Tweet
The differences in vacuum rectifiers and SS diodes have been worked over quite a bit, but I think I ran onto something new about them. The I-V curve of a vacuum rectifier tube is responsible for the voltage drop and sag in tube rectified power supplies. SS diodes have radically different I-V characteristics, so they produce a higher DC and much smaller sag than vacuum rectifiers.
It's possible to make SS diodes behave more like vacuum rectifiers by inserting an appropriately sized power resistor after the SS diodes to insert voltage drop and sag. This had been done many times, including in the Weber Copper Cap rectifiers, which many people find acceptable.
It occurred to me to look at the size and shape of the ripple voltage of model supplies in my simulator. I found a well thought of model for a 5U4 rectifier, and compared this to three other rectifiers; a 1N4007, a 1N4007 plus dropping./sag resistor, and a diode-zener-resistor network curve-matched to the 5U4 I-V curve. The results ticked all the boxes, in that the bare SS diode was highest DC, little sag; the diode plus resistor was lower DC, more sag; the 5U4 was as expected; and the curve-matched network came remarkably close to the 5U4. The ripple voltages were significantly different.
The single diode had sharp points on the peaks of the ripple voltage. The diode plus resistor rounded the points a little; and the curve-matched network was so close to the waveform of the 5U4 that it was hard to see the differences.
I ran a spectrum analyzer modeller on the ripples. The spectrum peaks were quite different. The single diode had the expected big 120Hz and 240Hz, and a very long train of harmonics that did not reduce quickly and extended to high frequencies. The diode plus resistor reduced the upper harmonics very noticeably. I was quite happy to see that the curve-matched network and 5U4 model produced almost exactly the same spectrum for ripple. It was hard to squint the differences.
This was actually a "well, duuuh!" moment. Of course sharp pointy waveforms have lots of harmonics. But seeing a diode-zener-resistor net produce less harmonics with no filtering caps or inductors was rewarding. And I think the spectra explain some of the preference for vacuum rectifiers. Beyond sag, it's possible for higher line harmonics to get into the amplifying circuit if the filtering network isn't well done. Higher line harmonics could well introduce a brittleness to the sound. Some of it is certainly going to get into the OT CT, and possibly into the PI supply and screens. Yes, those are filtered, but for people who claim to hear artifacts below the measurement threshold of modern instruments it might be possible...
In any case, I think that there's enough there to make a rectifier curve matching network a good place to dig.
Nota Bene: No, this does not match the warm up lag. There are ways to do that. but I-V matching doesn't simulate warm-up.
It's possible to make SS diodes behave more like vacuum rectifiers by inserting an appropriately sized power resistor after the SS diodes to insert voltage drop and sag. This had been done many times, including in the Weber Copper Cap rectifiers, which many people find acceptable.
It occurred to me to look at the size and shape of the ripple voltage of model supplies in my simulator. I found a well thought of model for a 5U4 rectifier, and compared this to three other rectifiers; a 1N4007, a 1N4007 plus dropping./sag resistor, and a diode-zener-resistor network curve-matched to the 5U4 I-V curve. The results ticked all the boxes, in that the bare SS diode was highest DC, little sag; the diode plus resistor was lower DC, more sag; the 5U4 was as expected; and the curve-matched network came remarkably close to the 5U4. The ripple voltages were significantly different.
The single diode had sharp points on the peaks of the ripple voltage. The diode plus resistor rounded the points a little; and the curve-matched network was so close to the waveform of the 5U4 that it was hard to see the differences.
I ran a spectrum analyzer modeller on the ripples. The spectrum peaks were quite different. The single diode had the expected big 120Hz and 240Hz, and a very long train of harmonics that did not reduce quickly and extended to high frequencies. The diode plus resistor reduced the upper harmonics very noticeably. I was quite happy to see that the curve-matched network and 5U4 model produced almost exactly the same spectrum for ripple. It was hard to squint the differences.
This was actually a "well, duuuh!" moment. Of course sharp pointy waveforms have lots of harmonics. But seeing a diode-zener-resistor net produce less harmonics with no filtering caps or inductors was rewarding. And I think the spectra explain some of the preference for vacuum rectifiers. Beyond sag, it's possible for higher line harmonics to get into the amplifying circuit if the filtering network isn't well done. Higher line harmonics could well introduce a brittleness to the sound. Some of it is certainly going to get into the OT CT, and possibly into the PI supply and screens. Yes, those are filtered, but for people who claim to hear artifacts below the measurement threshold of modern instruments it might be possible...
In any case, I think that there's enough there to make a rectifier curve matching network a good place to dig.
Nota Bene: No, this does not match the warm up lag. There are ways to do that. but I-V matching doesn't simulate warm-up.
Comment