Hi Mike,
 
Not trying to beat the poor dead horse, but...
 

mike_mccue
I guess what I was trying to say is along these lines; the extra 6dB you get with the balanced connection when you use all the wires is a direct comparison to when you use the same connections with only 2 of the wires.

 
That extra 6dB of signal is not guaranteed just because you use all three lines!!! The only way that happens is if the line is symmetrical! Balanced does not imply symmetrical, and symmetrical does not imply balanced.
 
Case in point - some popular A/D chips have symmetrical inputs, but if you read the fine print you'll discover that the signals applied to the two input pins must be absolutely equal, except for polarity. If there is any common mode signal then the two signals will not be equal and opposite, and the S/N ratio degrades - a LOT!
 

mike_mccueIf you had two discrete sets of connections, one that is unbalanced and one that is balanced you may find the comparison more favorable. Yes, there will probably be a a small amount of noise added to the SNR of the unbalanced send, and yes there may be a small amount of noise entering on the cable run, but you may, possibly, find a lot less additional noise than when you simply throw out half your signal by not using the extra wire on a connection that is designed to use one and then use make up gain to get to the nominal levels.

 
Not exactly... and you are somewhat mixing metaphors - I mean conditions...
 
Noise coming in on the shield is a specific problem, described as the "Pin 1" problem.
 
If you have a less noise when you "throw out half your signal" that indicates a design flaw somewhere. Reducing the signal level increases the S/N ratio unless you also reduce the noise. Make-up gain will always decrease the S/N ratio. And unless the equipment is really poorly designed, connecting a single-ended source to a balanced input as a balanced signal will always result in less noise. It has to, since the noise will be common mode, and that is what a differential input does - it cancels out the common mode signal.
 

mike_mccuePart of this idea includes the consideration that it is rare to run the device doing the sending at such a hot level that it is at the threshold of a SNR problem. If the device has a unbalanced send that was made to sound good at a comparable level as some balanced send we refer to then the differences in perceived results will be much slighter than if we simply throw half the signal away as in the examples cited in the first few posts.

 
I think you are confusing gain staging with other issues. It does not matter what the nominal operating level of a source is... that has no real effect on S/N ratio, it is simply the nominal operating level.

And herein lies the problem with mixing consumer, pro-sumer, and professional equipment! They have different nominal operating levels, different topologies, etc. Getting the best performance from this mix is a challenge, and that's why I prefer to avoid pro-sumer and consumer gear if I can.
 

mike_mccueIn general, I agree with everything you say about the benefit of balanced connections. :-) I'm just making a very small point about how the experience of observing a large difference in one circumstance does not lead to an absolute conclusion that the two technologies always provide such different results. 

 
But it will. If we assume competently designed equipment. Which is another problem - there is a lot of poorly designed equipment out there.

mike_mccue 
And now because I can't resist: :-)
 
"Anyway, the general consensus amongst folks that would be identified as authorities on the topic of electronic noise reduction is that single-ended inputs are inherently bad when applied to professional and/or commercial applications. This does not apply to single-ended sources (outputs) - these can be made to work with differential inputs."
 
I agree 99.9% but I can't help but think of the many times I have experienced huge SPL in commercial venues with low noise playback on a large PA system being fed by an iSomething and its unbalanced "1/8" output connection into the "tape in" on a Pacific rim mixer.

 
But SPL is not a function of the level feeding the Pacific rim mixer. SPL is a function of gain staging and the power available to drive the loudspeakers and the efficiency of the loudspeakers. As long as you can drive the amplifiers to their maximum output it is gonna get stupid loud no matter the source.

This is a very complex, and messy topic because it includes a lot of factors... and I guess I am not doing a very good job of explaining it here.

I just want to share my practical experiences. As a composer, connecting synthesisers to a mixer presents the variables that have been mentioned here. Even with 7 or so in my case they have different output arrangments. But they also have different amounts of quiescent noise that exist alongside the signal.
 
I am not referring to noise that is picked up on the line either between the synth and the mixer.  (I have never had a problem with that)  I am referring to the ambient or quiescent noise of the machine itself.  With something like the older Kurzweil K2000 that background noise is reasonably high and audible.  That output is single ended.  (Note: it is still mostly not an issue. The signal is still very loud compared to this noise and the noise will most often be masked. Noise gets heard more with sounds that fade away slowly and have long tails and start getting very quiet. With some effective downward expansion the synth can be made to be dead silent when not making any sound. Any downward expander VST will do a great job once you have printed this track in your DAW. I have a nice hardware one which kills any noise when not sounding, I patch it over that K2000 permamently. It does a lovely job of easing in the gain reduction as the signal almost disappears, Internal Reverbs will extend the tails of patches too..)
 
On the newer model that noise is considerably less and the signal level is much higher and all on a balanced output system. (No downward expander required)  There are many reasons why that background noise is there.  So you now choose to use the output options in two ways either single ended or balanced.  I was using the single ended connection for a while and it proved fine except the background noise was just present but very low.
 
However when you use the fully balanced interconnection between this and the mixer, the signal level jumps 6 dB but not only that any quiescent noise that was present before has now completely gone and I am now dealing with a machine that is dead silent when not making a sound.  Even with everything maxed out on the channel strip it is still dead slient.  I am amazed at how quiet they have got it. This is when using a balanced interconnection is doing something damn useful.
 
It is better to start with a dead silent machine if you can because many processes later down the line can bring up even low level quiescent noise.  But when things are dead silent many effects processes later won't be bringing up any noise because it not there. Saves time later cleaning up noisy tracks in your mix or tracks that are not sounding but still have low levels of noise present.
 
Some synths have only a single ended output and work perfectly with an unbalanced lead to a line input that is unbalanced.  This can work as well as any balanced system.  With some instruments the output is still very high + 4 to +8 dBu on average with a loud preset selected and when the sound comes to rest and you are just hearing the output quiescent noise then with some devices it is amazingly low and the machine sounds perfectly silent even with the channel fader up high.  Obviously those manufacturers had done a mightly fine job on their single ended output design.  I suspect some may have a form of downward expansion on the very output stage to create a form of noise reduction to eliminate any noise that has come from earlier stages. It works great because then you will never hear the noise of the machine itself, the output signal is so loud compared to it and masks any noise very effectively.
 
I did find out something else though. Be careful taking a balanced output from a synth on a TRS connector and using an XLR male plug on the other end to connect direct to a mic input. No issue from an audio perspective but if you apply phantom power for any reason you are now sending +48V to the outputs of your precious synth. I checked with Kurzweil on this and they said in no uncertain terms it would have blown the output stage. The Samson mixer I use has a balanced line input next door to the XLR input which feeds the same Mic pre anyway but it is DC isolated and safe. All balanced mixer line inputs will be the same.

Hi Bill,
 
It seems like keeping noise to a minimum is definitely factored by effective gain staging. So my general response to your post is that I am referring to gain staging but I don't think I'm that confused. :-)
 
"Not exactly... and you are somewhat mixing metaphors - I mean conditions..."
 
Yes, I am purposely mixing conditions. Two best case scenarios, one for balanced and unbalanced gear, and yes that definitely touches upon the idea of gain staging... gain staging as managed not only by the operator but also by the equipment designer.
 
 
"If you have a less noise when you "throw out half your signal"..."
 
I don't think I made this implication. Maybe it's a fair inference from something I said and I have not recognized it yet?
 
You do have me thinking though... is the system noise from any previous stage "common" or random? I have always thought of common mode rejection at an input as applying to noise that enters the cabling with the twisted pairs and not the noise emanating from the device sending the signal.
 
I think I may be confused on this point, but it seems as if you may be suggesting that CMMR can reduce system noise from a previous device... ???
 
 
 
I think your point about ineffective CMMR on some designs is very interesting but I suspect we are both straying far from the OPs interest. :-)
 
Finally,
 
"That extra 6dB of signal is not guaranteed just because you use all three lines!!! The only way that happens is if the line is symmetrical! Balanced does not imply symmetrical, and symmetrical does not imply balanced."
 
Yes! We might as well elaborate and point out that a good way to visualize a symmetrical waveform is to think of a sine wave test tone. In other words, musical content is unlikely to be symmetrical and so when I used the figure 6dB, for the sake of discussion, I was generalizing in a way that does not help to further an understanding of the points being made about balanced connections.
 
all the best,
mike

Woke up this morning... :-)
 
Or perhaps, when you mention the "term" symmetrical you are referring to the send devices capability to send a signal along the two lines so that they exactly reflect each other as polar opposites? In other words, the device may deviate in level, system noise, or distortion etc. and so when two signal arrive at the input the recombination does not add up to an exact doubling of level?
 
... always happy to learn! :-) Sometimes I act too big for my britches. :-)

I am learning so much from you guys on this issue.  I am dumbfounded as to how much punch and clean volume I now have without ANY detectable noise!  I brought my near field monitors in closer to me and at normal listening level they are dead silent when I hit stop.
 
Learning why this is so is important too but proof is in the pudding.
 
Mike - I am glad you woke up this morning!  

Bill, if you are out there... I've been waiting (and hoping) to learn my lesson... and find out what I said wrong about symmetry in the past few posts.
 
Thanks!
 
:-)

hmmm... not sure to what you are referring... seems like you get it.
Signal Symmetry refers to a condition where the signal on two conductors is equal but opposite in polarity. If it were a simple sine wave then the positive line would reach it's maximum at time t, and at that same time t the negative line would reach it's minimum.
 
When you take the difference you would twice the maximum excursion of either one of the lines - so if the sine wave was swinging from -1V to +1V on the "+" side, and +1V to -1V on the "-" side then the output of the difference stage would be 2V, which is 6dB greater than 1V.
 
At the same time, the noise that is induced into the system - or cable - will be equal, and of the same polarity on both sides of the line, so when we take the difference we'll end up with no noise voltage. That, of course, assumes a perfect system, which doesn't exist, but there will be a significant rejection of noise induced into the system when we do this. HOWEVER, and this is important, a differential stage works best when it is balanced.
 
Signal balance actually refers to impedance balance. The impedance from each source pin to ground must be very nearly equal. The impedance from each input pin to ground must be very nearly equal. And the impedance of each path must be very nearly equal, since the combination of a source impedance, line impedance, and input impedance is a series circuit.

Some input stages are remarkably tolerant of an impedance mis-match, think transformers. Others introduce an impedance mis-match - thing the single op-amp topology. Many really bright folks have spent considerably brain power figuring out tricks to make the inverting and non-inverting inputs of an op-amp 'look' the same to a signal. Some of those tricks work, some don't.

A single-ended source can drive a differential input - it won't have signal symmetry, and it may or may not be balanced, or rather the degree of balance may be compromised. At which point the ability of the input stage to tolerate imbalance becomes the key.

Last dead horse to beat - shielding vs twisted pairs... shielding can prevent RF energy from reaching the signal conductors. It can do nothing to prevent magnetic energy from reaching the pairs and inducing noise currents. If the shield is grounded at both ends then current can flow, and if the shield is improperly terminated at the input it can introduce noise.
 
This was the reason folks started connecting the shield at only one end (telescoping shields). However, a shield - or any conductor - that is grounded at only one end is an antenna. You have to pick your poison some times.
 
Twisting the signal pair helps to reject magnetic fields. If the twist rate is uniform then each conductor is exposed to the same magnetic field, which provides some additional rejection. And then of course there is the balanced, differential input stage to finish the job.

So isn't that basically what you said - only i used more words<G>???

Thanks for taking the time Bill!
 
The idea I mentioned about the sine wave seemed, in retrospect, to misapply to the point of seeming like a silly outburst, and then I felt like I was grasping to offer a pertinent explanation.
 
I appreciate the reminders to stay on point. :-)

The sine wave example is exactly on point, since it is by far the easiest way to visualize the physical symmetry.
 
When I taught this stuff that was one of the experiments we did in lab. Of course the poor students eventually had to use music, and it was always fun to watch them try to figure out how to prove the same thing was happening. I was a mean teacher!!
 
It sticks in my mind that you like to build things Mike - if you are still so inclined you really owe it to yourself to play with the InGenius(tm) chips from THAT Corp. They behave a LOT like a transformer, well, except for isolation<G>!
 

Hi Bill, Thanks for the info about the InGenius chip, I'll make a point of reading up about it and then see where that takes me. :-)