May 1, 2002 12:00 PM
i've always found Bill Whitlock's articles extremely helpful. [In his March 2002 column,] he cites line leakage currents in the shields of unbalanced cables as the principle cause of hum, but there's no mention of the infamous solution of breaking the ground connection of the cable at one end. If CI-coupling causes virtually all hum and buzz problems in unbalanced interfaces, then wouldn't breaking the path of flow solve the problem? Some claim that this compromises the shielding of the cable, [but] I have yet to see proof.
Whitlock also mentions that in typical (non-transformer) balanced output configurations, the actual CMRR is strongly affected by the output impedance, which is often poor due to great deviation in the output stage resistors (±5%) and capacitors (±20%). Does this mean that if manufacturers trimmed resistors and caps to a tighter tolerance, better performance would be achieved?
In the pseudo-balanced interconnection, Whitlock says the 470Ω impedance degrades the effectiveness of the CMRR. What about placing a second 470Ω resistor between the low conductor and ground to match impedance? Some manufacturers do this with some of their equipment by placing a matching resistor in series with pin 3 of an XLR connector on their output stage to ground and claim balanced outputs on their spec sheet.
Many thanks, and best wishes.
Evan T. Chen
The author responds:
Thanks for the kind words. As for your questions, remember that the signal current must have a complete path. Breaking the ground connection leaves an undefined path for signal return current. If both driving and receiving devices have 2-prong AC power connections, the ground-to-ground impedance would be very high and power-line leakage current would be forced to flow in the high impedances of the output port (about 500Ω ) and input port (about 50kΩ). Even a few microamps of leakage current would cause a full-scale 300mV hum signal to appear across the 50Ωinput impedance. Better that the leakage current flow through a wired ground connection where the resulting voltage drop would be in the range of microvolts (though still troublesome in many cases). Leaving the ground path open is simply not an option. Even if both devices had 3-prong AC power connections, the voltage difference between these connections in the AC power wiring can range from a few millivolts to more than a volt. This would be in series with the signal driver-to-receiver loop, again introducing a huge hum.
To your second question, CMRR is a system concept: It applies only to a connected driver and receiver. A balanced input or output stage can have a specified CMRR only when connected in a specified way, and CMRR depends on how a given driver and receiver interact. The two actually form a Wheatstone bridge in which impedance imbalances in either one will affect the “null” (elimination of common-mode response).
Most common transformer-less balanced input stages are specified for CMRR using a laboratory-precision signal generator as a source and specify CMRR numbers in the 70dB to 90dB range. Unfortunately, when driven by a source having only a few ohms of impedance imbalance, the CMRR of the interface drops to as little as 30 or 40 dB. It all depends on the ratios of the drive output impedances and the receiver input impedances on each balanced line. If the ratios are matched to 10%, CMRR is 20 dB; if matched to 1%, 40 dB; 0.1%, 60 dB; 0.01%, 80 dB; and so on. If driver common-mode output impedances are extremely low (difficult in real-world line drivers), or receiver common-mode input impedances are extremely high, the imbalance sensitivity problem is considerably eased. It is precisely the latter which enables a transformer to tolerate more imbalance in driver impedance than an ordinary transformer-less input. (See my 1996 AES paper at the Jensen Web site for full details.)
As for placing a second 470Ω resistor in a pseudo-balanced interconnection: It's a move in the right direction, and I recommend it (see AN003 at jensentransformers.com), but the 470Ω resistors must be precisely matched to allow ordinary transformer-less inputs to perform well. Even 5Ω of imbalance will degrade the CMRR.
Thanks for writing.