I just fixed up the Super Cuckoo converter diagram (for the Nth time).
The diode flux reset windings are now shown with less turns than the Mosfet drive windings for the faster reset time needed.
One note on the Cuckoo type converter(s) using 2 overlapped phases, the power transformers are only being 50% utilized as far as their power ratings go. This is likely a don't care issue for tube amps.
I would guess you can get 100 Watts out of a Cuckoo scheme using 4 xfmrs. and 4 Mosfets. Where as, a 4 phase scheme with the same 4 xfmrs. and 8 Mosfets (Berning, SCIC, SSC) can likely do 200 Watts.
Don
The diode flux reset windings are now shown with less turns than the Mosfet drive windings for the faster reset time needed.
One note on the Cuckoo type converter(s) using 2 overlapped phases, the power transformers are only being 50% utilized as far as their power ratings go. This is likely a don't care issue for tube amps.
I would guess you can get 100 Watts out of a Cuckoo scheme using 4 xfmrs. and 4 Mosfets. Where as, a 4 phase scheme with the same 4 xfmrs. and 8 Mosfets (Berning, SCIC, SSC) can likely do 200 Watts.
Don
Attachments
I've been doing a bit more research on the Cuk converter for use as an impednance converter, and its not looking so good.
It only maintains a low intrinsic (no feedback) damping factor up to a few hundred Hertz, due to resonances there between its L and C components. That rules it out for no feedback designs. Too bad, because I was looking forward to an impedance ratio control knob.
There are some ripple steering techniques that can be applied to other converters to make them ripple free, but they use L and C components too, so they probably have the same problem.
Well, on that note, I'm finished beating impedance converters to death, unless someone has some new ideas.
That leaves the Berning, SCIC, Super Cuckoo and SSC/Matrix, which all look pretty good.
Don
It only maintains a low intrinsic (no feedback) damping factor up to a few hundred Hertz, due to resonances there between its L and C components. That rules it out for no feedback designs. Too bad, because I was looking forward to an impedance ratio control knob.
There are some ripple steering techniques that can be applied to other converters to make them ripple free, but they use L and C components too, so they probably have the same problem.
Well, on that note, I'm finished beating impedance converters to death, unless someone has some new ideas.
That leaves the Berning, SCIC, Super Cuckoo and SSC/Matrix, which all look pretty good.
Don
Picture attached of the differential switched OTs and B+/2 xfmrs with HV windings on them for the SSC/Matrix 4 phase converter.
LV windings go on next. I used some fairly large ferrite cores as you can see, so that the HV windings would fit on in just one layer. I'm using thickly insulated wire to keep capacitance down. LV windings will be just a few turns, but will be a bunch of windings in parallel to fill the cores. Design flux levels have been kept conservatively low for operation up to 250KHz.
(a Pencil and a 9 pin tube are included in the picture for size comparison.)
Late note on the Cuckoo conv.:
On the Super Cuckoo converter, I earlier said that it didn't need an air gapped core with the flux reset windings added, but I see that Pressman recommends a 2 to 4 mil air gap still, for forward converter topologies (like the Cuckoo) in order to overcome core remanence. This would usually be done using a mylar film between halves of E cores, U cores or Pot cores. So don't try using a toroid for the Cuckoo design.
Don
LV windings go on next. I used some fairly large ferrite cores as you can see, so that the HV windings would fit on in just one layer. I'm using thickly insulated wire to keep capacitance down. LV windings will be just a few turns, but will be a bunch of windings in parallel to fill the cores. Design flux levels have been kept conservatively low for operation up to 250KHz.
(a Pencil and a 9 pin tube are included in the picture for size comparison.)
Late note on the Cuckoo conv.:
On the Super Cuckoo converter, I earlier said that it didn't need an air gapped core with the flux reset windings added, but I see that Pressman recommends a 2 to 4 mil air gap still, for forward converter topologies (like the Cuckoo) in order to overcome core remanence. This would usually be done using a mylar film between halves of E cores, U cores or Pot cores. So don't try using a toroid for the Cuckoo design.
Don
Attachments
I give...
I havn't understood theory of operation in your last ten
or so design revs. Can't say for certain you havn't found
some strange way around the following concerns???
1) Flyback designs that completely reset the core by
forcing all the stored energy out. Through the tube
or to ground, giving it no-where else to go. Do not
present a variable to the loudspeaker. The speaker
side (if powered) sees only that it must charge a
completely reset core each cycle.
This might work OK for tube side power switching,
But the loadspeaker load is then completely hidden
from interactions with the plate. The tube sees only
a dumbed down completely reset core each cycle...
This could be good or bad, but it definately ain't
normal...
2) Designs that escalate to infinate stress if the tube
fails to conduct (perhaps if the cathode might be cold,
biased or driven to cutoff).
3) Designs that reverse or reset the core thru any
non-audio path (such as a freewheeling diode) other
than thru a tube (or mirror) or the loudspeaker load.
The relationship between tube and load is in error by
whatever flux the non-audio path may drain away...
--------------------------------------------------------------------
Variants (such as Berning) that reverse, rather than
reset the core seem relatively immune to all the above
problems. I am not too sure yet about some of these
other cycles... I want to understand, but I am lost.
I havn't understood theory of operation in your last ten
or so design revs. Can't say for certain you havn't found
some strange way around the following concerns???
1) Flyback designs that completely reset the core by
forcing all the stored energy out. Through the tube
or to ground, giving it no-where else to go. Do not
present a variable to the loudspeaker. The speaker
side (if powered) sees only that it must charge a
completely reset core each cycle.
This might work OK for tube side power switching,
But the loadspeaker load is then completely hidden
from interactions with the plate. The tube sees only
a dumbed down completely reset core each cycle...
This could be good or bad, but it definately ain't
normal...
2) Designs that escalate to infinate stress if the tube
fails to conduct (perhaps if the cathode might be cold,
biased or driven to cutoff).
3) Designs that reverse or reset the core thru any
non-audio path (such as a freewheeling diode) other
than thru a tube (or mirror) or the loudspeaker load.
The relationship between tube and load is in error by
whatever flux the non-audio path may drain away...
--------------------------------------------------------------------
Variants (such as Berning) that reverse, rather than
reset the core seem relatively immune to all the above
problems. I am not too sure yet about some of these
other cycles... I want to understand, but I am lost.
"I havn't understood theory of operation in your last ten
or so design revs. Can't say for certain you havn't found
some strange way around the following concerns???"
Oh..., sorry I lost everyone I guess. The flux reset thing is just for the Cuckoo converter. I sorta modded away until I ended up with the Forward Converter inadvertantly, so was a messy explanation I guess. Wasn't sure where it was going myself, just sorta stumbled into Forward Converters eventually. Main thing was getting the ON duty cycle a little past 50%, so only 2 phases are necessary, instead of 4 for continuous coverage.
The other ones (SSC) are just standard P-P converters like the Berning effectively. The Cuk is kaput.... N.G. The Matrix version of the SSC is just a Mosfet demod., instead of the cross conduction diode bridge demod.
--------------------
If you have a book on switching converters, the Cuckoo is just a standard Forward converter except fixed up to do the flux reset faster by lowering the primary turns for the reset diode. In Abraham Pressman's book "Switching Power Supply Design" 2nd ed. on page 76, he explains the trick. The secondary HV side diode disconnects the tube during the flux reset (which is really just a flyback pulse).
During the next Mosfet ON cycle, the the secondary reflects the same Volt.Seconds as the primary again x turns ratio of course (ie, transformer action, volt.seconds = change in flux) and powers the tube again thru the HV diode. It's standard Forward Converter operation setting the output voltages. But in any case, the Cuckoo is really just a distraction here, mainly for low parts count.
----------------------------------
I thought the switched differential Output xfmr design was OK with everone (the center tapped plate windings, multiplexed by pulling up the B+) (then I added the LV cross conduction diode bridge de-mod to get the audio back for the SSC version). Is there some question on this part?
I'll probably get the SSC/Matrix one up and running this weekend, so I can show some waveforms later if that helps. I'm pretty confident that this scheme is workable, just a matter of getting the design details all fine tuned.
Don
edit: Just read thru your #s again. The Cuckoo/Forward conv. brings the core back to the same conditions as a normal P-P scheme for each ON pulse, and the tube and output are connected by the turns ratio xfmr action the same. Just the reset pulse is altogether missing as far as the tube is concerned. But the other xfmr., running at 180 degrees fills in then, alternating back an forth between xfmrs. Really just the same as the Berning P-P case, but every other cycle is on a different xfmr, and the cycles are allowed to exceed 50% slightly.
or so design revs. Can't say for certain you havn't found
some strange way around the following concerns???"
Oh..., sorry I lost everyone I guess. The flux reset thing is just for the Cuckoo converter. I sorta modded away until I ended up with the Forward Converter inadvertantly, so was a messy explanation I guess. Wasn't sure where it was going myself, just sorta stumbled into Forward Converters eventually. Main thing was getting the ON duty cycle a little past 50%, so only 2 phases are necessary, instead of 4 for continuous coverage.
The other ones (SSC) are just standard P-P converters like the Berning effectively. The Cuk is kaput.... N.G. The Matrix version of the SSC is just a Mosfet demod., instead of the cross conduction diode bridge demod.
--------------------
If you have a book on switching converters, the Cuckoo is just a standard Forward converter except fixed up to do the flux reset faster by lowering the primary turns for the reset diode. In Abraham Pressman's book "Switching Power Supply Design" 2nd ed. on page 76, he explains the trick. The secondary HV side diode disconnects the tube during the flux reset (which is really just a flyback pulse).
During the next Mosfet ON cycle, the the secondary reflects the same Volt.Seconds as the primary again x turns ratio of course (ie, transformer action, volt.seconds = change in flux) and powers the tube again thru the HV diode. It's standard Forward Converter operation setting the output voltages. But in any case, the Cuckoo is really just a distraction here, mainly for low parts count.
----------------------------------
I thought the switched differential Output xfmr design was OK with everone (the center tapped plate windings, multiplexed by pulling up the B+) (then I added the LV cross conduction diode bridge de-mod to get the audio back for the SSC version). Is there some question on this part?
I'll probably get the SSC/Matrix one up and running this weekend, so I can show some waveforms later if that helps. I'm pretty confident that this scheme is workable, just a matter of getting the design details all fine tuned.
Don
edit: Just read thru your #s again. The Cuckoo/Forward conv. brings the core back to the same conditions as a normal P-P scheme for each ON pulse, and the tube and output are connected by the turns ratio xfmr action the same. Just the reset pulse is altogether missing as far as the tube is concerned. But the other xfmr., running at 180 degrees fills in then, alternating back an forth between xfmrs. Really just the same as the Berning P-P case, but every other cycle is on a different xfmr, and the cycles are allowed to exceed 50% slightly.
I just thought of an issue that is maybe part of what you are concerned about with the Cuckoo conv. The flux in the core does not reset to exactly the same condition, as after a reverse P-P type cycle. So with the next ON cycle it starts with some residual remanance not seen in the P-P case. The effect of this is to increase the magnetizing current during ON cycles, but the voltage transfer is still unaffected, since volt seconds convert into flux change and vice versa. After all, SE xfmrs work fine for voltage transfer. They just have high magnetizing current.
These magnetizing currents in the Cuckoo converter end up cross conducting between +LV and -LV since both sides are causing magnetizing current. The power actually gets returned to the power supplies via the diode/reset currents, so isn't wasted.
But the audio voltage for top and bottom converters is not the same generally, so the two magnetizing currents are not generally equal. It would seem to be depending on the voltage transfer action being a low impedance to avoid affecting the speaker. But since the windings have some resistance, there will be some effect, but might be just a linear effect with audio voltage, so un-noticeable.
To fix this, I'm thinking now that the reset/diodes should be connected to the speaker load instead of ground, so as to null out the magnetizing currents directly. Avoiding cross conduction currents. (there will be a half clock cycle delay of course, but likely not much change in the audio in half a cycle) Opinions?
Don
These magnetizing currents in the Cuckoo converter end up cross conducting between +LV and -LV since both sides are causing magnetizing current. The power actually gets returned to the power supplies via the diode/reset currents, so isn't wasted.
But the audio voltage for top and bottom converters is not the same generally, so the two magnetizing currents are not generally equal. It would seem to be depending on the voltage transfer action being a low impedance to avoid affecting the speaker. But since the windings have some resistance, there will be some effect, but might be just a linear effect with audio voltage, so un-noticeable.
To fix this, I'm thinking now that the reset/diodes should be connected to the speaker load instead of ground, so as to null out the magnetizing currents directly. Avoiding cross conduction currents. (there will be a half clock cycle delay of course, but likely not much change in the audio in half a cycle) Opinions?
Don
Attachments
smoking-amp said:
You lost me for a while on the flyback variations. Not that I couldn't follow technically but I have a slug of work from my day job and I have little interest in flyback or SE converter topologies for some reason...
Not sure I follow the anti-deadtime idea either unless it's the overlap between the quadrature drive phases.
What I like so far is the SSC on the tube side and a synchronous bridge, or multiphase half bridge using multiple windings, on the load side. I guess that's what you're calling the SSC/Matrix?
Your prototype should be very interesting. Do you care to share your operating parameters? What core material, cross section, mag path length, turns, frequency (you said 250 KHz?) Bmax (I guess could be calculated from the others)?
Do you plan to build it into an output stage or use a current source and resistive load for testing? It looks like your cores are plenty big depending on the permeability of the material. What's the B+ voltage and power envelope you're designing for?
Good luck!
Michael
PS
I'll be starting the build of my "big" 2A3-based 4X current boosted current mirror amp. Maybe one day soon we'll have a current mirror amp with a switched converter output.
Throw in a MOSFET source follower drive stage and you have hybrid SS technology used in three places where it's least likely to impart a lot of it's own character to the sound. We can build a 2 stage SE or PP character amplifier with plenty of gain and no grid current blocking (single stage drive + MOSFET), good power and damping factor (current mirror output stage), and a "perfect" output transformer (switched converter).
"Not sure I follow the anti-deadtime idea either unless it's the overlap between the quadrature drive phases."
Yes, just overlapped phases, 50+% duty. But only two phases required per tube/converter then instead of 4 for 100% coverage.
"multiphase half bridge using multiple windings, on the load side. I guess that's what you're calling the SSC/Matrix?"
Yes, I managed to get rid of the diode (for blocking the reverse diode in the Mosfet) in series with the Mosfets that way.
"Do you care to share your operating parameters? What core material, cross section, mag path length, turns, frequency (you said 250 KHz?) Bmax (I guess could be calculated from the others)?
The larger cores I'm using for the differential P-P xfmr to the tube plates. It has two interleaved 32 turn, center tapped windings for the HV windings (using 600V insulated wire) that just fit on one layer. It uses two stacked Fair-Rite cores P/N 5977004203
( 2.7 in. OD, 1.5 in. ID, .5 in. thk per core) for a combined area of 3.87 cm2. Material #77 is a power matl. with Ui = 2000, Bmax = 4600, 100 OHm CM resistivity (measures 1000 Ohm across core).
Measuring across 32 turns gives 8.475 mH and .18 Ohm (on L/R meter) and between windings 200 pF (I think that means the tubes will see 1/4 of that between them, most of the interwinding cap. on both cores is just across the HF drive. Typical conventional OTs are more like 500 pF plate to plate). I'm assuming max acceptable flux density of 1600 G at 60 KHz, 1200 G at 125 KHZ and 800 G at 250 KHz.
Working this out at 125 KHz, 1200 G gives +/- 1009 V pk -pk on each 16 turn section, so that would be the maximum swing allowed for a tube plate. This clearly is well above what I need. So actual operation will be more like 400 G at 125 KHz. At 250 KHz, that would drop to 200 G being used. Should be nice low core losses.
The B+/2 cores are some unknown units I got from Electronic Goldmine. They measure 55 mm OD, 32 mm ID, 18 mm thk. Have a black coating on them. Area = 2.07 CM2, le = 136.6 mm. I measured .475 mH for 8 turns giving Ui about 3880. These have two interleaved 25 turn windings on them, filling a single layer.
Measure at 4.73 mH and .11 Ohm on L/R meter. Capacitance between windings is 113 pF.
Working out B+/2 at 1200 G and 125 KHz gives 310.5 Vrms or 439 Vpk. So max B+ would be 878 V which is well above what I plan to use, maybe up to 500V tops. So working Bmax would be around 683 G, 1/2 that at 250 KHz.
I plan to put together two more of the largest core units (identical to above) so as to have 4 of them available to do a Berning 4 phase setup. I will only be putting the audio AC pk-pk thru them, with a 150 V B+ DC (minimum voltage for plates) supply in series with the rectifiers. Don't have to put 2x peak plate voltage thru the xfmrs that way, although I guess my cores could handle it anyway.
For measurements, I think a simple test would be to just put delta DC on the tube side and load the output side to determine output Z.
Then I will test AC dist. (SSC conv.) with two SS audio amps to supply the "tube" signals (will have to bias them up +50V to allow HV diode multiplexing and reduce the B+/2 drive ) I guess the Berning conv. can be tested similarly. Can't really exercise the full voltage range that way however. Could try some DC power supplies and just measure loaded I/O voltages with the DVM if the dist. is large enough to see that way. Or run a SS amp backwards thru a tube P-P OT xfmr to simulate tubes. (maybe have to subtract dist. spectra then, won't work for freq. response though.)
Don
Yes, just overlapped phases, 50+% duty. But only two phases required per tube/converter then instead of 4 for 100% coverage.
"multiphase half bridge using multiple windings, on the load side. I guess that's what you're calling the SSC/Matrix?"
Yes, I managed to get rid of the diode (for blocking the reverse diode in the Mosfet) in series with the Mosfets that way.
"Do you care to share your operating parameters? What core material, cross section, mag path length, turns, frequency (you said 250 KHz?) Bmax (I guess could be calculated from the others)?
The larger cores I'm using for the differential P-P xfmr to the tube plates. It has two interleaved 32 turn, center tapped windings for the HV windings (using 600V insulated wire) that just fit on one layer. It uses two stacked Fair-Rite cores P/N 5977004203
( 2.7 in. OD, 1.5 in. ID, .5 in. thk per core) for a combined area of 3.87 cm2. Material #77 is a power matl. with Ui = 2000, Bmax = 4600, 100 OHm CM resistivity (measures 1000 Ohm across core).
Measuring across 32 turns gives 8.475 mH and .18 Ohm (on L/R meter) and between windings 200 pF (I think that means the tubes will see 1/4 of that between them, most of the interwinding cap. on both cores is just across the HF drive. Typical conventional OTs are more like 500 pF plate to plate). I'm assuming max acceptable flux density of 1600 G at 60 KHz, 1200 G at 125 KHZ and 800 G at 250 KHz.
Working this out at 125 KHz, 1200 G gives +/- 1009 V pk -pk on each 16 turn section, so that would be the maximum swing allowed for a tube plate. This clearly is well above what I need. So actual operation will be more like 400 G at 125 KHz. At 250 KHz, that would drop to 200 G being used. Should be nice low core losses.
The B+/2 cores are some unknown units I got from Electronic Goldmine. They measure 55 mm OD, 32 mm ID, 18 mm thk. Have a black coating on them. Area = 2.07 CM2, le = 136.6 mm. I measured .475 mH for 8 turns giving Ui about 3880. These have two interleaved 25 turn windings on them, filling a single layer.
Measure at 4.73 mH and .11 Ohm on L/R meter. Capacitance between windings is 113 pF.
Working out B+/2 at 1200 G and 125 KHz gives 310.5 Vrms or 439 Vpk. So max B+ would be 878 V which is well above what I plan to use, maybe up to 500V tops. So working Bmax would be around 683 G, 1/2 that at 250 KHz.
I plan to put together two more of the largest core units (identical to above) so as to have 4 of them available to do a Berning 4 phase setup. I will only be putting the audio AC pk-pk thru them, with a 150 V B+ DC (minimum voltage for plates) supply in series with the rectifiers. Don't have to put 2x peak plate voltage thru the xfmrs that way, although I guess my cores could handle it anyway.
For measurements, I think a simple test would be to just put delta DC on the tube side and load the output side to determine output Z.
Then I will test AC dist. (SSC conv.) with two SS audio amps to supply the "tube" signals (will have to bias them up +50V to allow HV diode multiplexing and reduce the B+/2 drive ) I guess the Berning conv. can be tested similarly. Can't really exercise the full voltage range that way however. Could try some DC power supplies and just measure loaded I/O voltages with the DVM if the dist. is large enough to see that way. Or run a SS amp backwards thru a tube P-P OT xfmr to simulate tubes. (maybe have to subtract dist. spectra then, won't work for freq. response though.)
Don
Update and new unexpected design issue
Well, I goofed on the center tapped plate xfmrs (SSC design) in a few ways, so I am rewinding them. Biggest goof, is not enough inductive reactance. These big cores make it too easy to meet the voltage specs with too few turns for the inductance spec.
As wound, I was going to use 2 turns for the primary (welding cable!) to get a 2000 Ohm p to p primary (500 Ohm per tube, for some horiz. type outputs), but this would give only 2.12 mH per tube, which translates to Xl = 1665 Ohm at 125 KHz, which is too low I think.
I say I think, because this magnetizing current is looking like DC current to the tubes due to the switching diodes demodulating it. (but see below) This seems to be just waisting power in the tubes. In any case it seems like a good idea to increase the primary inductance some, so I am going to rewind the cores with smaller wire, more turns. (this will further lower flux levels in the cores too, these big cores are really going to be loafing)
It's interesting that these switching designs not only get rid of the hysteresis, but even the primary inductance. But we still are left with DC current side effects.
This magnetizing current issue is different for the Berning design, since the power for the magnetizing current there comes from the LV side (burden on the Mosfets, not the tubes). An odd situation appears to occur in the Berning scheme though. The magnetizing currents are maximum there when the audio signal is zero or tiny. This means that reverse currents occur thru the Mosfets. So the reverse substrate diodes had better be matched or this stuff will cause effects in the audio during the quietest signal times. I'm wondering how well these substrate diodes are matched or controlled in production. I'm also a little uncomfortable now with the output signal having to jump from froward conduction thru the Mosfet to the backward threshold voltage of the diode, all during the quietest audio passages. Hmmm.
In the SSC case, the magetizing currents are proportional to audio signal level, so should never require backwards conduction thru the HV diodes, assuming they are smaller than the signal current. (more on this below) Providing that is the case, then the 90 degree phase shifted HF magnetizing current will end up adding and subtracting from the signal current during a clock cycle, with no net affect on the DC current. So no power will lost or waisted in the tubes fortunately.
But this brings up a new design issue:
It looks like the HV plate side primary reactance (at the switching freq.) must be at least greater than the reflected load resistance in all cases to keep magnetizing currents less than signal currents (or some reverse flyback currents will develop thru the opposite side HV diodes I guess). Bad speaker resonances could be an issue here. So I better get my xfmr. primary Xl up some more.
Also, I am re-winding the center tapped plate toroids as quadrafilar (there are two center tapped primaries, so 4 plate windings), not bifilar, this time. Each tube plate winding needs to cover the whole core.
This magnetizing current issue brings me back to the Cuckoo converter, which I was ready to forget about. It keeps the magnetizing currents out of the tubes, just like the Berning design does. And maybe even out of the Mosfets, due to the reset windings cancelling the magn. current for the opposite phase.
However, after winding some toroids, I have to say that # of xfmrs is really the criteria one wants to minimize, and the Cuckoo always requires 4 of them, and all of them are specifiic to the impedance ratio in case you want to change that. The SSC can get by on a single xfmr. if you really want to push things to the limit.
Don
note: I have decided that for testing purposes I don't really need to do the whole 4 phase setup, 2 phase should be sufficient. (4 phase is just two of these in parallel, so output Z will be 1/2, and input shunt C on the tubes will be twice) Will save some winding efforts till I feel comfortable with the final design. (although the controller will be full 4 phase capable from the start)
Well, I goofed on the center tapped plate xfmrs (SSC design) in a few ways, so I am rewinding them. Biggest goof, is not enough inductive reactance. These big cores make it too easy to meet the voltage specs with too few turns for the inductance spec.
As wound, I was going to use 2 turns for the primary (welding cable!) to get a 2000 Ohm p to p primary (500 Ohm per tube, for some horiz. type outputs), but this would give only 2.12 mH per tube, which translates to Xl = 1665 Ohm at 125 KHz, which is too low I think.
I say I think, because this magnetizing current is looking like DC current to the tubes due to the switching diodes demodulating it. (but see below) This seems to be just waisting power in the tubes. In any case it seems like a good idea to increase the primary inductance some, so I am going to rewind the cores with smaller wire, more turns. (this will further lower flux levels in the cores too, these big cores are really going to be loafing)
It's interesting that these switching designs not only get rid of the hysteresis, but even the primary inductance. But we still are left with DC current side effects.
This magnetizing current issue is different for the Berning design, since the power for the magnetizing current there comes from the LV side (burden on the Mosfets, not the tubes). An odd situation appears to occur in the Berning scheme though. The magnetizing currents are maximum there when the audio signal is zero or tiny. This means that reverse currents occur thru the Mosfets. So the reverse substrate diodes had better be matched or this stuff will cause effects in the audio during the quietest signal times. I'm wondering how well these substrate diodes are matched or controlled in production. I'm also a little uncomfortable now with the output signal having to jump from froward conduction thru the Mosfet to the backward threshold voltage of the diode, all during the quietest audio passages. Hmmm.
In the SSC case, the magetizing currents are proportional to audio signal level, so should never require backwards conduction thru the HV diodes, assuming they are smaller than the signal current. (more on this below) Providing that is the case, then the 90 degree phase shifted HF magnetizing current will end up adding and subtracting from the signal current during a clock cycle, with no net affect on the DC current. So no power will lost or waisted in the tubes fortunately.
But this brings up a new design issue:
It looks like the HV plate side primary reactance (at the switching freq.) must be at least greater than the reflected load resistance in all cases to keep magnetizing currents less than signal currents (or some reverse flyback currents will develop thru the opposite side HV diodes I guess). Bad speaker resonances could be an issue here. So I better get my xfmr. primary Xl up some more.
Also, I am re-winding the center tapped plate toroids as quadrafilar (there are two center tapped primaries, so 4 plate windings), not bifilar, this time. Each tube plate winding needs to cover the whole core.
This magnetizing current issue brings me back to the Cuckoo converter, which I was ready to forget about. It keeps the magnetizing currents out of the tubes, just like the Berning design does. And maybe even out of the Mosfets, due to the reset windings cancelling the magn. current for the opposite phase.
However, after winding some toroids, I have to say that # of xfmrs is really the criteria one wants to minimize, and the Cuckoo always requires 4 of them, and all of them are specifiic to the impedance ratio in case you want to change that. The SSC can get by on a single xfmr. if you really want to push things to the limit.
Don
note: I have decided that for testing purposes I don't really need to do the whole 4 phase setup, 2 phase should be sufficient. (4 phase is just two of these in parallel, so output Z will be 1/2, and input shunt C on the tubes will be twice) Will save some winding efforts till I feel comfortable with the final design. (although the controller will be full 4 phase capable from the start)
"I'm also a little uncomfortable now with the output signal having to jump from forward conduction thru the Mosfet to the backward threshold voltage of the diode, all during the quietest audio passages. Hmmm."
There will be a discontinuity in the output signal when the audio signal becomes large enough to eliminate the reverse magnetizing current thru the Mosfet substrate diodes. This is really NOT SO GOOD! If both tube converters go thru this transition at exactly the same time, then it would just cancel out in the output, but how likely are these to be that well matched. Probably one side goes thru transition then the other. The winding reactance of the xfmr LV windings will help leap the gap voltage-wise, so this is not going to be a 1st order problem. But it certainly would get my attention enough to put a spectrum analyzer on the output with an audio signal level just at the crossing threshold.
This is something that needs to be checked for the Berning type design and the Matrix design. I don't think this issue comes up for the SSC design except in the form of the primary reactance having to be larger than the load resistance (as mentioned in the above post).
------------------------------------------------------
Since we are now into design details, for any of these P-P converters ( Berning, SSC), DC magnetization of the cores is an issue that should be brought up. Normally, for switching power supplies, the use of an H bridge driver (used in the Berning design) is accompanied with a DC blocking capacitor in series with the winding. This keeps DC core balance. But a glance at the Berning schematics shows this capacitor is not used. Probably for "Audio" reasons, its not wanted in the audio path.
For the center tapped driven windings, that reduce the Mosfets from 4 down to 2 (that I have been suggesting), this capacitor is not an option anyway. So we are left in either case with the requirement that the Mosfets ON resistances must be matched (better than 10% recommended). Whether production lots of Mosfets are sufficiently matched is something that will need to be checked. One measures the match with a current probe on the xfmr winding leads. Opposite phase (alternating) current pulses need to be checked for match.
Don
error fix:
In post 109 above:
"As wound, I was going to use 2 turns for the primary (welding cable!)..."
this should read:
"As wound, I was going to use 2 turns for the LV winding (welding cable!)..."
There will be a discontinuity in the output signal when the audio signal becomes large enough to eliminate the reverse magnetizing current thru the Mosfet substrate diodes. This is really NOT SO GOOD! If both tube converters go thru this transition at exactly the same time, then it would just cancel out in the output, but how likely are these to be that well matched. Probably one side goes thru transition then the other. The winding reactance of the xfmr LV windings will help leap the gap voltage-wise, so this is not going to be a 1st order problem. But it certainly would get my attention enough to put a spectrum analyzer on the output with an audio signal level just at the crossing threshold.
This is something that needs to be checked for the Berning type design and the Matrix design. I don't think this issue comes up for the SSC design except in the form of the primary reactance having to be larger than the load resistance (as mentioned in the above post).
------------------------------------------------------
Since we are now into design details, for any of these P-P converters ( Berning, SSC), DC magnetization of the cores is an issue that should be brought up. Normally, for switching power supplies, the use of an H bridge driver (used in the Berning design) is accompanied with a DC blocking capacitor in series with the winding. This keeps DC core balance. But a glance at the Berning schematics shows this capacitor is not used. Probably for "Audio" reasons, its not wanted in the audio path.
For the center tapped driven windings, that reduce the Mosfets from 4 down to 2 (that I have been suggesting), this capacitor is not an option anyway. So we are left in either case with the requirement that the Mosfets ON resistances must be matched (better than 10% recommended). Whether production lots of Mosfets are sufficiently matched is something that will need to be checked. One measures the match with a current probe on the xfmr winding leads. Opposite phase (alternating) current pulses need to be checked for match.
Don
error fix:
In post 109 above:
"As wound, I was going to use 2 turns for the primary (welding cable!)..."
this should read:
"As wound, I was going to use 2 turns for the LV winding (welding cable!)..."
Hi Don,
Wow, thanks for all the details. It sounds like you have a good project going there.
I was a little shocked to read >1000V over 16 turns, but you figured it out soon enough!
You get fooled into thinking with big cores you can do less turns, but even with forward topologies you still need a certain Lpri...
What power level are you designing for? audio voltage, anode current?
Now my approach and that used in a number of more sophisticated H bridge converters would be to design for more Lpri than needed and then gap it down in inductance. This needs more turns though.
The gap then allows plenty of tolerance for mismatched Rdson across the diagonals of the bridge, solving that problem nicely.
But that's all harder to do with toroids. Are there "gapped" ferrite toroids as in the permalloy ones?
I like U-cores because they can easily be re-gapped and it's easy to wind low capacitance sectioned HV windings and achieve very good primary-secondary isolation. Easy to accomodate more turns.
Another thing to look at for under-power-utilized cores is where the eddy current losses are. Sometimes more magnetic material can hurt.
For the secondary it would probably need litz wire to reduce the heat losses at the higher power levels. At 50-100 watts bi- or quad-filar should be pretty good.
Looks like you're making great progress, and have a good test plan!
Michael
Wow, thanks for all the details. It sounds like you have a good project going there.
I was a little shocked to read >1000V over 16 turns, but you figured it out soon enough!
You get fooled into thinking with big cores you can do less turns, but even with forward topologies you still need a certain Lpri...
What power level are you designing for? audio voltage, anode current?
Now my approach and that used in a number of more sophisticated H bridge converters would be to design for more Lpri than needed and then gap it down in inductance. This needs more turns though.
The gap then allows plenty of tolerance for mismatched Rdson across the diagonals of the bridge, solving that problem nicely.
But that's all harder to do with toroids. Are there "gapped" ferrite toroids as in the permalloy ones?
I like U-cores because they can easily be re-gapped and it's easy to wind low capacitance sectioned HV windings and achieve very good primary-secondary isolation. Easy to accomodate more turns.
Another thing to look at for under-power-utilized cores is where the eddy current losses are. Sometimes more magnetic material can hurt.
For the secondary it would probably need litz wire to reduce the heat losses at the higher power levels. At 50-100 watts bi- or quad-filar should be pretty good.
Looks like you're making great progress, and have a good test plan!
Michael
Hi Michael,
I'm looking at doing 50 to 100 tube watts using some horizontal output tubes, probably around 400V B+. These big doubled up cores are probably big enough to do 1500 Watts in a switcher. But I just want to get the HV primary down to a single layer though, so core area helps. Looking do-able.
I also just want something available for easy construction of any primary impedance OT for DIY amps. Might be worthwhile to sell the controller PC boards on the WWW to other DIYers too, don't know if there is much market though. And could be a patent issue too, since the same controller will run a Berning type design.
Many tube types won't be interested in some SS switching gizmo, and the learning curve to apply this stuff is steep too. Might have to sell completed OTs of specific primary Z, then you have all the stocking of ten diff. version problems..... If one tried to sell complete amplifiers, like Berning, you end up with a huge new set of issues, would finish off my DIY activities......
(I also have some other BIG honker ferrite toroid cores I got a while back to try doing an audio frequency OT, with no switching. They are 6 inches in diameter, 1 inch thick. I was going to stack 4 of them up for an OT. But it still takes 1000s of turns and only puts out 50 W at 20 Hz. Nice heavy paper weights now. Maybe I'll build a particle accelerator or something someday!)
Someone also mentioned on the forum that ferrite OTs sounded bad, but I doubt whether anyone has really constructed one with sufficient core material, it needs to be at least four times bigger than a laminated steel one. The constant Ui down to low flux should be similar to permalloy too.
I found a box of big ferrite "U-I laminations" at the local junkyard a few years back that look more promising for winding 1000's of turns. They are machined 1/8 inch thick pieces, with mylar film on one side for additional insulation. Were used for RF ionizing of gas in quartz tubes for semiconductor processing. Hate to say how much they cost new, I sure won't be buying any new ones. But I got enough for two OTs.
I haven't actually seen any gapped toroids around, but I think the manufacturers mention they can do them by cutting a gap or outright cutting them in half. The gapped ones would be limited by the diamond saw blade thickness though, and not adjustable. I usually see E cores or pot cores (or U ones too) offered with machined gaps or film inserts for small gaps.
Berning uses pot cores, which are easy to wind. No idea if they have gapping. But I would think for this application one would want to avoid the gap if possible and go for manually selecting the Mosfets to a very good matching level. (At $30,000, I think Berning can afford to match a few Mosfets) No gap should give the least leakage inductance, which would lead to the best intrinsic damping factor (ie. low Zout of the conv.). And matching will keep HF residuals out of the audio output. Toroids being the best for low leakage L.
The gapped ferrite E core I mentioned earlier:
http://www.goldmine-elec-products.c...sp?number=G8912
has a 1/16 inch gap, measures 20 uH for 9 turns. I got some to try for the resonant switched capacitor converter. Too big a gap though for the SSC or Berning convs. They also should work well for a switched CCS. (most likely were DC output inductor cores for some switcher) I've seen some others around on the WWW with .1 mm gaps.
On the Litz wire, I guess you mean for the LV (low voltage) side? Since the wire thickness is greatest there. I have some Litz stuff around in storage, but I would have to make a long trip to get it. Maybe in a few months.
Don
I'm looking at doing 50 to 100 tube watts using some horizontal output tubes, probably around 400V B+. These big doubled up cores are probably big enough to do 1500 Watts in a switcher. But I just want to get the HV primary down to a single layer though, so core area helps. Looking do-able.
I also just want something available for easy construction of any primary impedance OT for DIY amps. Might be worthwhile to sell the controller PC boards on the WWW to other DIYers too, don't know if there is much market though. And could be a patent issue too, since the same controller will run a Berning type design.
Many tube types won't be interested in some SS switching gizmo, and the learning curve to apply this stuff is steep too. Might have to sell completed OTs of specific primary Z, then you have all the stocking of ten diff. version problems..... If one tried to sell complete amplifiers, like Berning, you end up with a huge new set of issues, would finish off my DIY activities......
(I also have some other BIG honker ferrite toroid cores I got a while back to try doing an audio frequency OT, with no switching. They are 6 inches in diameter, 1 inch thick. I was going to stack 4 of them up for an OT. But it still takes 1000s of turns and only puts out 50 W at 20 Hz. Nice heavy paper weights now. Maybe I'll build a particle accelerator or something someday!)
Someone also mentioned on the forum that ferrite OTs sounded bad, but I doubt whether anyone has really constructed one with sufficient core material, it needs to be at least four times bigger than a laminated steel one. The constant Ui down to low flux should be similar to permalloy too.
I found a box of big ferrite "U-I laminations" at the local junkyard a few years back that look more promising for winding 1000's of turns. They are machined 1/8 inch thick pieces, with mylar film on one side for additional insulation. Were used for RF ionizing of gas in quartz tubes for semiconductor processing. Hate to say how much they cost new, I sure won't be buying any new ones. But I got enough for two OTs.
I haven't actually seen any gapped toroids around, but I think the manufacturers mention they can do them by cutting a gap or outright cutting them in half. The gapped ones would be limited by the diamond saw blade thickness though, and not adjustable. I usually see E cores or pot cores (or U ones too) offered with machined gaps or film inserts for small gaps.
Berning uses pot cores, which are easy to wind. No idea if they have gapping. But I would think for this application one would want to avoid the gap if possible and go for manually selecting the Mosfets to a very good matching level. (At $30,000, I think Berning can afford to match a few Mosfets) No gap should give the least leakage inductance, which would lead to the best intrinsic damping factor (ie. low Zout of the conv.). And matching will keep HF residuals out of the audio output. Toroids being the best for low leakage L.
The gapped ferrite E core I mentioned earlier:
http://www.goldmine-elec-products.c...sp?number=G8912
has a 1/16 inch gap, measures 20 uH for 9 turns. I got some to try for the resonant switched capacitor converter. Too big a gap though for the SSC or Berning convs. They also should work well for a switched CCS. (most likely were DC output inductor cores for some switcher) I've seen some others around on the WWW with .1 mm gaps.
On the Litz wire, I guess you mean for the LV (low voltage) side? Since the wire thickness is greatest there. I have some Litz stuff around in storage, but I would have to make a long trip to get it. Maybe in a few months.
Don
I am favoring Berning for the top half and Anti-Triode for the bottom.
With the Berning end, pehaps a Tubelab SuperSE style beta mirror.
I will probably mirror-up to a small twin triode. 6N30Pi looks good.
Use the other triode half in the preamp.
But I might prefer an NPN Base-Emitter junction between Cathode
and RSense. Tie my collector off to the plate, or cascode up to meet
the plate with a power mosfet sandwiched inbetween to take the
thermal burden... I would most certainly follow Tubelab's lead to
superdrive my grid into A2. I am not exactly sure why Tubelab had
chosen topology requiring PNP? Merely whatever parts at hand???
I am not having luck with my Transonar... I can build a decent 1:1
resonant around 50Khz with the junk on my desk. But I don't have
convenient way to operate at much higher frequency, nor construct
Rosen or other Step-up types... Unless I Parallel-Series more than
one of them ~ Balun style... But with my frequency limit, even that
is not realistic. Might have to resort to a generic pot core switching
transformer for now.
I have some Class-D chips, but they all have signal processing
features such as dynamic compression I can't predictably defeat
with simple negative feedback... And are meant to drive a bridged
load anyway. Probably go with discrete switching Anti-Triode so I
know and control exactly whats is going on. Or keep it all a simple
linear Class-A circuit for now.
Wiped out my other schematic software (SwitchCadIII) to install
Cadence, but I can't seem to get it working yet. May not have
enough memory to properly run (or even install) Cadence on my
old laptop... Also tempted to try TI's free TINA which is supposed
to have a virtual scope and spectrum analyzer. I wasn't having
any special problems with SwitchCadIII, I liked it fine. But my work
wants me to learn Cadence reasonably soon. So I was trying to
force myself that way. Never done any Spice sims before, so that
stuff is gonna be new no matter which package I try to learn.
With the Berning end, pehaps a Tubelab SuperSE style beta mirror.
I will probably mirror-up to a small twin triode. 6N30Pi looks good.
Use the other triode half in the preamp.
But I might prefer an NPN Base-Emitter junction between Cathode
and RSense. Tie my collector off to the plate, or cascode up to meet
the plate with a power mosfet sandwiched inbetween to take the
thermal burden... I would most certainly follow Tubelab's lead to
superdrive my grid into A2. I am not exactly sure why Tubelab had
chosen topology requiring PNP? Merely whatever parts at hand???
I am not having luck with my Transonar... I can build a decent 1:1
resonant around 50Khz with the junk on my desk. But I don't have
convenient way to operate at much higher frequency, nor construct
Rosen or other Step-up types... Unless I Parallel-Series more than
one of them ~ Balun style... But with my frequency limit, even that
is not realistic. Might have to resort to a generic pot core switching
transformer for now.
I have some Class-D chips, but they all have signal processing
features such as dynamic compression I can't predictably defeat
with simple negative feedback... And are meant to drive a bridged
load anyway. Probably go with discrete switching Anti-Triode so I
know and control exactly whats is going on. Or keep it all a simple
linear Class-A circuit for now.
Wiped out my other schematic software (SwitchCadIII) to install
Cadence, but I can't seem to get it working yet. May not have
enough memory to properly run (or even install) Cadence on my
old laptop... Also tempted to try TI's free TINA which is supposed
to have a virtual scope and spectrum analyzer. I wasn't having
any special problems with SwitchCadIII, I liked it fine. But my work
wants me to learn Cadence reasonably soon. So I was trying to
force myself that way. Never done any Spice sims before, so that
stuff is gonna be new no matter which package I try to learn.
Hi KenPeter,
"Wake up and smell the diode"
Arrgh, you're right. Sorry I messed up on that one Ken. I seem to have forgotten about the inverted AC on the other plate to plate winding. Late night posts getting to me I guess. So the center tap DOES indeed have to switch between + and - full B+ for the SSC converter front end.
Darn, now I gotta fix the transformers too. The extra voltage swing makes distributed capacitance a much bigger issue now, so I'll have to separate the opposite phase center tap windings completely on the core.
Anatoliy's idea of rectifying with the tubes is looking more attractive now. If we doubled up the tubes, two per center tapped plate winding, and multiplexed by switching on the screen voltages for one pair or the other. Hmmm, we don't even need two center tapped windings, just wire the g1's in reverse on one pair. With half duty cycle on the tubes, they could be biased up for double the normal dissipation, might be more linear that way.
------------------------
Ahh, so you got one of those transonars. Any way to checkout it's bandwidth? Also Vout versus Vin amplitude linearity?
Don
"Wake up and smell the diode"
Arrgh, you're right. Sorry I messed up on that one Ken. I seem to have forgotten about the inverted AC on the other plate to plate winding. Late night posts getting to me I guess. So the center tap DOES indeed have to switch between + and - full B+ for the SSC converter front end.
Darn, now I gotta fix the transformers too. The extra voltage swing makes distributed capacitance a much bigger issue now, so I'll have to separate the opposite phase center tap windings completely on the core.
Anatoliy's idea of rectifying with the tubes is looking more attractive now. If we doubled up the tubes, two per center tapped plate winding, and multiplexed by switching on the screen voltages for one pair or the other. Hmmm, we don't even need two center tapped windings, just wire the g1's in reverse on one pair. With half duty cycle on the tubes, they could be biased up for double the normal dissipation, might be more linear that way.
------------------------
Ahh, so you got one of those transonars. Any way to checkout it's bandwidth? Also Vout versus Vin amplitude linearity?
Don
Unity - Coupled Switchless Front End
Well, I think I have fixed the distributed capacitance issue for the xfmrs. And no extra turns even.
All the xfmrs are wound as single layer windings on big cores, so that the HF voltage is distributed around the circumference of the core. The audio signal similarly is distributed around for low capacitance on T1.
(I'll use the biggest core for the T1 audio and two medium size cores for the T2,T3 B+ supplies.)
Its nice to get down to just one plate to plate winding on T1, since now I can devote twice as many turns to it in a single layer to get low magnetizing current (at HF) for the tubes. The T2,T3 xfmrs impose their magnetizing current on the Mosfets, so not as critical on turns.
Note that the ground at the center tap of T1 is the correct ground point for the V1, V2 grid drives. (The scheme is similar to a Circlotron, but using HF B+, DC B+ at the tubes and AC HF B+ for T1)
One is not stuck with 50% CFB drive here ( with high grid drive requirements). If the driver tubes are pentodes, and have bootstrapped load resistors, then the output stage appears as common cathode effectively. Only if the drivers are triode-like does the drive requirement become like 50% CFB.
With the ultra low distributed capacitance available this way, there is some hope of eliminating the HF magnetizing currents in the tubes altogether (for the 4 phase version) by placing a very small cap between the tube plates. This would be set to be resonant at twice the switching frequency with the T1 primary inductance. The magnetizing currents from one 2 phase setup, versus the other 2 phase setup, would then become 180 degree out of phase sine waves and cancel. (ie, the cap Xc would be nulling out the primare Xl) This requires VERY low distributed capacitance to achieve this. (Normally one worries about Cdist and Lleak setting the upper xfmr bandwidth, but here we are resonating with Lprimary which is MUCH bigger than Lleak) Can try anyway.
Don
Maybe I should call it Circlotron Switchless front end. I actually derived it with the McIntosh setup, then realized that half of it was redundant.
Well, I think I have fixed the distributed capacitance issue for the xfmrs. And no extra turns even.
All the xfmrs are wound as single layer windings on big cores, so that the HF voltage is distributed around the circumference of the core. The audio signal similarly is distributed around for low capacitance on T1.
(I'll use the biggest core for the T1 audio and two medium size cores for the T2,T3 B+ supplies.)
Its nice to get down to just one plate to plate winding on T1, since now I can devote twice as many turns to it in a single layer to get low magnetizing current (at HF) for the tubes. The T2,T3 xfmrs impose their magnetizing current on the Mosfets, so not as critical on turns.
Note that the ground at the center tap of T1 is the correct ground point for the V1, V2 grid drives. (The scheme is similar to a Circlotron, but using HF B+, DC B+ at the tubes and AC HF B+ for T1)
One is not stuck with 50% CFB drive here ( with high grid drive requirements). If the driver tubes are pentodes, and have bootstrapped load resistors, then the output stage appears as common cathode effectively. Only if the drivers are triode-like does the drive requirement become like 50% CFB.
With the ultra low distributed capacitance available this way, there is some hope of eliminating the HF magnetizing currents in the tubes altogether (for the 4 phase version) by placing a very small cap between the tube plates. This would be set to be resonant at twice the switching frequency with the T1 primary inductance. The magnetizing currents from one 2 phase setup, versus the other 2 phase setup, would then become 180 degree out of phase sine waves and cancel. (ie, the cap Xc would be nulling out the primare Xl) This requires VERY low distributed capacitance to achieve this. (Normally one worries about Cdist and Lleak setting the upper xfmr bandwidth, but here we are resonating with Lprimary which is MUCH bigger than Lleak) Can try anyway.
Don
Maybe I should call it Circlotron Switchless front end. I actually derived it with the McIntosh setup, then realized that half of it was redundant.
Attachments
Just for reference- Diode Bridge Demod
Here is the rest of the Switchless Converter, the diode bridge demodulator for the LV (load) side. Phases 3 and 4 would similarly connect to the load at points X and Y.
Ohh, on the Unity-coupled or Circlotron switchless front end above, the number of turns on the T1 xfmr (plate to plate xfmr) is actually 1/2 again of the previous dual CT winding version (beside the 1/2 gained from ditching one winding I already mentioned above) since the cathode and plate sides are using the same winding. (same as the normal Circlotron having half the turns of a P-P xfmr.)
So now I get 4 times as many turns available on a single layer for the primary versus before. This should bring the HF magnetizing current in the tubes down to an estimated less than 1% of the signal current now. Very nice. I'm happy with this now.
Don
Here is the rest of the Switchless Converter, the diode bridge demodulator for the LV (load) side. Phases 3 and 4 would similarly connect to the load at points X and Y.
Ohh, on the Unity-coupled or Circlotron switchless front end above, the number of turns on the T1 xfmr (plate to plate xfmr) is actually 1/2 again of the previous dual CT winding version (beside the 1/2 gained from ditching one winding I already mentioned above) since the cathode and plate sides are using the same winding. (same as the normal Circlotron having half the turns of a P-P xfmr.)
So now I get 4 times as many turns available on a single layer for the primary versus before. This should bring the HF magnetizing current in the tubes down to an estimated less than 1% of the signal current now. Very nice. I'm happy with this now.
Don
Attachments
Matrix Demod.
And here is the Matrix Demod. for the Switchless Circlotron front end. Can't really call it Switchless any more though, certainly are "real" switches in the audio path now.
This operates similarly as the Berning demod, I just had to matrix the signals around, by sumation of windings, to get the same voltages on the Mosfet switches. The T2 and T3 switched B+ supply xfmrs are now powered by these Mosfets for "free". (see next paragraph though)
And similarly to the Berning demod, I have some reservations about the magnetizing currents here (from T2 and T3 here) causing the substrate diodes in the Mosfets to conduct at low audio signal levels. (the magnetizing currents are 90 degree phase shifted with respect to the switching phase, and will cause reverse current to flow thru the substrate diodes when the audio signal current is not great enough to overrule it.)
Which then puts in a discontinuity at some signal level, when they quit conducting. (and they quit conducting at DIFFERENT times, since the audio signal affects top and bottom total current flows assymetrically, and in the Berning case too) It may be possible to null this magnetizing current out by placing a resonating cap across the T2 and T3 primaries and so stop substrate diode conduction.
Don
Although the Matrix Demod may SEEM simpler than the Switchless Demod, notice that TWO LV power supplies are now needed. Both schemes require a total of 4 Mosfets per 2 phases.
And here is the Matrix Demod. for the Switchless Circlotron front end. Can't really call it Switchless any more though, certainly are "real" switches in the audio path now.
This operates similarly as the Berning demod, I just had to matrix the signals around, by sumation of windings, to get the same voltages on the Mosfet switches. The T2 and T3 switched B+ supply xfmrs are now powered by these Mosfets for "free". (see next paragraph though)
And similarly to the Berning demod, I have some reservations about the magnetizing currents here (from T2 and T3 here) causing the substrate diodes in the Mosfets to conduct at low audio signal levels. (the magnetizing currents are 90 degree phase shifted with respect to the switching phase, and will cause reverse current to flow thru the substrate diodes when the audio signal current is not great enough to overrule it.)
Which then puts in a discontinuity at some signal level, when they quit conducting. (and they quit conducting at DIFFERENT times, since the audio signal affects top and bottom total current flows assymetrically, and in the Berning case too) It may be possible to null this magnetizing current out by placing a resonating cap across the T2 and T3 primaries and so stop substrate diode conduction.
Don
Although the Matrix Demod may SEEM simpler than the Switchless Demod, notice that TWO LV power supplies are now needed. Both schemes require a total of 4 Mosfets per 2 phases.
Attachments
Looks like one can push the magnetization current glitch out of the operating range by going to Class A operation of the tubes. Then the tube current (or T2,T3 current in our case) can always be greater than the magnetization current.
But for Class B or AB there will be some signal level where a tube cuts off, and the magnetization glitch then re-appears. With screen grid drive and high efficiency, like used in the Berning, this glitch would be near the audio zero crossing. Need to check this with an spec. analyzer to see if it is a problem or not.
Don
But for Class B or AB there will be some signal level where a tube cuts off, and the magnetization glitch then re-appears. With screen grid drive and high efficiency, like used in the Berning, this glitch would be near the audio zero crossing. Need to check this with an spec. analyzer to see if it is a problem or not.
Don
Glitch banished
Well, I think I have determined that the reverse magnetization current (it is 90 degrees out of phase with the load current, so it can at times reverse the drain current, if its greater than the load current) can NOT cause a glitch in the Mosfet demodulated converters.
This is because the Mosfet itself will conduct in the reverse drain voltage direction as long as it is still turned on by gate voltage. Which it is in this case. The Mosfet reverse conduction prevents the voltage from rising to the 0.6V threshold of the reverse substrate diode. So it would appear there is NO glitch problem for the Matrix converter or the Berning converter.
Mosfet four quadrant diagram attached.
An interesting observation by the way, is that the Mosfet, with reverse drain voltage, looks a lot like a triode. Might be worth checking out a few common types to see if there is any "triode" potential here.
Don
Well, I think I have determined that the reverse magnetization current (it is 90 degrees out of phase with the load current, so it can at times reverse the drain current, if its greater than the load current) can NOT cause a glitch in the Mosfet demodulated converters.
This is because the Mosfet itself will conduct in the reverse drain voltage direction as long as it is still turned on by gate voltage. Which it is in this case. The Mosfet reverse conduction prevents the voltage from rising to the 0.6V threshold of the reverse substrate diode. So it would appear there is NO glitch problem for the Matrix converter or the Berning converter.
Mosfet four quadrant diagram attached.
An interesting observation by the way, is that the Mosfet, with reverse drain voltage, looks a lot like a triode. Might be worth checking out a few common types to see if there is any "triode" potential here.
Don
Attachments
There is a little triode characteristic in the reverse drain curve but it's equivalent to a 2V anode voltage. It could be the basis of some kind of tube emulator. One would have to use a current multiplier or some such circuit to get the signal level up to usable levels but it might be an interesting analog model of a triode...
Michael
Michael
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.