“In the old days”—oh, here he goes again, someone pull out the liver pills—audio equipment manufacturers paid a great deal of attention to the input and output impedances of their products.  Modern operational amplifier circuit design has made a mockery of some of the lengths to which these manufacturers went.  Try looking at a schematic for an old CBS Audimax, with input and output level controls made up of three-section potentiometers configured as “T”-pads.  Or for real humour, try any of the old Ampex tape machine audio circuit designs….


The reason for all this attention to detail was obvious at the time:  unless balanced lines were sourced and terminated at the proper impedance, “bad” things would happen.  “Bad” being unpredicted and undesired results.  We were all taught to just terminate everything properly, and life would be good.  The real smart guys, we knew, could sometimes make miracles happen by selectively breaking the rules, but beware to the mere mortal that tried it.  So we carried on the tradition of pads, pads, pads everywhere.  All broadcast circuits started and ended at 600 ohms.


The real culprit, mainly, was all that iron in the circuit—the input and output transformers.  And then opamps came along… suddenly, all input impedances are bridging, and output impedances are so close to zero that few worry anymore about the evils of double- and treble-loading.  Split pads and bridging pads are no longer necessary… and generally you just hook up the inputs and the outputs and plug the whole works in.  Very forgiving, and a hell of a lot simpler than wiring everything up with pads.


Today I want to talk about a situation that you might find yourself in where you’ll need to start worrying about impedance again, or you’ll rue the consequences:  the good old telephone program line.  Hey, those phone guys invented the balanced line… 


Even if the program line to your transmitter is a digital circuit, as is likely, it most likely has an analog loop between your studio and the nearest telephone central office.  At the C.O., the phone company will equalize the circuit, then it’ll go into some kind of A/D, and from there it could go on a microwave carrier, or fibreoptic link, or copper T1 or HDSL, or a combination of all of these to get to your transmitter site.  Here in British Columbia, the phone company likes to run HDSL or T1 right to the transmitter site.  There, it runs through a D/A converter (600 ohms out of course!), and  to your equipment.  But we’re getting ahead of ourselves—back to the analog loop and the telco equalizer.


For the first 150m or so, a twisted pair at 600, or 150, or even 50 ohms just looks like a pair of wires.  Beyond that, in addition to copper resistance, we have series inductance and parallel capacitance, which give us your typical low-pass filter.  See Figure 1.  The audio response of the analog loop rolls off at the high end.  The telco equalizer is adjusted to extend and smooth the passband response.  But how is the response at the input of the equalizer affected by the value of R, the source impedance of the signal generator?


Here’s the key:  telco engineers will adjust the equalizer at the C.O. for flat response using a 600-ohm source.  When you connect your modern processor with its 30-ohm buildout resistors to the line, it acts like a 60-ohm source.  I have personally seen circuits that measured +12 dB at 10 kHz (ref 0 dB at 400 Hz) because of this!!  The amount of the effect is determined by the length of the analog loop before the equalizer, with longer loops causing larger HF peaks.  Many engineers will run the processor output through a repeat coil, or 600:600 ohm transformer before leaving the studio.  Very nice, but it won’t do you a bit of good here—the repeat coil, true to its name, presents the 60-ohm source with an image of what it sees.  The really nefarious element of this problem is that if you suspected the line was poorly equalized, you’d likely patch in a 600-ohm generator, which would measure this circuit as flat.  Only if you fed the tone into the processor, with the processor in the proof position, would you truly see the problem.


The solution is quite simple:  either add buildout resistors in series with the processor output tip and ring connections, to make up a total of 300 ohms/leg (when added to the internal buildout resistors inside the processor, see Figure 2), or run the output through a 600-ohm pad of 10 dB or so.  (Ahh, once again we witness the universal curative properties of pads!)  And keep the repeat coil in the circuit—it does provide some protection from that 48V battery that the telco types are so fond of, and keeps the processor side of the coil ground-referenced.