A servo is an opamp or comparator which takes a reading from the output of an amp and feeds a compensation signal into either the bias or feedback IIRC. This causes the amp to swing more towards the direction off the offset and null it.
Think of a servo as a mathematical building block that INTEGRATES the small amount of DC signal on the amp's output. This is continuously added up over a time period of several seconds or more and stored on a capacitor in the feedback loop of a simple opamp circuit. The voltage on the servo's feedback loop capacitor is fed back into the negative input (or negatively into the input bias) of the amp.
Music waveforms play on the output, of course, but since it has no DC on average, it essentially gets ignored by the servo.
Originally Posted by a1rocketpilot Hmm, does anyone have a basic schematic of this just to get a general idea of how the servo is implemented? I vaguely see how it works in the Dynalo, but otherwise, I'm confused.
Just think of a low pass filter with a very low cutoff point and a difference amp. The low pass filter passes only the very low (essentially DC) part of the signal, which subtracts from the unfiltered signal.
Does anyone know of any literature that goes into the nitty gritty of DC Servos? I haven't yet seen them mentioned in any general purpose textbook and searching it on Google had been a huge faliure. I would definately like to educate myself more on the topic given the chance.
The opamp is taking the output from the amplifier through a 1M resistor. This prevents the loading of the output by the opamp and effectively "removes" it from the output of the signal path. The signal is then amplified but in the feedback loop there is a capacitor. IIRC the 15k resistor, 7.5k resistors and the 0.33uf Cap forms a low pass filter to ensure that nothing in the audible range is is run through the servo. This signal is then fed directly into the point where the LED, transistor and 500ohm resistor bias the amplifier.
If any DC appears on the output the opamp will integrate it and adjust the bias of the amplifier to null it. Compared to Dan's PCBs it's almost identical to adjusting the POTs to reduce output DC except that this is done on the fly.
The following diagram shows four classic DC servo topologies. The "Amp" is a generic representation and could be anywhere from an IC opamp to a fully-discrete design. The "Servo" on the other hand is usually implemented with an IC opamp (preferably with a precision FET device with low intrinsic DC offset). The servo, as mentioned in posts above, is configured as an integrator (low-pass filter with very low corner frequency, well below the audio band) such that it only has an effect near DC (0Hz). Any DC offset at the output is detected by the servo and fed back to the amplifier as an inverse correction to keep the DC offset locked to (near) 0 at all times.
The "A" and "B" topologies are for non-inverting amplifiers, while "C" and "D" are for inverting amplifiers. "A" and "C" uses a non-inverting servo, and a second-order filter (two RC combinations) for a steeper low-pass slope, and feeds the offset-nulling correction at the inverting input of the amp. "B" and "D" on the other hand uses an inverting servo, a simpler first-order filter, and feeds the offset-null correction at the non-inverting input of the amp.
The Kevin Gilmore dynalo and dynahi amps use a variation that is most similar to "B", except that the offset-nulling is not done at either of the amp's inputs. Rather, the correction signal is fed to the amp's input stage constant current source bias junction, where it shifts the DC balance up and down to accomplish the same purpose.