How could I miss this thread... I´ll throw in some thoughts:
A friend of mine who´s one of the hot mastering guys here in Germany talked me into experimenting with transformers many years ago. He was reading a lot of talk about the sound of transformers in the internet and wanted me to DIY him a box with switchable transformers just to find out for himself if and what it´s all about. I was breadboarding something modular based on Eurocards in a Schroff Eurocard tray with a relay switching system to be able to chose any combination of transformers in series. My transformers of choice for first experiments were the Edcor line from their smallest core size upto XSM series. These should represent the steel core transformers. Other contenders were a few OEPs, representing the 50/50 Ni/steel core. I had some more trannies (Lundahl with MuMetal Core and Beyers with pure Nickel core), but they never found their way into that thingy because I stopped working on it for some reason. But the results from early tests made me chose the Edcor PC600/600 and OPE A262A2E in the switchable output stage of the Filterbox 500 project I started posting a few weeks ago in the Filters/Equalizers section.
In a mastering scenario there´s a few things to consider with a transformer switcher as mentioned above.
1. Insertion loss, which is different for each transformer. Mastering guys don´t want to have level changes due to inserting gear into their chain. Not to say that they are anal about it. They are neurotic control freaks w.r.t. levels. So a gainstage is necessary to compensate for the loss of single or multiple transformers in the signal path. In an ideal world the gain compensation is done automatically when a transformer is switched into the signal path. And since I frequently have to calibrate gear levels in various studios it makes sense to be able to calibrate each transformer individually due to their tolerances.
2. A passive transformer box will vary results depending on the gear connected up- or downstream due to impedances of that gear. In mastering often the order of gear in a chain will change depending on the engineer´s goals. That´d give different sonic results for a certain transformer combination, hence is not a very predictable system. This calls not only for a gain compensation but an input and output buffer to interface the transformers without retroactive effects from the circuits around the transformer box. The sonic result will be much more predictable when the transformers "see" the ever same impedances.
3. The signal levels through the transformers are the key to how the sonic imprint of the transformers is percieved. Now if you switch multiple transformers in series then inevitably certain transformers in the chain will not work in their sweetspot (swampy terrain, I know), while others may do. So if one would like to go the extra mile then this could be addressed by adding an input buffer in front of every transformer which raises (or lowers) the level sent into the iron. It should be trimmable of course, because the sweetspot for sure is a subjective thing. And behind each tranny should be a pad or eventually another gainstage, to bring the signal level back to a nominal level before it either leaves the box or enters the next transformer stage.
When my thought process reached point 3. I basically lost interest in DIYing that thing because at that time I was breadboarding only. That for sure would be more work than my DIY enthusiasm tells me to do.
Very interesting is Jakob´s approach to add something chaotic to the system. I was brainstorming that, too. Because as you can see in 3. a transformer switcher (in stereo!) quickly piles up a lot of electronics as soon as you want to have a reasonable choice of transformers, hence choice of sounds. To keep that on a level which can be handled my thoughts went into how a single transformer can be manipulated to create different sounds by a flick of a switch. Following ideas came to my mind, but all is theory and untested:
- superimposing DC bias in a (gaped) transformer to saturate the core to a certain degree so that virtually the core diameter turns into an adjustable parameter.
- using a tertiary winding to suck some of the power frequency dependent out of the core. This could be done by any form of passive circuitry, e.g. something RLC + various types of diodes come to mind.
Both ideas will change level and thus call for gain compensation.
- using some EQ to boost a frequency band before hitting the tranny. This will create harmonics dominated by the chosen frequency band. Right behind the tranny another EQ with reversed characteristics to the first one will equal out the earlier boost, but leaving the harmonics created in the transformer alone. That´d give us following parameters and controls: the combined boost/cut control will be responsible for the amount of harmonics while the chosen frequency band will be responsible for the frequency from which the harmonics set in.