Slew Rate & Transformer coloration

[jbuntz]

I have a few questions:

How does a transformer distort the low end more than the high end? I almost seems like you would have to have a compressor with a say a 15 ms attack and release to simulate this.

Does slew rate distortion increase gradually as it approaches maximum mA/s or does it do a hard "clip" once it reaches the maximum. If it is gradual, is it a linear increase?

Is slew rate somehow related to the distortion that is more weighted on the low end?

[Geoff_T]

Originally posted by jbuntz
I have a few questions:

How does a transformer distort the low end more than the high end? I almost seems like you would have to have a compressor with a say a 15 ms attack and release to simulate this.

Does slew rate distortion increase gradually as it approaches maximum mA/s or does it do a hard "clip" once it reaches the maximum. If it is gradual, is it a linear increase?

Is slew rate somehow related to the distortion that is more weighted on the low end?

Hi

In simple terms, transformers work on a.c. signals... they don't like dc but can be wound to handle 100KHz at t'other end.

As the frequency drops below about a 100Hz you won't notice much difference at regular +4dBu line levels and it will happily go down to 20Hz with a -3dB around 4Hz before the level drops like a stone.

If you crank the level up to around +20dBu and scroll down from 100Hz a whole different picture appears because the core starts to saturate and it gets worse and worse as the frequency drops. Even a chunky Neve transformer will be nudging whole per cent figures down around 20Hz.

This is a function of decreasing the frequency and increasing the level... I think tossing in "slew rate" clouds the issue a little, especially at 20Hz.

[Paul Frindle]

Originally posted by jbuntz
I have a few questions:

How does a transformer distort the low end more than the high end? I almost seems like you would have to have a compressor with a say a 15 ms attack and release to simulate this.

Does slew rate distortion increase gradually as it approaches maximum mA/s or does it do a hard "clip" once it reaches the maximum. If it is gradual, is it a linear increase?

Is slew rate somehow related to the distortion that is more weighted on the low end?

The business surrounding transformer related distortion is very complex indeed and is also highly dependent not only on the transformer design but also the conditions in the circuits and loads that surround it. The following is a copy of a post on this stuff I did on another forum some time ago. I hope it is helpful:

>>>>
Oh dear - how much time have you got? I resisted the tube versus transistor discussion because it would go forever, but this is worse by miles :-( I have been designing and winding all sorts of transformers since the 1960's and the story is long indeed. I'll put some issues down of the top of my head.

Winding resistance loses your signal level depending on load.

Input parallel inductance causes loading to increase as freq reduces - potentially messing up bass performance and causing LF phase shift.

The core of the transformer increases inductance (and impedance) by hundreds or even thousands of times - but core materials are not linear. They exhibit B/H curves that has hysteresis at low magnetisation (history dependent remnance), a partially linear(ish) portion in the middle and then reducing magnetic efficiency as we get closer to saturation. So the impedance is non-linear over the whole range and is heavily related to freq and level. This is effectively in parallel with your signal and can cause loads of complex distortion at all levels particularly if driven from a circuit with higher source impedance that in susceptible to being loaded down (600R audio systems come to mind). Some designers actually use this effect to produce pleasing distortion profiles that give warmth and what they call 'LF precision'.
The HF bias in analogue tape machines was developed to overcome these problems in magnetic tape and produce recordings with lower distortion than the DC bias schemes that preceded it. It also doubled the max level you could record onto the tape because you were using both +ve and -ve parts of the B/H curve.

The increasing primary current as freq goes down (due to inductance) increases magnetic field and can cause the core to saturate. Saturating the core radically reduces the inductance further (therefore increasing current even further) so the transformer gives up suddenly and 'clips' causing massive LF distortion - mostly odd orders often with strange shapes and overshoots. The only way to avoid this is to exponentially increase the core mass as freq requirements go down. But apart from being large and heavy, having more core material increases hysteresis errors and other HF losses - so it can be self-defeating in some cases.

Residual DC in the signal will cause this potential for saturation and non-linear B/H curve to be biassed one way or the other - causing premature clipping and even order harmonics as well. This can happen even with AC coupled systems if the waveform is assymetrical. DC injection was the biggest problem that plagued early single ended power amps, and also affects push-pull designs that have unbalanced output tubes etc. Typically single ended designs have air gaps to reduce the total u of the core, but this means a drastic cut in inductance - so bigger then ever transformers are needed - or you have to roll-of the LF - or put up with it being rolled-off by the tranny anyway.

BTW some early original Neve designs used a current source to drive the other half of the transformer to null out long term DC (kind of push-pull but with only one stage driven with signal). This had the advantage of allowing historically popular single ended style distortion and overload characteristics without needing a vastly oversized tranny - part of the much loved Neve sound.

Capacitively coupling into signal transformers to lose the DC produces a series tuned circuit between the cap and the transformers input inductance. This can put a boost bump in the response at LF which is load dependent as well. And can even cause a level related freq response bump that rises in centre freq as the inductance decreases when the core becomes saturated as LF levels increase etc. I have seen this happen to drastic levels of 10dB or more - but at the time the sound of this was wanted and integral to the reported 'punch and warmth' in the console sound. So even after I discovered it I had to leave it in place for a while anyway. What they had was a freq and level selective dynamically expanding LF boosting EQ on every channel!!

Incomplete primary to secondary coupling causes leakage inductance that ends up effectively in series with your signal - potentially (inevitably in fact) causing HF roll-off and/or ringing as well HF phase shift. You either have to chose ringing or some roll-off which is load dependent. This is the main cause of the 'HF boost when unterminated/HF loss when terminated' problem we used to see with Ampex tape machines like the AG440 etc.. The Jensen transformers had a multiple interleave of several segmented primary and secondary windings to increase mutual coupling.

Distributed winding capacitance loads your signal as freq goes up and rings with the leakage inductance even without external capacitive loads.

Conductance in the core material caused losses from 'eddy currents - partially shorted turn syndrome' and loads your signal down increasingly as freq increases causing HF losses. Non-conducting material such as ferrite can be used to avoid this (especially for HF and RF work) but generally it has a much lower magnetic advantage (because it is partially loaded with non-magnetic material), meaning you need loads more turns for audio band use, therefore yet more resistance loss and larger leakage inductances etc..

If you try to put negative feedback around the transformer to reduce all the above mentioned issues you find that only small amounts are possible before instability sets in, because of all the phase shifting etc.. In fact performance my even get worse because of tendancies to oscillation caused by both LF and HF phase shifts. Sustained LF instability my occur during overload saturation that cannot recover when the signal is removed and HF instabilty may occur from loading by long interconnects etc..

Making transformers do audio even acceptably over a freq range of 4 decades is a very very tedious business.
>>>>

[jason]

WOW!

Does anyone know of any good internet resources or texts that go in depth about audio transformers? The texts I have either completely skip them or have no more than a few pages about the basics of them.

[Paul Frindle]

Originally posted by jason
WOW!

Does anyone know of any good internet resources or texts that go in depth about audio transformers? The texts I have either completely skip them or have no more than a few pages about the basics of them.

I'm sorrry that I don't know of any really comprehensive resources of this info. The above is just some of what I have found out over the years trying to design these things.

[Bernd G]

There is a good introduction by a variety of experts incl Bill Whitlock and Steve Dove found in the "New Audio Encyclopedia", BETTER yet in THE HANDBOOK FOR SOUND ENGINEERS, Third Edition, Glenn Ballou, ed.

[jbuntz]

Where can that be found?

[Bernd G]

Originally posted by jbuntz
Where can that be found?

Amazon - not cheap ($ 120), but probably one of the only few books on audio you will ever need. Or, find it in your nearest library and xerox the relevant pages. Another thing: I think most of the information on Audio Transformers by Bill Whitlock in the Handbook for Sound Engineers is actually available on the Jensen website.

Cheers