Characteristics of Guitar Amplifier Output Transformers, and criteria for design

[Gregg]

Don't get scared from the theory talk. It may be interesting (to some) but most of it has no practical use whatsoever. When you start calculating an actual OT you'll see what I mean.

[Malcolm Irving]

… What the hell does "'teaching my grandmother to suck eggs" mean?? …

It means 'trying to teach something to someone who probably already knows more about it than you do'.

It is harder to say why anybody would want to suck an egg though. On the internet somewhere I found that in Europe and Russia, in an earlier century, kids were taught to put a pin-hole at each end of an egg and suck out (and consume) the contents as a 'health tonic'. Not recommended these days due to fears of Salmonella etc.

Probably there are old-fashioned practices in guitar amp design which fall into a similar category.

[J M Fahey]

It means 'trying to teach something to someone who probably already knows more about it than you do'.

It is harder to say why anybody would want to suck an egg though. On the internet somewhere I found that in Europe and Russia, in an earlier century, kids were taught to put a pin-hole at each end of an egg and suck out (and consume) the contents as a 'health tonic'. Not recommended these days due to fears of Salmonella etc.

Probably there are old-fashioned practices in guitar amp design which fall into a similar category.

Well, MY Grandmother used to do that when she was young, tired, and needing a little Energy boost without stopping her farm work. We are talking 1928 or so.

I am quite certain that Salmonella was not such a big threat way back then, since chickens were free to roam in the open, all around the farm house, eating insects, "natural" seeds and vegetables and whatever food scraps were thrown at them.

The real "infection factories" are modern concentration camp type chicken farms: 5000 or 10000 chickens growing inside a huge "barn", with lights on 24/7 and as I was told, subject to noise so they never sleep, running all over and trampling each other, and of course over chicken poop covered floor.
That´s why they are pumped with tons of antibiotics, they are in such crowded contact that "one sick bird means **all* sick in a matter of hours" and so losing the whole batch.

Eggs , specially raw ones, were considered a quick and gfood source of Protein.

I didn´t personally see my Grandma suck raw eggs from the shell although it´s Famiy tradition and one of my old Aunts mentioned "she kept the nail Grandma used to punch eggs in her youth" , go figure.

What I DO clearly remember is visiting "Milk Bars" which where very popular when I was a kid (late 50´s) , and seeing mid to old aged guys ordering "Candéal", which was made on the counter by beating inside a glass a raw egg (or was it just the yolk?), a "measure" (the standard sized ration, probably a fluid ounce or so) of Brandy and a teaspoonful of sugar.
Might also have a drop of vanilla or a pinch of cinnamon.
Distantly related to Eggnog but without even a drop of milk and always with alcohol.

Besides the Health angle, I guess it was probably a loophole to allow "officially non drinking" people to get a couple shots now and then.

[SoulFetish]

should we designing for a some DC content due to stage imbalance (as is common in guitar output stages)? From some of my reading, if even a small 3% of DC offset is present it can result in a large error, increasing the magnetizing current to 135%. Umm,... what kind of things can we do about that?

see article below:

7s_110-34.pdf

[Gregg]

No. With articles like this you're about to set foot onto the very well oiled high slope HiFi surface

[eschertron]

From my quick skim through the article, the 3% appears to refer to % of full load current. This would require a 20..25% imbalance at idle. At which point there are other, more audible, side effects to worry about. I'm thinking heater hum.

[Helmholtz]

should we designing for a some DC content due to stage imbalance (as is common in guitar output stages)? From some of my reading, if even a small 3% of DC offset is present it can result in a large error, increasing the magnetizing current to 135%. Umm,... what kind of things can we do about that?

see article below:

[ATTACH]48939[/ATTACH]

Thanks for the interesting link. One of the most important aspects of transformer design is to avoid core saturation under worst case conditions. And these have to include a reasonable amount of DC offset in the peak value of H. For a given transformer design saturation depends on the peak magnetizing current Imag. Increasing the number of primary turns decreases Imag and thus increases saturation headroom.
Saturation does not depend on the load current/power.

[Gregg]

In practice OT saturation in a PP guitar amp under it's usual operating conditions is a non issue and very unlikely to happen (despite current imbalances) unless it's a poorly designed cheap iron crappy OT.

[Helmholtz]

In practice OT saturation in a PP guitar amp under it's usual operating conditions is a non issue and very unlikely to happen (despite current imbalances) unless it's a poorly designed cheap iron crappy OT.

Yes, but I thought this thread was about ab initio DIY design of OTs. Saturation easily occurs when Nprim is chosen too low.

[R.G.]

It is not that simple. The observed effect is larger inductance at larger flux levels (higher transformer power). This requires an analysis that takes into account the movement around the hysteresis loop. I do not think that it is so easy to get a good intuitive understanding of why the inductance is higher at higher levels.

Of course it's not that simple. However, when you're trying to introduce complex, slippery subjects to someone new to the concepts, you start with the simplest views first, then add on modifications as they get the grosser concepts under control. I didn't get into the quantum mechanical view of how magnetism arises, either.

I have on old textbook: 'Transformers for Electronic Circuits' by Nathan R. Grossner, 2nd Edition, 1983.

Hang on to that book. It's one of the few textbooks I still refer to. Solid gold.

Don't get scared from the theory talk. It may be interesting (to some) but most of it has no practical use whatsoever. When you start calculating an actual OT you'll see what I mean.

And we have another winner. Practical considerations of how much iron and copper you can afford and get into the available space will be much larger than highly detailed theory. I do find that theoretical considerations about the iron are much more necessary in the design of a SE output trannie, where you're actually relying on the core to store energy sufficient for one half cycle of the biggest outputs.

should we designing for a some DC content due to stage imbalance (as is common in guitar output stages)? From some of my reading, if even a small 3% of DC offset is present it can result in a large error, increasing the magnetizing current to 135%. Umm,... what kind of things can we do about that?
see article below:

As a practical matter, don't sweat designing for DC offset in a push-pull design. The way DC offsets are always handled is to introduce some air gap into the iron core. In non-toroidal cores, it is impossible to NOT introduce slight air gaps. There's a great deal of effort expended in stacking E-I laminations to reduce these air gaps as much as possible.

Most "toroidal" transformers are made by winding a coil of transformer iron into the toroidal core. There is still some effect of air gap because the iron is not really continuous, but is in thin layers with mostly air (or varnish) in the gap. It's much less air gap than E-I laminations. The article you found looked at offset effect in punched rings of iron; this is done so the testing can actually have a single loop of real iron to test the iron completely without air gaps. It is and is intended to be completely free of air gaps and the real material --can--- be tested. This is most definitely not a real-world condition.

As a practical matter, no transformer outside some lab setups is a true toroid. This is because (1) it's so very difficult to get a true, solid-iron toroid to wind and (2) solid iron toroids are impractical because they are solid. You want the core iron to be divided into very thin laminations so that eddy currents in the iron itself are thwarted by being forced to run in circles in a thin piece of iron. Eddy current losses act like a nonlinear resistance in parallel with the primary. The magnetic field in the iron itself causes current to flow in loops in the iron, which is conductive itself. Lamination breaks up the electrical path in the iron layers and makes the path for eddy currents be thin and higher resistance, cutting the losses.

As a side note, reducing eddy current losses is one reason that transformer iron is made with silicon in the alloy. The silicon makes the iron be much higher in electrical resistivity, further reducing the losses to eddy currents in the iron itself.

Eddy current losses and lamination is another of those cross-purposes things so common in transformers. Assuming you could get your core made from a solid lump of iron in the right shape and still thread your copper into the right places, you'd have horrible losses to eddy currents from current loops in the iron. Splitting the iron into laminations allows you to (1) wind the transformer coils then stack the iron around it, saving a whole lot of labor and (2) dramatically reduce eddy current losses. So the more laminations, approaching iron foil, the better, right?

Wrong. The thinner the laminations, the harder they are to stack and the more fragile they are, and also the more air gaps you introduce between laminations. The transformer industry homed in on a narrow range of lamination thicknesses as the relative optimums for most transformers. You can get super thin special material laminations, but the price, difficulty in stacking, and other ugly issues get out of hand rapidly. That's one reason you don't see iron foil cores, or pure-nickel OTs.

DC offset, air gaps, eddy current losses, the list of competing things to optimize just goes on and on.

Thanks for the interesting link. One of the most important aspects of transformer design is to avoid core saturation under worst case conditions. And these have to include a reasonable amount of DC offset in the peak value of H. For a given transformer design saturation depends on the peak magnetizing current Imag. Increasing the number of primary turns decreases Imag and thus increases saturation headroom.
Saturation does not depend on the load current/power.

That brings up another of those sideways steps. We all (should have) learned that V = L * di/dt. Or, put another way, di = V*dt/L. This is a way of saying that the magnetic field change in the iron depends on the voltage applied and the time it's applied. That assumes a constant value for L, which we know is not strictly true, but let's start simple.

The current in windings of an inductor is equal to the volt-time integral divided by a constant (L in this case, which isn't completely constant...) and this current is the current that drives the ampere-turns in the B-H curve. It has almost no relation to the current which merely goes THROUGH the magnetic field and out to the secondaries. As a sidelight, this is also the reason that transformers are frequency-sensitive - the core has to be wound to NOT go into saturation on a full half-cycle of applied primary voltage from zero crossing to the next zero crossing. Looked at another way, the core plus windings give you a certain volt-time "endurance" before you hit saturation. If you double the frequency, you halve the time, and you can then apply twice the voltage and still stay out of saturation.

[Malcolm Irving]

.... We all (should have) learned that V = L * di/dt. Or, put another way, di = V*dt/L. This is a way of saying that the magnetic field change in the iron depends on the voltage applied and the time it's applied. That assumes a constant value for L, which we know is not strictly true, but let's start simple.
...

Yes. I seem to remember from a theory class, in the dim and distant past, that the full formula is:

V = L di/dt + i dL/dt

But if inductance L is constant, the second term vanishes, of course.

In iron-cored coils, where reluctance is varying, the inductance of the coil would also vary, making the situation complicated.

[Helmholtz]

Yes. I seem to remember from a theory class, in the dim and distant past, that the full formula is:

V = L di/dt + i dL/dt

But if inductance L is constant, the second term vanishes, of course.

In iron-cored coils, where reluctance is varying, the inductance of the coil would also vary, making the situation complicated.

The OTs' primary inductance is not constant at all. As my measurements (above) and manufacturers' information show, L rises with increasing current typically by around a factor of more than 4 (even though there is some unavoidable airgap) up to a maximum from where it decreases steeply.

In my transformer and choke designs, avoiding saturation always was a major concern.

[Gregg]

In my transformer and choke designs, avoiding saturation always was a major concern.

Chokes and SE OTs are a different story. PP OTs however are very unlikely to saturate. Over the years I was able to saturate guitar PP OTs only with 30-40Hz signals which they were not designed to handle anyway. In rare cases I've seen some of them go down to 35Hz at -3dB.
Contrary to the popular belief even a toroidal OT won't saturate with DC imbalance. I've tried and seen that myself with one tube running at 30mA the other at 40mA and I'm talking full power not just couple of Watts.

[Helmholtz]

Chokes and SE OTs are a different story. PP OTs however are very unlikely to saturate. Over the years I was able to saturate guitar PP OTs only with 30-40Hz signals which they were not designed to handle anyway. In rare cases I've seen some of them to go down to 35Hz at -3dB.
Contrary to the popular belief even a toriodal OT won't saturate with DC imbalance. I've tried and seen that myself with one tube running at 30mA the other at 40mA.

Short and simple: For a given core, saturation determines the minimum number of primary turns for max. peak current to be expected.

[R.G.]

Contrary to the popular belief even a toroidal OT won't saturate with DC imbalance. I've tried and seen that myself with one tube running at 30mA the other at 40mA and I'm talking full power not just couple of Watts.

Whether they will saturate with DC imbalance or not greatly depends on the size of the imbalance and whether it's a current limited imbalance or a voltage imbalance that can ramp up over time. Practical toroids necessarily have some air gap no matter how you try to reduce it. It's just much smaller than you can get with EI stacks.

And I did see a write up of a fellow whose high-dollar hifi setup started humming. A lot. But only after dark. The power amps had toroidal PTs. A great deal of fire drill activity ensued, and only ended when he remembered putting those light-dimmer life-extending pellets in his outdoor garage lights. Got the pellets in the same electrical direction for both of them, and this introduced a diode's worth of voltage offset in his AC ///voltage///. There's a lot of current available there, at least the total current offset made by the garage lights and the resistance of the toroid primary windings, so it walked the toroids up to saturation. Removing the diode pellets fixed it.

Toroids are one of those be-careful-what-you-wish-for things.

Short and simple: For a given core, saturation determines the minimum number of primary turns for max. peak current to be expected.

Not exactly. Core saturation determines the number of turns and core area needed to withstand a certain amount of volt-time integral. That's not directly related to the peak output current to be expected. Primary magnetizing current is not the same as reflected secondary current. The two are only very indirectly linked. It's not in general possible to saturate a transformer core from the secondary. There is an exception to this for half wave rectification in small transformers with high resistance primary windings, and I have never seen this exception demonstrated.

You pick wire sizes for the secondary and primary so that the full load current (or max peak rms average ) can be managed without spending all your power heating transformer wires. Given that wire cross section, you proceed to pick a transformer with enough core area and window area to get your wires inside the window and still get enough turns in to get a primary inductance that does not let you get into saturation (very far... saturation is a soggy, soft slippery slope with most transformer iron), then you go back and see if you need to increment wire size because you've wound so many turns, and then so see if you have to add more stack to reduce turns by adding core area or jump to the next core size.

If an AC-line transformer is running a no-load magnetizing current more than a few percent of its full load current, it's defective or of low quality in materials and/or design. If your employee turns in such a design, you educate him/her or fire her/him. Given that no-load current is and must be so small, the current to magnetize the core is inconsequential in dealing with the peak loads.

The OTs' primary inductance is not constant at all. As my measurements (above) and manufacturers' information show, L rises with increasing current typically by around a factor of 10 (even though there is some unavoidable airgap) up to a maximum from where it decreases steeply.

That is correct. Core inductance is very, very much not a constant. The nature of the iron and the BH curve it traverses, including minor loops and offsets make sure of that. A conservative OT designer gets familiar with his iron and picks some minimum value to be used, then designs as though the core will always show that minimum inductance, all the while knowing he may and almost certainly will get more at some operating conditions.

Yes. I seem to remember from a theory class, in the dim and distant past, that the full formula is:
V = L di/dt + i dL/dt
But if inductance L is constant, the second term vanishes, of course.
In iron-cored coils, where reluctance is varying, the inductance of the coil would also vary, making the situation complicated.

That is correct. Most transformer practice was codified long before we had machine ways to handle systems of partial differential equations. The transformer designers of that time went off and made up charts and graphs of measure quantities of every transformer they built and use that as a way to iterate in on a design. The history book of transformer designs made and measured was a highly treasured intellectual asset of every transformer design shop. A great deal of transformer practice depends on knowing some kind of bounds on what you might see, then going off and designing with small variants and extrapolations of what you did see.