> I can't seem to read any of the text on the linked page.
It is one of the Asian languages. My browser somehow picked-up a font to display it properly, but I can't read it. If your browser has not installed the right Asian font, you probably get a lot of boxes or ASCII gibberish.
> the downswing looks terribly different than the upswing.
Yeah, but not so very different for signals up to a volt or two. To a first approximation: who cares what happens after that? The output stage will be in great distress. (In fact, asymmetric overload does affect sound at high levels....)
> Non-linear input capacitance in the ‘170 is the reason to prefer the 6DJ8
Ah, quite correct.
> I'm building a class-D power amp
Then you are doing great violence to the audio already, and have a major debugging chore ahead. Personally, if the Class-D needs a buffer, I'd throw in a NE5532 and move along to the big problems of Class-D. After the smoke clears, if the sound is good, I'd explore other linear input buffers.
> if it is absolute rubbish
If you can accept a volt or two peak output, it is fine. Resistor-loaded, it is mildly sweet. CCS loaded it will be simply clean. When taken past a volt or so, it overloads, and asymmetrically.
> why self bias?? Maybe to dial in more voltage across the tube than the CCS?
The tube has to take most of the voltage: electrons won't cross vacuum as easily as doped sillycon.
My worry is that, of all the possible biases you could arrive at with the trimmer, only some will work. And with higher voltages, some will melt the JFET.
> I checked the circuit with a Tesla E88CC tube. Results are: Va=44V Ia=5mA Vgc=-0.5V The AC signal at output looks fine up to 6Vpp. source impedance... 10k resistor
Seems reasonable. Me, I would take a little less volts on the tube, maybe 3/4 of available supply voltage. That would give Va=37V, Ia=4mA, Vgk~ 0.4V. There is a trade-off between cathode bottoming on negative swings and grid-current limiting on positive swings.
> The AC signal at output looks fine up to 6Vpp.
My two-thumbs estimate was "gross distortion" at about 4V peak on one side. With B+=48V, Va=44V, it can't hit 4V either way, so 6V p-p seems likely. THD will be about 0.5%, so it will "look fine". IF you can live within 6V p-p, it should be OK, either resistor-loaded for a little sweetness or CSS loaded for very-clean.
> a class-D power amp
Then a real question is: how "solid" does the signal have to be? Some A/C converters need a SOLID driver to absorb great switching spikes and damp them before the next sampling moment. Note that 99% rejection of switch-transient leaves 1% garbage on the next sample, so you often need source impedances near a few ohms for frequencies well into the MegaHertz. One nice things about the fat tube: its output impedance will be "low" and flat to many MHZ, but it will be around 100-200Ω, which may not be low enough for some digitizers, even the comparators used in triangle conversion.
It is one of the Asian languages. My browser somehow picked-up a font to display it properly, but I can't read it. If your browser has not installed the right Asian font, you probably get a lot of boxes or ASCII gibberish.
> the downswing looks terribly different than the upswing.
Yeah, but not so very different for signals up to a volt or two. To a first approximation: who cares what happens after that? The output stage will be in great distress. (In fact, asymmetric overload does affect sound at high levels....)
> Non-linear input capacitance in the ‘170 is the reason to prefer the 6DJ8
Ah, quite correct.
> I'm building a class-D power amp
Then you are doing great violence to the audio already, and have a major debugging chore ahead. Personally, if the Class-D needs a buffer, I'd throw in a NE5532 and move along to the big problems of Class-D. After the smoke clears, if the sound is good, I'd explore other linear input buffers.
> if it is absolute rubbish
If you can accept a volt or two peak output, it is fine. Resistor-loaded, it is mildly sweet. CCS loaded it will be simply clean. When taken past a volt or so, it overloads, and asymmetrically.
> why self bias?? Maybe to dial in more voltage across the tube than the CCS?
The tube has to take most of the voltage: electrons won't cross vacuum as easily as doped sillycon.
My worry is that, of all the possible biases you could arrive at with the trimmer, only some will work. And with higher voltages, some will melt the JFET.
> I checked the circuit with a Tesla E88CC tube. Results are: Va=44V Ia=5mA Vgc=-0.5V The AC signal at output looks fine up to 6Vpp. source impedance... 10k resistor
Seems reasonable. Me, I would take a little less volts on the tube, maybe 3/4 of available supply voltage. That would give Va=37V, Ia=4mA, Vgk~ 0.4V. There is a trade-off between cathode bottoming on negative swings and grid-current limiting on positive swings.
> The AC signal at output looks fine up to 6Vpp.
My two-thumbs estimate was "gross distortion" at about 4V peak on one side. With B+=48V, Va=44V, it can't hit 4V either way, so 6V p-p seems likely. THD will be about 0.5%, so it will "look fine". IF you can live within 6V p-p, it should be OK, either resistor-loaded for a little sweetness or CSS loaded for very-clean.
> a class-D power amp
Then a real question is: how "solid" does the signal have to be? Some A/C converters need a SOLID driver to absorb great switching spikes and damp them before the next sampling moment. Note that 99% rejection of switch-transient leaves 1% garbage on the next sample, so you often need source impedances near a few ohms for frequencies well into the MegaHertz. One nice things about the fat tube: its output impedance will be "low" and flat to many MHZ, but it will be around 100-200Ω, which may not be low enough for some digitizers, even the comparators used in triangle conversion.
The problem is nonlinearity from grid current. With a lot of small signal tubes, stray below about a volt cathode-to-grid and you start seeing a strong dependence of distortion on source impedance, well before the traditional zero volt grid current onset. That's fine if you know your circuit is always being driven by a low source impedance, but lots of circuits aren't, especially when a volume control rears its head. I don't have too much ECC88 data at low grid voltages, but I have done a bunch on ECC81/12AT7 at low grid bias, and the distortion varies enormously from tube to tube and with varying source Z. One 12AT7 test circuit that I have data at hand for showed 0.07% THD at 1kHz with a 600 ohm driver, but nearly 1% at the same frequency and level with a 100K resistor interposed. That's a 20dB difference, pretty significant.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.