Output Transformer Saturation

[Steve Conner]

we were getting DC through a transformer with no copper connection between primary and secondary. We did file patents on that one....

Sounds like you made half of a saturable reactor.

why not two different pieces of iron? One saturates positive, the other negative, and then we combine those.. er, somehow...

Sounds like you want two halves of a saturable reactor.

Let me explain further. To me, saturable reactor means an inductor whose inductance can be varied with a DC bias. The things they used to use as lamp dimmers and welding current controllers before SCRs were invented.

But the variation is done by saturating the core more or less. And since its behaviour is symmetrical, and the flux due to the load current must add to the flux from the control current, then it must produce that symmetrical distortion.

They're made exactly how you say: a different part of the core saturates on each half-cycle, and in fact you can make a saturable reactor from two regular transformers. Primaries in parallel as the load winding, secondaries in series as the control winding. The load windings pass alternate half-cycles, so each transformer is actually outputting a net DC current, the reason for my comment above.

So, for instance, you could connect a saturable reactor across the speaker terminals of an oversized OT, and get variable saturation.

As a professional magnetic designer, I'm sure you can figure out how to combine the OT and the saturable reactor into a multi-legged magnetic monstrosity.

true core saturation can be a very ugly thing.

I think that the better the quality of the OT, the more bass it'll handle before saturating, but the uglier the saturation sounds when it finally happens. Toroid OTs are the worst. You can make a saturable reactor out of two toroid cores, but the saturation is so hard that it practically switches like a triac instead of varying its inductance smoothly.

If you want to hear true core saturation, you can hook a guitar amp up to a bass cabinet and play bass through it.

[R.G.]

I think the saturation people talk about in OTs is generally a bunch of misinformed hype. I doubt saturation is as prevalent as people think, as true core saturation can be a very ugly thing.

I think you're absolutely correct. "Output transformer saturation" is another of those terms which sounds cool to say if the person you're talking to doesn't know what it means.

Sounds like you made half of a saturable reactor.

Actually, it was two halves of two saturable reactors plus a flyback transformer on the same core.

Let me explain further. To me, saturable reactor means an inductor whose inductance can be varied with a DC bias. The things they used to use as lamp dimmers and welding current controllers before SCRs were invented.

Yep. Been there, got the T-shirt. Not many people know that before saturable reactors, theaters used the leftover brine in the barrels that pickles were shipped in to dim lights. One copper plate in the bottom, one on a variable height-setting rig, and voila! a 5kW dimmer resistor.

But the variation is done by saturating the core more or less. And since its behaviour is symmetrical, and the flux due to the load current must add to the flux from the control current, then it must produce that symmetrical distortion.

Actually, it's done by making the AC flux much smaller than the DC flux so the variation in reactance because of the AC signal varying the saturation can be ignored. Magamps, the AC version of saturable reactors, often use the squarest-loop material they can; the "gain" of a magamp in terms of change in output versus change of input is biggest when the smallest control signal pushes it over the saturation edge, and this is easiest where the saturation edge is sharpest.

So, for instance, you could connect a saturable reactor across the speaker terminals of an oversized OT, and get variable saturation.

Yep - as long as you don't insist on partial-saturating (for that nice compression on both positive and negative peaks that is what people probably mean but don't understand when they say) for both directions of signal at the same time. You can't work on both saturation edges of the BH curve at the same time.

As a professional magnetic designer, I'm sure you can figure out how to combine the OT and the saturable reactor into a multi-legged magnetic monstrosity.

I'll work on that. Just as yet, I haven't come up with a good way. The simple ways aren't good.

true core saturation can be a very ugly thing.
I think that the better the quality of the OT, the more bass it'll handle before saturating, but the uglier the saturation sounds when it finally happens.

Probably true.

Toroid OTs are the worst. You can make a saturable reactor out of two toroid cores, but the saturation is so hard that it practically switches like a triac instead of varying its inductance smoothly.

I used that effect in a flux-gate magnetometer I made once. Pretty neat trick: a few windings on and around a toroidal core and you can have an electronic meter which tells you degrees from true local magnetic north in a 0-360 sense. I think Hall effect devices have gotten sensitive enough to use those for compasses now though.

If you want to hear true core saturation, you can hook a guitar amp up to a bass cabinet and play bass through it.

You can for sure do that!

I think people mistake secondary overloading for saturation. It's not widely appreciated that you simply cannot saturate a transformer from the secondary. Power out of the secondary was never in the core in the way that people normally think of it, so secondary loading cannot cause magnetic saturation.

[Old Tele man]

Sounds like you made half of a saturable reactor.

Sounds like you want two halves of a saturable reactor.

..yep! that's the term I was mentally stumbling around searching for (been too many years since I studied and worked with magnetic-amplifiers).

1) Here's the wiki illustration:



2) Here's the USN NEETS illustrations: http://www.tpub.com/content/neets/14.../14180_136.htm

3) Here's a Toshiba Application Note: http://www.toshiba.com/taec/components/Generic/M1.pdf

[R.G.]

..yep! that's the term I was mentally stumbling around searching for (been too many years since I studied and worked with magnetic-amplifiers).

1) Here's the wiki illustration:



2) Here's the USN NEETS illustrations: Figure 3-32.—A simple saturable-core reactor circuit

Good illustration of both the concept, and the problem I'm getting at. The current from battery B through resistor R (and any wire resistance, even though that's usually negligible in setups like this) sets up a flux B in the core proportional to N*I. The flux is single-directional, as suits the DC current which sets it up. The current through the signal coil is AC, which means it alternates between directions which either aid or oppose the flux set up by the direct current its winding.

=> The flux in the single core is the resultant of the sum of the two MMFs, one from the DC coil, and one from the AC coil. <=

We can have several cases.
1. AC-created flux is much, much smaller than the DC flux.
In this case, the circuit works as the generic proposition suggests. The impedance of the AC coil side is modulated by the level of the DC current, and only trivially changes the core's flux density. So the inductance as seen by the AC source is modulated by the DC.
2. The AC created flux is much, much bigger than the DC flux.
In this case, the DC has the effect of making one of the AC peaks go over the saturation edge. In that case, the AC is largely unaffected by the DC until it hits the saturation threshold, then the core saturation decreases the incremental inductance for only that part of the wave, and you get a current spike from the AC source for the saturated part of the wave.
3. The AC and DC created fluxes are similar in magnitude.
In this case, like (2), only the part of the AC wave that pushes the core into saturation gets the reduced reactance limiting.

All of this works, but none of them are the output transformer saturation effect the way the "common knowledge" wrongly perceives it.

[lowell]

Ok, so, back to why I originally posted. One channel of my amp is a 5E3, however the power amp has a larger OT and I'm using a ss rectifier. I have lowered the plate voltages to 250v, and rebiased to 90% Wa so as to attempt at getting "the sound." My amp is not clipping nearly as much as a stock 5E3 does. I also changed the Concertina PI load resistors to 100k as to allow more signal to reach the 6V6 grids. (I have read that the PI clipping contributes to the 5E3 distortion character so maybe this is a step in the wrong direction) My conclusion is that I must be missing "the sound" due to the larger OT. You guys are saying that OT saturation sounds ugly and is undesirable, however if that's what I'm missing then I think it's very desirable. Anyone know how much OT saturation contributes to the 5E3 sound?

[R.G.]

Anyone know how much OT saturation contributes to the 5E3 sound?

I'm guessing that the answer to that one is probably "no". Just a guess.

[teemuk]

Some people claim that this whole "OT saturation" is actually just an effect caused by reduction of negative feedback during clipping. Since NFB can increase bandwidth at the cost of reduced gain (gain-bandwidth product) the decrease in negative feedback also means the overall bandwidth will decrease towards the bandwidth of an open loop amp, in which case it is limited by the "deficiencies" of OT, such as poor bass response.

If the OT performs excellently even in open loop condition (e.g. Partridge transformers of Hiwatt amps) the effect of bandwidth reduction will be negligible. Hence bigger OTs are mistakenly thought to introduce less "saturation".

Given that I've never seen any kinds of scope captures of this supposedly smooth OT saturation (quite contrary) and very few can even seem to explain what it's actually about, I'm more inclined to believe the NFB theory.

[ThermionicScott]

Ok, so, back to why I originally posted. One channel of my amp is a 5E3, however the power amp has a larger OT and I'm using a ss rectifier. I have lowered the plate voltages to 250v, and rebiased to 90% Wa so as to attempt at getting "the sound." My amp is not clipping nearly as much as a stock 5E3 does. I also changed the Concertina PI load resistors to 100k as to allow more signal to reach the 6V6 grids. (I have read that the PI clipping contributes to the 5E3 distortion character so maybe this is a step in the wrong direction) My conclusion is that I must be missing "the sound" due to the larger OT. You guys are saying that OT saturation sounds ugly and is undesirable, however if that's what I'm missing then I think it's very desirable. Anyone know how much OT saturation contributes to the 5E3 sound?

Did you drop the voltage for the whole B+ rail or just to the output stage? Lowering voltages to the preamp will drop the gain -- that might be why you're having trouble driving the output tubes.

- Scott

[Diablo]

Well, Bruce is the king of the 5E3, so you should listen to his experience on transformer choices. I'm sure he has tried more than a few, and knows the relationship between the iron and the tone.

[Diablo]

Back to the theory on transformer saturation.
So how can you change the saturation point for the iron?
Change the permeability of the iron by changing the stress or temperature of the iron. Is it practical to do this? Probably not.

[ThermionicScott]

I just had another thought (sorry about the burning smell in here!) The Hammond 1600 series OT's have two secondary windings -- the idea is that you can wire the two together to make any impedance and use the windings more effectively. If you just hooked up a speaker to one of the secondaries, you'd be able to run DC through the other one and get core saturation that way, right?

(Not that it's the path to tonal Nirvana or anything. )

- Scott

[lowell]

Did you drop the voltage for the whole B+ rail or just to the output stage? Lowering voltages to the preamp will drop the gain -- that might be why you're having trouble driving the output tubes.

- Scott

I only dropped the power amp plate and screen supplies (excluding the PI).

[wizard333]

For all this discussion of this, which I admit is fascinating from a theoretical standpoint, I have to wonder:

How would it sound?

And would it make enough of a difference to justify the hoops?

[R.G.]

Back to the theory on transformer saturation.
So how can you change the saturation point for the iron?
Change the permeability of the iron by changing the stress or temperature of the iron. Is it practical to do this? Probably not.

The short answer is that it's impractical. Iron does have a "Curie temperature", that being the temp where all ferromagnetic effects vanish, rather abruptly. It is abrupt as I mentioned, and is up around 500-600C, over 900F, depending on the alloy. It's glowing red at those temps, so it's not very practical to use that for saturation modification. You can include gaps in the iron path. This makes it much harder to saturate because it takes much more N*I to get enough MMF through the gap+iron magnetic field path; however, the iron has the same saturation flux density in a gapped structure that it always had.

One thing you *could* do is a modification to get the "saturation of the thousand cuts", which I just invented. Note that this isn't all that practical, either.

If you had a way to cross-cut the magnetic path so you could get many - like a thousand, two-to-the tenth, current paths through it without sacrificing too much of its magnetic goody-ness, you could progressively saturate pieces of the magnetic path, forcing the flux density in the remaining path higher for the normal working conditions. Like in the flux-gate magnetometer, you can cross off pieces of the magnetic path by saturating the small areas, increasing the net drive in the remaining signal area and changing the signal-to-saturation sense. There are some warts with this, not least that it has to be done symmetrically to cancel the transformer action from the control signal that is saturating individual bits.

It might be possible to mechanically modify the magnetic path by mechanically sliding a plug of iron in and out to form a smaller area of magnetic path that would cause lower saturation for the same MMF by reducing the effective area of the iron path. This could get mechanically tricky to do, as well as being not all that reliable.

I just had another thought (sorry about the burning smell in here!) The Hammond 1600 series OT's have two secondary windings -- the idea is that you can wire the two together to make any impedance and use the windings more effectively. If you just hooked up a speaker to one of the secondaries, you'd be able to run DC through the other one and get core saturation that way, right?

Yes, you could. But it would be the one-sided saturation I've been talking about.

Here's another analogy. I have here in my hypothetical hand an amplifier. It's a high quality amplifier within its limitations. It can reproduce any audio signal with negligible distortion as long as it's not driven outside it's power supply limits. For the moment, let's assume that it has a power supply of +/-11V, and it can swing to +/-10V distortion-free. When it hits +10V, it clips flat. When it hits -10V, it clips flat.

This amplifier has two inputs. One input is for the output bias. The output goes to exactly the DC level set on the bias input. The other input is for AC signal. From this input, the amplifier amplifies the signal by 1 and puts that on the output on top of the DC level that the bias input sets.

We have an analogy to the single-ended versus double ended saturation. When the amplifier receives instructions from its inputs to go to a specific DC output level, and then wiggle the AC signal about that level it does. If we want the maximum undistorted output, About two seconds of thought shows us that the bias has to be set dead in the middle between the power supplies because in that position the signal can swing the maximum to either positive or negative without banging into the power supply limit and clipping.

If we have an AC signal that is +/-10V peak, the signal will be amplified without distortion. But what if we *like* the sound of the signal clipping? We like the sound of a +/-12V signal into the AC input, because it clips on both peaks. How can we do this with a smaller signal by jiggering around with the bias input?

The short answer is - we can't. We can make smaller signals clip on either the positive half cycle or the negative half cycle by offsetting the bias to one side or the other. But there is no setting of bias that will make a smaller signal clip on both positive and negative peaks at the same time. The amplifier's output has only one DC output level. It can't be two different output DC voltages at the same time.

This is not to say that the sound of clipping on only one side but not the other won't be pleasant too. But it's a different thing. That is analogous to what you get in single ended tube power amps.

This whole discussion is about how you would analogously make the amplifier's power supplies smaller to make smaller signals clip. You can't do that with the two inputs you have. You have to modify the power supply voltages (analogous to the inherent saturation properties of the core metal) to do that. And there's no way to dial down both flux saturations of a chunk of iron at the same time.

[ThermionicScott]

It might be possible to mechanically modify the magnetic path by mechanically sliding a plug of iron in and out to form a smaller area of magnetic path that would cause lower saturation for the same MMF by reducing the effective area of the iron path. This could get mechanically tricky to do, as well as being not all that reliable.

Ooh, now we've delved into gun-type uranium bomb design.

We have an analogy to the single-ended versus double ended saturation. When the amplifier receives instructions from its inputs to go to a specific DC output level, and then wiggle the AC signal about that level it does. If we want the maximum undistorted output, About two seconds of thought shows us that the bias has to be set dead in the middle between the power supplies because in that position the signal can swing the maximum to either positive or negative without banging into the power supply limit and clipping.

If we have an AC signal that is +/-10V peak, the signal will be amplified without distortion. But what if we *like* the sound of the signal clipping? We like the sound of a +/-12V signal into the AC input, because it clips on both peaks. How can we do this with a smaller signal by jiggering around with the bias input?

The short answer is - we can't. We can make smaller signals clip on either the positive half cycle or the negative half cycle by offsetting the bias to one side or the other. But there is no setting of bias that will make a smaller signal clip on both positive and negative peaks at the same time. The amplifier's output has only one DC output level. It can't be two different output DC voltages at the same time.

This is not to say that the sound of clipping on only one side but not the other won't be pleasant too. But it's a different thing. That is analogous to what you get in single ended tube power amps.

I get it now. This has been quite the informative thread!

- Scott