-triode splitting action. Alternatively, if only a single 8 Ohm secondary winding is available, it could be connected in series with R6 for effective CFB.
Don
Some interesting variants would be to try connecting the Schade feedback to just the 6AU6 screen grid instead. This makes for an odd situation though where the 6AU6 internal "triode" (ie, screen grid) is running the sound show and the left PL84 is running in anti-triode mode to the 6AU6 effectively. (and the right PL84 would be anti-anti, or effectively directly emulating the 6AU6 "triode".)
Leaping even further, an AC Schaded feedback network from C1,C2,R5 junction back to the 6AU6 screen might remove just odd harmonics generated in the output stage. But this would not triode-ify the 6AU6 pentode. So either a higher value cathode resistor for the 6AU6 would be needed as well, or another Schade network enclosing just the 6AU6.
If the OT xfmr secondary has 0,4,16 Ohm taps, it could be inserted in at the junction of C1, C2, R5 to give CFB for the PL84s and still preserve the anit-triode splitting action. Alternatively, if only a single 8 Ohm secondary winding is available, it could be connected in series with R6 for effective CFB.
Don
You now got a strapped triode's plate resistance in series with that 15K...
I prolly wouldn't have chosen the driver cathode to be bypassed either.
It really doesn't matter the drive is high impedance, low impedance,
or something inbetween. As long as its flat, and with frequency too.
Whatever impedance, computes part of the Schade feedback divider...
May have faked more like a real triode, not to use any triodes at all...
-------------------------------
You plannin to garter those lm317's? Would that be going overboard?
Last edited:
The pentodes at the left and middle move PSR common to the top rail.
Shouldn't grid of the pentode on the right similarly bypassed upward?
I don't think the current sources will object.
Similar in SE effect to ultrapath, but with pentode as cap multiplier....
C5 bypass to ground may better bypass to B+, or omit completely.
Shouldn't grid of the pentode on the right similarly bypassed upward?
I don't think the current sources will object.
Similar in SE effect to ultrapath, but with pentode as cap multiplier....
C5 bypass to ground may better bypass to B+, or omit completely.
Last edited:
OOPs!!
On my previous suggestion of connecting the Schade Neg. feedback back to the 6AU6 screen grid (post 29) the phasing is wrong, if taken from the left PL84. And while taking the FDBK from the right side PL84 plate would give the right phasing for Neg. Fdbk, it probably would oscillate after so many stages. Scratch that idea.
Could connect the Schade N. Fdbk to the 6AU6 cathode resistor though to linearize the 6AU6 and output stage.
On my previous suggestion of connecting the Schade Neg. feedback back to the 6AU6 screen grid (post 29) the phasing is wrong, if taken from the left PL84. And while taking the FDBK from the right side PL84 plate would give the right phasing for Neg. Fdbk, it probably would oscillate after so many stages. Scratch that idea.
Could connect the Schade N. Fdbk to the 6AU6 cathode resistor though to linearize the 6AU6 and output stage.
Feedback to 6AU6 cathode only complicates the PSRR situation.
Unless you are going for the full Loftin White treatment...
(Don't bother, I've already spent weeks on it. You can cancel
the hum this way in this application, but never the IMD...)
I think it better to encourage the highest possible plate Z here,
and keep the Schade feedback entirely common to any noise on
the high ref side... Hybrid cascode might even be the ticket.
Unless you are going for the full Loftin White treatment...
(Don't bother, I've already spent weeks on it. You can cancel
the hum this way in this application, but never the IMD...)
I think it better to encourage the highest possible plate Z here,
and keep the Schade feedback entirely common to any noise on
the high ref side... Hybrid cascode might even be the ticket.
Last edited:
Keep in mind that any unresolved PSRR problems are gonna be twice worsified by
the anti-triode in this topology. So take extra heed to keep all hums common.
By this I mean: Don't mix signals to the opposing rail without a high impedance
inbetween. Both CCS anodes, and 6AU6 plate already serve as such impedance...
Just don't sabotage that goodness with C5 or similar cross-rail hum referencing
mistake. The grid of rightmost pentode wants a bypass to B+ common hum.
Where plate, grid, screen and cathode all float the same hum, there is no hum.
the anti-triode in this topology. So take extra heed to keep all hums common.
By this I mean: Don't mix signals to the opposing rail without a high impedance
inbetween. Both CCS anodes, and 6AU6 plate already serve as such impedance...
Just don't sabotage that goodness with C5 or similar cross-rail hum referencing
mistake. The grid of rightmost pentode wants a bypass to B+ common hum.
Where plate, grid, screen and cathode all float the same hum, there is no hum.
Last edited:
Hey KP,
I think you should bypass a triodes cathode-resistor to keep its Zout low. Can not see why rising Zout of it would help. A good low Ri triode will have no problem driving the 15k load.
In the case with a pentode and no resistor, Zin will be in the ballpark of one or a few kohms. This will not be a good scenario for a triode even with an unbypassed cathoderesistor:
Rires=Ri+(mu*Rk+1)
Pentode or triode (with 15k or whatever) is off course a matter of taste.
Without series-resistor the output pentode can be seen as an I/V-amp and in the more Scadeish approach with resistor it can be seen as an invertingamp with its Zin defined by the same resistor. Schade recommended an IT though.
Ken,
I'm seeing your points on the PSRR with the pentode input.
In the case of the CCS'd triode input that Lars has shown, I think the right side PL84 grid will still want to be ground referenced, since the triode driver is ground referenced. Or possibly an AC voltage divider reference between B+ and ground in the same ratio as the Schade feedback divider roughly (100K/15K). This would then allow the speaker output winding to be series connected with R6 to ground for CFB if wanted.
I'm seeing your points on the PSRR with the pentode input.
In the case of the CCS'd triode input that Lars has shown, I think the right side PL84 grid will still want to be ground referenced, since the triode driver is ground referenced. Or possibly an AC voltage divider reference between B+ and ground in the same ratio as the Schade feedback divider roughly (100K/15K). This would then allow the speaker output winding to be series connected with R6 to ground for CFB if wanted.
Last edited:
Seems like a gyrator or a DC bias referenced anti-triode/gyrator load should work fine for the 6AU6 load. They make the pentode look more High Z out-ish for a V to I converter. But neither addresses any 3/2 power distortion generated by the 6AU6's V to I transfer function (it's current will still be varying in either case due to the feedback loading). I think Shoog needs to determine where the distortion he suggested is originating from.
Have you really done the maths?
Did some calculations on the load the 6AU6 sees. It is in the ballbark of 0,7k//14k//680k! Guess it is a little hefty for the 6AU6 even if I haven´t checked the loadline. Changing Rfb from 41k to 100k makes it a little better at ca 2k//Ra//Rg.
Still think a better idea is to use a triode like a strapped 6E5P and add a input feedback resistor to make it an inverting feedback amp.
Did some calculations on the load the 6AU6 sees. It is in the ballbark of 0,7k//14k//680k! Guess it is a little hefty for the 6AU6 even if I haven´t checked the loadline. Changing Rfb from 41k to 100k makes it a little better at ca 2k//Ra//Rg.
Still think a better idea is to use a triode like a strapped 6E5P and add a input feedback resistor to make it an inverting feedback amp.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.