What does "transformer coupled" mean?

Here are a few things. If you want a little more discussion on ITs I suggest looking for stuff Lynn Olson has written (some of his posts at AAdiytubes are very informative) as he has been working on P-P IT non feedback amps for quite a while and also Kevin at the the Lundahl forum. ( Im sure there are others too) Much of below is from them.

1) One design advantage is that you can reduce the voltage output of a stage and also get a reduction in driving impedance. The impedance (like OPTs) is reduced by the square of the voltage reduction, so it can be significant.

2) There has been lots of talk about caps and their sonic signature. One way of thinking is that any cap is a bad cap. Using IT coupling may allow you to get rid of a cap in the signal path and the “sonic degradation” some associate with it.

3) I have heard the sound of transformers (we are not talking output but IT) discussed but I have not seen the sound associated with any measurements. In fact, the “sound” seems to buck the measurements rather than follow any as far as I’ve read. This is maybe one of those areas where measurements are not yet helpful or the improvement is imaginary(depending on your way of thinking about it).
4) ITs are gaining popularity. IT-transformer coupling largely disappeared at the same time that feedback became standard practice in the late Thirties as a way of increasing power at low cost. ITs do not do well with lots of non local negative-feedback.
5) As noted in above posts Trannies have their weaknesses. even the best transformers have at least a 12dB/octave roll-off at both ends of the frequency range - they are always bandpass filters. And a smooth rolloff (on both ends) is fairly unusual. Usually you see ripples in the time and frequency domains. When the transformer "sees" a high impedance on the primary, secondary, or much worse, both, the bandwidth decreases. This is part of the reason that IT's are notoriously harder to design ($$$$$$) than output transformers: at least the OPT sees a low impedance on the secondary ( the speaker load.) With an IT, the secondary is near-infinite, and the source Z on the primary is controlled by the Rp of the driver tube(s).
6) They are expensive to design make.
7.) Other things: They break the signal ground, isolate RF trash from the input stage, they conveniently filter off ultrasonic distortion components from preceding stages (preventing IM crossmodulation), they give very precise phase-splitting and re-summation (if designed for the task), and protect the speakers from nasty DC offsets that could destroy them.
Cheers
I remain

Denis, It's an interesting question indeed and so far very interesting answers as well.
Clueless, I'd like to clarify if you mean IT as an Ideal Transformer?...

I was curious about transformers since my early youth and was devoted to electro-magnetizm that is widely applied in our turntable cartridges, tape heads, speakers and transformers thanks to Faraday's electro-magnetic force emf.

Secondary circuit induced emf opposes direction of primary circuit's source voltage(or signal) and so is their reactances(mainly inductive). In ideal transformer inductive reactances cancel out each other thus making such transformer free of any magnetic losses and independed of ANY freequency changes.

A real transformer have magnetic losses due to resulting inductive load and electric losses due to the wire resistance. More powerful transformers have larger coiled wire lengths, thickness of ferromagnetic core and so is larger magnetic and electric losses especially on higher freequencies due to resulting inductive load i.e. acting as a low-pass filter.

Audio freequencies considered to be technically low freequencies. A transformer will always be very close to ideal in lower freequencies than slowly loose its responce as the freequency goes higher but almost flat(compared to the capacitor) to the tolarable freequency range.

If we compare it to the coupling capacitor, well,
caps are having an issue of correct selection of the resulting reactance to fit tolerable dynamic and freequency range for the further stages and in general can bring much more distortions and/or roll-offs to the audio circuit than transformer. The main advantage of the cap is the SIZE and PRICE compared to transformer.

The advantage of transformer as a coupling device is to the degree of dynamic stability in its tolerable freequency range with further presence of roll-offs in upper region. Many of us now can realize that these roll-offs are the main reason why transformers "sound" so damn good as they're devices with naturaly built-in garbage filter.
If roll-off touches too much an audiable freequency we can apply a small feedback and make tranny's load responce much wider.

Sorry, IT=interstage transformer.

I remain,

Marakanetz stated: "If roll-off touches too much an audiable freequency we can apply a small feedback and make tranny's load responce much wider."

This sounds like you are taking one negative ( transformer induced frequency response errors ) and trying to correct it with another negative ( increased negative feedback ). In most instances, at least when it comes to audio, two negatives don't make a positive.

Audio Xpress had an interesting article about an amp that was somewhat similar in design to the Marantz 8 / 8B and the trade-offs that they had to make when building that amp. The unit built & described in this article showed the differences in bandwidth / linearity using various amount of negative feedback, etc.. while making use of what were supposed to be "high tech" wide-bandwidth toroidal output transformers. None of the results looked real good to me as each approach seemed to solve one problem and create another. As a side note, the measured distortion on this amp was pretty attrocious too.

Having said all of that, much of this does not really apply to interstage or "coupling" transformers due to the low current levels in use.

Clueless stated: " 2) There has been lots of talk about caps and their sonic signature. One way of thinking is that any cap is a bad cap. Using IT coupling may allow you to get rid of a cap in the signal path and the “sonic degradation” some associate with it.

My thoughts about this are that you have simply traded one sonic signature for another. While every cap made will have its' own characteristics, so will a transformer. The ease of changing a cap within a circuit is FAR easier than doing the same with a transformer. This is not to mention that one literally has dozens upon dozens of caps to choose from / manipulate whereas the options with transformers are rather limited ( in most cases ).

Clueless stated: "4) ITs do not do well with lots of non local negative-feedback.

Personally, i don't know of any gear that uses a lot of global feedback that sounds real good. It might measure quite well in specific areas, but sonically, it is a mess. Can you say "sterile & lifeless" ???

Clueless also stated: "7.) Other things: They break the signal ground, isolate RF trash from the input stage,"

My thoughts about this are that any transformer that is not well designed / shielded becomes a source to pick up / pass on RFI. Transformers work by creating a magnetic field and varying the energy transferred within that field. RF based signals are nothing more than "floating fields" and can easily find their way into such designs. Obviously, a poor design combined with high levels of RF would become pretty obvious and HIGHLY annoying. Once again, it is not so much the design as it is the quality of parts used and how well the design is implimented.

Other than that, i thought that some of the points that he covered in parts 4, 5, 6 & 7 were very fair and even-handed. All designs / products have trade-offs and i thought he did a very nice job of highlighting both the pro's and con's of IT's. Sean
>

Hi Sean:

You Said>>
>>My thoughts about this are that you have simply traded one sonic signature for another.

I agree. That is why I use the phrase "one way of thinking...' and talked about "“sonic degradation” some associate with it."
I've not had the chance to listen and test Ts and caps in all different circuits myself (obviously)and I just withhold judgment. Lots of people really do not like caps though and some like the sound of ITs. You are just trading one sonic signature for another..but that is what it's all about, eh?

>>Personally, I don't know of any gear that uses a lot of global feedback that sounds real good.

I do not use NFB myself. I like SETS for sound and simplicity to make. I was just making an observation as to why transformer coupling was not popular from the outset. 1930s-40s. IT was not used largely because NFB-PP amps became really popular and NFB is very hard (impossible) to use with ITs. It was just an observation. Not made in support of NFB. I also said it because the a thought sometimes heard is, as M stated, "If roll-off touches too much an audible frequency we can apply a small feedback and make tranny's load response much wider." This is true (well.. kinda true) for output Ts but will get you into trouble with interstage transformers using nfb for this fix especially on the low side.
(At the same time I've listened to some old vintage stuff (dyna-Eico-Scott) that sounds very good (not great) for the dollar with nfb.”

>>>>My thoughts about this are that any transformer that is not well designed / shielded becomes a source to pick up / pass on RFI. Transformers work by creating a magnetic field and varying the energy transferred within that field. RF based signals are nothing more than “floating fields” and can easily find their way into such designs.

Anything that is “not well designed” or poorly implemented is going to cause trouble in a circuit. If the layout is poor a simple wire is going to be antenna so your statement is impossible to disagree with. Most everything is a design compromise too. Anywaay, thatis why I said IT's are “notoriously harder to design ($$$$$$) than output transformers.” If designed and implemented well they can do good things too. This is true with all things audio.

Cheers
I remain