Magnetic hysteresis is a complicating factor. When you apply a magnetic field to a magnetic material, it is partially magnetized. When you saturate it, it is strongly magnetized. When it is magnetized, it opposes the application of a reverse magnetic field, and it takes a certain reverse applied field to overcome the residual field in the material, and it takes a little more reverse applied field to reverse the residual field. This causes a timing delay in the magnetic transitions. The amount of delay depends on the applied signal level.
Even below saturation level, hysteresis causes increasing distortion as level increased, because the residual magnetization is greater at higher levels of applied magnetic field. At the same time, the magnetic transfer of the core falls off at high levels, introducing more distortion. This is why bigger iron and better iron are favored. But the windings are important too. Issues with transformer windings can really affect high frequency response. Magnetic coupling of outer windings is not as good as the coupling of the inner windings, and different "leakage" inductance results. This, and different amounts of stray coupling between windings and to ground, cause high frequency rolloff, and it causes different windings to have slightly different frequency responses and throughput time delays. In class-B operation, these effects cause distortion of high frequency signals. Half of the waveform is delayed more than the other half, and you get asymmetrical crossover distortion. This distortion sounds terrible, especially on sibilants. The two halves of the waveform may have different amplitudes as well. I have improved this distortion by adding some capacitance to the faster side at the driver level. This tends to balance the delay and eliminate this effect. When a driver transformer is used, reversal of the grid connections may help by making the unequal delays of the driver cancel those of the output stage. The plate connections to the driver transformer can also be reversed to maintain the original audio polarization, if necessary. Eddy currents in the core absorb high frequency energy. This is why we 'laminate' the core, slicing it into thin layers, and coating them with insulating paint. The thinner the layers, the less high frequency eddy current loss there will be. This is why audio transformer laminations are usually thinner than power transformer laminations. In analog magnetic recording, hysteresis causes a huge crossover gap that must be compensated, or unbelievable crossover distortion results. Modern analog recorders use about 100 KHz AC bias, adjusted slightly higher than required for maximum recorded output, to achieve minimum crossover distortion. Typically this overbias reduces sensitivity just a little bit at 1 KHz , about 0.25dB. However, it reduces higher frequencies more strongly, about 2dB at 10 KHz. This is equalized out during recording. Bacon, WA3WDR ----- Original Message ----- From: "Larry Will" <[EMAIL PROTECTED]> To: "Discussion of AM Radio" <amradio@mailman.qth.net> Sent: Tuesday, March 14, 2006 3:12 PM Subject: Re: [AMRadio] Class AB and B audio XFMRS > Bob and all, > > Both of our discussions are correct but deal with a different part of > the problem. Core saturation is definitely one of the problems but > hysteresis, the inability of the magnetic core to instantly change > polarity is another. The core saturation causes a spiking due to > field collapse as Bacon said but the hysteresis causes the "crossover > distortion". Both contribute to the overall distortion. Higher > quality iron and enough of it helps LF but the higher interwinding > capacitance of the larger transformers can kill HF response. Thats > one reason many audio transformers can't go even to 10 or 20 kcs. > > > > > Nothing is simple. > > Larry > > > At 03:47 PM 3/13/2006, you wrote: > >Electromagnetism really confused the early scientists. > >They thought it should behave symmetrically. That is, > >if DC passing through a coil produces a fixed magnetic > >field, they thought that a similar fixed magnetic field > >should produce DC from a coil. This would have been > >every cool, because they had permanent magnets from > >which free power could have been derived. > > > >The problem was, it didn't work that way. There is a > >story about how this problem was solved. Michael > >Faraday was trying everything; he held a magnet in > >every possible place around a coil, he tried holding > >the magnet at every possible angle and direction, etc. > >But no matter what he did, no DC came out of his coil. > > > >Finally the great scientists had had enough. I imagine > >him standing up, cursing, and throwing the magnet > >violently at the coil, in anger. > > > >But something happened when he did that. The > >galvanometer twitched when the magnet passed through > >the coil! Faraday had discovered that the magnetic > >field needed to be changing in order to produce a > >voltage from the coil, and the output voltage would > >alternate. (And this sort of comedy has been entirely > >typical of the process of scientific discovery from > >earliest antiquity.) > > > >OK, now about an audio transformer. The flux must be > >changing in one direction to produce a steady dc output > >from the winding. That means that the longer a square > >wave needs to hold positive, the more flux there has to > >be in the core. Even without unbalanced DC in the > >windings, the core will saturate at some point. This > >places a limit on the lowest frequency square wave that > >can be produced at any given power level. > > > >The situation with sine waves is similar. At high > >frequencies, the alternating flux does not have to > >build up to very high levels to produce a given amount > >of output power. But as the frequency decreases, the > >magnetic flux needs to go higher and higher to maintain > >the necesssary rate of change over the slower cycles, > >in order to produce the necessary voltage and power > >output. And at some point, the core runs out of > >magnetic capability. > > > >When that happens, the flux can not continue to rise. > >It can only hold steady until the applied current > >falls. The coil can not produce DC in this situation, > >and the output voltage falls to zero and sits there > >until the current falls, which happens at the next > >crossover. At the crossover, the magnetic flux changes > >and then saturates in the opposite direction. This > >produces a pulse, followed by a drop to zero volts and > >another flatline. So we get a flat line where the > >signal should have had a positive peak, we get a > >negative peak where we should have seen the signal > >waveform falling, and we get another flat line where > >the signal should have had a negative peak, and we get > >a positive pulse where we should have seen the signal > >waveform rising. And unbalanced DC tends to make this > >happen with an offset. > > > >So. A given transformer can handle more power at > >medium and higher frequencies than it can at low > >frequencies, and the situation gets worse when > >unbalanced DC is applied. Unbalanced DC is bad news, > >because it builds the core up to significant magnetic > >flux levels. > > > >It turns out that for a given amount of DC magnetic > >flux, there is an optimum "gap" that produced the > >maximum efect a given core can produce. More gap than > >that or less gap than that is not as good. This gives > >less inductance than no gap, but the inductance > >survives unbalanced DC better, so it's a winning > >compromise. But if there will be no unbalanced DC in > >the winding, then we want to eliminate the gap. That > >gives us more inductance from a given winding, which > >gives better low frequency response. But remember, > >those lows will saturate the core all by themselves at > >some point. > > > > Bacon, WA3WDR > > > > > >----- Original Message ----- > >From: "Larry Will" <[EMAIL PROTECTED]> > >To: "Discussion of AM Radio" <amradio@mailman.qth.net>; > >"'Discussion of AM Radio'" <amradio@mailman.qth.net> > >Sent: Monday, March 13, 2006 1:24 PM > >Subject: RE: [AMRadio] Class AB and B audio XFMRS > > > > > > > Hi all, > > > > > > As I dimly remember from Motors and Machines 1 and 2, > >THE TRANSFORMER > > > IS A HIGHLY NON-LINEAR DEVICE. WE USE THESE SIMPLE > >FORMULAS for TR > > > and ZR but in reality YOU NEED ADVANCED CALCULUS TO > >adequately > > > explain BOTH THE Hysteresis and eddy current losses > >and > > > distortions. The open circuit and short circuit > >tests can get that > > > info out of a particular device, I haven't done these > >in years but it > > > is a valuable tool for analyzing any transformer. > >You need power and > > > the correct frequencies however. Its the hysteresis > >losses that > > > result in the poor LF response. The secondary > >voltage lags the > > > primary current (I think I remember that correctly) > >more and more as > > > the frequency is decreased. The problem is > >especially acute at the > > > polarity change, ie the zero crossing where the > >magnetic field must > > > reverse instantly. The non linear effects generate a > >discontinuity > > > in the waveform and the harmonic components and odd > >phases are the > > > result. What is happening the primary power is > >converted to a > > > magnetic flux which is then converted back to power > >in the > > > secondary. Its the medium - the IRON that causes the > >problems. This > > > cannot be analysed except by non-linear mathematics > >Messy at > > > best. The DC current, if present polarizes the > >magnetic field making > > > the effects worse. MacIntosh got around this with > >his patented > > > transformer and circuit which greatly minimized the > >magnetic > > > non-linearity, circuits which are still used in > >MacIntosh audio amps today. > > > > > > > > > BTW audio analog tape recorders minimized this > >problem with the head > > > by using a high frequency bias, say 22 kcs, to keep > >the flux > > > constantly changing and allowing good LF response > >while making the > > > recording. The HF signal is filtered out on playback > >either by the > > > playback head or immediately before the first preamp. > > > > > > Larry W3LW > > > > > > Some folks on here surely can amplify this and > >correct my fuzzy > > > memory if needed. > > > > > > > > > The problem is the energy transfer medium - THE IRON. > > > > > > > > > > > > At 01:02 PM 3/13/2006, John E. Coleman (ARS WA5BXO) > >wrote: > > > > Perhaps I should clarify one point that we > >may all be forgetting > > > >here. A XFMR will only transfer energy during the > >movement of the magnetic > > > >field (EXPANDING OR CONTRATING). If the magnetic > >field becomes stationary > > > >then no energy will be transferred to the secondary > >regardless of the amount > > > >of iron. But if the magnetic movement is fast > >enough then transfer > > > >efficiency can be high. As the frequency is lowered > >the magnetic movement > > > >is slowed down then the efficiency drops off. I'm > >not sure if this is the > > > >proper term mathematically but it is as if the > >coefficiency of coupling is > > > >not as good when the frequency becomes too low. > > > > > > > > I hope I'm not boring folks with this and > >some may say I am making a > > > >mountain of a mole hill. I just find it > >fascinating. I guess it is just my > > > >type of thing. > > > > > > > >John, WA5BXO > > > >----Original Message----- > > > >From: [EMAIL PROTECTED] > > > >[mailto:[EMAIL PROTECTED] On Behalf > >Of John E. Coleman (ARS > > > >WA5BXO) > > > >Sent: Monday, March 13, 2006 11:36 AM > > > >To: 'Discussion of AM Radio' > > > >Subject: RE: [AMRadio] Class AB and B audio XFMRS > > > > > > > >The Band pass and energy Xfer of a transformer with > >no load is one thing but > > > >it all changes depending on the load and the DC > >involved. In class A > > > >balanced PP circuits the XFMR will still saturate at > >some frequency and load > > > >even if it is perfect balance on the PP circuit. > >XFMR saturation distortion > > > >in class A single ended service has a trapezoidal > >shape if the quiescent > > > >current is too high but in PP class A the shape is > >weird because the XFMR > > > >remains balance as for as DC is concerned but yet > >the XFMR will not produce > > > >the sign wave on the output if the frequency is too > >low. It resembles cross > > > >over distortion even though there is no cross over > >in class A push-pull. > > > > > > > > > > > > > > > > > > > > > >_________________________________________________ _____ > >________ > > > >AMRadio mailing list > > > >Home: > >http://mailman.qth.net/mailman/listinfo/amradio > > > >Help: http://mailman.qth.net/mmfaq.html > > > >Post: mailto:AMRadio@mailman.qth.net > > > >AMfone Website: http://www.amfone.net > > > >AM List Admin: Brian Sherrod/w5ami, Paul > >Courson/wa3vjb > > > > > > > >_________________________________________________ ______ > >_______ > > > AMRadio mailing list > > > Home: http://mailman.qth.net/mailman/listinfo/amradio > > > Help: http://mailman.qth.net/mmfaq.html > > > Post: mailto:AMRadio@mailman.qth.net > > > AMfone Website: http://www.amfone.net > > > AM List Admin: Brian Sherrod/w5ami, Paul > >Courson/wa3vjb > > > > > > >_________________________________________________ _____________ > >AMRadio mailing list > >Home: http://mailman.qth.net/mailman/listinfo/amradio > >Help: http://mailman.qth.net/mmfaq.html > >Post: mailto:AMRadio@mailman.qth.net > >AMfone Website: http://www.amfone.net > >AM List Admin: Brian Sherrod/w5ami, Paul Courson/wa3vjb > > __________________________________________________ ____________ > AMRadio mailing list > Home: http://mailman.qth.net/mailman/listinfo/amradio > Help: http://mailman.qth.net/mmfaq.html > Post: mailto:AMRadio@mailman.qth.net > AMfone Website: http://www.amfone.net > AM List Admin: Brian Sherrod/w5ami, Paul Courson/wa3vjb >
