As Jakob said, it is arguable if there's any advantage for balancing. I once asked Bill Whitlock about this - I hope it is OK to post his answer for your reference:
Q:
I had the pleasure to read an article ("Practical Circuits For Quiet Audio Transmission") by Walter Jung; in this article, the author shows that a JT-11P-1 has better CMRR if it looks into a balanced load, in this case an instrumentation amplifier in the usual three-opamp-configuration. However, in your application notes I do not see balanced loading of input transformers. Is there a distinct disadvantage of doing so (except higher cost etc.)? Will every input transformer benefit from a balanced interface? Is a simple one-opamp-instrumentation amplifier sufficent to improve the CMRR?
My questions are related to a microphone preamp design I'm working on and which will use a JT-13K7-A, but I would be glad if you could answer them in a more general fashion.
Jung's article may be downloaded from his site
(http://home.comcast.net/~walt-jung/wsb/html/view.cgi-home.html-.html).
A:
Thanks for writing Jensen. We do not generally recommend balanced loading of our input transformers. This is especially true for microphone input transformers. In order to properly balance the distributed capacitances of the secondary, the necessary design tradeoffs would compromise bandwidth and time-domain response. Since noise is a prime consideration for mic preamps, and single-ended preamp topologies will always have lower noise than balanced ones, we intend the secondary to be (ac) grounded at the indicated end.
In the case of the single JT-11P-1 that Mr. Jung's experimented with, a significant CMR improvement was indeed noted, but I hasten to add that he did not measure frequency or phase response in that configuration. I believe it would be compromised somewhat. I also believe, because of the lower impedances involved, that a balanced secondary load will have far less effect on the JT-13K7-A. I don't know if a simple differential amplifier (single op-amp and 4 resistors) would give comparable results. It would probably depend largely on the input impedances of the circuit.
Note that his answer is related to Jensen transformers. There are other transformer manufacturer that recommend balanced loading (Lundahl comes to my mind) - it all depends on the specific transformer.
Basically I see two advantages for balanced configurations:
* higher output level
* very high CMRR both for the input- and outputtransformer (may depend on type)
Both are prone to be completely irrelevant in practical use.
Specific thoughts on the given circuit:
* I would distribut the ICs differently - make U1A and U2A to U1A and U1B etc. This way, the output offset of the line driving pair tends to match and thus you can skip the output coupling caps (C9/C10).
* The input transformer needs a very high turns ratio for best noise performance. The OPA604 is a voltage-noisy coveal and wants a high source impedance, which is doubled due to the balanced nature of the circuit. Without taking the pocket calculator out, I'd say a 1:10 is the lowest reasonable turns ratio if you care about noise.
* Where does the first stage pair get their + input bias current from? It is low with JFET inputs, but not zero. Relying on leake is not good. Replace R5 with two grounded 11k (you'll need much higher values for the suggested 1:10 transformer).
* At low gains, the first opamp pair has to drive a rather hefty load - if you use the 1:10 transformer, I would double the impedance level of the feedback network (i.e. R1=440, R3=860 etc.).
* If you intend to drive 600 ohm loads, you should use a 2:1 output transformer, as with a 1:1 each opamp will see 300 ohms which is too low. You'll loose the high output level with this.
* Just to make sure: you know that the dual pot should be of linear law? A big advantage of this topology.
Sorry for the extended writing - I hope it is for any use!
Samuel