I guess I've been bitten...
Hoping to find a rail voltage and choke combination that would keep everything happy direct coupled without incurring huge dissipation in the FETs.
Potential chokes:
Hammond 156C 3k7 DCR 150H (used in the past and measured 3k25 actual DCR)
Hammond 155C 2k75 DCR 60H
Lundahl LL1667 2k4 DCR 270H
Lundahl LL1668 680 DCR 100H
Hoping to find a rail voltage and choke combination that would keep everything happy direct coupled without incurring huge dissipation in the FETs.
Potential chokes:
Hammond 156C 3k7 DCR 150H (used in the past and measured 3k25 actual DCR)
Hammond 155C 2k75 DCR 60H
Lundahl LL1667 2k4 DCR 270H
Lundahl LL1668 680 DCR 100H
Nice design Sodacose. The TL431's work well for this situation. I am not sure finding a Choke will be as easy as using a cheap IRF240 and a slab of aluminum fin (with a $1 fan).
I am working on a 10w SE class A design right now and using IRFP240 as current sink also. But finding it's cooking maybe 32w.
I am working on a 10w SE class A design right now and using IRFP240 as current sink also. But finding it's cooking maybe 32w.
Cool. Got a schematic?
The choke would be for the tube anode load, btw, not the MOSFET. Playing with numbers, I don't think there's a real practical way to direct couple, keep the MOSFET dissipation at a sane level, and provide a decent output.
Two 7370s in parallel with a Hammond 156C load and 40V B+ would only net around 5W because it would be limited by the grid cutoff. If adding a second power rail for the tube and cap coupling, it would be very simple (heatsinks allowing) to quadruple the power.
I'm thinking around a 25.2V power rail, allowing two 12.6V heater tubes in series without an extra heater winding. This should do almost 10W. Requires an extra B+ for the tubes, but the current requirement would be very small.
The choke would be for the tube anode load, btw, not the MOSFET. Playing with numbers, I don't think there's a real practical way to direct couple, keep the MOSFET dissipation at a sane level, and provide a decent output.
Two 7370s in parallel with a Hammond 156C load and 40V B+ would only net around 5W because it would be limited by the grid cutoff. If adding a second power rail for the tube and cap coupling, it would be very simple (heatsinks allowing) to quadruple the power.
I'm thinking around a 25.2V power rail, allowing two 12.6V heater tubes in series without an extra heater winding. This should do almost 10W. Requires an extra B+ for the tubes, but the current requirement would be very small.
Sodacose,
Your design looks good. May I offer some suggestions to improve power and efficiency?
Class A is inefficient, and you might like to reduce the voltage across the sense resistor which here dissipates 5W.
We are using a mosfet CCS, and over 1A they slip into negative tempco. So if you use a 0.22R sense resistor, and string a 10k resistor from Vcc, then a 1k trimmer, then two green LEDs in series, you can drive the gate via a 220R stopper from the wiper of the trimmer. It is rudimentary, but it's all you need, and offers indication and fool proof operation at any temperature. You can put in a 10uF cap from wiper to ground to make it utterly silent.
If you use a large mosfet like a FQA40N25 from Fairchild, you get a 280W rating and very high transconductance at 2A; S is 29, but at 2A it would be around 20 and this is maybe four times that of the IRFP240. Higher transconductance means least compressive distortion at higher outputs, and higher power means you can run it at around 40V Vcc.
Your inductor will have to be made custom. I use a 100mH running at 4A standing current, DCR of 0.47R and gapped with GOSS to 0.5mm in a 4" x 3" x 5" lump. This will cost you plenty, and it uses 276 turns of 1.2mm DTE enamel, but it doubles the efficiency of the Class A SE hybrid. At 4A, this inductor drops just under 2V, dissipates 8W, never gets hot, but will swing down 80% of the Vcc; that's down to-32V, total 64Vpp into 8R. That is over 60W from a 40Vcc, not too shabby. It does dissipate at 160W from the supply, a heavy load, but to deliver 60W of glorious SE sound it's worth it....... it is still much more efficient than a gasoline engine in a car.
Note that I'm suggesting a choke in place of the CCS, actually, but there is no requirement from the tube and in fact I'd suggest you use a much higher B+ (see Broskie for a 6DJ8 circuit) and cap connect to the source follower.
I do like your moniker......
HD
Your design looks good. May I offer some suggestions to improve power and efficiency?
Class A is inefficient, and you might like to reduce the voltage across the sense resistor which here dissipates 5W.
We are using a mosfet CCS, and over 1A they slip into negative tempco. So if you use a 0.22R sense resistor, and string a 10k resistor from Vcc, then a 1k trimmer, then two green LEDs in series, you can drive the gate via a 220R stopper from the wiper of the trimmer. It is rudimentary, but it's all you need, and offers indication and fool proof operation at any temperature. You can put in a 10uF cap from wiper to ground to make it utterly silent.
If you use a large mosfet like a FQA40N25 from Fairchild, you get a 280W rating and very high transconductance at 2A; S is 29, but at 2A it would be around 20 and this is maybe four times that of the IRFP240. Higher transconductance means least compressive distortion at higher outputs, and higher power means you can run it at around 40V Vcc.
Your inductor will have to be made custom. I use a 100mH running at 4A standing current, DCR of 0.47R and gapped with GOSS to 0.5mm in a 4" x 3" x 5" lump. This will cost you plenty, and it uses 276 turns of 1.2mm DTE enamel, but it doubles the efficiency of the Class A SE hybrid. At 4A, this inductor drops just under 2V, dissipates 8W, never gets hot, but will swing down 80% of the Vcc; that's down to-32V, total 64Vpp into 8R. That is over 60W from a 40Vcc, not too shabby. It does dissipate at 160W from the supply, a heavy load, but to deliver 60W of glorious SE sound it's worth it....... it is still much more efficient than a gasoline engine in a car.
Note that I'm suggesting a choke in place of the CCS, actually, but there is no requirement from the tube and in fact I'd suggest you use a much higher B+ (see Broskie for a 6DJ8 circuit) and cap connect to the source follower.
I do like your moniker......
HD
Last edited:
Hi Aksa,
Thanks for the input! I'll take a look at that Fairchild part. Not married to the IRFP240 at all; I just see them used here and there and figured it would be a decent option. I'm really a tube builder and just starting to dip my toes more into solid state. I'm still a little in the dark on preferred parts.
I'm coming to the same conclusion as you are about cap coupling the tube stage. I'd really love to direct couple, but it would play best with at least 50V anode-cathode on the tube stage and that would lead to huge dissipation in the FETs if it's all on a single power rail. I'll be redesigning the input, probably using a Mu follower topology. I'm hoping to use a Vcc that's a multiple of tube heater voltage so that I can at least eliminate one extra power supply winding (25.2V for two 12.6V heaters in series, 40V for 20/40V heaters, etc). B+ for the tube can then be a multiplied power winding or a daisy chain power transformer. Only issue would be to keep Vhk below maximum if the heaters are powered off of the Vcc.
To be sure I understand your bias scheme, this is for the CCS correct? Two greens are around 6.6V and variable with the 1k pot. So this would provide a very stable gate voltage reference and some feedback via the sense resistor. But can't current draw still vary with Vgs? I readily admit that thermal considerations and high power FETs are a new topic for me and I really need to do more reading.
The TL431 has a solid Vref of 2.5V and so I was counting on that and the sense resistor to set a stable current. Variation in current through the sense resistor causes the TL431 to compensate with current draw through the resistor from Vcc, pushing the gate voltage up or down until the Vref is 2.5V. Seems like it should be a really high impedance CCS with this arrangement. I was basing the schematic on how I had seen TL431 used in shunt regulators, but low and behold the datasheet has exactly the same thing on it (Figure 39, Page 29):
http://www.ti.com/lit/ds/symlink/tl431.pdf
I'd love to choke load everything. I just don't want to pay for the privilege
Thanks for the input! I'll take a look at that Fairchild part. Not married to the IRFP240 at all; I just see them used here and there and figured it would be a decent option. I'm really a tube builder and just starting to dip my toes more into solid state. I'm still a little in the dark on preferred parts.
I'm coming to the same conclusion as you are about cap coupling the tube stage. I'd really love to direct couple, but it would play best with at least 50V anode-cathode on the tube stage and that would lead to huge dissipation in the FETs if it's all on a single power rail. I'll be redesigning the input, probably using a Mu follower topology. I'm hoping to use a Vcc that's a multiple of tube heater voltage so that I can at least eliminate one extra power supply winding (25.2V for two 12.6V heaters in series, 40V for 20/40V heaters, etc). B+ for the tube can then be a multiplied power winding or a daisy chain power transformer. Only issue would be to keep Vhk below maximum if the heaters are powered off of the Vcc.
To be sure I understand your bias scheme, this is for the CCS correct? Two greens are around 6.6V and variable with the 1k pot. So this would provide a very stable gate voltage reference and some feedback via the sense resistor. But can't current draw still vary with Vgs? I readily admit that thermal considerations and high power FETs are a new topic for me and I really need to do more reading.
The TL431 has a solid Vref of 2.5V and so I was counting on that and the sense resistor to set a stable current. Variation in current through the sense resistor causes the TL431 to compensate with current draw through the resistor from Vcc, pushing the gate voltage up or down until the Vref is 2.5V. Seems like it should be a really high impedance CCS with this arrangement. I was basing the schematic on how I had seen TL431 used in shunt regulators, but low and behold the datasheet has exactly the same thing on it (Figure 39, Page 29):
http://www.ti.com/lit/ds/symlink/tl431.pdf
I'd love to choke load everything. I just don't want to pay for the privilege
So the previous schematic just wasn't leaving any headroom for the tube without creating huge dissipation in the FETs (due to high Vcc). One could add a B+ for the tube and cap couple (also drew up a schematic for that), but I was thinking about how to keep it all direct coupled and came up with the above incorporating a relatively low voltage negative supply. The 12B4A Mu of 6.5 seems workable here (max output at under 1.5Vrms input).
The two 12B4 heaters could be wired in series on the 25V power rail. The -50V supply should be easy to get, maybe with the same power transformer.
The 12.5mA current through the choke drops .0125 * 680 = 8.5V (easy to tweak up and down with the CCS).
One thing I'm not sure of is how exactly to calculate power output. If assuming it can swing the whole 25Vptp, it's close to 10W into 8 ohms. But I think we'd have to subtract the 2.5V Vref for the TL431 at least. Does anyone know how to ballpark the available voltage swing from the FETs in an amp like this?
I think you have about 4v drop in each mosfet and 2.5v in TLM431. So 10.5v off the rail so you are at 14.5v and maybe subtract a couple more for headroom before clipping. So maybe 12v swing peak-peak. Or 8.5v rms which is about 9w into 8ohm load. Does that sound about right? These SE amps are not very efficient.
I like how simple this is and it reminds me of my pocket headamp. Maybe I can just replace the tube with a cascoded JFET and get a nice 8w headamp?
Although a 1k drain resistor would probably protect a BF862 sufficiently to brushed without a cascode. Or just use an 2sk170.
I just tried Hugh's string of 10k, 1k pot with wiper to gate of CCS, and 2 LEDs - it works very well to control bias on Class A from 1A to 2A.
I like how simple this is and it reminds me of my pocket headamp. Maybe I can just replace the tube with a cascoded JFET and get a nice 8w headamp?
Although a 1k drain resistor would probably protect a BF862 sufficiently to brushed without a cascode. Or just use an 2sk170.
I just tried Hugh's string of 10k, 1k pot with wiper to gate of CCS, and 2 LEDs - it works very well to control bias on Class A from 1A to 2A.
Last edited:
Hmm, 12V ptp is about 4Vrms. That's only 2W output into 8 ohms.
Thinking about it another way, the source of the driven FET follows the gate. If the gate is at about 16V, it can swing up to 25V and down to about 7V. That's 18V ptp or around 6.3Vrms. This would be about 5W into 8 ohms.
Good to know about the LED bias scheme! Could easily buy a watt of extra power by saving a couple volts in output swing.
Thinking about it another way, the source of the driven FET follows the gate. If the gate is at about 16V, it can swing up to 25V and down to about 7V. That's 18V ptp or around 6.3Vrms. This would be about 5W into 8 ohms.
Good to know about the LED bias scheme! Could easily buy a watt of extra power by saving a couple volts in output swing.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.