Power Supply Sag
    This "noise" is the collapsing and expanding of the rail voltage in response to the demands made on the power supply during the playing of music. The sag is caused by the greatly varying current demands the output stage makes on the power supply, which when multiplied against the effective series resistance of the power supply, create a varying power supply rail voltage. At full output a 100 watt amplifier draws a peak current of 5A, which results in a 5 volt sag in power supply voltage, if the power supply's ESR is 1 ohm; and 25 volts, if the power supply's ESR is 5 ohms. (The latter figure is more likely in an OTL amplifier.) The ESR is the sum of the diode and transformer winding resistance; and the higher the power supply voltage, the greater the ESR.
    So even a perfectly quiet power supply can cause trouble. If the fluctuations in power supply voltage are not matched in the drive signal for the bottom device, gross distortions occur. Imagine a loud passage collapsing the negative power supply rail, but the DC value of the drive remains at the idle value. In this scenario, the cathode or source or emitter has effectively been made much more positive than the grid or gate or base, which in turn turns off the bottom device altogether. Yes feedback will work to correct the problem, but this means that the feedback must ease up on its correction of bandwidth limitations and general distortion cancellation.

    Now if the power supply rails were truly noise free, grounding the input to the bottom device would not result in any noise at the output, as the ground and the negative rail would both be noise free. So wouldn't our efforts be better spent in trying to cancel the power supply at its source rather than trying to cleverly cancel it through circuitry? If the amplifier is run in pure Class-A mode, the brute force approach has some merit. But hybrid power amplifier are seldom run in anything close to pure Class-A and pure tube OTL amplifier never are, in spite of advertising copy.  What does mode of operation have to do with noise transmission? More than most tube gurus know it turns out.
    Hybrid and OTL output stages usually run an idle current of 50 to 200 mA, which translates into .8 to 3.2 volts into an 8 ohm load. Beyond these voltages, only one output device delivers the current into the load. Thus any scheme that seeks to eliminate power supply noise at the output by having the noise currents through bottom output devices cancel ceases to work beyond this point.  Fortunately, the music signal will tend to mask the noise contribution to the mix. Now, let's imagine a power supply that had no hum. Such a power supply would be greatly valued when no music was playing, but it would still give rise to a different type of "noise," one that would not show up at idle, but would manifest itself when play music.

< PREVIOUS

NEXT >