A tangent of "Another Kind of Hybrid", possibly deserving its own thread. Or not...
http://www.diyaudio.com/forums/showthread.php?s=&threadid=118725
I think as drawn, this is only gonna deliver 1 or 2 watts SE...
Not quite sure how the voltage swing at the RP multiplier will be
affected by the primary of the transformer??? Somebody else
here surely has a better grasp of the math than I do. Wouldn't
be at all surprised if I screwed up other things too.
I think the 100 Mu of the input stage might have been twice what
was needed to wag the diode end to full swing on the opposite of
the OPT... Again not at all sure in this instance...
Anyways, the idea here was to explore "perfect" triode behavior.
By imposing a perfect (constant current) triode Mu onto the plate
curve of a Diode... Thus producing an emulated triode with forced
parallel curves... (no lean to the right whatsoever...)
Does it succeed? I dunno, not sure its anything but a thought
exercise to explore a dead end to its full and complete extinction.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=118725
I think as drawn, this is only gonna deliver 1 or 2 watts SE...
Not quite sure how the voltage swing at the RP multiplier will be
affected by the primary of the transformer??? Somebody else
here surely has a better grasp of the math than I do. Wouldn't
be at all surprised if I screwed up other things too.
I think the 100 Mu of the input stage might have been twice what
was needed to wag the diode end to full swing on the opposite of
the OPT... Again not at all sure in this instance...
Anyways, the idea here was to explore "perfect" triode behavior.
By imposing a perfect (constant current) triode Mu onto the plate
curve of a Diode... Thus producing an emulated triode with forced
parallel curves... (no lean to the right whatsoever...)
Does it succeed? I dunno, not sure its anything but a thought
exercise to explore a dead end to its full and complete extinction.
Attachments
Even More Noodling
Here I abuse the other half of the 12AX7 as a diode...
But why use two crummy triodes to emulate a much
bigger one, compatible with Hammond 300b SE OPT???
An analog computer, emulating a big triode such a
very roundabout way. Why not just current multiply
the triode in the output stage and be done with it?
Cause that would only create a big crummy triode.
With curves that lean to the right on account of
internal variances. I got something else in mind...
Here I isolate the computation of Mu from the load.
Mu is "perfect" as it gets, working a constant current.
However, I deliberately DO NOT isolate the load from
the Plate resistance and the 3/2 power law. Here the
diode (or triode with fixed grid) creates a single plate
curve, that is duplicated over the entire SOA by the
action of the Mu Starver.
I get a virtual triode with "perfectly parallel" curves.
I'm sure the 12AX7's microphonics are still an issue...
Can't fake away everthing...
-------------------------------------------------------------------
Anyways, I'm not entirely sure the current multiplier
section is right. Are On-Semi's ThermalTrak diodes
matched transistors as I have drawn them here?
Or are they just diodes that do some biasing trick
deliberately not the same as the transistor VBE?
I tried various multipliers on my breadboard with
seriously unmatched RadioShack 2n3904's. Had
about the same chance of finding two identical
people on a crowded bus.
Anyways, the "plain" mirror didn't work worth a
darn, totally inacurate.... 3 Transistor Wilson was
much better, but I could only get it to mirror, not
multiply. 4 Transistor would Multiply, not sure why
as its the same as the 3 transistor with an extra
diode drop.
I havn't tried such wildly different emitter values.
Not sure I totally understand Shockley's relation
in regard to emitter currents. This mirror may or
may not work as drawn, guess thats all I'm saying.
--------------------------------------------------------------
Even with 155mA DC and 230VPP swing (or is it
more on account of the primary?) I still only figure
some 2 or 3 Watts. Is that just the way it goes
with all true SE?
Mind you, the schematic would "look" a lot more
SE if I put the OPT under the reference diode
with one end grounded... Just that might not be
good for the filament, thus I draw it in a confusing
way in the middle, but its still SE....
The output is pseudo P-Channel. Starvation between
the input and the anode, the load between cathode
and B-. Its a total basakwarz mess...
Here I abuse the other half of the 12AX7 as a diode...
But why use two crummy triodes to emulate a much
bigger one, compatible with Hammond 300b SE OPT???
An analog computer, emulating a big triode such a
very roundabout way. Why not just current multiply
the triode in the output stage and be done with it?
Cause that would only create a big crummy triode.
With curves that lean to the right on account of
internal variances. I got something else in mind...
Here I isolate the computation of Mu from the load.
Mu is "perfect" as it gets, working a constant current.
However, I deliberately DO NOT isolate the load from
the Plate resistance and the 3/2 power law. Here the
diode (or triode with fixed grid) creates a single plate
curve, that is duplicated over the entire SOA by the
action of the Mu Starver.
I get a virtual triode with "perfectly parallel" curves.
I'm sure the 12AX7's microphonics are still an issue...
Can't fake away everthing...
-------------------------------------------------------------------
Anyways, I'm not entirely sure the current multiplier
section is right. Are On-Semi's ThermalTrak diodes
matched transistors as I have drawn them here?
Or are they just diodes that do some biasing trick
deliberately not the same as the transistor VBE?
I tried various multipliers on my breadboard with
seriously unmatched RadioShack 2n3904's. Had
about the same chance of finding two identical
people on a crowded bus.
Anyways, the "plain" mirror didn't work worth a
darn, totally inacurate.... 3 Transistor Wilson was
much better, but I could only get it to mirror, not
multiply. 4 Transistor would Multiply, not sure why
as its the same as the 3 transistor with an extra
diode drop.
I havn't tried such wildly different emitter values.
Not sure I totally understand Shockley's relation
in regard to emitter currents. This mirror may or
may not work as drawn, guess thats all I'm saying.
--------------------------------------------------------------
Even with 155mA DC and 230VPP swing (or is it
more on account of the primary?) I still only figure
some 2 or 3 Watts. Is that just the way it goes
with all true SE?
Mind you, the schematic would "look" a lot more
SE if I put the OPT under the reference diode
with one end grounded... Just that might not be
good for the filament, thus I draw it in a confusing
way in the middle, but its still SE....
The output is pseudo P-Channel. Starvation between
the input and the anode, the load between cathode
and B-. Its a total basakwarz mess...
Attachments
Hi Ken,
I just noticed this thread.
One thing I noticed about both schematics is that the plate Rp emulations give lower Rp when the input triode is turning off instead of on. So Rp effects and Mu effects will be inverted with respect to each other. May still be fine. No idea, would have to build. But I'm wondering about the separation of Mu and Rp idea, since the normal triode produces its Mu with the Rp taken into account. Here it is an after effect, so may be just putting a diode in series with the load.
On the Wilson mirror in the 2nd diagram, doesn't the ratio of the emitter resistors need to agree with the multiplier ratio? Maybe I'm not reading the numbers right.
I think I heard that the biasing diodes in the On-Semi Thermal Traks are just MUR120 or similar diodes, so not matched to the transistor base emitter voltage drop. There are curves given for both the Vbe and the Vdiode on the data sheets (versus temp.).
So may have to use a Schottky or something somewhere to get the voltages matched up.
Don
I just noticed this thread.
One thing I noticed about both schematics is that the plate Rp emulations give lower Rp when the input triode is turning off instead of on. So Rp effects and Mu effects will be inverted with respect to each other. May still be fine. No idea, would have to build. But I'm wondering about the separation of Mu and Rp idea, since the normal triode produces its Mu with the Rp taken into account. Here it is an after effect, so may be just putting a diode in series with the load.
On the Wilson mirror in the 2nd diagram, doesn't the ratio of the emitter resistors need to agree with the multiplier ratio? Maybe I'm not reading the numbers right.
I think I heard that the biasing diodes in the On-Semi Thermal Traks are just MUR120 or similar diodes, so not matched to the transistor base emitter voltage drop. There are curves given for both the Vbe and the Vdiode on the data sheets (versus temp.).
So may have to use a Schottky or something somewhere to get the voltages matched up.
Don
Shockley and junk...
I need to change up the whole current mirror thing...
Especially if Thermaltraks are merely bias diodes.
Where do I find matched pair power BJT's in one
thermal package for current mirror apps?
As for the Mu reversal: If Mu is linear, does it matter?
If even slightly nonlinear, then I must assume it would
matter, and I should address that in some future rev.
Probably more important is which end of Mu clamps
to zero cathode voltage....
And the ratio of the mirror weighting would be correct
if Shockley's Relation sais the emitter drop will always
be 0.7V+26ohms. I don't fully grasp Shockley yet, and
I might have misapplied the intrinsic emitter resistance.
You are dead right about it being merely a glorified
diode in series with the load. SE Diode, I claimed no
more or less... Or did I?
In any case, Diode + Starver output stage is EXACTLY
the same as if a "Perfect Triode" (by the book theory
only) with a Mu of precisely 1. And Pseudo P-Channel
to boot.... Additional Mu has to come from somewhere
further up the chain.
Thats not saying a "Perfect" emulated triode would
sound any better than a real one, where Mu varies
somewhat with Cathode>Plate current.
Just that the more famous audio triodes trend toward
the perfect model with parallel curves that barely vary.
The ones with bad reps, all have the lean to the right
or even less explainable non-linearities....
Even with parallel curves and no lean to the right, any
resistive load line will still compress somewhat at the
cutoff by the 3/2 power law which applies equally to
Vacuum Diodes.... Linearity, but only up to proximity
of clipping.
Clipping is evil. But a gracefully non-linear evil is better
than pretending some truly linear amp will never clip.
I need to change up the whole current mirror thing...
Especially if Thermaltraks are merely bias diodes.
Where do I find matched pair power BJT's in one
thermal package for current mirror apps?
As for the Mu reversal: If Mu is linear, does it matter?
If even slightly nonlinear, then I must assume it would
matter, and I should address that in some future rev.
Probably more important is which end of Mu clamps
to zero cathode voltage....
And the ratio of the mirror weighting would be correct
if Shockley's Relation sais the emitter drop will always
be 0.7V+26ohms. I don't fully grasp Shockley yet, and
I might have misapplied the intrinsic emitter resistance.
You are dead right about it being merely a glorified
diode in series with the load. SE Diode, I claimed no
more or less... Or did I?
In any case, Diode + Starver output stage is EXACTLY
the same as if a "Perfect Triode" (by the book theory
only) with a Mu of precisely 1. And Pseudo P-Channel
to boot.... Additional Mu has to come from somewhere
further up the chain.
Thats not saying a "Perfect" emulated triode would
sound any better than a real one, where Mu varies
somewhat with Cathode>Plate current.
Just that the more famous audio triodes trend toward
the perfect model with parallel curves that barely vary.
The ones with bad reps, all have the lean to the right
or even less explainable non-linearities....
Even with parallel curves and no lean to the right, any
resistive load line will still compress somewhat at the
cutoff by the 3/2 power law which applies equally to
Vacuum Diodes.... Linearity, but only up to proximity
of clipping.
Clipping is evil. But a gracefully non-linear evil is better
than pretending some truly linear amp will never clip.
Since I can't seem to figure an accurate current multiplier...
Back to Diode Voltage and Plate Resistance multiplication.
I think I have come up with a cheat that combines
Plate Starver, Plate Multiplier, and a small gain for Mu...
All with very few components.
Here Mu is an emitter follower (mostly). then gain comes
from multiplying the plate. Its screwy, and maybe more
sand than I wanted in the gain path... But it has enough
local feedback I don't think sand linearity will be a prob.
The non-linearities will come from interaction between
the plate multiplied vacuum diode and the loadline. The
diode sees the real load through the resistor network,
summed with input and whatever... Again should "ACT"
like a SE Triode (in A1 only) with perfectly parallel plate
curves, all duplicated from the Reference Diode.
I am not feeling well, so didn't figure all the component
values, caps and such. Values should ballpark obvious.
And with only 5Mu, an input gain stage is still needed,
and one with a fairly low interstage impedance.
I am thinking this one is 25 or 30 Watts, but the 1K SE
iron that can handle 350mA will be the tough shopping.
I'll probably figure some sort of Anti-Diode SEPP in the
near future to boost the power even further keep the
costs reasonable.
I am interested in critique of the mixing network that
combines the three functions. If I screwed something,
or doesn't work as advertised, I would want to know.
I am leaning away from DHD's as I am unsure if the
filament(s) are balance-able or not... With an indirect
Cathode, maybe one less thing to worry about.
And I just happen to have a 5V4GT in my junk box!
Back to Diode Voltage and Plate Resistance multiplication.
I think I have come up with a cheat that combines
Plate Starver, Plate Multiplier, and a small gain for Mu...
All with very few components.
Here Mu is an emitter follower (mostly). then gain comes
from multiplying the plate. Its screwy, and maybe more
sand than I wanted in the gain path... But it has enough
local feedback I don't think sand linearity will be a prob.
The non-linearities will come from interaction between
the plate multiplied vacuum diode and the loadline. The
diode sees the real load through the resistor network,
summed with input and whatever... Again should "ACT"
like a SE Triode (in A1 only) with perfectly parallel plate
curves, all duplicated from the Reference Diode.
I am not feeling well, so didn't figure all the component
values, caps and such. Values should ballpark obvious.
And with only 5Mu, an input gain stage is still needed,
and one with a fairly low interstage impedance.
I am thinking this one is 25 or 30 Watts, but the 1K SE
iron that can handle 350mA will be the tough shopping.
I'll probably figure some sort of Anti-Diode SEPP in the
near future to boost the power even further keep the
costs reasonable.
I am interested in critique of the mixing network that
combines the three functions. If I screwed something,
or doesn't work as advertised, I would want to know.
I am leaning away from DHD's as I am unsure if the
filament(s) are balance-able or not... With an indirect
Cathode, maybe one less thing to worry about.
And I just happen to have a 5V4GT in my junk box!
Attachments
Arg! need a blocking diode in series with that Zenier.
I don't want it to forward conduct if the node swings
below ground (or do I?) Rather my large signals clamp
against the +55V Zenier (as if it were a real grid) and
squash against vacuum diode to ground (without any
ground clamping pump the signal back the other way).
And to clarify that network (if I got it right). Multiplies
plate resistance by 10, plate impedance by 20, and
Mu by 5. Assuming zero input impedance driving...
I don't want it to forward conduct if the node swings
below ground (or do I?) Rather my large signals clamp
against the +55V Zenier (as if it were a real grid) and
squash against vacuum diode to ground (without any
ground clamping pump the signal back the other way).
And to clarify that network (if I got it right). Multiplies
plate resistance by 10, plate impedance by 20, and
Mu by 5. Assuming zero input impedance driving...
Miscalculated the resistor matrix slightly. Multipled voltage by 11.
I needed x10 to set Diode currrent at the best operating point.
The 100K starver bias resistor becomes 90K? Everything figures
correct now except Mu, changes to 4.5 , still perfectly acceptable.
And Z=950... DC Resistance 710ohms, 350mA @ 250V...
I needed x10 to set Diode currrent at the best operating point.
The 100K starver bias resistor becomes 90K? Everything figures
correct now except Mu, changes to 4.5 , still perfectly acceptable.
And Z=950... DC Resistance 710ohms, 350mA @ 250V...
In attempting to clarify and correct minor component errors.
90K was intended to multiply plate voltage, Z, and Resistance.
Unintentionally may be NFB to the input, linearizing the circuit.
I may have to re-separate plate multiplier and Starver stages
to preserve the intended non-linearity...
Back to square one...
I was hoping by now to have anti-diode SEPP to show, but still
slagging the maths of this silly resistor network. Combines too
many interdependent functions for easy analysis in my head.
Prior "more complicated" circuit with extra parts was far easier
to explain...
New -15V clamping Zenier is just protect the gate from excessive
negative swings, as the IGBT emitter cannot follow below +0.5V ...
The gate should idle around 30V quiescent, if no recent events
have pumped it away from either the +55V "Virtual Grid Zero" or
-15V safety clamp. Hoping the -15 clamp should never need to
conduct, Can't imagine it does anything "good" for the sound
if it ever does. Pretending it will never happen, the naive wishing
a real signal might never reach the rail. Yeah, right...
90K was intended to multiply plate voltage, Z, and Resistance.
Unintentionally may be NFB to the input, linearizing the circuit.
I may have to re-separate plate multiplier and Starver stages
to preserve the intended non-linearity...
Back to square one...
I was hoping by now to have anti-diode SEPP to show, but still
slagging the maths of this silly resistor network. Combines too
many interdependent functions for easy analysis in my head.
Prior "more complicated" circuit with extra parts was far easier
to explain...
New -15V clamping Zenier is just protect the gate from excessive
negative swings, as the IGBT emitter cannot follow below +0.5V ...
The gate should idle around 30V quiescent, if no recent events
have pumped it away from either the +55V "Virtual Grid Zero" or
-15V safety clamp. Hoping the -15 clamp should never need to
conduct, Can't imagine it does anything "good" for the sound
if it ever does. Pretending it will never happen, the naive wishing
a real signal might never reach the rail. Yeah, right...
Attachments
6L6 as a Square Law Diode
Reading "Schade 1938 Beam Power Tubes" from:
http://www.one-electron.com/Misc_Docs.html
It would seem that the Diode plate curve of Beam Power tubes
are deliberately constructed for the square power law (as are
modern FETs), not the 3/2 power law often quoted for vacuum
tubes in general.
Done deliberately to emphasize the 2nd harmonic and suppress
all other distortions. Given that the 2nd could easily be canceled
in push-pull. The old PDF explains better than I could.
Given within, is a plate to grid feedback scheme that seems to
emulate a pretty good triode without the power loss of wiring
the screen to the plate. And not requiring an UL tap...
I may want to investigate a 6L6 beam power Diode as an
alternative reference with square law curve... See how the
two power laws compare for SE sound in otherwise perfect
virtual Triodes...
Yeah, I could just wire the 6L6 as Triode, or copy the 1938
emulation, or abuse an UL tap. But all that has been done
before, what would be the point?
------------------------------------------------------------------------------
Another observation I've made regarding SE Triodes, is that
when clipping does occurs, it is always to a high impedance.
Disconnecting the load and letting it do whatever it pleases.
The voltage may change, but the current does not...
Most Push Pull schemes, Clip to low impedance. Tying a load
hard to one power rail or the other. The current may change,
but the voltage does not...
I am not sure through an OPT if the speaker can tell the diff
which direction (over or under) the impedance mismatched...
But it probably makes a huge diff in OTL.
Reading "Schade 1938 Beam Power Tubes" from:
http://www.one-electron.com/Misc_Docs.html
It would seem that the Diode plate curve of Beam Power tubes
are deliberately constructed for the square power law (as are
modern FETs), not the 3/2 power law often quoted for vacuum
tubes in general.
Done deliberately to emphasize the 2nd harmonic and suppress
all other distortions. Given that the 2nd could easily be canceled
in push-pull. The old PDF explains better than I could.
Given within, is a plate to grid feedback scheme that seems to
emulate a pretty good triode without the power loss of wiring
the screen to the plate. And not requiring an UL tap...
I may want to investigate a 6L6 beam power Diode as an
alternative reference with square law curve... See how the
two power laws compare for SE sound in otherwise perfect
virtual Triodes...
Yeah, I could just wire the 6L6 as Triode, or copy the 1938
emulation, or abuse an UL tap. But all that has been done
before, what would be the point?
------------------------------------------------------------------------------
Another observation I've made regarding SE Triodes, is that
when clipping does occurs, it is always to a high impedance.
Disconnecting the load and letting it do whatever it pleases.
The voltage may change, but the current does not...
Most Push Pull schemes, Clip to low impedance. Tying a load
hard to one power rail or the other. The current may change,
but the voltage does not...
I am not sure through an OPT if the speaker can tell the diff
which direction (over or under) the impedance mismatched...
But it probably makes a huge diff in OTL.
Schrade (RCA 1938)
http://www.diyaudio.com/forums/showthread.php?s=&postid=1520934#post1520934
As I used no vacuum reference this spin,
I posted the schematic to Solid State...
http://www.diyaudio.com/forums/showthread.php?s=&postid=1520934#post1520934
As I used no vacuum reference this spin,
I posted the schematic to Solid State...
- Status
- Not open for further replies.