I only wish I had my tubes with me to try it.
Should tolerate inductive loads (e.g. PWM filter choke) while driving substantial output current. Dunno what saturation voltage is going to be though.
Wonder if the high side (even with peaking coil) will have enough dV/dt. It could be quite slow to turn on.
Tim
An externally hosted image should be here but it was not working when we last tested it.
Should tolerate inductive loads (e.g. PWM filter choke) while driving substantial output current. Dunno what saturation voltage is going to be though.
Wonder if the high side (even with peaking coil) will have enough dV/dt. It could be quite slow to turn on.
Tim
Hmmmm... What is a good resource to get up to speed on the basics of class D amplification? I don't have a clue how this thing works, but I'd like to.
Basically, drive it with PWM. The average value of a square wave is easy to calculate, as it's on for one voltage (say +V) for part of the time, then off the rest. So the average is just +V * duty. A filter averages out the switching nasties, leaving you with the average again. Pipe the averaged signal over to your output transformer and there you go.
The best part comes when making the square wave. Because the tubes are only ever conducting ~50V at 1A or 500V at 0.001A, the average power dissipation is moderate, while the total output is huge (500V * 1A = 500VA, figure 1/2 of that usable as real undistorted audio, or 250W!).
The hard part is making tubes switch without toastergrid. Triodes don't "switch", they just make crummy resistors. Pentodes switch, but screen current goes into the stratosphere when plate voltage is low. And you'll notice the damper diodes, which are required because, not only will plate voltage be low, it will be *negative* for part of the cycle!
Tim
The best part comes when making the square wave. Because the tubes are only ever conducting ~50V at 1A or 500V at 0.001A, the average power dissipation is moderate, while the total output is huge (500V * 1A = 500VA, figure 1/2 of that usable as real undistorted audio, or 250W!).
The hard part is making tubes switch without toastergrid. Triodes don't "switch", they just make crummy resistors. Pentodes switch, but screen current goes into the stratosphere when plate voltage is low. And you'll notice the damper diodes, which are required because, not only will plate voltage be low, it will be *negative* for part of the cycle!
Tim
Here's an interesting variation on the tube class D approach by David Berning. He calls it OTL impedance matching, which I guess is correct. But it seems to be more accurately a class D amp that uses a standard class D output stage in which a vacuum tube is used to imprint a tube's characteristic curve, and subsequently the sound of a tube amp, into a class D amp. Very interesting. 
United States Patent: 5612646
Tube amplifiers for high-end audio by The David Berning Company
Edit: Seems I was beaten to the punch yet again.
http://www.diyaudio.com/forums/class-d/69538-tubes-class-d.html
United States Patent: 5612646
Tube amplifiers for high-end audio by The David Berning Company
Edit: Seems I was beaten to the punch yet again.
http://www.diyaudio.com/forums/class-d/69538-tubes-class-d.html
Last edited:
Berning's technology is quite interesting. It's essentially the reciprocal of what I'm doing: I propose using the tubes directly in class D. Instead, he uses switching transistors to build a "DC transformer", which saves on iron (not eliminating it, since it still requires a little ferrite transformer), allowing the tubes to operate in class A or AB, linear.
This circuit still requires a monster OPT, and indeed it will be quite monstrous given the number of tubes behind it. Since it has a low impedance output, it's excellently capable of low frequency performance, suggesting a transformer suitable for full power down to 20Hz.
There is one advantage to OPT designers: the primary impedance is dramatically reduced. For instance, Hammond doesn't even make an OPT low enough for a typical class D amp (e.g., 500V supply, PP 6LQ6 outputs). The impedance required is about 600 ohms!
To use Hammond's lowest impedance: the 1650W, 1.9ka-a and 280W, would require close to 1200V supply (and since we're using sweep tubes here..... who's gonna stop us??!!
), and easily deliver more like 500W output (while keeping plate dissipation constant, placing efficiency in the upper 80% range!). (The 1640SEA, at 1250 ohms, might be more suitable, being this circuit is "SEPP", but with only 30W capacity, it would burn up easy under power, and saturate easily even if the core were restacked for AC operation.)
Tim
This circuit still requires a monster OPT, and indeed it will be quite monstrous given the number of tubes behind it. Since it has a low impedance output, it's excellently capable of low frequency performance, suggesting a transformer suitable for full power down to 20Hz.
There is one advantage to OPT designers: the primary impedance is dramatically reduced. For instance, Hammond doesn't even make an OPT low enough for a typical class D amp (e.g., 500V supply, PP 6LQ6 outputs). The impedance required is about 600 ohms!
To use Hammond's lowest impedance: the 1650W, 1.9ka-a and 280W, would require close to 1200V supply (and since we're using sweep tubes here..... who's gonna stop us??!!
), and easily deliver more like 500W output (while keeping plate dissipation constant, placing efficiency in the upper 80% range!). (The 1640SEA, at 1250 ohms, might be more suitable, being this circuit is "SEPP", but with only 30W capacity, it would burn up easy under power, and saturate easily even if the core were restacked for AC operation.)Tim
The oldest schematic of a class-d amp that I have ever seen was indeed using tubes. It wasn't an audio amp but a DC/AC converter for 400 Hz (=aircraft onboard power grid). It was a US patent dating around WW 2.
Regards
Charles
Regards
Charles
I've been thinking about this when I was trying to figure out details for my idea and realized that sliding screen mechanism could be used to prevent screen meltdown. It will reduce gain slightly but so would any ower limit on g1 swing imposed from the outside (zener, neon lamp, diode + external reference, varistor ...). You can tweak stopper's value to get the maximum output until glow appears, then increase its size slightly again.
Yup, hence the screen resistors.
Offhand, it looks like something around 150V, 100mA (voltage and current limits) will drive a 6KD6 into saturation quite nicely while keeping screen dissipation fairly low, even when plate voltage is zero or negative. In fact, because the screen becomes such an excellent diode, it shorts out the CCS, minimizing screen dissipation.
A screen choke might be handy, too (low current when the screen starts yanking, then rising as plate voltage rises), but an RCD snubber is needed to keep peak screen voltage down. You end up wasting as much power as a CCS, though you're only using another diode, so it might not be too bad.
Tim
Offhand, it looks like something around 150V, 100mA (voltage and current limits) will drive a 6KD6 into saturation quite nicely while keeping screen dissipation fairly low, even when plate voltage is zero or negative. In fact, because the screen becomes such an excellent diode, it shorts out the CCS, minimizing screen dissipation.
A screen choke might be handy, too (low current when the screen starts yanking, then rising as plate voltage rises), but an RCD snubber is needed to keep peak screen voltage down. You end up wasting as much power as a CCS, though you're only using another diode, so it might not be too bad.
Tim
Even a resistor may be handy. This looks like a useful graph:
http://myweb.msoe.edu/williamstm/Images/6KD6_Graph1.jpg
Compare this plot and the plate curves (screen voltage as parameter) from which it was derived. At low plate voltages (20-50V), screen current increases abruptly with respect to plate voltage. At high plate voltages (above this knee), the screen is almost resistive. The resistive region shows up as the bunching of curves at low current, i.e. the 50-150V, 0-100mA region. At low plate voltages (or higher screen voltages), the screen has more of a power-law diode type characteristic (indeed, the Vg = 0 curve shows Ig2 = 0.1819*(Vg2)^1.54, an excellent representation of the 3/2 power law). The knee between them seems to be approximately Vg2 = 3*Vp.
Now, with this graph, you can draw screen load lines. This might be of value to the Steve Bench types, who might be interested in plate-to-screen mu and transconductance (ballpark values look like 10mmho, mu = 3.5). The bizzare feature of this graph is it shows screen-output characteristics, yet seems to represent values similar to the screen driven, plate-output graph!
For the class D amp, we can decide what kind of current limiting could be used. Ideally, a foldback current limit should start at maybe 100mA (so the screen drops to about 60V minimum peak), rising to perhaps 200mA in the 100-150V range, then dropping off again as voltage rises over 175V, dropping to maybe 0mA at 175-200V. The nice part about foldback is, because we know what the plate voltage is, we could clamp it with a diode -- a tube "baker clamp".
Tim
http://myweb.msoe.edu/williamstm/Images/6KD6_Graph1.jpg
Compare this plot and the plate curves (screen voltage as parameter) from which it was derived. At low plate voltages (20-50V), screen current increases abruptly with respect to plate voltage. At high plate voltages (above this knee), the screen is almost resistive. The resistive region shows up as the bunching of curves at low current, i.e. the 50-150V, 0-100mA region. At low plate voltages (or higher screen voltages), the screen has more of a power-law diode type characteristic (indeed, the Vg = 0 curve shows Ig2 = 0.1819*(Vg2)^1.54, an excellent representation of the 3/2 power law). The knee between them seems to be approximately Vg2 = 3*Vp.
Now, with this graph, you can draw screen load lines. This might be of value to the Steve Bench types, who might be interested in plate-to-screen mu and transconductance (ballpark values look like 10mmho, mu = 3.5). The bizzare feature of this graph is it shows screen-output characteristics, yet seems to represent values similar to the screen driven, plate-output graph!
For the class D amp, we can decide what kind of current limiting could be used. Ideally, a foldback current limit should start at maybe 100mA (so the screen drops to about 60V minimum peak), rising to perhaps 200mA in the 100-150V range, then dropping off again as voltage rises over 175V, dropping to maybe 0mA at 175-200V. The nice part about foldback is, because we know what the plate voltage is, we could clamp it with a diode -- a tube "baker clamp".
An externally hosted image should be here but it was not working when we last tested it.
Tim
- Status
- Not open for further replies.