Here's a report on my own thinkings and experience with the 1p24b. As part of a larger project, I wanted to make a push-pull audio amplifier using Russian rod pentodes, after having built, but scarcely used, a voltage amplifier and driver stage using 1j24b's with their super-low 11mA filament current - less than half a DAF96.
Big thanks to Joe Sousa who has put up data sheets on Radiomuseum (and to Trevor who I see also posted in Post #7), these were a great help, especially with their inclusion of the absolute maximum ratings. I wasn't out to break any records (50W from a pair of 1p24b's when cooled with liquid nitrogen and having a lifetime of 2 hours anyone?), rather, I wanted to use them relatively conservatively so that they would run for thousands of hours.
Intuitively, a pair of rods as a 'grid' is never going to be as effective at controlling electron streams as a pair of rods with fine wires running between, as per DL92, 96 etc. The data sheet, and practical experience, confirms this.
The circuit arrived at is as below. I wanted to set the quiescent currents equal, hence the balance preset pot. The output transformer has separate primaries, so as to be able to meter each valve separately. I also wanted to operate in Ultra-Linear configuration, both to reduce distortion, and also to be able to utilise the otherwise wasted screen-grid current and make it contribute to output power - it is battery-operated, after all! The ultra-linear tap was arrived at, looking at the curves, and working out how low the screen-grid could go, without needing so much drive that the control grids go positive. It's 20% away from HT+. As the maximum average cathode current is 20mA, I wanted to operate in Class AB to keep average current well below (and save that battery power). In fact quiescent current can be reduced to 6mA (standing anode dissipation 0.72W, well below the 2.4W maximum) before crossover distortion becomes noticeable on the scope.
The screen stoppers turned out essential! Not fitted at first, the amplifier worked well in UL mode, but switched to pentode took off with little provocation. After suspecting the driver, or feedback through the LT line, I AC-grounded the 1p24b grids and found that reducing bias turned it into an oscillator, frequency unknown - current shot up and I didn't like leaving it running - but probably low 10's of MHz. The values are not optimised, and I'd like to see how low I can get away with, but the very low screen-grid current means voltage drop even in 750Ω is only 0.2V.
For the same reason, contribution of the screens to overall power output, is actually infinitesimal. With anodes switched to HT+ and screens to taps, nothing can be heard in the loudspeaker.
Both filament power (400mA for the pair) and grid bias (-21V) are a culture shock after working with DL96's (50mA each and about 6V respectively).
Waveforms, with a decent sine-wave input, at onset of clipping, are as below. Output power is 2W. Unfortunately I miscalculated by a factor of 2 the secondary turns for the output transformer, and it is thus a perfect match for a notional 32Ω loudspeaker, rather than the aimed-for 8Ω twin cone. Negative feedback is, as yet, unconnected, that will be the next job after a neaten-up!
Big thanks to Joe Sousa who has put up data sheets on Radiomuseum (and to Trevor who I see also posted in Post #7), these were a great help, especially with their inclusion of the absolute maximum ratings. I wasn't out to break any records (50W from a pair of 1p24b's when cooled with liquid nitrogen and having a lifetime of 2 hours anyone?), rather, I wanted to use them relatively conservatively so that they would run for thousands of hours.
Intuitively, a pair of rods as a 'grid' is never going to be as effective at controlling electron streams as a pair of rods with fine wires running between, as per DL92, 96 etc. The data sheet, and practical experience, confirms this.
The circuit arrived at is as below. I wanted to set the quiescent currents equal, hence the balance preset pot. The output transformer has separate primaries, so as to be able to meter each valve separately. I also wanted to operate in Ultra-Linear configuration, both to reduce distortion, and also to be able to utilise the otherwise wasted screen-grid current and make it contribute to output power - it is battery-operated, after all! The ultra-linear tap was arrived at, looking at the curves, and working out how low the screen-grid could go, without needing so much drive that the control grids go positive. It's 20% away from HT+. As the maximum average cathode current is 20mA, I wanted to operate in Class AB to keep average current well below (and save that battery power). In fact quiescent current can be reduced to 6mA (standing anode dissipation 0.72W, well below the 2.4W maximum) before crossover distortion becomes noticeable on the scope.
The screen stoppers turned out essential! Not fitted at first, the amplifier worked well in UL mode, but switched to pentode took off with little provocation. After suspecting the driver, or feedback through the LT line, I AC-grounded the 1p24b grids and found that reducing bias turned it into an oscillator, frequency unknown - current shot up and I didn't like leaving it running - but probably low 10's of MHz. The values are not optimised, and I'd like to see how low I can get away with, but the very low screen-grid current means voltage drop even in 750Ω is only 0.2V.
For the same reason, contribution of the screens to overall power output, is actually infinitesimal. With anodes switched to HT+ and screens to taps, nothing can be heard in the loudspeaker.
Both filament power (400mA for the pair) and grid bias (-21V) are a culture shock after working with DL96's (50mA each and about 6V respectively).
Waveforms, with a decent sine-wave input, at onset of clipping, are as below. Output power is 2W. Unfortunately I miscalculated by a factor of 2 the secondary turns for the output transformer, and it is thus a perfect match for a notional 32Ω loudspeaker, rather than the aimed-for 8Ω twin cone. Negative feedback is, as yet, unconnected, that will be the next job after a neaten-up!
