In , rickman wrote (edited for space):
How about LM386 and add in series with your board's output the lowest value and wattage resistor that according to the datasheet with your supply voltage will cause no problem if that resistor from LM386 output (after any series capacitor) goes to ground?
As I said a bit earlier, I would be wary of positive feedback. If the ultimate load has impedance varying with frequency and is on a line of length no more than a few percent of a wavelength of the highest frequency that matters (I would guess about 15-20 KHz), positive feedback will at least largely reinforce the gain variations that correspond to impedance variations with frequency. Your ultimate load may be a loudspeaker preceded by a matching transformer.
If you use positive feedback or even substantial output impedance, be prepared to explain how your device meets the spec but fails to satisfy the customer. You will be better off if the customer is satisfied even if the device meets the spec. Prime example: Many, probably most loudspeakers (especially in unported enclosures) that have frequency response peaks in low or lower-middle audio frequencies have corresponding impedance peaks. Significant output impedance and positive feedback will worsen these. There are other loudspeakers with flatter frequency response, especially at lower and lower-middle frequencies, but with impedance varying greatly with frequency over some frequencies where ratio of output SPL to input voltage is more constant. Having significant output impedance will make frequency response peaks out of load impedance (as a function of frequency) peaks. Keep in mind that loudspeaker frequency response (ratio of listener SPL to input voltage as a function of frequency) is generally determined with zero or approaching-zero source impedance.
- Don Klipstein ( snipped-for-privacy@misty.com)