Some thoughts for your consideration:
Balanced interfaces are used in a professional environment because of the flexibility and reliability that they offer. These qualities are more important than ultimate audio performance. The professional environment requires designs that can accommodate virtually any less than ideal interfacing task with a certain degree of grace. That does not mean perfect signal quality, but perfectly
usable signal. This is a working environment and not a lab.
I have worked on the design and implementation of systems with hundreds of thousands of patch points and some million+ feet of twisted pair, and balanced interfacing provides a platform that makes this kind of thing work.
Now, if you have a situation that is a little less wild and doesn't require handling whatever-the-cat-dragged-in with grace, then you
can design a system which is less flexible, but higher quality; i.e., less distortion and noise. This is done all the time within the confines of, say, a mixing console, where there are sub-systems which interface with one another under restricted and well defined conditions. Of course ultimate quality is not always achieved due to restrictions on the cleverness of the designer, cost of components, space restrictions, power consumption restrictions, etc.
A facility like a mastering room normally does not need to deal with a lot of signal routing issues and can focus more closely on idealizing the signal path. They have the luxury of idealizing their interface specifications and then rearranging the world to fit the spec., rather than the usual, make the spec fit the world you find yourself in. They can, therefore, take advantage of unbalanced interfacing that is difficult and expensive to beat using balancing, provided they avoid the problems that balancing is meant to address.
The best of both worlds could be had, but would require everyone to agree to use the same I/O topology. That is not going to happen.
Next rant:
The world is not constructed according to your intuition!
We tend to associate complexity with 'dirtiness'. It is a natural, albeit, simplistic tendency to think that simpler is more accurate, less distorted, cleaner. This is not always so. It might be useful to think of the development of timekeeping technology. Simple mechanisms were deployed initially and found to lack accuracy. In other words, the signal was distorted. Refinements were contrived that added complexity, but removed the distortion! I don't think that anyone is going to be able to produce a magical simple clock that is accurate (distortion free).
Audio is much the same, only we have a more emotional attachment to our opinions and conclusions. A few watch collecting fetishists aside, we don't really get emotional about the accuracy of a timepiece. None the less, it appears that audio signals also respond to the addition of the the right kind of complexity to weed out factors contributing to distortion.
Therefore, counter to intuition, a balanced input with a simple design with one opamp and a handful of resistors is less noisy than a passive circuit contrived to do the same job, and this is much noisier than a complex arrangement of 12 opamps! In fact, adding exponentially more opamps would lower noise further, albeit with diminishing returns that make it an academic exercise in short order.
It turns out that many of the undesirable characteristics of passives also diminish when complexity is added. You will see parallel arrangements of capacitors that demonstrate this. This is all due to fundamental differences between signal and noise. Independent noise sources generate uncorrelated noise, i.e., the noise signals do not add algebraically. A duplicate of the signal does add in simple algebraic fashion, however, and therefore, the signal increases at a rate that is faster than the noise when the noise sources are uncorrelated (independently generated). That means that if you feed a summing node by sourcing from a common noisy component, then you add noise at the same rate as you add signal. But, if you source them from independent noisy components, then the signal will trump the noise.
So, what has this all got to do with the OP's question?
This thread has gone pretty far afield in attempt to answer a relatively simple question (I have done my best to follow the trend
). Some of the salient points have been raised though. Let's have a look at the issues and solutions:
First we've got to get a signal to the unbalance inputs hot. OK, choose the in-phase signal from the balanced output.
Second, we've got to establish a reference at the unbalanced inputs reference input (usually called ground). This may exist by default in the form of the chassis grounds being tied together through the mains and the signal grounds being tied to their respective chassis.
I have wired unbalanced mixers into patchbays that interface to balanced and unbalanced gear, with nothing more than the hot connected at the mixer. It can work, but is unusual.
Without documentation for the balanced output, I would start by connecting the cold to the unbalanced reference input and leave the shield connected at the source and dropped at the destination (sometimes a small capacitor is used to drain RF here, but it is most likely not needed). Depending on the type of output, the shield may want to be tied to the reference pin at either end. This is easy to experiment with if you leave the shells unseated at both ends and short the shield as needed. There are some circuits that can be damaged by having the cold shorted to ground, but in practice I have never encountered this, and most gear that might have a problem with this is explicitly labeled.
OK, so we've the signal connected and have played with the shield to establish good screening and referencing, so interference in the form of hum, RF, etc. have been minimized as best they can be, given what we have to work with. If we have been unsuccessful at achieving an acceptable result, then some intermediate form of interfacing will be required to address the specific shortcomings. This is not likely unless the requirements are stringent or the situation is unusually bad.
That leaves only level matching to deal with. In most cases you will have reduced the level of the output by virtue of unbalancing, though this is not always true. If your source has an output level control, then simply knock of another 10db or so, and your good. If not, you may need to construct a pad which can be installed in the connector which is plugged into the output. Keep in mind that in terms of signal to noise ratio, if you need to add gain after this, the noise comes up with it.
That's a lot of verbiage for what comes down to: Hook it up, play with it a little and call it day.
Seriously! F-it, if you're needing to come up with a good solution for hundreds or thousands of connections, then do take the trouble to think it through... Otherwise go at it guerrilla style and move on.
If your ears are not satisfied with the result, then let them argue with your wallet. I think you can live with less than perfect when you know you've done the best with what you've got. These days, less than perfect is usually pretty good anyway.
Cheers ;D
My Daughter informs me that it is standard protocol to offer cookies to anyone who has made it through an epic rant. Here they are... freshly baked... chocolate chip... mmmmm...
