They hysteresis is the time delay for the magnetic field to establish itself in the secondary after a signal is passed into the primary. I think when the magnetic flux is polarized as in class A , s.e.t. this hysteresis is less...
I have a picture of the particular response of magnetic material to magnetic fields, this is the decision for the different core materials. Alnico is very linear and neodymium better than Alnico, ill get the picture when I get time
I have a picture of the particular response of magnetic material to magnetic fields, this is the decision for the different core materials. Alnico is very linear and neodymium better than Alnico, ill get the picture when I get time
Alnico and Neodymium are usually associated with permanent magnets. Not working too well in an OT.
Speed of electromagnetic propagation = 1/SQRT[permeability x permittivity]
ie 1/SQRT[Mu x Eo]
Might get a few nanoseconds across the OT. The greater primary wire length for SE could be a time killer though.
Speed of electromagnetic propagation = 1/SQRT[permeability x permittivity]
ie 1/SQRT[Mu x Eo]
Might get a few nanoseconds across the OT. The greater primary wire length for SE could be a time killer though.
Just out of curiosity:
For a 20KHz signal, the wavelength is 48387 feet in vacuum, and a 1/2 wavelength winding would null out the signal. Permeability of the core and insulation on the wire would slow the EM wave by another factor of 4 or so down the wire. So effective 1/2 wavelength in primary core: about 6000 feet.
And then 6000 turns of 6 inches length per turn for a big SE OT primary, would give 3000 feet for the primary.
3000/6000 = .5
So the SE primary is smearing the 20 KHz waveform by 50%, Yikes.
You need to just get rid of the steel core altogether for SE premium performance. Those big hi primary Z SE OTs for transmitting tubes are probably frequency challenged no matter how they are designed.
It could be that since the core magnetization is common to all turns, it does not slow the EM propagation by Sqrt (Mu) in that configuration. Only insulation permittivity would factor in then. About 80% C then instead of 1/4 C. Even then, 15% smearing. So much for "detail".
For a 20KHz signal, the wavelength is 48387 feet in vacuum, and a 1/2 wavelength winding would null out the signal. Permeability of the core and insulation on the wire would slow the EM wave by another factor of 4 or so down the wire. So effective 1/2 wavelength in primary core: about 6000 feet.
And then 6000 turns of 6 inches length per turn for a big SE OT primary, would give 3000 feet for the primary.
3000/6000 = .5
So the SE primary is smearing the 20 KHz waveform by 50%, Yikes.
You need to just get rid of the steel core altogether for SE premium performance. Those big hi primary Z SE OTs for transmitting tubes are probably frequency challenged no matter how they are designed.
It could be that since the core magnetization is common to all turns, it does not slow the EM propagation by Sqrt (Mu) in that configuration. Only insulation permittivity would factor in then. About 80% C then instead of 1/4 C. Even then, 15% smearing. So much for "detail".
Last edited:
Remember that the first turn and the last turn (and all in between) are tightly coupled together so the signal does not have to propagate through all that length of wire. It can take a short cut via the magnetic field.
Agreed.
However, RF transmission line xfmrs do have to take into account the transmission line length within the core. So not clear how to handle the audio case. There is a difference in functionality however between the two cases. Transmission line xfmrs use the core to isolate input from output, while audio xfmrs couple input to output through the core. The transmission line xfmrs do reach greater bandwidths. But maybe not greater freq. ratios.
For the audio case, the magnetic coupling near instantly couples all the turns. But the copper winding path is still a valid signal path, and its got delay. Assuming the ends of the winding are correctly phased from the drive source, then 1/2 way through the winding, the delayed copper signal pathes will be equally affecting the core as the end turns, but with a 180 degree phase difference if 1/2 wavelength. So my guess is at 1/2 wavelngth delay to the 1/2 way winding length point, they start nulling each other out as to affecting the core. So effective permeability maybe heads to zero. Some measurements would be useful to see what happens. That would only give another factor of 2 extension for max wire length then. It would effectively become an air core xfmr at that 1/2 lambda frequency.
However, RF transmission line xfmrs do have to take into account the transmission line length within the core. So not clear how to handle the audio case. There is a difference in functionality however between the two cases. Transmission line xfmrs use the core to isolate input from output, while audio xfmrs couple input to output through the core. The transmission line xfmrs do reach greater bandwidths. But maybe not greater freq. ratios.
For the audio case, the magnetic coupling near instantly couples all the turns. But the copper winding path is still a valid signal path, and its got delay. Assuming the ends of the winding are correctly phased from the drive source, then 1/2 way through the winding, the delayed copper signal pathes will be equally affecting the core as the end turns, but with a 180 degree phase difference if 1/2 wavelength. So my guess is at 1/2 wavelngth delay to the 1/2 way winding length point, they start nulling each other out as to affecting the core. So effective permeability maybe heads to zero. Some measurements would be useful to see what happens. That would only give another factor of 2 extension for max wire length then. It would effectively become an air core xfmr at that 1/2 lambda frequency.
Last edited:
Transmission line transformers have two different things going on at the same time. For differential mode signals they are simply a piece of transmission line which doesn't know it is wrapped around a core. For common-mode signals they are a choke, which doesn't really know that its wire has two cores so it only chokes the net current. The propagation speed affects the former.
You also need to remember that at audio frequencies the speed of propagation in a wire is much slower than at RF, as resistance dominates over inductance. Hence if the issue you raise was a real problem it would be much worse than you think.
You also need to remember that at audio frequencies the speed of propagation in a wire is much slower than at RF, as resistance dominates over inductance. Hence if the issue you raise was a real problem it would be much worse than you think.
Yes, agree on the transmission line xfmr scenario. It might be interesting to make an audio version of a transmission line xfmr using common mode chokes and unity bifilar couplers (series coupled primaries, parallel coupled secondaries), but getting a sufficient frequency range ratio would need careful analysis of L and C parasitics.
On the standard audio xfmr, the resistive HF loss could be what saves it from doom. If the signal propagating with delay along the copper winding pathes is attenuated before reaching a lambda/2 point, then opposing phase effects on the magnetic core may be avoided. The active transformer just "shrinks" toward the winding ends as the frequency goes up, leaving in-phase components. In that case, making a high Z xfmr "better", RF wise, with flat ribbon conductor or Litz wire might ruin the frequency response. Seems to me I have heard of Litz wound OTs occasionally. Never hear any follow up on them. Could be useful for the secondary at least.
On the standard audio xfmr, the resistive HF loss could be what saves it from doom. If the signal propagating with delay along the copper winding pathes is attenuated before reaching a lambda/2 point, then opposing phase effects on the magnetic core may be avoided. The active transformer just "shrinks" toward the winding ends as the frequency goes up, leaving in-phase components. In that case, making a high Z xfmr "better", RF wise, with flat ribbon conductor or Litz wire might ruin the frequency response. Seems to me I have heard of Litz wound OTs occasionally. Never hear any follow up on them. Could be useful for the secondary at least.
Last edited:
Because of their high hysteresis
Hysteresis is a text book term for the energy required to reverse a magnetic flux, in a S.E.T. there is no Hysteresis because the flux is never reversed.
By time-delay I mean that you will lose part of the AC signal in class AB transformers while the energy is required to reverse the magnetic field ( realign magnetic domains 180 deg.) this energy is the music signal and is lost without any transmission and feedback can't bring it back. Sorry to use that term delay which should have been instead 'time& energy dependent signal loss' .
"in a S.E.T. there is no Hysteresis because the flux is never reversed. "
This is a common mis-perception fostered by SET advertising. (ie, standard audio lies) You don't get a voltage induced in the secondary unless flux is changing. The DC in SET just holds down half the magnetic domains and the other half perform the AC field reversals to induce output voltage. No difference from the usual P-P core operation. A hysteresis curve develops around the new operating point. It has a more elliptical shape because the easy to influence magnetic domains are pinned down and cannot avalanche to give high permeability slopes.
There is no loss of the signal in the OT due to hysteresis. None whatsoever. The voltage across the primary determines the rate of flux change, and the rate of flux change determines the output voltage at the secondary. The hysteresis parasitically determines how the magnetizing current on the primary input will vary (non sinusoidally with sine wave input), and it can cause distortion if the driving tube does not present a low source impedance to prevent voltage errors from the tube. SET amps typically do NOT provision for low source impedance, and as a result, produce abundant distortion of the signal. It's easy to hear. Fortunately for SETs some like it.
This is a common mis-perception fostered by SET advertising. (ie, standard audio lies) You don't get a voltage induced in the secondary unless flux is changing. The DC in SET just holds down half the magnetic domains and the other half perform the AC field reversals to induce output voltage. No difference from the usual P-P core operation. A hysteresis curve develops around the new operating point. It has a more elliptical shape because the easy to influence magnetic domains are pinned down and cannot avalanche to give high permeability slopes.
There is no loss of the signal in the OT due to hysteresis. None whatsoever. The voltage across the primary determines the rate of flux change, and the rate of flux change determines the output voltage at the secondary. The hysteresis parasitically determines how the magnetizing current on the primary input will vary (non sinusoidally with sine wave input), and it can cause distortion if the driving tube does not present a low source impedance to prevent voltage errors from the tube. SET amps typically do NOT provision for low source impedance, and as a result, produce abundant distortion of the signal. It's easy to hear. Fortunately for SETs some like it.
Last edited:
No it isn't. Energy loss is a consequence of hysteresis, but it is not the definition of hysteresis.gabdx said:
No. Hysteresis is seen whenever you change the flux. Nothing special happens at zero.
Pure nonsense. I assume you don't know how ferromagnets work?
No, its not a mis perception, maybe you think it is. I never saw an advertising about this.
You can try very low signal levels for a pp transformer and i will be curious how much signal you have to put in before there is something comming out. Then compare it with a se transformer. Take low a frequency otherwise the core will do the job instead of the core.
You can try very low signal levels for a pp transformer and i will be curious how much signal you have to put in before there is something comming out. Then compare it with a se transformer. Take low a frequency otherwise the core will do the job instead of the core.
esltransformer will probably come back with a reply something like "MC transformers are smaller so need less signal to drive them than a PP OPT". Dogged persistence in the face of overwhelming evidence is strong in this one.
Well, its easy to see where the many mis-perceptions about SET mode come about, and hysteresis is truly something that no one wants, P-P or SET. But one needs to arrive at a clear picture of what is actually going on to make some design progress.
Hysteresis in an OT will typically (indirectly) cause distortion since the tube (especially a pentode) has high enough output impedance, that it can be (Voltage) corrupted by the OT magnetizing current. Low Mu triodes are used in SET for a reason, they have lower Rp.
Magnetization curves showing square hysteresis lead to the first big mis-conception: that one can bias the field with DC (like a tube) to use the high permeability (high linear slope) region. This doesn't work. The DC just pins down the most easily reversed domains, lowering the permeability and removing the high slope region. The remaining domains still perform the flux reversals needed for voltage induction, producing a thinner elliptical local hysteresis loop. A somewhat more astute analysis then concludes that the hysteresis (area within the loop) is smaller, so the hysteresis is at least reduced.
Unfortunately, these curves are plotted versus magnetizing CURRENT, not drive signal VOLTAGE (or flux change). With the reduced permeability, a larger (magnetizing) current span is required to induce the same voltage (flux height of the loop). So in reality, the nice elliptical loop expands to cover similar (if not greater) area. So the hysteresis losses are similar (it is using the stickier, more frictional domains after all). The magnetizing current is at least lower harmonic inducing on average than the previous square loop with avalanching, but it is a much bigger current (especially with an air gap). The level of higher harmonic reduces some due to the loop shape change (and air gap dilution), but is also better hidden within the higher low harmonic current component.
Now it might still seem like progress, to have traded off some higher harmonic magnetizing current for more lower harmonic magnetizing current, but now the tube comes into play. The load line becomes an ellipse from the significant 90 degree out of phase magnetizing current, instead of a straight line. This causes a type of hysteresis within the tube, since different gain regions are traversed for increasing and decreasing signal. Fortunately, a low output Z tube, or local feedback (or even global Fdbk) can reduce (or near eliminate) this, by keeping all load pathes at more equal gain.
So a better (lower hysteresis) magnetic alloy will help either SET or P-P, definately. Higher permeability material will help P-P by lowering magnetizing current further (already low). But high perm material will just saturate with DC more easily, and usually has less Bmax flux to play with (ie nickel alloys). So choosing material to upgrade SET is tricky. Mostly determined by listening effects.
The payoff (in linearity) is far greater to just lower the Zout of the driving stage for any configuration (to reduce effects of magnetizing current), rather than paying for exotic materials. However, strict linearity is not usually the real goal for SET.
Hysteresis in an OT will typically (indirectly) cause distortion since the tube (especially a pentode) has high enough output impedance, that it can be (Voltage) corrupted by the OT magnetizing current. Low Mu triodes are used in SET for a reason, they have lower Rp.
Magnetization curves showing square hysteresis lead to the first big mis-conception: that one can bias the field with DC (like a tube) to use the high permeability (high linear slope) region. This doesn't work. The DC just pins down the most easily reversed domains, lowering the permeability and removing the high slope region. The remaining domains still perform the flux reversals needed for voltage induction, producing a thinner elliptical local hysteresis loop. A somewhat more astute analysis then concludes that the hysteresis (area within the loop) is smaller, so the hysteresis is at least reduced.
Unfortunately, these curves are plotted versus magnetizing CURRENT, not drive signal VOLTAGE (or flux change). With the reduced permeability, a larger (magnetizing) current span is required to induce the same voltage (flux height of the loop). So in reality, the nice elliptical loop expands to cover similar (if not greater) area. So the hysteresis losses are similar (it is using the stickier, more frictional domains after all). The magnetizing current is at least lower harmonic inducing on average than the previous square loop with avalanching, but it is a much bigger current (especially with an air gap). The level of higher harmonic reduces some due to the loop shape change (and air gap dilution), but is also better hidden within the higher low harmonic current component.
Now it might still seem like progress, to have traded off some higher harmonic magnetizing current for more lower harmonic magnetizing current, but now the tube comes into play. The load line becomes an ellipse from the significant 90 degree out of phase magnetizing current, instead of a straight line. This causes a type of hysteresis within the tube, since different gain regions are traversed for increasing and decreasing signal. Fortunately, a low output Z tube, or local feedback (or even global Fdbk) can reduce (or near eliminate) this, by keeping all load pathes at more equal gain.
So a better (lower hysteresis) magnetic alloy will help either SET or P-P, definately. Higher permeability material will help P-P by lowering magnetizing current further (already low). But high perm material will just saturate with DC more easily, and usually has less Bmax flux to play with (ie nickel alloys). So choosing material to upgrade SET is tricky. Mostly determined by listening effects.
The payoff (in linearity) is far greater to just lower the Zout of the driving stage for any configuration (to reduce effects of magnetizing current), rather than paying for exotic materials. However, strict linearity is not usually the real goal for SET.
Last edited:
Cores for low signal level transformers are often "permalloy" or other materials chosen for higher initial permeability and a flatter curve of flux density vs. field intensity. Some fancy output transformers from the Golden Age would include some nickel lams for the same reason (Peerless, for example). The flat spot around zero crossing observed in B/H magnetization curves deserves more attention than simply ragging on how stupid everyone interested in SETs must be.
Sorry to be so blunt, but several folks here are capable of much better. Thank you all, as always,
Chris
Sorry to be so blunt, but several folks here are capable of much better. Thank you all, as always,
Chris
The flat spot at zero crossing is just the reversion to initial permeability domain wall motion before avalanching occurs. This is mostly all one is left with for the entire operating curve for SET with 1/2 Bmax DC flux present.
High permeability nickel alloy is useful when the source is high impedance.
I didn't say anyone was stupid, magnetics is a complex and subtle area to understand. But the endless parade (over the years) of people spouting the same mis-information about SET needs some counter balance to the SET Cool-Aid.
I'm out of here. No point in wasting effort on this area.
High permeability nickel alloy is useful when the source is high impedance.
I didn't say anyone was stupid, magnetics is a complex and subtle area to understand. But the endless parade (over the years) of people spouting the same mis-information about SET needs some counter balance to the SET Cool-Aid.
I'm out of here. No point in wasting effort on this area.
Last edited:
Calling it a "flat spot" implies that permeability disappears.Chris Hornbeck said:
But it doesn't disappear, it just reduces.smoking-amp said:
People want simple answers. If the true answer is complex they will often invent their own simple 'answer' and proclaim that as the truth.
Well, it reduces a lot ! So much that i called it disappear......
And i already showed in post 21 that just a simple 1:100 reduction has a hugh impact on some transformers going from 6Hz -3 dB to 40Hz -3dB.
And i already showed in post 21 that just a simple 1:100 reduction has a hugh impact on some transformers going from 6Hz -3 dB to 40Hz -3dB.
Last edited:
- Status
- Not open for further replies.