i followed that, but tehre is no conlusion.
like you, i am investigating a 200+ SE. i am just lookin for a solution
like on tubecad.com i might stack SE's
if i find a lot of big watt se's at reasonable price i might try it
8 of those hammonds would be too much of a wallet drain!
Bas
like you, i am investigating a 200+ SE. i am just lookin for a solution
like on tubecad.com i might stack SE's
if i find a lot of big watt se's at reasonable price i might try it
8 of those hammonds would be too much of a wallet drain!
Bas
Hi Bas,
"what would go terribly wrong if you's just loas ane side of the primaries with currens. Keeping it far away from the signal path. Just a way to compensate the average current of the tube stage and eleiminating saturation and air gap."
This works fine using a current source to drive the compensator winding as Tubelab commented. But "Keeping it far away from the signal path." with a loosely coupled coil will not remove the requirement for a high impedance current source drive. This approach also suffers from poor efficiency since power dissipation is doubled but output power is the same from the triode.
Hi Joel,
"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."
Yes it is indeed a transformer loaded differential stage topology, I said so already. (if you haven't read the thread thru from the beginning, I will explain the details again)
The dissimilar devices is the key to what is going on here. The second device has LARGE transconductance so as to freeze the cathode voltage on the triode. It is used to produce a true complementary current to the triode current. Think of it as a complementary current mirror.
In a NORMAL matched device diff. amp. stage the second side does NOT produce true complementary currents to the input signal (for LARGE or POWER signal models where distortion is relevent) (here, I emphasize to the INPUT SIGNAL, obviously the two outputs must be complementary to each other due to the CCS tail, unless some parasitic capacitances to ground interferes). This mechanism involves the voltage swing on the common cathode or "tail" section. The voltage swing on the common cathode modifies the transfer curve of the 1st device, the triode in our case. This modification of the transfer curve is what allows the diff. amp. stage to cancel even distortion harmonics. This is still true in a single input diff. amp. stage as well. SETs are known for there even harmonic signature, and so OBVIOUSLY no normal diff. amp. stage could be used to copy a single ended triode current faithfully.
In this design, however, we effectively freeze the tail voltage swing by the use of the LARGE transconductance of the 2nd device. By picturing this circuit as an ordinary diff. amp. stage you are missing the different action here.
The present circuit is much more clearly pictured as a complementary current mirror stage attached to the triode's cathode. (current mirrors have very low impedance at their inputs like we have here) Carefull design of the complementary current mirror section is required here so as to preserve the original harmonic structure of the triode current. Hence the on-going concern about screen currents for a pentode mirror device or base currents for a bipolar device or gate/source capacitance for a Mosfet device. Again, I emphasize, if you look at this with the usual diff. amp. stage model you will miss all the relevance. The correct model to use here is the triode with a complementary current mirror attached.
We COULD, for example, move the complementary current mirror section to the plate side of the triode (between the top side of the xfmr's triode primary winding and B+, unfortunately requiring P channel devices to implement) and still use the complementary current derived to drive the second inverted phase xfmr winding. This would accomplish the same DC current balance in the P-P xfmr but would no longer have ANY resemblence to a diff. ampl. stage. Its function to the present circuit design is still identical however. Hope that helps!
Don
"what would go terribly wrong if you's just loas ane side of the primaries with currens. Keeping it far away from the signal path. Just a way to compensate the average current of the tube stage and eleiminating saturation and air gap."
This works fine using a current source to drive the compensator winding as Tubelab commented. But "Keeping it far away from the signal path." with a loosely coupled coil will not remove the requirement for a high impedance current source drive. This approach also suffers from poor efficiency since power dissipation is doubled but output power is the same from the triode.
Hi Joel,
"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."
Yes it is indeed a transformer loaded differential stage topology, I said so already. (if you haven't read the thread thru from the beginning, I will explain the details again)
The dissimilar devices is the key to what is going on here. The second device has LARGE transconductance so as to freeze the cathode voltage on the triode. It is used to produce a true complementary current to the triode current. Think of it as a complementary current mirror.
In a NORMAL matched device diff. amp. stage the second side does NOT produce true complementary currents to the input signal (for LARGE or POWER signal models where distortion is relevent) (here, I emphasize to the INPUT SIGNAL, obviously the two outputs must be complementary to each other due to the CCS tail, unless some parasitic capacitances to ground interferes). This mechanism involves the voltage swing on the common cathode or "tail" section. The voltage swing on the common cathode modifies the transfer curve of the 1st device, the triode in our case. This modification of the transfer curve is what allows the diff. amp. stage to cancel even distortion harmonics. This is still true in a single input diff. amp. stage as well. SETs are known for there even harmonic signature, and so OBVIOUSLY no normal diff. amp. stage could be used to copy a single ended triode current faithfully.
In this design, however, we effectively freeze the tail voltage swing by the use of the LARGE transconductance of the 2nd device. By picturing this circuit as an ordinary diff. amp. stage you are missing the different action here.
The present circuit is much more clearly pictured as a complementary current mirror stage attached to the triode's cathode. (current mirrors have very low impedance at their inputs like we have here) Carefull design of the complementary current mirror section is required here so as to preserve the original harmonic structure of the triode current. Hence the on-going concern about screen currents for a pentode mirror device or base currents for a bipolar device or gate/source capacitance for a Mosfet device. Again, I emphasize, if you look at this with the usual diff. amp. stage model you will miss all the relevance. The correct model to use here is the triode with a complementary current mirror attached.
We COULD, for example, move the complementary current mirror section to the plate side of the triode (between the top side of the xfmr's triode primary winding and B+, unfortunately requiring P channel devices to implement) and still use the complementary current derived to drive the second inverted phase xfmr winding. This would accomplish the same DC current balance in the P-P xfmr but would no longer have ANY resemblence to a diff. ampl. stage. Its function to the present circuit design is still identical however. Hope that helps!
Don
this is probably very stupid, but theoretically:
if you would wind a pp opt.
but make on primary winding suitable for a triode and one winding suitable to hook up to a 12v car battery in a way that itoutbalances the dc of the triode. A car battery can't be blamed for any sonic interferences nor signal amplification.
So i burn up a huge amount of energy...if i didn't care, would it matter?
Bas
if you would wind a pp opt.
but make on primary winding suitable for a triode and one winding suitable to hook up to a 12v car battery in a way that itoutbalances the dc of the triode. A car battery can't be blamed for any sonic interferences nor signal amplification.
So i burn up a huge amount of energy...if i didn't care, would it matter?
Bas
Hi Bas,
Ummm, if you use a car battery you will need something to limit the current, either a series resistor or the internal resistance of the winding. In either case, the resistance is effectively shunted across the output signal (by some turns ratio, but will come out effectively to something like the Rp of the triode at idle current to get DC balance in the xfmr.) This resistance is effectively a heavy dummy load on the triode's output, and will cut the output power by at least 50%. Working, yes, but not recommended.
Don
Ummm, if you use a car battery you will need something to limit the current, either a series resistor or the internal resistance of the winding. In either case, the resistance is effectively shunted across the output signal (by some turns ratio, but will come out effectively to something like the Rp of the triode at idle current to get DC balance in the xfmr.) This resistance is effectively a heavy dummy load on the triode's output, and will cut the output power by at least 50%. Working, yes, but not recommended.
Don
guessed that.
i was just wondering. is it really true that xfrmr limitations are so that a SE amp never gets past 200watt without paralelling transformers?
Could pp-xfrmrs be wound really big (1000w or so)?
Just filosofical here, don't blame anyone if not answereing these wanderings..
Bas
i was just wondering. is it really true that xfrmr limitations are so that a SE amp never gets past 200watt without paralelling transformers?
Could pp-xfrmrs be wound really big (1000w or so)?
Just filosofical here, don't blame anyone if not answereing these wanderings..
Bas
Hi Don,
You keep telling me not to look at this as a differential stage, yet that's exactly what it is - albeit a poor one, and not at all balanced, to spite the CCS tail load.
I don't doubt at all that you can set up the pentode to have the same standing DC current as the triode, but once we apply an AC signal to the triode's input, the two devices will behave quite differently, I assure you.
I don't understand what you mean by "freezing the cathode voltage". The CCS will not force balance with AC currents in this circuit as the two devices have completely different AC impedances and transfer functions.
It seems that you think the pentode will just sit there, happily passing a DC current only, and not reacting to an AC voltage present at its cathode. If you agree that it will react, than we have a real problem: two halves of a push-pull primary, driven by devices with completely different transfer characteristics.
Joel
You keep telling me not to look at this as a differential stage, yet that's exactly what it is - albeit a poor one, and not at all balanced, to spite the CCS tail load.
I don't doubt at all that you can set up the pentode to have the same standing DC current as the triode, but once we apply an AC signal to the triode's input, the two devices will behave quite differently, I assure you.
I don't understand what you mean by "freezing the cathode voltage". The CCS will not force balance with AC currents in this circuit as the two devices have completely different AC impedances and transfer functions.
It seems that you think the pentode will just sit there, happily passing a DC current only, and not reacting to an AC voltage present at its cathode. If you agree that it will react, than we have a real problem: two halves of a push-pull primary, driven by devices with completely different transfer characteristics.
Joel
With the pentode's grid held at a contant voltage (i.e., AC ground), what is the looking-in impedance at the pentode's cathode?
Hi Joel,
"I don't understand what you mean by "freezing the cathode voltage". "
With a LARGE transconductance 2nd device, the voltage on its input (source, emitter, or cathode for Mosfet, bipolar or pentode respectively) will hardly vary with respect to its AC grounded gate, base or grid respectively for any variation in current. Large transconductance means large current variation for little input voltage variation by definition of transconductance. This is particulary evident in the bipolar case where the base-emitter voltage sticks very closely to 0.6 volt no matter the current level.
Since the input or emitter ... in these case(s) is relatively fixed in voltage with respect to the AC grounded base, I am referring to it as being "frozen" AC wise to ground.
"The CCS will not force balance with AC currents in this circuit as the two devices have completely different AC impedances and transfer functions. "
The sum of the two devices "cathode" (or equiv.) currents must sum to the CCS tail current by Kirchoffs Laws.
So Idevice1 +Idevice2 = Iccs.
Since Iccs is constant, then any delta in Idevice1 must equal minus delta Idevice2.
delta Idevice1 = - delta Idevice2 ( currents)
This is what I mean by complementary AC currents.
The difficult part comes in when we look at the drain, collector or plate current of device2 being sent to the xfmr winding. If device2 is acting like an ideal Mosfet (no gate capacitance), then the drain current must equal the source current (grounded gate amplifier with no gate current lost, Kirchoffs Laws make drain current exactly equal to source current. No place else for the source current to go!). Pentode is similar to Mosfet if we can ignore screen current loss. A Bipolar needs recapture of the base current loss as explained earlier.
"It seems that you think the pentode will just sit there, happily passing a DC current only, and not reacting to an AC voltage present at its cathode. If you agree that it will react, than we have a real problem: two halves of a push-pull primary, driven by devices with completely different transfer characteristics."
The second device is acting as a grounded grid / grounded gate / or grounded base stage. So it is not just sitting there passing only DC current, it is an AC amplifer with hopefully accurate unity AC current gain and substantial voltage gain. The important factor is getting the unity current gain very accurate, hence the consideration of fixups to base current loss or screen current loss.
These devices (device 2 a Mosfet or bipolar or pentode) all are high impedance output devices, especially in grounded gate / base/ grid mode. So voltage gain of device 2 is set exclusively by the xfmr coupling from the device 1 (triode) side. Equal primary winding halves of the P-P xfmr mandate equal voltage gain on the device2 side as on the device1 (triode) side. The triode being the only one that can set the voltage gain, it controls voltage gain for both sides.
Since the P-P primaries are phase inverted with respect to each other, the complementary current from device 2 gets inverted and acts to effectively double the device 1 current (minus a minus). Now the complementary current appearing on the P-P winding (device 2 side) WILL affect the load impedance seen by the triode, so the turns ratio of the P-P xfmr primaries will need adjusting to 0.707 as many turns (compared to the original SET model) to get the same reflected load impedance on the triode plate.
Don
"I don't understand what you mean by "freezing the cathode voltage". "
With a LARGE transconductance 2nd device, the voltage on its input (source, emitter, or cathode for Mosfet, bipolar or pentode respectively) will hardly vary with respect to its AC grounded gate, base or grid respectively for any variation in current. Large transconductance means large current variation for little input voltage variation by definition of transconductance. This is particulary evident in the bipolar case where the base-emitter voltage sticks very closely to 0.6 volt no matter the current level.
Since the input or emitter ... in these case(s) is relatively fixed in voltage with respect to the AC grounded base, I am referring to it as being "frozen" AC wise to ground.
"The CCS will not force balance with AC currents in this circuit as the two devices have completely different AC impedances and transfer functions. "
The sum of the two devices "cathode" (or equiv.) currents must sum to the CCS tail current by Kirchoffs Laws.
So Idevice1 +Idevice2 = Iccs.
Since Iccs is constant, then any delta in Idevice1 must equal minus delta Idevice2.
delta Idevice1 = - delta Idevice2 ( currents)
This is what I mean by complementary AC currents.
The difficult part comes in when we look at the drain, collector or plate current of device2 being sent to the xfmr winding. If device2 is acting like an ideal Mosfet (no gate capacitance), then the drain current must equal the source current (grounded gate amplifier with no gate current lost, Kirchoffs Laws make drain current exactly equal to source current. No place else for the source current to go!). Pentode is similar to Mosfet if we can ignore screen current loss. A Bipolar needs recapture of the base current loss as explained earlier.
"It seems that you think the pentode will just sit there, happily passing a DC current only, and not reacting to an AC voltage present at its cathode. If you agree that it will react, than we have a real problem: two halves of a push-pull primary, driven by devices with completely different transfer characteristics."
The second device is acting as a grounded grid / grounded gate / or grounded base stage. So it is not just sitting there passing only DC current, it is an AC amplifer with hopefully accurate unity AC current gain and substantial voltage gain. The important factor is getting the unity current gain very accurate, hence the consideration of fixups to base current loss or screen current loss.
These devices (device 2 a Mosfet or bipolar or pentode) all are high impedance output devices, especially in grounded gate / base/ grid mode. So voltage gain of device 2 is set exclusively by the xfmr coupling from the device 1 (triode) side. Equal primary winding halves of the P-P xfmr mandate equal voltage gain on the device2 side as on the device1 (triode) side. The triode being the only one that can set the voltage gain, it controls voltage gain for both sides.
Since the P-P primaries are phase inverted with respect to each other, the complementary current from device 2 gets inverted and acts to effectively double the device 1 current (minus a minus). Now the complementary current appearing on the P-P winding (device 2 side) WILL affect the load impedance seen by the triode, so the turns ratio of the P-P xfmr primaries will need adjusting to 0.707 as many turns (compared to the original SET model) to get the same reflected load impedance on the triode plate.
Don
Hi Sy,
"With the pentode's grid held at a contant voltage (i.e., AC ground), what is the looking-in impedance at the pentode's cathode?"
Should be Z = 1 / gm of the device. Thats why we need a large gm device for the complementary current mirror device (device 2).
Don
"With the pentode's grid held at a contant voltage (i.e., AC ground), what is the looking-in impedance at the pentode's cathode?"
Should be Z = 1 / gm of the device. Thats why we need a large gm device for the complementary current mirror device (device 2).
Don
A true SE transformer in the 100 watt plus range that covers the complete audiophile frequency range (20Hz to 50HKz) is very difficult to do. Getting enough primary inductance (with a gapped core) for good LF response usually requires a lot of wire and a lot of iron. This generates a lot of winding capacitance and leakage inductance, making good HF response difficult.
The same restrictions apply to a P-P transformer, but the gapped core requirement goes away, allowing for much higher power levels before the same limitations begin to restrict the bandwidth to less than the complete audiophile range.
1000 watt plus P-P transformers are possible, and 5000 watt transformers were commonly used for modulation in AM radio transmitters, although they didn't exactly have audiophile specs.
If you really want to build a 200 watt plus SE amp, my first inclination would be something like SRPP using 833A's or A 4-400 CCS on top of an 833A. Capacitor couple the whole thing to a big P-P transformer. I plan to go down this path, on a smaller scale. I have an amp working with a triode wired KT-88 on the bottom and a 6LW6 based CCS on the top, cap coupled to a UTC LS-33 P-P transformer. It sounds pretty good. A lot of this was covered in this thread:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=67437
Next: An 813 based CCS on top of an 845. I don't think that I will build anything bigger along these lines, but you never know.
The ultimate design is a swithcmode CCS that is modulated by the audio signal such that it supplies the plate of a triode with exactly the voltage that it needs at any given instant. Efficiency of the power supply could be in the 90+% region. This would make an efficient triode SE amp that uses a P-P transformer. Anyone know how to build it? I have built a similar design for a radio transmitter. It worked well, but was all solid state and ran on 28 volts. 1KV would be a challenge.
The same restrictions apply to a P-P transformer, but the gapped core requirement goes away, allowing for much higher power levels before the same limitations begin to restrict the bandwidth to less than the complete audiophile range.
1000 watt plus P-P transformers are possible, and 5000 watt transformers were commonly used for modulation in AM radio transmitters, although they didn't exactly have audiophile specs.
If you really want to build a 200 watt plus SE amp, my first inclination would be something like SRPP using 833A's or A 4-400 CCS on top of an 833A. Capacitor couple the whole thing to a big P-P transformer. I plan to go down this path, on a smaller scale. I have an amp working with a triode wired KT-88 on the bottom and a 6LW6 based CCS on the top, cap coupled to a UTC LS-33 P-P transformer. It sounds pretty good. A lot of this was covered in this thread:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=67437
Next: An 813 based CCS on top of an 845. I don't think that I will build anything bigger along these lines, but you never know.
The ultimate design is a swithcmode CCS that is modulated by the audio signal such that it supplies the plate of a triode with exactly the voltage that it needs at any given instant. Efficiency of the power supply could be in the 90+% region. This would make an efficient triode SE amp that uses a P-P transformer. Anyone know how to build it? I have built a similar design for a radio transmitter. It worked well, but was all solid state and ran on 28 volts. 1KV would be a challenge.
smoking-amp said:
That was actually a Socratic question for Joel, sorry.
I think what you meant to do was bypass the screen to the cathode, not ground.
Joel said:
I don't think anyone claimed it is SE, the point was to balance the transformer, keeping the harmonic structure generated by a triode, and not using simple DC current balance, in order to make at least part of the wasted power useful. The fact that in a differential amplifier, the leg that has it's input grounded can be a different device, while still preserving the current symetry was used as an advantage. What makes it interesting is that having the 'pentode' device have MUCH larger gm than the triode, splits the differential voltage at the inputs unevenly between the two amplifying devices and the tail. If the 'pentode' device had infinite gm, ALL the differential input voltage (between 'grids' of both devices) would appear appear as the input to the triode only, yet the whole thing would still behave as a differential amplifier, but the input to output transfer characteristic would be dictated by the triode only.
What you should concentrate is exactly the easymetric nature of this differential amp. Of course, in order to get a higher gm difference, you not only get to use a high gm 'pentode' device, you can also shift things in your favor by using a low gm triode (although drive requirements get to be more difficult).
They are all SE schematics. The point is that an ordinary push pull type transformer can be used for SE if you keep the DC out of it. I used the entire winding for my experiments, the CT was not connected. That is the idea behind parafeed, but parafeed has its own requirements because there are two inductances (one being the OPT) connected together by a capacitor. This usually requires a matched choke - transformer set to avoid resonant effects in the audio band. We replaced the parafeed choke with a current source.
In fact once you remove the DC component completely there are several possibilities for output transformers. Even some power transformers can be used for relatively low powered SE amps. Toroidal mains transformers sound reasonably good, although you need a 50 VA transformer to get good bass on an 8 watt amp.
The point of this thread is to let the DC flow through the transformer, just cancel out the magnetic field with an equal but opposite current through the other half of the secondary, just as a push pull amp does. Both approaches yield the same result, although each may have different advantages and drawbacks.
In fact once you remove the DC component completely there are several possibilities for output transformers. Even some power transformers can be used for relatively low powered SE amps. Toroidal mains transformers sound reasonably good, although you need a 50 VA transformer to get good bass on an 8 watt amp.
The point of this thread is to let the DC flow through the transformer, just cancel out the magnetic field with an equal but opposite current through the other half of the secondary, just as a push pull amp does. Both approaches yield the same result, although each may have different advantages and drawbacks.
The several mentions of the bandwidth of output transformers dropping off with power got me interested in analyzing some models to see what goes wrong. I worked up the results for a 2x linear dimension scaling of a transformer with the same number of turns and a 2x scaling with half the number of primary turns versus a 1x linear scale transformer. Unfortunately my computer froze up due some Windows mouse bug and I lost the whole writeup. But here is a brief summary of the results I got. By linear scaling, I mean the magnetic core linear dimensions as well as linear wire dimensions. (like taking a magnifying glass to the whole transformer) Insulation dimension turns out to need a 4x linear thickness scaling due to the resultant 4x turns per volt increase.
The 2x scaled xfmr with the same number of turns achieves the same low frequency limit with twice the primary (and secondary) impedance. Primary inductance doubles. Upper bandpass halves. Power handling goes up by 8x. The bandpass penalty is due to increased leakage inductance and distributed capacitance. Varying the insulation thickness to accomodate the 4x turns per volt increase only shifts levels of leakage inductance and distrib. capacitance with no affect on upper bandpass frequency.
The 2x scaled transformer with half the turns achieves the same low frequency limit with half the primary impedance. Primary inductance halves. Upper bandpass stays the same. Power handling goes up by 8x. Its well known that lower primary impedance increases bandwidth, so here it is being played off successfully against the size- power penalty.
The same analysis works for P-P as well as SE designs if the gap width is linearly scaled the same amount too.
So it would seem that to design a high power SE xfmr, one needs to scale the primary impedance as 1/ cube root of the power handling capability to get the same bandwidth as a low power version. (this assumes the same optimum interleaving and or sectioning in all the models)
Don
The 2x scaled xfmr with the same number of turns achieves the same low frequency limit with twice the primary (and secondary) impedance. Primary inductance doubles. Upper bandpass halves. Power handling goes up by 8x. The bandpass penalty is due to increased leakage inductance and distributed capacitance. Varying the insulation thickness to accomodate the 4x turns per volt increase only shifts levels of leakage inductance and distrib. capacitance with no affect on upper bandpass frequency.
The 2x scaled transformer with half the turns achieves the same low frequency limit with half the primary impedance. Primary inductance halves. Upper bandpass stays the same. Power handling goes up by 8x. Its well known that lower primary impedance increases bandwidth, so here it is being played off successfully against the size- power penalty.
The same analysis works for P-P as well as SE designs if the gap width is linearly scaled the same amount too.
So it would seem that to design a high power SE xfmr, one needs to scale the primary impedance as 1/ cube root of the power handling capability to get the same bandwidth as a low power version. (this assumes the same optimum interleaving and or sectioning in all the models)
Don
Konnichiwa,
If you do that, why not go for something akin to the Berning Output design? Sync up two conventional low Voltage SMPS one operating (say) -24V and the other from +24V.
One drives the various Valve heaters (and maybe driver circuit) plus a shunt "ballast" to balance the DC and the other one supplies the anode circuit of the output valve. If without signal both supplies draw the same current (you have the ballast servoed for that) you can connect a speaker between the join of the two supplies and the center tap of the +/-24V supply.
The output circuit now varies the load on the suitable supply and thus modulates the supply currents, resulting in a suitable change of current through the speaker, obviously the idea works also PP.
Sayonara
tubelab.com said:
If you do that, why not go for something akin to the Berning Output design? Sync up two conventional low Voltage SMPS one operating (say) -24V and the other from +24V.
One drives the various Valve heaters (and maybe driver circuit) plus a shunt "ballast" to balance the DC and the other one supplies the anode circuit of the output valve. If without signal both supplies draw the same current (you have the ballast servoed for that) you can connect a speaker between the join of the two supplies and the center tap of the +/-24V supply.
The output circuit now varies the load on the suitable supply and thus modulates the supply currents, resulting in a suitable change of current through the speaker, obviously the idea works also PP.
Sayonara
Actually I have experimented along the lines of David Berning's ZOTL circuit. I blew up a lot of mosfets, but it did sound good when it worked.
I wanted to do a conventional SE amp with a conventional OPT, just feed it with a dynamic switching CCS to improve the efficiency over that of a conventional vacuum tube or mosfet CCS that I have now. I will get around to it someday. With circuits of this type I will start with a small low powered amp, and get that working before trying to build the big one.
As far as transformers go, I have contacted several vendors about large SE transformers, and have learned the following generalizations:
In the 10K impedance range 30 to 50 watts is about as big as any good transformer winder wants to make.
In the 5K impedance range 50 to 75 watts seems to be the comfortable limit.
I did not ask about lower impedances.
To build a big SE amp, you can use one big tube at a relatively high voltage 1200 - 2500 volts. This needs current in the 150 to 300 mA range and a transformer in the 5K to 10K ohm range. The transformer gap must be sized for the current.
You can also run several smaller tubes in parallel. Now the voltage and the required impedance goes down, but the current goes up, way up. This requires a larger gap, and therefore more iron to get enough inductance.
I don't know where the optimum lies to build a 100 Watt SE amp, but I have several 833A tubes, so I am trying this route (5K ohm). I had a transformer winder who builds good 10 - 20 watt SE transformers build me a 100 watt transformer according to his computer models, and it was not good enough for HiFi.
I wanted to do a conventional SE amp with a conventional OPT, just feed it with a dynamic switching CCS to improve the efficiency over that of a conventional vacuum tube or mosfet CCS that I have now. I will get around to it someday. With circuits of this type I will start with a small low powered amp, and get that working before trying to build the big one.
As far as transformers go, I have contacted several vendors about large SE transformers, and have learned the following generalizations:
In the 10K impedance range 30 to 50 watts is about as big as any good transformer winder wants to make.
In the 5K impedance range 50 to 75 watts seems to be the comfortable limit.
I did not ask about lower impedances.
To build a big SE amp, you can use one big tube at a relatively high voltage 1200 - 2500 volts. This needs current in the 150 to 300 mA range and a transformer in the 5K to 10K ohm range. The transformer gap must be sized for the current.
You can also run several smaller tubes in parallel. Now the voltage and the required impedance goes down, but the current goes up, way up. This requires a larger gap, and therefore more iron to get enough inductance.
I don't know where the optimum lies to build a 100 Watt SE amp, but I have several 833A tubes, so I am trying this route (5K ohm). I had a transformer winder who builds good 10 - 20 watt SE transformers build me a 100 watt transformer according to his computer models, and it was not good enough for HiFi.
tubelab.com said:
IIRC Davd uses a resonant SMPS topology, which IMO may well be the key to making it work. Using a MOSFET at the limits of it's breakdown voltage in a SMPS implies that you have dealt with the leakage inductance. The latter is 'not coupled' so when the SMPS switches, especially when it switches off, it generates huge voltage spikes. Having designed a few SMPS, leaving these unchecked is a sure way to MOSFET death. With a resonant topology, you could say leakage inductance works for you rathr than against you.
I looked at the ZOTL patent some time ago... briliant idea, the principle reminds me of AC bias in tape recorders...
Hello Tubelab.com,
"You can also run several smaller tubes in parallel. Now the voltage and the required impedance goes down, but the current goes up, way up. This requires a larger gap, and therefore more iron to get enough inductance."
You should be able to use the same gap no matter the primary Z, the turns go down for lower Z and the ampere-turns (product) is what matters to the core saturation.
Example: use twice the current and half the voltage (for same wattage) gives 1/4 Z, turns drop by 1/2 to get 1/4 Z and so 2xcurrent times 1/2 turns gives same ampere-turns. So gap stays the same.
The SMPS idea with a ferrite choke to balance the DC current sounds like a nice idea. Can get excellent HF response with a cheap choke and use a P-P output xfmr for good bandwidth. Maybe need a servo on the PWM to hold DC balance accurately. Need snubber circuits to protect the switching Mosfet. Resonant topology is just a glorified snubber circuit anyway.
Don
"You can also run several smaller tubes in parallel. Now the voltage and the required impedance goes down, but the current goes up, way up. This requires a larger gap, and therefore more iron to get enough inductance."
You should be able to use the same gap no matter the primary Z, the turns go down for lower Z and the ampere-turns (product) is what matters to the core saturation.
Example: use twice the current and half the voltage (for same wattage) gives 1/4 Z, turns drop by 1/2 to get 1/4 Z and so 2xcurrent times 1/2 turns gives same ampere-turns. So gap stays the same.
The SMPS idea with a ferrite choke to balance the DC current sounds like a nice idea. Can get excellent HF response with a cheap choke and use a P-P output xfmr for good bandwidth. Maybe need a servo on the PWM to hold DC balance accurately. Need snubber circuits to protect the switching Mosfet. Resonant topology is just a glorified snubber circuit anyway.
Don
Just wanted to add a note to clarify the sizing of the magnetic gap for SE xfmrs. As the linear dimension scaling modeling shows, doubling the linear size of the xfmr requires twice the gap width for the same magn. saturation level. This doubled xfmr size will then handle 8x the power.
So gap width varies as the cube root of power ratio. But gap width is insensitive to primary impedance if turns number are correctly adjusted versus Z. Optimum primary impedance selection (for bandwidth) would then vary as 1/(cube root of power ratio).
Don
So gap width varies as the cube root of power ratio. But gap width is insensitive to primary impedance if turns number are correctly adjusted versus Z. Optimum primary impedance selection (for bandwidth) would then vary as 1/(cube root of power ratio).
Don
- Status
- Not open for further replies.