Poor sound quality. Ugh.
I wish you had access to a scope. Distortion might be due to oscillation and/or due to crossover nonlinearity arising from too-low bias current.
I'm bothered by the adjustment range of the trim pots. Try repeating the experiment of #35, but with the goal of driving that meter reading to a null of 0V, using a combination of both the 4.7k and 470k pots. If the amp can reach that state, I would be more confident the amp isn't oscillating. I'm not advocating these pot positions as the final settings, only demonstration that the circuit is behaving correctly.
Don't trust any bias current adjustments until the amp is free of oscillation.
Addressing bias current compensation: I wasn't able to find any temperature data on 1N5819 vs. temp, biased at 3mA where D304 operates [~50V/(10k+5.6k)], but I found data on NSR20F30-D. A data sheet is posted below and an expanded version of Figure 1.
I estimate forward voltages at 3mA as:
25C, 3mA: 230mV
75C, 3mA: 135mV
125C, 3mA: 45mA
The linearity vs. temperature looks reasonable, roughly -1.85mV/C, and forward voltage is about 230mV at 25C.
In contrast, silicon diodes are roughly -2mV/C and forward voltage is about 0.6V at 25C. It appears that Schottky diodes might be preferable temp sensors, as they seem to deliver better temperature sensitivity per forward volt than ordinary P/N diodes.
I don't know how well 1N5819 devices would follow the above data. You could measure 1N5819 vs. temp at 3mA, or devise a method to mount NSR20F30-D diodes. I would suggest three or four Schottky's in series. The combined tempco should be enough to compensate the three Vbe junctions in the output stage and leave some available adjustment range on the 470R bias trimmer for setting bias current.
Monitor bias current at power-on and be be prepared to cut power if current becomes alarming. Current should be higher than desired at power-on, then decrease as the Schottky's settle to the heatsink temperature.
Good luck!
I wish you had access to a scope. Distortion might be due to oscillation and/or due to crossover nonlinearity arising from too-low bias current.
I'm bothered by the adjustment range of the trim pots. Try repeating the experiment of #35, but with the goal of driving that meter reading to a null of 0V, using a combination of both the 4.7k and 470k pots. If the amp can reach that state, I would be more confident the amp isn't oscillating. I'm not advocating these pot positions as the final settings, only demonstration that the circuit is behaving correctly.
Don't trust any bias current adjustments until the amp is free of oscillation.
Addressing bias current compensation: I wasn't able to find any temperature data on 1N5819 vs. temp, biased at 3mA where D304 operates [~50V/(10k+5.6k)], but I found data on NSR20F30-D. A data sheet is posted below and an expanded version of Figure 1.
I estimate forward voltages at 3mA as:
25C, 3mA: 230mV
75C, 3mA: 135mV
125C, 3mA: 45mA
The linearity vs. temperature looks reasonable, roughly -1.85mV/C, and forward voltage is about 230mV at 25C.
In contrast, silicon diodes are roughly -2mV/C and forward voltage is about 0.6V at 25C. It appears that Schottky diodes might be preferable temp sensors, as they seem to deliver better temperature sensitivity per forward volt than ordinary P/N diodes.
I don't know how well 1N5819 devices would follow the above data. You could measure 1N5819 vs. temp at 3mA, or devise a method to mount NSR20F30-D diodes. I would suggest three or four Schottky's in series. The combined tempco should be enough to compensate the three Vbe junctions in the output stage and leave some available adjustment range on the 470R bias trimmer for setting bias current.
Monitor bias current at power-on and be be prepared to cut power if current becomes alarming. Current should be higher than desired at power-on, then decrease as the Schottky's settle to the heatsink temperature.
Good luck!
Today, by pure chance I found the STV-3H that was mounted into the amp. I checked it and then installed it. I am now able to get to 50.1V on the center voltage and to 5.1 mV on the idle current. Messing around with the waveform, I couldn't hear any difference. I used a console as preamp and a decent speaker, and the sound was really not bad. It still doesn't get warm but then again I did not use it for a prolonged time.
