Output Transformer driven hard, What does I sound Like ?

[Helmholtz]

The real ‘proof of the pudding’ would be to see an OT input current waveform with spiky current peaks at the instants in time where saturation is occurring.

I agree, but I can't imagine how a constant magnetizing inductance could cause the observed voltage distortion, even taking into account increased magnetizing current at that low frequency.

I don't think that transformer steel has much hysteresis loop area as that would mean high core losses.

[Malcolm Irving]

…
I don't think that transformer steel has much hysteresis loop area as that would mean high core losses.

Yes true. Maybe the shoulder in the downward-going side of the positive part of the wave is a symmetrical match to the shoulder in the corresponding upward-going side of the negative part of the wave. Perhaps, both are caused by the loading effect of an approximately sinusoidal magnetizing current and hysteresis is not necessary to explain it?

[Malcolm Irving]

... Perhaps, both are caused by the loading effect of an approximately sinusoidal magnetizing current …

I think I'm wrong there (what? again? ). If the excitation voltage is single-frequency sinusoidal and all the components are linear, then all the voltages and currents have to be sinusoidal.

[Malcolm Irving]

The following is based on a single-ended (class A) stage, which is easier to explain, but similar logic applies to class AB push-pull.

The Output Transformer, looking into the primary side, can be quite accurately modelled by an ideal transformer which reflects the secondary load, and a magnetizing inductance in parallel with that reflected load. Let’s say the reflected load is 2K ohms (resistive) and the magnetizing inductance (also called the primary inductance) is 70H.

When the output stage is at idle, DC current passes through the magnetizing inductance, but nothing through the reflected resistive load.

If we now drive the output pentode between full-conduction and cut-off, the plate current is the sum of the AC current through the reflected load plus the DC current through the magnetizing inductance. The DC current through the magnetizing inductance stays practically constant, because the inductance is very large.

If the tube enters cut-off for a period, the DC current in the magnetizing inductance tries to stay constant, but can now only complete its circuit through the reflected resistive load. This current will decay with a time constant L/R (which for our example values is 70/2000 = 35 mS).

For low frequencies the exponentially decaying current (which produces a voltage across the resistive load) becomes noticeable and accounts for the ‘missing chunks’, or ‘shoulders’, in the output waveform.

If this explanation is correct (don’t hold your breath ) a shoulder in the scope plot would be the result of a period of cut-off of an output tube, rather than OT saturation.

[Helmholtz]

Not sure if this picture is complete and comparable to a class B, PP stage, where magnetizing current is pure AC.

I still think that beginning saturation is the most plausible explanation of the voltage distortion observed. But I can't completely exclude that high magnetizing current at low frequencies and non-linear interaction with the tube could produce a similar effect without saturation. The vector sum of load and magnetizing current might have a phase shift relative to voltage between 45° and 90° and produce this kind of distortion.

When I eventually find the time I might do some saturation test with OTs at line frequency with step-up transformer + variac and determine critical voltage-time integral values.

[Malcolm Irving]

Not sure if this picture is complete and comparable to a class B, PP stage, where magnetizing current is pure AC.
…
When I eventually find the time I might do some saturation test with OTs at line frequency with step-up transformer + variac and determine critical voltage-time integral values.

Yes, true for class B, but for class AB there is a DC current component through each tube that corresponds to the bias current. Not sure if we can call those 'magnetization currents' during normal operation, as the magnetizing effect of the bias current on the push side is cancelled by that on the pull side. However, while one side is in cut-off, the current through the magnetizing inductance on that side will decay, the DC balance will be lost and magnetization will occur.

Test results, or simulations in which all the relevant factors are included, would be very informative.

[Gregg]

This is a video of a 50W toroidal OT going into saturation in real time. Frequency swept and amplitude are shown in the window below (link will expire in 30 days).

https://ufile.io/yv6ab20v

[Tony Bones]

This is a video of a 50W toroidal OT going into saturation in real time. Frequency swept and amplitude are shown in the window below (link will expire in 30 days).

https://ufile.io/yv6ab20v

Very interesting. Thank you!

[teemuk]

Yes, a good video. Few years ago I posted some links to oscilloscope plots of a saturating output transformer to another one of these threads. The videos displays pretty much the same effects.

Some of my observations (and they haven't changed at all within the years):
It is not “nice” or “musical” distortion by a long shot. I can only think that this kind of distortion produces very high order harmonics of distortion, both even and odd. Also saturation does not produce the typical "pattern" of generic clipping distortion with gradually diminishing upper order harmonics. Instead the "pattern" is something totally different. ...And it gets worse the deeper the saturation state is. I can’t imagine the results combined with IM would be any better. On a contrary. Basically, looks similar to a SS power amp going to current limit protect mode and instead of traditional peak clipping its “chopping off pieces of the waveform”. It will sound horrible. As is, terms like “fizzy farting” would likely describe that tone most adequately.

Luckily, in order to produce the horrible tone of a saturating OT we need a bass or generic “audio” input stretching down to about 30 Hz and below. Not viable with guitar, maybe with bass. Maybe. Most bass amps don’t try to produce the fundamental in full amplitude either, by the way. It just would not sound that good.

Any well-designed OD channel of a guitar preamp that doesn’t sound like mushy farting, on the other hand, will start to roll off lower frequencies already around 600 – 1 kHz (!!!) so there’s no chance that such signals could saturate a generic OT. Even cutoff set at @80Hz to produce full bandwidth of the instrument could not produce a signal that could saturate a generic OT.
Personally, I’m not familiar with anyone who would actually prefer “full range” distortion tones over distortion tones with pre-enhancement that drastically cuts off lows.

So, saturation is a phenomenon that exists but we don’t really have to care about it with usual signal inputs and due to overall channel voicing typical to guitar preamps. I think the glorious tone of a saturating OT is nothing but a myth and in practice we don’t hear any of such effects in generic amps working properly.