Just to be complete, here is the original Switchless Switchmode Conv. front end. SSC (T2 and T3 can be combined on one xfmr.)
---------------------------
I'm experimenting with different ferrite toroid winding configurations lately to reduce distributed capacitance. Latest is winding two wires bifilar in one layer, then removing one to leave an air gap between turns. Hoping to get the winding self resonance up to 250 KHz or better without having to reduce the turns too much.
-----------------------------
It appears that some similar swapping around of xfmrs on the LV demod. side can give some new configurations there too. More work to do on that.
Don
---------------------------
I'm experimenting with different ferrite toroid winding configurations lately to reduce distributed capacitance. Latest is winding two wires bifilar in one layer, then removing one to leave an air gap between turns. Hoping to get the winding self resonance up to 250 KHz or better without having to reduce the turns too much.
-----------------------------
It appears that some similar swapping around of xfmrs on the LV demod. side can give some new configurations there too. More work to do on that.
Don
Attachments
Oops, correction.
The B+ xfmrs generally have a similar number of turns on them as the full plate to plate center tapped xfmrs. So the main consideration between versions is the CM (common mode) capacitance issues.
Don
The B+ xfmrs generally have a similar number of turns on them as the full plate to plate center tapped xfmrs. So the main consideration between versions is the CM (common mode) capacitance issues.
Don
Not so constant beta
I am finding that beta may be flat in ~some~ select transistors
over some useful range of base currents. But most NPN curves
rise with current and then gain falls like stone near saturation.
The majority of the minority of devices that are truly flat, most
seem to be PNP. I am not sure the reason why. Very few NPN's
aproximate flatness in Beta.
The big gotcha seems to be temperature. From -40 to +125C,
the beta varies as much as 3x. Even more for Darlingtons...
Beta vs Current may remain flat while Beta vs Temperature
sails away into la-la land....
Now Anti-Triode on the flip side would match any variance.
Duplicating a balance against it is not a problem. Yet why
would I accept the current gain of my output stage mirrors
to drift 3:1 with temperature? Even in balance, I would not.
A beta mirror would have to be uber-heatsinkified, and then
cascode protected to hide it from doing any real work. Or jus
forget the whole beta thing as a not so simple simplification,
and go back to the Op-FET mirrors for both Triode and Anti.
Comparing voltage drop across resistors of the same temp
seems a far more stable way to multiply current.
I am finding that beta may be flat in ~some~ select transistors
over some useful range of base currents. But most NPN curves
rise with current and then gain falls like stone near saturation.
The majority of the minority of devices that are truly flat, most
seem to be PNP. I am not sure the reason why. Very few NPN's
aproximate flatness in Beta.
The big gotcha seems to be temperature. From -40 to +125C,
the beta varies as much as 3x. Even more for Darlingtons...
Beta vs Current may remain flat while Beta vs Temperature
sails away into la-la land....
Now Anti-Triode on the flip side would match any variance.
Duplicating a balance against it is not a problem. Yet why
would I accept the current gain of my output stage mirrors
to drift 3:1 with temperature? Even in balance, I would not.
A beta mirror would have to be uber-heatsinkified, and then
cascode protected to hide it from doing any real work. Or jus
forget the whole beta thing as a not so simple simplification,
and go back to the Op-FET mirrors for both Triode and Anti.
Comparing voltage drop across resistors of the same temp
seems a far more stable way to multiply current.
Looked at KSA1220A or MJL4302A ?
True, the temperature has a significant effect on Beta, thats why bipolars have thermal runaway problems. Can just do the same fixup that current mirrors with gain do: put a resistor in the emitter to linearize and then a diode/resistor input load to correct for junction voltage and temperature. Or use a Mosfet with a source resistor.
True, the temperature has a significant effect on Beta, thats why bipolars have thermal runaway problems. Can just do the same fixup that current mirrors with gain do: put a resistor in the emitter to linearize and then a diode/resistor input load to correct for junction voltage and temperature. Or use a Mosfet with a source resistor.
Found a real nice NPN and PNP pair for constant Beta. Only 7 Watt, 300 V though. 2SA1353 2SC3417 Must be some others, these are Sanyo parts.
An interesting idea comes up for the NJL4302D (230 Watt, 350 V), this is the Thermaltrak version of the MJL4302A, and it has a diode mounted on the same substrate for thermal tracking. Maybe the diode could be used in a current mirror like setup for temperature compensation. (the diode and a resistor as load for the input signal, and an approx. 1/Beta scaled resistor in the emitter path.) This part has a nice flat Beta curve.
For big NPN match: NJL4281D (350V, 230 W) and NJL3281D (260V 200 W)
and PNP match to NJL3281D is: NJL1302D
Don
An interesting idea comes up for the NJL4302D (230 Watt, 350 V), this is the Thermaltrak version of the MJL4302A, and it has a diode mounted on the same substrate for thermal tracking. Maybe the diode could be used in a current mirror like setup for temperature compensation. (the diode and a resistor as load for the input signal, and an approx. 1/Beta scaled resistor in the emitter path.) This part has a nice flat Beta curve.
For big NPN match: NJL4281D (350V, 230 W) and NJL3281D (260V 200 W)
and PNP match to NJL3281D is: NJL1302D
Don
oven regulated beta
With tubes and power transistors providing heat, it should be
possible to stabilize the temp of the beta transistor at ~100C.
Possibly some low tech method regulating flow of fresh air?
There's no saying it has to give full boost cold at 25C, only
that it warms up to a regulated stable temperature fairly
soon, and does not invite further runaway to meltdown...
Hmmmm. I wonder if varying the cascode blocking voltage
could be abused to influence the temperature of the beta
transistor? To make it warm up quickly, and then reduce
the blocking voltage (and dissipation) to hold that temp
and beta steady.
But once you have to add a cascode to keep the temp in
control, the beta method seems no great savings in parts
or complexity compared to other mirrors that are inherently
temperature compensated.
NJL4302D sure does look very tempting for a simple mirror.
Thanks for bringing that one to my attention.
With tubes and power transistors providing heat, it should be
possible to stabilize the temp of the beta transistor at ~100C.
Possibly some low tech method regulating flow of fresh air?
There's no saying it has to give full boost cold at 25C, only
that it warms up to a regulated stable temperature fairly
soon, and does not invite further runaway to meltdown...
Hmmmm. I wonder if varying the cascode blocking voltage
could be abused to influence the temperature of the beta
transistor? To make it warm up quickly, and then reduce
the blocking voltage (and dissipation) to hold that temp
and beta steady.
But once you have to add a cascode to keep the temp in
control, the beta method seems no great savings in parts
or complexity compared to other mirrors that are inherently
temperature compensated.
NJL4302D sure does look very tempting for a simple mirror.
Thanks for bringing that one to my attention.
Ken,
Since you mentioned the idea of adjusting the cascode voltage to control the temp. How about an amplified diode circuit, as used for a SS amplifier's totem pole biasing, here using the amplified Thermaltrak diode to set the cascode voltage.
Or, could use an Op. Amp. to monitor the Thermaltrak diode and control the cascode voltage to keep the diode at constant volts, ie, const. temp.
Don
Since you mentioned the idea of adjusting the cascode voltage to control the temp. How about an amplified diode circuit, as used for a SS amplifier's totem pole biasing, here using the amplified Thermaltrak diode to set the cascode voltage.
Or, could use an Op. Amp. to monitor the Thermaltrak diode and control the cascode voltage to keep the diode at constant volts, ie, const. temp.
Don
cascoded diodes...
I was aslo thinking of ways to abuse a hollow state diode as
a reference to emulate a really big triode (or anti-triode). By
using a solid state cascode to modulate % of virtual plate
voltage as seen by the real plate of the reference diode.
Basically fake the effect of a grid upon the transfer curve...
Also possibility of current mirroring a hollow diode to make
lower impedance power supply with some tube character.
Gotta be careful how the sand would mirror the zero cut
off and recovery. Might not be worth the trouble.
I was aslo thinking of ways to abuse a hollow state diode as
a reference to emulate a really big triode (or anti-triode). By
using a solid state cascode to modulate % of virtual plate
voltage as seen by the real plate of the reference diode.
Basically fake the effect of a grid upon the transfer curve...
Also possibility of current mirroring a hollow diode to make
lower impedance power supply with some tube character.
Gotta be careful how the sand would mirror the zero cut
off and recovery. Might not be worth the trouble.
I think Anatoliy mentioned earlier an Op amp. configuration with a therm. diode in the neg. feedback and another therm. diode as the input resistor to the inv. input. (needs some DC biasing fixups) Ratio of the diode Gm's gives the Mu of the "triode" formed. Should give a nice triode response without the usual g1 "island effect" problem. (maybe like a nice 300B?)
edit:
I was thinking of post #15:
http://www.diyaudio.com/forums/showthread.php?postid=1446357#post1446357
but now that I look at it again, it is something totally different.
Should be some other Op. amp. configuration with therm. diodes that would give an anti-triode function. Maybe a therm. diode shunt from inv. input to ground and a resistor from inv. input to the input signal source. Then a resistor from the output to a high Rp diode to ground, with a resistor from the midpoint back to the inv. input. Something like that anyway.
Don
edit:
I was thinking of post #15:
http://www.diyaudio.com/forums/showthread.php?postid=1446357#post1446357
but now that I look at it again, it is something totally different.
Should be some other Op. amp. configuration with therm. diodes that would give an anti-triode function. Maybe a therm. diode shunt from inv. input to ground and a resistor from inv. input to the input signal source. Then a resistor from the output to a high Rp diode to ground, with a resistor from the midpoint back to the inv. input. Something like that anyway.
Don
Re: Triode Clone
Looks like STC 😉
http://www.gem.hi-ho.ne.jp/katsu-san/audio/what_STC1_english.html
Kind regards,
Darius
Looks like STC 😉
http://www.gem.hi-ho.ne.jp/katsu-san/audio/what_STC1_english.html
Kind regards,
Darius
"Looks like STC"
There is certainly a strong similarity in the circuit appearance between the two. And in the first post I made some years ago about the triode clone idea, I also linked to Kamijo's STC site.
But the two are actually quite distinct. The triode clone operates the triode at constant current (versus signal), giving a true gain of Mu. (notice the grounded grid versus the auto-biased grid between the two models) The STC operates with varying current (versus signal) thru the triode. Giving the same effect as a plate resistor loaded triode.
You will notice that the load lines given for the STC always slope rather than being flat. So, it depends on how you want to operate the tube. But the STC idea is certainly a valid tube emulation idea when combined with SS. Since it leaves some fraction of an actual load on the triode, some may find it to be a more "realistic" "triode" sound. (mostly 2nd harmonic dist. a little 3rd)
One problem I find with the STC, however, is that this "load" on the triode is NOT the actual load. Rather the triode current is programmed to increase with INPUT signal instead of the OUTPUT signal. (now for a fixed R load, the two will track, but not for a real reactive load) So this STC load is simulated in effect. This "simulated load" can be added to the triode clone circuit too, by just putting some resistance across the CCS to make current vary with input signal likewise. The triode emulations that Ken and Michael have done on the other hand, I believe, offer true loading by a fraction of the real load.
On the other hand, the triode clone has virtually ONLY 2nd harmonic distortion, which some may find more pleasing. One can also select how much 2nd harmonic by the setting of the CCS current and a fixed bias voltage on the "grounded" grid (to bring the cathode voltage on the inverting Op. amp. input back to zero volts). (the diagram really should show a bias pot. on the grid to select a voltage between ground and some minus voltage)
Don
There is certainly a strong similarity in the circuit appearance between the two. And in the first post I made some years ago about the triode clone idea, I also linked to Kamijo's STC site.
But the two are actually quite distinct. The triode clone operates the triode at constant current (versus signal), giving a true gain of Mu. (notice the grounded grid versus the auto-biased grid between the two models) The STC operates with varying current (versus signal) thru the triode. Giving the same effect as a plate resistor loaded triode.
You will notice that the load lines given for the STC always slope rather than being flat. So, it depends on how you want to operate the tube. But the STC idea is certainly a valid tube emulation idea when combined with SS. Since it leaves some fraction of an actual load on the triode, some may find it to be a more "realistic" "triode" sound. (mostly 2nd harmonic dist. a little 3rd)
One problem I find with the STC, however, is that this "load" on the triode is NOT the actual load. Rather the triode current is programmed to increase with INPUT signal instead of the OUTPUT signal. (now for a fixed R load, the two will track, but not for a real reactive load) So this STC load is simulated in effect. This "simulated load" can be added to the triode clone circuit too, by just putting some resistance across the CCS to make current vary with input signal likewise. The triode emulations that Ken and Michael have done on the other hand, I believe, offer true loading by a fraction of the real load.
On the other hand, the triode clone has virtually ONLY 2nd harmonic distortion, which some may find more pleasing. One can also select how much 2nd harmonic by the setting of the CCS current and a fixed bias voltage on the "grounded" grid (to bring the cathode voltage on the inverting Op. amp. input back to zero volts). (the diagram really should show a bias pot. on the grid to select a voltage between ground and some minus voltage)
Don
Some interesting extensions to the triode and anti-triode emulations in posts 150/151 would be to use a square law Mosfet (gate controlled by a resistive divider between source and drain) in place of ONE of the devices (or even both maybe too). That way we would be playing off 3/2 power (thermionic diode) versus 4/2 power (Mosfet) laws between the two non-linear devices, might give some interesting distortions.
Putting in the square law Mosfets for BOTH devices would nominally return the overall response to a linear one. But Mosfets change their characteristic power law versus current from a square law thru 3/2 law down to near linear law at high current. So by biasing one device at a different DC operating current than the other, some intentional mistracking could be generated. Maybe could be used as a low order corrector for the whole amplifier. Fun, fun...
Don
Putting in the square law Mosfets for BOTH devices would nominally return the overall response to a linear one. But Mosfets change their characteristic power law versus current from a square law thru 3/2 law down to near linear law at high current. So by biasing one device at a different DC operating current than the other, some intentional mistracking could be generated. Maybe could be used as a low order corrector for the whole amplifier. Fun, fun...
Don
Ohh..., for the Anti-Triode tube clone (variant of post 152) all you have to do is just put the input signal thru a resistor to the inverting input, instead of to the non-inverting input as shown. (anti-triode is just a true complement to triode) Should be a high value resistor if you want to maintain near constant current thru the triode however. Or, could put a feedback resistor in parallel with the triode to compensate for the inverting input signal current.
Basakwerd Triodieodieodie Upsidasium
Lets clarify that the Anti-Triode has a triode curve flipped upside
down , and dynamically re-centered below 20KHz to balance out
in the current domain. Nothing about the curve's reference shape
in regard to plate or grid voltage is otherwise altered.
A gyrated triode, trading plate voltage for plate current. Flips the
curve along the diagonal, is something else completely different.
Not be confused or blindly substituted for the Anti-Triode.
I am not accusing anyone else of getting this wrong yet, though
I almost just did. An easy oops to make with these unintuitive
(for me) feedback circuits...
smoking-amp said:
Lets clarify that the Anti-Triode has a triode curve flipped upside
down , and dynamically re-centered below 20KHz to balance out
in the current domain. Nothing about the curve's reference shape
in regard to plate or grid voltage is otherwise altered.
A gyrated triode, trading plate voltage for plate current. Flips the
curve along the diagonal, is something else completely different.
Not be confused or blindly substituted for the Anti-Triode.
I am not accusing anyone else of getting this wrong yet, though
I almost just did. An easy oops to make with these unintuitive
(for me) feedback circuits...
Yet Another Anti-Triode
I have a feeling that Smoking may have suggested this before.
Funny thing, if you were to drive this from the other side, and
the triode were "good" enough to keep up. It would become
an Anti-Darlington...
To behave as an anti-device, the anti-side must not react to
input or B+ variations, only to current in the tail. It must also
exceed the performance envelope of the virtual device to be
emulated.
Clearly in case of a differential twin triodes, each side attempts
with only partial success to drive the other as an anti-triode.
The result is some blended curve inbetween. Neither end ever
entirely dominating the overall behavior...
Here the Anti-Triode is entirely dominated by the behavior of
the reference, it has no say of its own....
I have a feeling that Smoking may have suggested this before.
Funny thing, if you were to drive this from the other side, and
the triode were "good" enough to keep up. It would become
an Anti-Darlington...
To behave as an anti-device, the anti-side must not react to
input or B+ variations, only to current in the tail. It must also
exceed the performance envelope of the virtual device to be
emulated.
Clearly in case of a differential twin triodes, each side attempts
with only partial success to drive the other as an anti-triode.
The result is some blended curve inbetween. Neither end ever
entirely dominating the overall behavior...
Here the Anti-Triode is entirely dominated by the behavior of
the reference, it has no say of its own....
Attachments
- Status
- Not open for further replies.