18-02-2021, 09:38 AM
Here is the final circuit of my 1p24b push-pull, ultra-linear amplifier. A second output transformer has been wound, incorporating the new feedback windings, and it works well (albeit that the C cores are held in place with a strong elastic band for now). It's generously-sized, weighing in at about 720g, and secondaries are sectioned, either side of primary, for reduction in leakage inductance and better HF response. After the first transformer, I took the opportunity to change the UL tapping points from 20% to 22%. The grid drive requirements are correspondingly increased, but there is still a couple of volts safety margin away from the critical positive-grid region at full output, which would seriously upset the driver stage.
The amplifier is built over two pieces of plain matrix board, the first-stage and phase-inverter being on one board, and the output stage on the other. Sharp-eyed readers may note a heatsink with a TO220 device bolted on, with a trimmer beside it - that's a TL783 regulator to regulate HT at 120V. With fixed battery bias (the array of lithium coin cells), the circuit needs a similarly stable HT voltage. Eventually, it will be battery-operated from a stabilised switchmode converter.
When commissioning, although the amplifier was stable, there was some (not unexpected) ringing with square-waves. So, I added an RC roll-off (47pF and 27kΩ, optimised empirically) from anode of the voltage amplifier pentode to 0V. The phase inverter, of course, just follows what it gets fed with.
To give the same 6mA quiescent current, one of the 1p24b's needs 19.5V bias, the other 20.4V, so there is a bit of variation between specimens. As there's no self-adjusting action whatsoever, unlike cathode bias, the balance trimmer is needed for best balance. At full output, current rises from 2 x 6mA to 2 x 16mA. Reducing the no-signal current further starts to show evidence of cross-over distortion, and although connecting feedback compensates for this, I was happy with the no-signal current drain.
Distortion is unknown, although the output just below onset of clipping looks pretty good (and it sounds good, too).
Frequency response, with 8Ω resistive load, is within 0.5db from 28Hz to 20kHz. The 3db point is 60kHz, so it's quite capable of annoying the odd bat! The low-frequency response isn't really useable down to 28Hz because I designed the transformer for 50Hz minimum at full output, so saturation-induced distortion starts to set in below 40Hz unless output is reduced, even though there is sufficient inductance to maintain small-signal response.
A close look at the driver board will show there's an unaccounted-for 1j24b, that's part of the project's next stage, which will be an infinite-impedance detector for an as-yet unstarted superhet front-end!
The amplifier is built over two pieces of plain matrix board, the first-stage and phase-inverter being on one board, and the output stage on the other. Sharp-eyed readers may note a heatsink with a TO220 device bolted on, with a trimmer beside it - that's a TL783 regulator to regulate HT at 120V. With fixed battery bias (the array of lithium coin cells), the circuit needs a similarly stable HT voltage. Eventually, it will be battery-operated from a stabilised switchmode converter.
When commissioning, although the amplifier was stable, there was some (not unexpected) ringing with square-waves. So, I added an RC roll-off (47pF and 27kΩ, optimised empirically) from anode of the voltage amplifier pentode to 0V. The phase inverter, of course, just follows what it gets fed with.
To give the same 6mA quiescent current, one of the 1p24b's needs 19.5V bias, the other 20.4V, so there is a bit of variation between specimens. As there's no self-adjusting action whatsoever, unlike cathode bias, the balance trimmer is needed for best balance. At full output, current rises from 2 x 6mA to 2 x 16mA. Reducing the no-signal current further starts to show evidence of cross-over distortion, and although connecting feedback compensates for this, I was happy with the no-signal current drain.
Distortion is unknown, although the output just below onset of clipping looks pretty good (and it sounds good, too).
Frequency response, with 8Ω resistive load, is within 0.5db from 28Hz to 20kHz. The 3db point is 60kHz, so it's quite capable of annoying the odd bat! The low-frequency response isn't really useable down to 28Hz because I designed the transformer for 50Hz minimum at full output, so saturation-induced distortion starts to set in below 40Hz unless output is reduced, even though there is sufficient inductance to maintain small-signal response.
A close look at the driver board will show there's an unaccounted-for 1j24b, that's part of the project's next stage, which will be an infinite-impedance detector for an as-yet unstarted superhet front-end!
