John, do you mind explaining why this procedure does ensure correct input bias cancelation? Thanks!
The 990 discrete op-amp has, according to the specs, 2.2 microamps of current flowing from each of its inputs. This is the input bias current. This causes a small voltage to appear across the available DC paths to ground at each input. There are two DC paths for the non-inverting input: the 6.19k resistor that is across the secondary winding of the input transformer, and the secondary winding of the input transformer itself. This is a fixed resistance.
Meanwhile, the inverting input has two DC paths: the gain control path to ground, and the feedback loop back to the output of the 990. Since the gain control path varies from about 20 ohms to 10,020 ohms, it is NOT a fixed resistance. Therefore, the DC voltage that is caused by the flow of input bias current through this path will vary.
Since the 990 is DC-coupled in this circuit, and it is a "differential" circuit, any difference in the currents/voltages that appear at the two inputs will be amplified by as much as a factor of about 500 (54.4dB of gain for the 990 at maximum gain), or as little as a factor of 2 (6dB of gain for the 990 at minimum gain).
If the input bias currents are properly compensated for, or nulled, it has the effect of causing little or no voltage to appear at either input of the 990, so there will be very little voltage showing up at the output of the 990 regardless of gain setting. There is no DC voltage at the inputs to be amplified. Therefore, the DC voltage readings at the output of the 990 will be about the same at high gain as they are at low gain.
Traditional audio designs get around this by putting a fairly large value electrolytic capacitor in the gain control path to keep the bias current of the inverting input from reaching the gain control where it might cause noise during adjustment. The capacitor also causes the gain at DC to be unity, so there will be no amplification of any DC differential signal at the inputs. But capacitors have various problems, which I won't go into at this point (unless someone really wants to know). Thanks.
John Hardy
The John Hardy Co.
www.johnhardyco.com