AnalogSteph
Major Contributor
Audible tweeter hiss is a problem that could be solved quite easily by including a resistive dropper or L-pad from the factory - much like what you find in the tweeter section of a passive speaker's crossover. Usually max SPL will be limited by the driver rather than amplifier output anyway, so you can bring up gain before that as needed. (Higher load impedance also reduces amplifier crossover distortion, in case that's a factor.)
Plan B would be using external compensation components to bring amplifier closed-loop gain even below specified minimums if needed. An extra series RC in the feedback network's ground leg for an ultrasonic gain increase to the level required for stability is one possible approach. Shouldn't matter much if you have to EQ / filter anyway.
The fundamental problem is that a bare tweeter in a waveguide can be a lot more sensitive than a finished speaker - sensitivities of around 92 dB / W / m are not untypical for a 1" even outside a waveguide, and inside one of the larger kind, sensitivity can reach 100 dB / W / m levels towards the lower end.
For 0 dB SPL of noise from a 90 dB / 2.83 V / m speaker, amplifier noise needs to be at 90 µV max. At 96 dB this limit decreases to 45 µV, at 100 dB to 28 µV.
The TDA2052 mentioned earlier exhibits 3 µV of input noise over a 20 kHz bandwidth and has a minimum stable gain of 30 dB. Implemented as suggested by the datasheet (Gv = 32.1 dB), output noise becomes 120 µV. Note how this value exceeds all the numbers given above. (BTW: It would still be considered very good for an integrated amplifier. These generally have substantially less sensitive speakers to contend with, of course - 85-90 dB / 2.83 V / m would be a typical range.)
LM3886 input noise is given as 2.0 µV, A-wtd (to which you can add about 2.2 dB for a 20 kHz flat equivalent, so maybe 2.5-2.6 µV flat). Minimum stable gain for this IC is 20 dB, though it can get rather finicky down there and is mostly used at 26 dB, sometimes more. So let's use the suggested standard values of 20k and 1k (Gv = 26.4 dB), which gives ~55 µV flat (44 µV A-wtd). That's more like it. Could be a tad lower still, but should yield acceptable results.
And that's purely power amplifier noise, assuming everything before is sufficiently low in noise that it doesn't matter (definitely not always true as proper gain staging must be observed, and the limits of converters in inexpensive DSP constructions are not making things any easier).
Plan B would be using external compensation components to bring amplifier closed-loop gain even below specified minimums if needed. An extra series RC in the feedback network's ground leg for an ultrasonic gain increase to the level required for stability is one possible approach. Shouldn't matter much if you have to EQ / filter anyway.
The fundamental problem is that a bare tweeter in a waveguide can be a lot more sensitive than a finished speaker - sensitivities of around 92 dB / W / m are not untypical for a 1" even outside a waveguide, and inside one of the larger kind, sensitivity can reach 100 dB / W / m levels towards the lower end.
For 0 dB SPL of noise from a 90 dB / 2.83 V / m speaker, amplifier noise needs to be at 90 µV max. At 96 dB this limit decreases to 45 µV, at 100 dB to 28 µV.
The TDA2052 mentioned earlier exhibits 3 µV of input noise over a 20 kHz bandwidth and has a minimum stable gain of 30 dB. Implemented as suggested by the datasheet (Gv = 32.1 dB), output noise becomes 120 µV. Note how this value exceeds all the numbers given above. (BTW: It would still be considered very good for an integrated amplifier. These generally have substantially less sensitive speakers to contend with, of course - 85-90 dB / 2.83 V / m would be a typical range.)
LM3886 input noise is given as 2.0 µV, A-wtd (to which you can add about 2.2 dB for a 20 kHz flat equivalent, so maybe 2.5-2.6 µV flat). Minimum stable gain for this IC is 20 dB, though it can get rather finicky down there and is mostly used at 26 dB, sometimes more. So let's use the suggested standard values of 20k and 1k (Gv = 26.4 dB), which gives ~55 µV flat (44 µV A-wtd). That's more like it. Could be a tad lower still, but should yield acceptable results.
And that's purely power amplifier noise, assuming everything before is sufficiently low in noise that it doesn't matter (definitely not always true as proper gain staging must be observed, and the limits of converters in inexpensive DSP constructions are not making things any easier).
Last edited: