Why Single-Ended?

The Tim Paravicini-designed EAR-Yoshino tube circuits are all single ended, the pre-amps and power amps having transformers on the XLR/balanced inputs and outputs.

Some one please explain to me how an SET amp can be fully differential. I see some SET amps that are offered as balanced. Could one of you explain the circuitry in the amp that makes the use of XLR connections advantageous?

All amplifying devices are differential in nature; a tube or a transistor. This is so since if the signal is the same at both inputs to the device (in the case of a tube, the grid and  the cathode) the device will not amplify. This is because it looks at what is different between its inputs.

Most designers don't take advantage of this or don't realize it, so on nearly all SETs the XLR input is not balanced (although some, like the Viva, have input transformers and so can operate either way). However it should be obvious that it is possible to set up the amplifier to use the other input (which will be the cathode of the input tube) by tying it to pin 3 of the XLR while the grid of the tube is tied to pin 2 of the XLR. This technique is not balanced, but it is certainly differential and retains many of the advantages of balanced operation (such as noise rejection). The cathode input is relatively low impedance and some preamps may not be able to drive it (although ours have no difficulty in this regard).

This technique was originally used by George Philbrick who is generally credited with designing the first practical opamps, which were vacuum-tube (as a side note he was not the inventor of opamps although he often gets credit for that too).  

You can feed a single ended signal into a fully balanced preamplifier, ground the negative (inverting) amplifiers positive input - connect the two negative inputs of both amplifiers (non inverting and inverting) together and a balanced signal will appear at the output of the preamplifier and the signal would be fully balanced from here on.

This is certainly true- we do it with our preamps all the time, but its a simple fact that the preamp is accepting the input as a single-ended signal, with the weakness that the cable becomes part of the sound. The preamp then converts the signal (via its differential operation) to a balanced output. So you can see that in this example that the signal was either single-ended or balanced, but never both at the same time.

Phase has to be inverted somewhere, either for TRUE balanced (XLR cable) or FULLY balanced design.  It involves extra circuitry that does not make sound more transparent.  For short connection, in electrically quiet environment, single ended design might be a better choice.  As for the FULLY balance design - it offers slightly* better common mode electrical noise rejection at very high frequencies and cancellation of even harmonics produced by the amps.  I don't care for both since my connections are short, while shields and twisted pair, in my XLR cable, work really well.  Cancellation of even harmonics, produced by the amp, makes it sounding colder, while odd harmonics, responsible for brightness, are left intact.  Also, FULLY balanced amp has practically two amps inside and costs much more.  I could buy much better single ended amp instead. That's why I would never buy FULLY balanced amp ("Fully" is not always "Better").

* In order to provide good common mode rejection two halves of FULLY balanced amp cannot be independent.  Negative cross-feedback has to be used to equalize gains of each half and that might be far from perfect.

Hello Kijanki,
That was a very interesting and well reasoned post describing some of the inherent advantages of single ended topology. Just as Ralph presents a compelling case for fully differential balanced circuit.   It is quite clear why both topologies  can result in excellent sounding audio components. 
Charles 

+1 Charles.

Regarding this comment in Kijanki’s post:

In order to provide good common mode rejection two halves of FULLY balanced amp cannot be independent. Negative cross-feedback has to be used to equalize gains of each half and that might be far from perfect.

With some fully balanced architectures the need for negative cross-feedback, or even any feedback, can be minimized or avoided. For example, Ralph’s designs employ an architecture based on differential stages, and are typically spec’d as using just 1 or 2 db of feedback. And in the solid state domain many fully balanced Ayre amps employ zero feedback, and I believe most or all fully balanced Pass amps use very minimal amounts of feedback.

Also, as stated in one of the papers at Ralph's site, "for a given number of stages of gain, differential amplifiers have about 50% more parts," rather than being closer to the equivalent of two single-ended amps that are otherwise comparable. Although with some other balanced architectures the parts count may indeed be close to double.

Best regards,
-- Al