So which is it.................. 4% OR .4%
4% is actually a large tolerance. Statistically, if you take a dozen of 10% resistors from the same batch, you'll probably find they are within 1% of each other, so, although the average value may be quite different than nominal, matching will probably be very good.
Now, why do these resistors need to be matched?
There are two factors there:
One is the presence of DC current in the secondary. Supposing 1% tolerance* on the phantom-feed resistors (6.8k) and 4% for the 2.2k, the maximum DC current flowing through the secondary is less than 10uA. This is not enough to skew significantly the BH curve in the magnetic core.
The other is the loss of CMRR. With the same 1% tolerance for the 6.8k and 4% for the 2.2k, worst-case CMRR is 49dB.
With 0.4% tolerance for the 2.2k, worst-case CMRR goes up to 64dB.
None of these figures can be deemed good or bad, because it depends on many factors. In a clean environment with little EMI/RFI, proximity micing a decent spl source, there should be no problem with 49dB , when 64dB in a heavily EMI/RFI polluted environment, trying to record fly farts from a distance may prove unworkable.
Resistors are so cheap and matching them by hand is not a difficult task. I bet you take only 10 resistors, you can find pairs that match better than 0.4%.
I don't know where these figures (0.4 or 4%) come from, but I think the person who offers them should also provide an explanation of the basis for this assertion.
*The 1% figure I have retained for the 6.8k is 10 times the P48 standard (0.1%). Since the 2.2k are dominant over the 6.8k, it makes sense that, in oder to protect CMRR, the tolerance on the 2.2k should be about 3 times better than the standard (0.03%), which is quite difficult to achieve in practice because of component aging and temperature gradients.
Finally, remember that CMRR is pertinent to a
connection, i.e. to a combination of a source and a receiver.
It is perfectly possible to achieve excellent CMRR with significantly unbalanced legs, as long as the unbalance in the source is compensated by a complementary unbalance in the receiver.
Specifying CMRR for a source or a receiver separately is a way of achieving some kind of standard that guarantees a minimum level of performance and interchangeability.
Bill (CMRR here) may chime on it, as he is the one that helped defining the standard measurement, based on assuming an imperfectly balanced source.
For TMI, there are units that include a CMRR adjustment, quite often limited to HF, in the guise of a trimmer capacitor, but there are a few that also have an adjustment of wide band CMRR. Crown had that on their Micro/Macrotech series. It was not a user-adjustable control. I guess had it been user adjustable, some idiot would have cranked it all the way up and claimed it sounded better.
