insight
Just doing a little thinking on the case of using a triode in place of the pentode current source in CCT1 (ie, jumper out the Mosfet in CCT4 above).
The impedance modulation caused by the 2nd triode, V2, has exactly the right form of a diode conduction curve to cancel out the even harmonics from V1. So using just the triode V2 in place of the pentode is N.G. One MUST have the Mosfet as in CCT4 (above) to preserve the SET sound accurately. The Mosfet's high gm effectively pins the cathode voltage of the main triode V1 so it operates in a SET like environment. The Mosfet still accurately conveys the complementary current to V2, and V2 conveys it accurately to the xfmr. with power gain.
So JUST running a dual triode LTP pair with a single input will NOT give you SET sound. Use CCT4 ( figure 4 above).
Don
Just doing a little thinking on the case of using a triode in place of the pentode current source in CCT1 (ie, jumper out the Mosfet in CCT4 above).
The impedance modulation caused by the 2nd triode, V2, has exactly the right form of a diode conduction curve to cancel out the even harmonics from V1. So using just the triode V2 in place of the pentode is N.G. One MUST have the Mosfet as in CCT4 (above) to preserve the SET sound accurately. The Mosfet's high gm effectively pins the cathode voltage of the main triode V1 so it operates in a SET like environment. The Mosfet still accurately conveys the complementary current to V2, and V2 conveys it accurately to the xfmr. with power gain.
So JUST running a dual triode LTP pair with a single input will NOT give you SET sound. Use CCT4 ( figure 4 above).
Don
Re: Kevlar Balloon!
And this magically eliminates its distortion...?
It has nothing to do with plate to screen current ratio and everything to do with the Gm vs. Ip characteristic of the tube (which is different for different types, styles, loads and so forth). Once that is fixed, we can argue such subtleties as screen current.
All these circuits are *approximately* what they claim to be, but aren't as they are claimed. The claimed increase in power output only works if you don't count transistors or pentodes as amplifiers, which sounds distinctly pro-triode racist to me...
Tim
smoking-amp said:
And this magically eliminates its distortion...?
It has nothing to do with plate to screen current ratio and everything to do with the Gm vs. Ip characteristic of the tube (which is different for different types, styles, loads and so forth). Once that is fixed, we can argue such subtleties as screen current.
All these circuits are *approximately* what they claim to be, but aren't as they are claimed. The claimed increase in power output only works if you don't count transistors or pentodes as amplifiers, which sounds distinctly pro-triode racist to me...
Tim
clarification called for
Hi Tim,
The signal we are interested in reproducing accurately at the opposite side to the triode is the complementary CURRENT. Not voltage. The grounded grid or grounded gate circuits are indeed linear for current input to current output (ignoring screen current of course for the moment). In fact they are EXACTLY unity current gain, otherwise Kirchoffs Laws would be violated, since no current flows out of the grid or gate (at least not before the component fails anyway).
The gm factor only comes in when we figure the variation of the CCS voltage required to get the demanded current. This variation of the tail voltage can cause a distortion problem for the main triode by varying its cathode voltage. Often referred to as impedance modulation (Z=1/gm of the pentode or Mosfet at the tail point) By using a LARGE gm we make this impedance near zero, so it doesn't matter how badly it behaves around zero. Hence the preference for using a Mosfet there. (so if using a pentode, we would want one with high gm) The Mosfet effectively pins the tail voltage. We could also use a bipolar transister, which has even higher gm, but it has base current that subtracts from the emitter current, affecting our unity current gain or linearity somewhat in grounded base op. (only a tiny bit actually since it is mainly a constant tiny fraction).
So, I guess the answer is it DOES magically eliminate distortion, because it does. Its called optimum circuit topology.
On the power gain figure, I did not mean to imply some magic gain of power from the triode. The pentode or Mosfet is indeed supplying the extra power. The point is that if you have some triode that can produce X amount of watts in a SET circuit, this circuit can give you 2X amount of watts of the same sound quality (and DC compensation for the xfmr). Some people would be very happy to achieve that without having to parallel more triodes. (4X with the H bridge circuit)
One spot where the circuit does magically save on power waisted, is efficiency, versus just using a CCS load on a triode to get DC current compensation for the xfmr. (50% max eff. versus 25% max eff.) But it is the same efficiency obtained using an inductor for parafeed.
Don
Hi Tim,
The signal we are interested in reproducing accurately at the opposite side to the triode is the complementary CURRENT. Not voltage. The grounded grid or grounded gate circuits are indeed linear for current input to current output (ignoring screen current of course for the moment). In fact they are EXACTLY unity current gain, otherwise Kirchoffs Laws would be violated, since no current flows out of the grid or gate (at least not before the component fails anyway).
The gm factor only comes in when we figure the variation of the CCS voltage required to get the demanded current. This variation of the tail voltage can cause a distortion problem for the main triode by varying its cathode voltage. Often referred to as impedance modulation (Z=1/gm of the pentode or Mosfet at the tail point) By using a LARGE gm we make this impedance near zero, so it doesn't matter how badly it behaves around zero. Hence the preference for using a Mosfet there. (so if using a pentode, we would want one with high gm) The Mosfet effectively pins the tail voltage. We could also use a bipolar transister, which has even higher gm, but it has base current that subtracts from the emitter current, affecting our unity current gain or linearity somewhat in grounded base op. (only a tiny bit actually since it is mainly a constant tiny fraction).
So, I guess the answer is it DOES magically eliminate distortion, because it does. Its called optimum circuit topology.
On the power gain figure, I did not mean to imply some magic gain of power from the triode. The pentode or Mosfet is indeed supplying the extra power. The point is that if you have some triode that can produce X amount of watts in a SET circuit, this circuit can give you 2X amount of watts of the same sound quality (and DC compensation for the xfmr). Some people would be very happy to achieve that without having to parallel more triodes. (4X with the H bridge circuit)
One spot where the circuit does magically save on power waisted, is efficiency, versus just using a CCS load on a triode to get DC current compensation for the xfmr. (50% max eff. versus 25% max eff.) But it is the same efficiency obtained using an inductor for parafeed.
Don
Linearity in the CCT2/CCT3 circuits (figs. 2 & 3)
Lets look at the top devices in the fig. 2 and 3 circuits now. The R2 resistor has to be sized a little larger than the R1 resistor to provide drive signal to the Mosfet. If the Mosfet had a linear gm this would not affect linearity. But since the Mosfet does have a non-linear gm (same for a pentode there too) it will potentially produce distortion. The accuracy of the complementary current generation depends on a constant voltage across R1 + R2 (ignoring the extra delta R for the linear drive signal).
The non-linear portion of the drive required by the Mosfet (or pentode) makes this voltage across R1+R2 vary a small amount, distorting our complementary current generation. So it is important to minimize it. The way to do this is to use a large gm device so its drive voltage requirements will be small compared to the IR drops on R1+R2. We can also increase R1+R2 in value, but this eventually affects our amplifier output impedance and damping factor. So there is some practical limit to the value of the R1 and R2 current sampling resistors.
A mosfet (or even a bipolar transistor could be used, has even higher gm) has very large gm. The Mosfet may require around a volt or so of drive signal. Of this, some few % of that will be the non-linear component, say 10 mV. If the R1 and R2 resistors are dropping say 10 V, then our accuracy will be off by 10 mV/10 V or 0.1%. Which is at about the threshold of audibility. So this is in the ballpark of useability.
10V drop on the resistors with a 300V plate swing would be about 10/300 times the 8 Ohm output or .26 Ohm equivalent in the secondary. This would give a 30 to 1 damping factor if the triode's output impedance were zero. (Actually, R1 and R2 are in parallel as far as AC output goes, so would be more like 1/4 of .26 Ohm equiv. or 120 to 1 damping) So we can twiddle the R values some and use a higher gm part if necessary. Not quite as clean as the CCT1 (fig. 1) design, but probably acceptable with a little tweaking.
One could, by the way, use an extra Mosfet to make a Mosfet current mirror configuration to fix this non-linearity problem altogether, but that gets into selecting matched parts. Might be more feasible using bipolar transistors. (hard to match Mosfets up well) One could also put a floating Op Amp. in the Mosfet drive loop to push the effective drive level requirement down to maybe nanovolts, a little overkill I think.
Don
Lets look at the top devices in the fig. 2 and 3 circuits now. The R2 resistor has to be sized a little larger than the R1 resistor to provide drive signal to the Mosfet. If the Mosfet had a linear gm this would not affect linearity. But since the Mosfet does have a non-linear gm (same for a pentode there too) it will potentially produce distortion. The accuracy of the complementary current generation depends on a constant voltage across R1 + R2 (ignoring the extra delta R for the linear drive signal).
The non-linear portion of the drive required by the Mosfet (or pentode) makes this voltage across R1+R2 vary a small amount, distorting our complementary current generation. So it is important to minimize it. The way to do this is to use a large gm device so its drive voltage requirements will be small compared to the IR drops on R1+R2. We can also increase R1+R2 in value, but this eventually affects our amplifier output impedance and damping factor. So there is some practical limit to the value of the R1 and R2 current sampling resistors.
A mosfet (or even a bipolar transistor could be used, has even higher gm) has very large gm. The Mosfet may require around a volt or so of drive signal. Of this, some few % of that will be the non-linear component, say 10 mV. If the R1 and R2 resistors are dropping say 10 V, then our accuracy will be off by 10 mV/10 V or 0.1%. Which is at about the threshold of audibility. So this is in the ballpark of useability.
10V drop on the resistors with a 300V plate swing would be about 10/300 times the 8 Ohm output or .26 Ohm equivalent in the secondary. This would give a 30 to 1 damping factor if the triode's output impedance were zero. (Actually, R1 and R2 are in parallel as far as AC output goes, so would be more like 1/4 of .26 Ohm equiv. or 120 to 1 damping) So we can twiddle the R values some and use a higher gm part if necessary. Not quite as clean as the CCT1 (fig. 1) design, but probably acceptable with a little tweaking.
One could, by the way, use an extra Mosfet to make a Mosfet current mirror configuration to fix this non-linearity problem altogether, but that gets into selecting matched parts. Might be more feasible using bipolar transistors. (hard to match Mosfets up well) One could also put a floating Op Amp. in the Mosfet drive loop to push the effective drive level requirement down to maybe nanovolts, a little overkill I think.
Don
Hi G,
Sorry about the headaches!
Whenever a new idea crops up, it inevitably goes thru some flux and change as weaknesses are discovered or re-thinking occurs. I also would like to express my appreciation for all critical remarks, one often finds weaknesses in design or better ways to do things, when having to critically re-examine how things are working. Let the bullets fly!! (putting on my Kevlar suit hurriedly)
Having noted earlier that the figure 2 and 3 (CCT2 and CCT3) totem pole circuits were a little weak on precision, I have come up with a more precision design, see attached figure 5. Q1 and Q2 matched pair are used as a mini-op amp to control Q3 more accurately. The constant voltage across R1 and R2 will be maintained quite accurately now, so will be enforcing the complementary current generation. (I1*R1 + I2*R2 = constant, R1 = R2, so I1 = constant/R1 - I2 ) Q4 is used just to keep high voltage off of Q1's collector.
This circuit can also be easily changed to perform precision SRPP or Mu follower by simply moving the transformer connection to the top of R1 or bottom of R2 respectively.
Don
Sorry about the headaches!
Whenever a new idea crops up, it inevitably goes thru some flux and change as weaknesses are discovered or re-thinking occurs. I also would like to express my appreciation for all critical remarks, one often finds weaknesses in design or better ways to do things, when having to critically re-examine how things are working. Let the bullets fly!! (putting on my Kevlar suit hurriedly)
Having noted earlier that the figure 2 and 3 (CCT2 and CCT3) totem pole circuits were a little weak on precision, I have come up with a more precision design, see attached figure 5. Q1 and Q2 matched pair are used as a mini-op amp to control Q3 more accurately. The constant voltage across R1 and R2 will be maintained quite accurately now, so will be enforcing the complementary current generation. (I1*R1 + I2*R2 = constant, R1 = R2, so I1 = constant/R1 - I2 ) Q4 is used just to keep high voltage off of Q1's collector.
This circuit can also be easily changed to perform precision SRPP or Mu follower by simply moving the transformer connection to the top of R1 or bottom of R2 respectively.
Don
Attachments
analyzin away
Been up to a little more analyzin of the earlier CCT2/3 (totem pole type) circuits (figs. 2 & 3) using a pentode on the top part.
We looked at the general magnitudes of distortion earlier. But we also know the general nature of these gm non-linearities as well.
The nature of the distortion from the non-linear gm of the pentode is one of a thermionic diode conduction curve (3/2 power law), and is of the correct polarity to REMOVE even harmonics from the main triode's even harmonics when in the top totem pole position in CCT2/3.
(Notice that putting a pentode or triode of COMPARABLE gm to the main triode in the BOTTOM right section of CCT1/3 circuits also tends to cancel even harmonics from the triode's even harmonics, actually the pentode may likely overwhelm the triode's even harmonics)
If the gm of the top totem pentode is comparable to the main triode you will also get even harmonic cancellation and maybe even generation of inverted harmonics from the pentode overwhelming the triode's harmonics. The triode's harmonic generation is rather less than on a 1 to 1 gm comparison since it operates with plate feedback and so depends on the load impedance as to how much 3/2 power distortion is present. So you really need a very high gm pentode to preserve the SET sound.
Using a Mosfet on top in the CCT2/3 (totem pole) circuits will likely preserve 99% of the even harmonics due to the much higher gm of the Mosfet. (the Mosfet also has a gm non-linearity somewhat like the thermionic diode curve, square law instead of 3/2 power law) So using the last CCT6 version (figure just above) is probably unnecessary overkill just to get that last 1% restored.
Don
Been up to a little more analyzin of the earlier CCT2/3 (totem pole type) circuits (figs. 2 & 3) using a pentode on the top part.
We looked at the general magnitudes of distortion earlier. But we also know the general nature of these gm non-linearities as well.
The nature of the distortion from the non-linear gm of the pentode is one of a thermionic diode conduction curve (3/2 power law), and is of the correct polarity to REMOVE even harmonics from the main triode's even harmonics when in the top totem pole position in CCT2/3.
(Notice that putting a pentode or triode of COMPARABLE gm to the main triode in the BOTTOM right section of CCT1/3 circuits also tends to cancel even harmonics from the triode's even harmonics, actually the pentode may likely overwhelm the triode's even harmonics)
If the gm of the top totem pentode is comparable to the main triode you will also get even harmonic cancellation and maybe even generation of inverted harmonics from the pentode overwhelming the triode's harmonics. The triode's harmonic generation is rather less than on a 1 to 1 gm comparison since it operates with plate feedback and so depends on the load impedance as to how much 3/2 power distortion is present. So you really need a very high gm pentode to preserve the SET sound.
Using a Mosfet on top in the CCT2/3 (totem pole) circuits will likely preserve 99% of the even harmonics due to the much higher gm of the Mosfet. (the Mosfet also has a gm non-linearity somewhat like the thermionic diode curve, square law instead of 3/2 power law) So using the last CCT6 version (figure just above) is probably unnecessary overkill just to get that last 1% restored.
Don
correction
Oops, got the decimal point wrong on the final paragraph above. For a typical SET triode stage we would expect around a couple % of distortion, and the Mosfet on top will be contributing around -.1 or -.2% cancelling even harmonics, so will preserve about 90% of the SET even harmonics.
However, the CCT3 design (figure 3, H bridge, totem pole design) has the top Mosfet stages on BOTH sides of the xfmr and they will cancel each others even harmonics out. This leaves 100% of the SET even harmonics preserved. (although, the bottom Mosfet may still remove about 5%) Maybe would be worthwhile to use a bipolar transistor on the bottom section to knock that 5% loss down to maybe .5% loss of even harmonics.
For the H bridge design, using the more precision fig. 5 (CCT5) design on each side doesn't gain anything as far as the even harmonics go (nothing to fix), but would knock out the extra <.1% odd harmonic distortion from the Mosfets (should actually be well below the .1% total distortion figure, since its a 3rd order product or higher, and the device is a near square law curve with little odd order contribution). Maybe some golden ears can hear these <.01% effects?
(one might take a more detailed look at the odd order distortions here, the triode changes over from expansive 3/2 power effect to a compressive odd order effect due to saturation as signal level increases, whereas the top Mosfets continue as an expansive odd order distortion effect, this means the odd order distortions add at low levels and change over to cancelling at higher signal levels. The bottom section Mosfet only contributes even order, primarily 2nd harmonic, about -.05% with respect to the triode's, distortion.)
Don
Oops, got the decimal point wrong on the final paragraph above. For a typical SET triode stage we would expect around a couple % of distortion, and the Mosfet on top will be contributing around -.1 or -.2% cancelling even harmonics, so will preserve about 90% of the SET even harmonics.
However, the CCT3 design (figure 3, H bridge, totem pole design) has the top Mosfet stages on BOTH sides of the xfmr and they will cancel each others even harmonics out. This leaves 100% of the SET even harmonics preserved. (although, the bottom Mosfet may still remove about 5%) Maybe would be worthwhile to use a bipolar transistor on the bottom section to knock that 5% loss down to maybe .5% loss of even harmonics.
For the H bridge design, using the more precision fig. 5 (CCT5) design on each side doesn't gain anything as far as the even harmonics go (nothing to fix), but would knock out the extra <.1% odd harmonic distortion from the Mosfets (should actually be well below the .1% total distortion figure, since its a 3rd order product or higher, and the device is a near square law curve with little odd order contribution). Maybe some golden ears can hear these <.01% effects?
(one might take a more detailed look at the odd order distortions here, the triode changes over from expansive 3/2 power effect to a compressive odd order effect due to saturation as signal level increases, whereas the top Mosfets continue as an expansive odd order distortion effect, this means the odd order distortions add at low levels and change over to cancelling at higher signal levels. The bottom section Mosfet only contributes even order, primarily 2nd harmonic, about -.05% with respect to the triode's, distortion.)
Don
Hi Gavin, Nigel, and John,
Thank you all for the kind words of encouragement!
I would throw something together quickly with bench power supplies except I don't really have a suitable P-P transformer available at the moment. I have a couple of Hammond 1650T available from another project, but at 1950 Ohms P-P it's a bit low on impedance for any of the small triodes I have on hand at the present: 6EW7, 6DE7. I do have a 2000VCT power xfmr with 120 V primary that might do for a simple proof of concept, but will likely have bad frequency response. I also have an MTM2N90 Mosfet, 900V 2A, that should work fine opposite the triode. Maybe I should order a big voltage regulator tube for a low plate Rp triode.
What one should do is build a bench SET amp and the CCT (triode and Mosfet on a CCS tail and P-P xfmr) amp with similar load impedances and compare harmonic spectrums with similar loads. I have TEK spectrum analysis equipment. But I don't have a suitable SET transformer to use for the comparison amplifier. I do have a Hammond 1640SE xfmr at 1250 Ohms, so I COULD..I suppose.. make both amplifiers with way too low an output Z and compare TWO bad apples I guess! Certainly would have lots of harmonics to compare that way. Don't know how convincing that would be.
Don
Thank you all for the kind words of encouragement!
I would throw something together quickly with bench power supplies except I don't really have a suitable P-P transformer available at the moment. I have a couple of Hammond 1650T available from another project, but at 1950 Ohms P-P it's a bit low on impedance for any of the small triodes I have on hand at the present: 6EW7, 6DE7. I do have a 2000VCT power xfmr with 120 V primary that might do for a simple proof of concept, but will likely have bad frequency response. I also have an MTM2N90 Mosfet, 900V 2A, that should work fine opposite the triode. Maybe I should order a big voltage regulator tube for a low plate Rp triode.
What one should do is build a bench SET amp and the CCT (triode and Mosfet on a CCS tail and P-P xfmr) amp with similar load impedances and compare harmonic spectrums with similar loads. I have TEK spectrum analysis equipment. But I don't have a suitable SET transformer to use for the comparison amplifier. I do have a Hammond 1640SE xfmr at 1250 Ohms, so I COULD..I suppose.. make both amplifiers with way too low an output Z and compare TWO bad apples I guess! Certainly would have lots of harmonics to compare that way. Don't know how convincing that would be.
Don
Yeah!
Build a prototype!
Can you imagine what happens if this works that good? An if you could wrap the design in comprehensive formula's and even simlifie it a bit?
You would save the tube people about 50% in their xformer expenses, the 300bla freaks will be able to play rock music at descent levels and, as i think this was the main issue, we can make bigger SET's!
I might go for that gu81 sub driver!
god job, i really hope this gets to be a real amp and not just a wacky thought we can hyperlink to in other topics
Bas
Build a prototype!
Can you imagine what happens if this works that good? An if you could wrap the design in comprehensive formula's and even simlifie it a bit?
You would save the tube people about 50% in their xformer expenses, the 300bla freaks will be able to play rock music at descent levels and, as i think this was the main issue, we can make bigger SET's!
I might go for that gu81 sub driver!
god job, i really hope this gets to be a real amp and not just a wacky thought we can hyperlink to in other topics
Bas
I've recently come back to this topic in a rather roundabout way.
Having had a look at circuit 1 again, which is perhaps the most interesting in it's simplicity, I noticed a rather serious error in it, namely, the screen supply must not be ground referenced. It should be a floating supply between screen grid of the pentode and it's cathode - which immediatly presents a dissadvantage.
In the way it is drawn now, the screen grid current contributes to total cathode current and throws the balance off, because i'ts not constant, yet goes through the CCS. Another possibility would be to use a suitable MOSFET but this also presents a problem - you are limited to supply voltages of about 500V or so (keeping in mind the MOSFET drain swings over the positive rail, and it is extremely difficult to find MOSFETs for >1000V, with 800V parts being fairly common), and, unlike the tube, the MOSFET has a signifficant gate-suorce capacitance, which is connected to the CCS on one side, and effectively grounded on the gate side. This presents a capacitive load for the CCS which reduces it's impedance at HF, making it less of a CCS, and therefore upsetting balance at HF. For larger tubes I doubt this is a problem, but for flea-size amps it may get signifficant. And of course there is the unthinkable of using more sand in a tube amp
Having had a look at circuit 1 again, which is perhaps the most interesting in it's simplicity, I noticed a rather serious error in it, namely, the screen supply must not be ground referenced. It should be a floating supply between screen grid of the pentode and it's cathode - which immediatly presents a dissadvantage.
In the way it is drawn now, the screen grid current contributes to total cathode current and throws the balance off, because i'ts not constant, yet goes through the CCS. Another possibility would be to use a suitable MOSFET but this also presents a problem - you are limited to supply voltages of about 500V or so (keeping in mind the MOSFET drain swings over the positive rail, and it is extremely difficult to find MOSFETs for >1000V, with 800V parts being fairly common), and, unlike the tube, the MOSFET has a signifficant gate-suorce capacitance, which is connected to the CCS on one side, and effectively grounded on the gate side. This presents a capacitive load for the CCS which reduces it's impedance at HF, making it less of a CCS, and therefore upsetting balance at HF. For larger tubes I doubt this is a problem, but for flea-size amps it may get signifficant. And of course there is the unthinkable of using more sand in a tube amp
I have returned from the near dead!
Hi beamnet - Bas,
I would very much like to get back to work on a prototype soon, but I just completed moving to Hickory NC from Connecticut. I had to put 20,000 lbs of equipment into storage in the boondocks of Ct using PODs to ship it. (PODs being those shipping containers you load and unload yourself.) Then I used PODs to ship my furniture etc to NC. My body is nearly trashed after all the work, much of it in snow and freezing weather in Ct including unloading it all again into self storage where they didn't provide heat or lighting. I thought I sent a minumum lab setup to NC but after I got here, I see I missed some critical things. I won't be getting back to Ct to retrieve stuff from storage until Spring arrives. :-(
Hi ilimzn,
If the plate voltage of the pentode stays above the assigned screen voltage, one could use a solid state CCS from the plate to the screen with a zener (or VR tube) from screen to cathode. This way the cathode sees a constant current component due to the partitioning between screen and zener. Not an ideal solution, but better.
One would like to freeze the voltage seen by the triode on its cathode for a true SET like environment, so a very large transconductance on the other side of the diff. pair would be helpful. A bipolar transistor is perhaps best here, but suffers from "losing" some of the balanced current for the P-P transformer's benefit thru its base connection. This "lost" base current (due to transistor alpha being less than 1, like .995 maybe) could be recovered by using a grounded gate mosfet to hold the base voltage. Ie, the Mosfet source connects to the bipolar base, and the Mosfet drain connects the lost current back to the bipolar collector. Mosfet gate gets grounded to a bias voltage. But this is likely all overkill.
Just using a Mosfet with large transconductance for the other side of the diff. pair to the triode will do a pretty good job of freezing the triode's cathode voltage. The Mosfet source to gate capacitance only affects the small voltage variation remaining on the triode's cathode, a small second order affect. The Mosfet itself does not lose high frequency gain in grounded gate configuration (at least in the audio range, being driven by the low impedance cathode), so the P-P transformer will still see a nicely balanced full bandwidth complementary current.
Don
Hi beamnet - Bas,
I would very much like to get back to work on a prototype soon, but I just completed moving to Hickory NC from Connecticut. I had to put 20,000 lbs of equipment into storage in the boondocks of Ct using PODs to ship it. (PODs being those shipping containers you load and unload yourself.) Then I used PODs to ship my furniture etc to NC. My body is nearly trashed after all the work, much of it in snow and freezing weather in Ct including unloading it all again into self storage where they didn't provide heat or lighting. I thought I sent a minumum lab setup to NC but after I got here, I see I missed some critical things. I won't be getting back to Ct to retrieve stuff from storage until Spring arrives. :-(
Hi ilimzn,
If the plate voltage of the pentode stays above the assigned screen voltage, one could use a solid state CCS from the plate to the screen with a zener (or VR tube) from screen to cathode. This way the cathode sees a constant current component due to the partitioning between screen and zener. Not an ideal solution, but better.
One would like to freeze the voltage seen by the triode on its cathode for a true SET like environment, so a very large transconductance on the other side of the diff. pair would be helpful. A bipolar transistor is perhaps best here, but suffers from "losing" some of the balanced current for the P-P transformer's benefit thru its base connection. This "lost" base current (due to transistor alpha being less than 1, like .995 maybe) could be recovered by using a grounded gate mosfet to hold the base voltage. Ie, the Mosfet source connects to the bipolar base, and the Mosfet drain connects the lost current back to the bipolar collector. Mosfet gate gets grounded to a bias voltage. But this is likely all overkill.
Just using a Mosfet with large transconductance for the other side of the diff. pair to the triode will do a pretty good job of freezing the triode's cathode voltage. The Mosfet source to gate capacitance only affects the small voltage variation remaining on the triode's cathode, a small second order affect. The Mosfet itself does not lose high frequency gain in grounded gate configuration (at least in the audio range, being driven by the low impedance cathode), so the P-P transformer will still see a nicely balanced full bandwidth complementary current.
Don
Depending on what triode and pentode you chose, the G2 voltage on the pentode may end up relatively low, which in turn lowers the gm of the pentode... I would vote for a MOSFET in it's place. BJT has higher gm in principle, but it also has secondary breakdown problems, or, if you get HV BJTs that are sufficiently immune to them, then their gm is low - and you still have to deal with base current. So, MOSFET looks the most promising. It's gm is likely to be MUCH higher than that of the tube, anyway. Completely agree on the effects of MOSFET Cgs, I should have remembered that using higher gm MOSFETs (i.e. with higher Id rating) does increase Cgs well into the nanoFarads, but it also proportionally reduces voltage shifts on the CCS due to increased gm, these two should largely cancel out.
Hi Don,
I suggest you do more reading on differential tube circuits. What you have "designed" here is nothing more than a transformer loaded differential output stage. The fact that the two devices are dissimilar is irrelevant. There is nothing "SE" about this, and the fact that you'd get twice the power should be a strong clue.
The pentode in your circuit is, in fact, amplifying the signal voltage present at its cathode. The fact that the grid is grounded does not change this - it merely makes it a single-input differential amp.
So, since we know that the signal voltage is getting to the pentode, and we know that the pentode and triode both have greatly differing transconductances, load requirements, and output impedances, we will have a grossly non-linear reproduction of the signal.
Building a prototype is a great idea - it will at least prove to you that this is about as far from hi-fi as you can get.
Joel
Joel
I suggest you do more reading on differential tube circuits. What you have "designed" here is nothing more than a transformer loaded differential output stage. The fact that the two devices are dissimilar is irrelevant. There is nothing "SE" about this, and the fact that you'd get twice the power should be a strong clue.
The pentode in your circuit is, in fact, amplifying the signal voltage present at its cathode. The fact that the grid is grounded does not change this - it merely makes it a single-input differential amp.
So, since we know that the signal voltage is getting to the pentode, and we know that the pentode and triode both have greatly differing transconductances, load requirements, and output impedances, we will have a grossly non-linear reproduction of the signal.
Building a prototype is a great idea - it will at least prove to you that this is about as far from hi-fi as you can get.
Joel
Joel
Nothing would go wrong, I have tried it and it works. I used an IXYS 10M45 chip to load the unused side of a P-P transformer with a 45 on the other side. I tried several types of transformers, but none of them sounded as good as an Electra-Print SE transformer made for a 45.
This concept has been discussed in the thread:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=70125
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.