As a little speculation, I was thinking, what would happen if one built a 3 stage P-P amp this way.
Starting with Shoog's configuration as the driver/splitter (resistor loads instead of OT, or interstage xfmr), then a subsequent output stage with a similar Schaded triode and anti-triode into the OT.
Ie, use the first Schaded triode/anti-triode as a new type of splitter/driver stage. Curios as to the effect of such a splitter on a conventional P-P output stage too.
Don
Starting with Shoog's configuration as the driver/splitter (resistor loads instead of OT, or interstage xfmr), then a subsequent output stage with a similar Schaded triode and anti-triode into the OT.
Ie, use the first Schaded triode/anti-triode as a new type of splitter/driver stage. Curios as to the effect of such a splitter on a conventional P-P output stage too.
Don
We seem to be concluding that if the gm is equal, it's not an anti-triode...
I don't think it's strictly a matter of equal or not equal gm. I do see how the gm
nonlinearity of 2 identical tubes will cancel. And I suppose that a pentode
slaved to a triode can cancel a lot, if not all, of the triode's nonlinearity.
But I also suspect if a slave device is operated at relatively constant gm but
equal to the master device gm (for example by using a MOSFET with the
appropriate unbypassed Rs), some triode character will show at the output.
Ken; I have thought about but not yet tried the FET-pentode cascode. I
thought about using it for a selfsplit drive stage and as an output anti-triode
along with an 807 in series mode (think 1200V B+
) but the IXGH6N170 calls.
Don; how about a 2 stage amp with self-splitting pentode/antipentode driver
and push-pull Schadeode outputs? How about an input stage that gives the
amp a push-pull signature when driven with a balanced signal and a SE signature
when driven with a SE signal?
Time indeed for a spice model...
I don't think it's strictly a matter of equal or not equal gm. I do see how the gm
nonlinearity of 2 identical tubes will cancel. And I suppose that a pentode
slaved to a triode can cancel a lot, if not all, of the triode's nonlinearity.
But I also suspect if a slave device is operated at relatively constant gm but
equal to the master device gm (for example by using a MOSFET with the
appropriate unbypassed Rs), some triode character will show at the output.
Ken; I have thought about but not yet tried the FET-pentode cascode. I
thought about using it for a selfsplit drive stage and as an output anti-triode
along with an 807 in series mode (think 1200V B+
) but the IXGH6N170 calls.Don; how about a 2 stage amp with self-splitting pentode/antipentode driver
and push-pull Schadeode outputs? How about an input stage that gives the
amp a push-pull signature when driven with a balanced signal and a SE signature
when driven with a SE signal?
Time indeed for a spice model...
I wish I understood just half of what you guys are on about!
Re: Michael
Yes, I see what you are saying. I think you mean to say SE character here, not just triode. Equal non-linearities lead to a P-P sort of dist. cancellation. A near constant gm in the anti device would eliminate this cancellation. Using a high gm device also achieves the same effect by operation over a smaller portion of its non-linear transfer.
"how about a 2 stage amp with self-splitting pentode/antipentode driver
and push-pull Schadeode outputs? How about an input stage that gives the
amp a push-pull signature when driven with a balanced signal and a SE signature
when driven with a SE signal?"
Lets see, could we just duplicate Shoog's setup for both output tubes, providing two inputs. Seems that if both sides are identical we will just end up with the equal non-linearities case, leading to a P-P signature regardless of which input is driven. Unless one input to each device (cathode versus grid) can be more linear than the other. Some asymmetry in the Schade feedback perhaps? Seems like it affects both input terminals similarly.
How about duplicated but opposite dissimilar (each t/anti-t setup with unmatched gm's, but on opposite sides, and only one input driven to each setup) "triode"/anti-triode stages. Then sum the outputs. Complicated. Some way to simplify?
Maybe parallel devices on each side of just one t/anti-t setup, and each input only drives one of the parallel devices on each side? Each input then has less gm than the anti-side taken together in parallel. Wild speculation here no doubt. Some problems driving only one of a bunch of followers, this would need some cathode summing/degen. resistors likely.
"I wish I understood just half of what you guys are on about!"
I'm not sure we really know what we're talking about here either. A strange new landscape. Lots of hand waving theory, fortunately Lars and others put some real data onto the table periodically. But when there's no map, well .....
Don
Yes, I see what you are saying. I think you mean to say SE character here, not just triode. Equal non-linearities lead to a P-P sort of dist. cancellation. A near constant gm in the anti device would eliminate this cancellation. Using a high gm device also achieves the same effect by operation over a smaller portion of its non-linear transfer.
"how about a 2 stage amp with self-splitting pentode/antipentode driver
and push-pull Schadeode outputs? How about an input stage that gives the
amp a push-pull signature when driven with a balanced signal and a SE signature
when driven with a SE signal?"
Lets see, could we just duplicate Shoog's setup for both output tubes, providing two inputs. Seems that if both sides are identical we will just end up with the equal non-linearities case, leading to a P-P signature regardless of which input is driven. Unless one input to each device (cathode versus grid) can be more linear than the other. Some asymmetry in the Schade feedback perhaps? Seems like it affects both input terminals similarly.
How about duplicated but opposite dissimilar (each t/anti-t setup with unmatched gm's, but on opposite sides, and only one input driven to each setup) "triode"/anti-triode stages. Then sum the outputs. Complicated. Some way to simplify?
Maybe parallel devices on each side of just one t/anti-t setup, and each input only drives one of the parallel devices on each side? Each input then has less gm than the anti-side taken together in parallel. Wild speculation here no doubt. Some problems driving only one of a bunch of followers, this would need some cathode summing/degen. resistors likely.
"I wish I understood just half of what you guys are on about!"
I'm not sure we really know what we're talking about here either. A strange new landscape. Lots of hand waving theory, fortunately Lars and others put some real data onto the table periodically. But when there's no map, well .....
Don
Last edited:
I have used toroidals in at least three amps now - all working excellently. These are the conditions I think you need to forfill in order to achieve best results. The design should have low output impedance to bite on the lower primary inductance of transformer - this design forfills that because the partial feedback produces lower output Z than a triode. Secondly all toroidals I have used have excellent bandwidth with response down to 10hz and and out to beyond 60khz. I use at least 100VA transformers to get the response down past their designed 60/50hz. They will not tolerate any standing DC and that is why I have the CCS arrangement in the tails - it is essential.
You should be able to get universal transformers with split primary windings of 110 or 120V per primary and this is what you need. In terms of manufacturer - I haven't come across a bad one - theres very little to go wrong with how they are wound. I have found that interwinding capacitance comes into play depending on the polarity of the primary winding and you should do listening tests to see which one sounds smoother. Generally if there is high interwinding capacitance then the response peaks at about 60khz. This initially sounds nice and lively but quickly becomes fatiguing. So what you are looking for is the less bright sounding response.
Finally remember the 8V is specified for 4ohm speakers and if you have 8ohms you will need 12V.
Overall you can expect the toroidals to cost about a third of convential transformers
Hope that all helps you make a decision.
Shoog
I have just done a few hours of listening tests with my hardist core techno.
All hint of the fatiguing harshness has indeed gone so its not distorting badly. However the overall balance is tilted towards the treble a little. I am thinking possibly that the coupling cap is a little small but would welcome any other input as to what might be the source.Other than that it sounding very good.
Shoog
.All hint of the fatiguing harshness has indeed gone so its not distorting badly. However the overall balance is tilted towards the treble a little. I am thinking possibly that the coupling cap is a little small but would welcome any other input as to what might be the source.Other than that it sounding very good.
Shoog
This toroidals transformers is the part I like most in Shoog's circuit, actually the whole circuit is a masterpiece indeed. Who says there is not much progress in the tube amp design ???
Other things came into my mind when I read Shoog's circuits.
1, Baby Huey amp (Yves, Gingertube), has similar partial feedback circuit.. it is a "voltage feedback"..the partial feedback can also be implemented by pentode loaded with CCS or Gyrator,which is "current feedback" (I/V convertion).. Broskie's 300B amp is a good referrence in DC and AC calculation.
2. Triodington, (OldEurope), anti-triode circuit. again, very similar concept.
I find it hard to understand concepts discussed in this thread.. a good learning though..
I would say adding more of filtering capacitance is an elegant solutions. It lowers PS's dynamic resistance as well, lowering parasitic feedbacks, that is good for the amp.
The elegant solution would have used smaller caps and abuse the output stage
(cathodes) as its own (negative) regulators or cap multipliers or whatever....
This is one design (properly implemented) that shouldn't have cared how noisy
the power rail might be, within reason...
Well, at least brute force clean power you won't have a residual IMD of hum.
Can't swear that my solution to float away the hum would have been totally
free of IMD side effects.
No rule saying you can't combine both brute force and elegant techniques.
I suspect the benefits would be cumulative.
(cathodes) as its own (negative) regulators or cap multipliers or whatever....
This is one design (properly implemented) that shouldn't have cared how noisy
the power rail might be, within reason...
Well, at least brute force clean power you won't have a residual IMD of hum.
Can't swear that my solution to float away the hum would have been totally
free of IMD side effects.
No rule saying you can't combine both brute force and elegant techniques.
I suspect the benefits would be cumulative.
.SUBCKT EL86 A G2 G1 K
BGG GG 0 V=V(G1,K)+0.38874904
BEP EP 0 V=URAMP(V(A,K))+1e-10
BEG EG 0 V=URAMP(V(G1,K))+1e-10
BEG2 EG2 0 V=URAMP(V(G2,K))+1e-10
BM1 M1 0 V=(0.09419804*(URAMP(V(EG2)-1e-10)+1e-10))**-1.560546
BM2 M2 0 V=(0.49010862*(URAMP(V(GG)+V(EG2)/5.4129723)+1e-10))**3.060546
BP P 0 V=0.0055371733*(URAMP(V(GG)+V(EG2)/11.044434)+1e-10)**1.5
BIK IK 0 V=U(V(GG))*V(P)+(1-U(V(GG)))*0.0058898823*V(M1)*V(M2)
BIG IG 0 V=0.0027685867*V(EG)**1.5*(V(EG)/(V(EP)+V(EG))*1.2+0.4)
BIK2 IK2 0 V=V(IK,IG)*(1-0.4*(EXP(-V(EP)/V(EG2)*15)-EXP(-15)))
BIG2T IG2T 0 V=V(IK2)*(0.945717092*(1-V(EP)/(V(EP)+10))**1.5+0.054282908)
BIK3 IK3 0 V=V(IK2)*(V(EP)+2245)/(V(EG2)+2245)
BIK4 IK4 0 V=V(IK3)-URAMP(V(IK3)-(0.0035373923*(V(EP)+URAMP(V(EG2,EP)))**1.5))
BIP IP 0 V=URAMP(V(IK4,IG2T)-URAMP(V(IK4,IG2T)-(0.0035373923*V(EP)**1.5)))
BIAK A K I=V(IP)+1e-10*V(A,K)
BIG2 G2 K I=URAMP(V(IK4,IP))
BIGK G1 K I=V(IG)
* CAPS
CGA G1 A 0.6p
CGK G1 K 6.9p
C12 G1 G2 4.6p
CAK A K 5.4p
.ENDS
BGG GG 0 V=V(G1,K)+0.38874904
BEP EP 0 V=URAMP(V(A,K))+1e-10
BEG EG 0 V=URAMP(V(G1,K))+1e-10
BEG2 EG2 0 V=URAMP(V(G2,K))+1e-10
BM1 M1 0 V=(0.09419804*(URAMP(V(EG2)-1e-10)+1e-10))**-1.560546
BM2 M2 0 V=(0.49010862*(URAMP(V(GG)+V(EG2)/5.4129723)+1e-10))**3.060546
BP P 0 V=0.0055371733*(URAMP(V(GG)+V(EG2)/11.044434)+1e-10)**1.5
BIK IK 0 V=U(V(GG))*V(P)+(1-U(V(GG)))*0.0058898823*V(M1)*V(M2)
BIG IG 0 V=0.0027685867*V(EG)**1.5*(V(EG)/(V(EP)+V(EG))*1.2+0.4)
BIK2 IK2 0 V=V(IK,IG)*(1-0.4*(EXP(-V(EP)/V(EG2)*15)-EXP(-15)))
BIG2T IG2T 0 V=V(IK2)*(0.945717092*(1-V(EP)/(V(EP)+10))**1.5+0.054282908)
BIK3 IK3 0 V=V(IK2)*(V(EP)+2245)/(V(EG2)+2245)
BIK4 IK4 0 V=V(IK3)-URAMP(V(IK3)-(0.0035373923*(V(EP)+URAMP(V(EG2,EP)))**1.5))
BIP IP 0 V=URAMP(V(IK4,IG2T)-URAMP(V(IK4,IG2T)-(0.0035373923*V(EP)**1.5)))
BIAK A K I=V(IP)+1e-10*V(A,K)
BIG2 G2 K I=URAMP(V(IK4,IP))
BIGK G1 K I=V(IG)
* CAPS
CGA G1 A 0.6p
CGK G1 K 6.9p
C12 G1 G2 4.6p
CAK A K 5.4p
.ENDS
Agh... And I'd already redone with 6V6 last night...
Well, quite a few PSRR mods here to chew on. Comments???
The biggest problem (for me) in getting full cancellation turned
out to be screen currents. I ended up ditching the pentode up
front and gone to cascode. Hybrid, but only cause the voltage
I had to work with (for direct coupling) wants to be quite low.
There is still quite a bit of IMD when hum (2VPP Simmd) is present.
So, Shoog's solution is still the better one. But you can do both.
Well, quite a few PSRR mods here to chew on. Comments???
The biggest problem (for me) in getting full cancellation turned
out to be screen currents. I ended up ditching the pentode up
front and gone to cascode. Hybrid, but only cause the voltage
I had to work with (for direct coupling) wants to be quite low.
There is still quite a bit of IMD when hum (2VPP Simmd) is present.
So, Shoog's solution is still the better one. But you can do both.
Attachments
Makes my brain hurt. You guys move too fast to follow casually, so there's a lot I don't completely grasp. But where are you making up the current through L2 that feeds the input tube?
Sheldon
Last edited:
Making up (the balance?) for driver current, I'm not. Its a goof I may need to address.
One could fidget with the emitter resistors in the mirror I suppose...
If its single ended in behavior, then we might even want the quiescent to be off the
center. That the opt might be better centered saturationwise at full signal... But this
is no excuse, I simply missed it.
I did deliberately imbalance the cathode voltages. That only one polarized cap might
be needed in the bridge... Feedback biasing inspired by something I saw Waveborn do.
One could fidget with the emitter resistors in the mirror I suppose...
If its single ended in behavior, then we might even want the quiescent to be off the
center. That the opt might be better centered saturationwise at full signal... But this
is no excuse, I simply missed it.
I did deliberately imbalance the cathode voltages. That only one polarized cap might
be needed in the bridge... Feedback biasing inspired by something I saw Waveborn do.
Last edited:
I'm in the process of testing this amplifier. One channel is sucking all juice from my 100mA HV supply, probably because of the 160 ohm cathode resistors for starters. One thing bugs me and that's the voltage I measure between R17 and R12. Peculiar isn't it?
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.