You're very welcome! By "laboratory" source I mean one that uses source resistors of extreme (better than 0.1%) precision. In fact, the IEC test (based on a BBC test) for CMRR guaranteed absolute precision using a clever method. One of the two source resistors was trimmable. A DPDT switch was put between generator and DUT (device under test) and the switch swapped the inputs. The trimmable source resistor was tweaked until the CMRR reading was identical - which guaranteed a match as good as the resolution of the test instruments! Clever, but that's not even close to what real-world sources are like. Most gear uses "build-out" resistors of 1% to 5% tolerance (and often electrolytic coupling caps of far looser tolerances), so imbalances of 10 Ω are not unsual (and used in my test that the IEC adopted in 2000). It does little good to laser-trim the diff-amp resistors to ±0.005% when a ±5% resistor will be put in series with it - as it is with real-world sources. So my point is that the CMRR numbers quoted for input stage ICs are essentially LIES, because that number is achievable only in a lab setup with PERFECT source resistance matching. With a typical device connected, CMRR may be some 20 to 30 dB less. Real input transformers, and the InGenius IC, reduce this dependence on the source by a factor of roughly 1,000! I started digging into all this back in 1994 because Jensen customers would report that adding an input transformer at an equipment input always made noise disappear. At the time, I couldn't explain why, so I started researching and experimenting to find out ... and it all boils down to common-mode input impedances. I've been trying to "spread the gospel" since ...Thanks a lot for this lengthy post. I am familiar with your work and published articles on AES.org where I am a member myself. I was just confused with what you meant by "laboratory signal sources".
I’ve always wondered why this InGenius topology wasn’t used in the mic preamp ICs that THAT produces, and where a large CMRR would be far more necessary because the cables can be long and the signal very small.
I tried to interest THAT in doing a mic-pre using the technology but, rightfully, they raised two issues. First, to do it properly would require DC-coupled inputs for the preamp - and its common-mode voltage range would need to include +48 V. The IC would still need a negative rail to support bi-polar output signal swings. This would require a 70 V manufacturing process for the IC ... not trivial! Second, the amount of common-mode noise in mic signals is generally very low - largely because no one in their right mind would make a second ground connection at the microphone itself. Significant common-mode voltages almost always result from a "ground" connection at each end of a signal cable. With a mic, one end invariably "floats" - so the only current flow (and resulting voltage drop) in the shield is very small. Only when something unusual happens (like a mic on the floor rolling into contact with, say, the flange of some stage lighting gear) are there common-mode noise voltages to reject. But, in theory, the pair of 6.81 kΩ resistors for phantom power can be "bootstrapped away," greatly reducing their negative effect on common-mode input impedance. But, for the reason just cited, it's an academic exercise. In practice, RF suppression and avoidance of a "pin 1" problem at mic input jacks is a far bigger issue than CMRR.