Folks,
I'm having an interesting issue with my high voltage, positive Maida regulator. For schematic see attachment to this post.
In the schematic you'll notice a positive and a negative Maida regulator. The negative works fine. It delivers a rock solid -160 V as adjusted.
But the positive regulator has a few issues as built. In its original state (as drawn in the schematic), it has trouble regulating above 380~385 V at light load. I figured it needed more current to regulate properly. I also needed to change the output voltage to 470 V, so I needed to make changes anyway...
To get 470 V out, I use an Antek AS-4T430. That gives me 600 V rectified DC (no load). All the caps have been replaced by two caps in series with a 680k in parallel with each cap to allow the voltages across the caps to split evenly.
I used 100 kOhm for the resistor from ADJ to ground. This should allow 4.7 mA to flow through the regulator. According to the National LM317 datasheet, this should be plenty. However, the regulator failed to regulate. I had to increase the current to over 15 mA before it would regulate properly. I'm wondering if the high resistances cause the LM317 to run out of loop gain or something... Anyway. I finally got the stinkin' thing to regulate.
I plugged it into my amp (see this thread for a rough schematic). And the silicon in the regulator went *poof*. I'm not sure it likes starting up into a 100 uF "short circuit".
My question is two-fold:
1) Is the Maida regulator suited for these high voltages at all? 600 V in, 470 V out. Is it normally capable of starting up into a relatively high capacitive load (100 uF) at these voltages?
2) If I'm required to burn 10 W in the ADJ-ground resistor to make the Maida regulator work, I might as well use a zener regulator. I'm thinking a 10x 47 V stack, each diode with its own 1200 uF/63 V cap (because I have a bunch) and a source follower on top. Are there any common pitfalls to this solution?
I will give the Maida another try. I'll replace the 6.2 V zener with a 15 V to give the LM317 more headroom. We'll see where this leads...
~Tom
I'm having an interesting issue with my high voltage, positive Maida regulator. For schematic see attachment to this post.
In the schematic you'll notice a positive and a negative Maida regulator. The negative works fine. It delivers a rock solid -160 V as adjusted.
But the positive regulator has a few issues as built. In its original state (as drawn in the schematic), it has trouble regulating above 380~385 V at light load. I figured it needed more current to regulate properly. I also needed to change the output voltage to 470 V, so I needed to make changes anyway...
To get 470 V out, I use an Antek AS-4T430. That gives me 600 V rectified DC (no load). All the caps have been replaced by two caps in series with a 680k in parallel with each cap to allow the voltages across the caps to split evenly.
I used 100 kOhm for the resistor from ADJ to ground. This should allow 4.7 mA to flow through the regulator. According to the National LM317 datasheet, this should be plenty. However, the regulator failed to regulate. I had to increase the current to over 15 mA before it would regulate properly. I'm wondering if the high resistances cause the LM317 to run out of loop gain or something... Anyway. I finally got the stinkin' thing to regulate.
I plugged it into my amp (see this thread for a rough schematic). And the silicon in the regulator went *poof*. I'm not sure it likes starting up into a 100 uF "short circuit".
My question is two-fold:
1) Is the Maida regulator suited for these high voltages at all? 600 V in, 470 V out. Is it normally capable of starting up into a relatively high capacitive load (100 uF) at these voltages?
2) If I'm required to burn 10 W in the ADJ-ground resistor to make the Maida regulator work, I might as well use a zener regulator. I'm thinking a 10x 47 V stack, each diode with its own 1200 uF/63 V cap (because I have a bunch) and a source follower on top. Are there any common pitfalls to this solution?
I will give the Maida another try. I'll replace the 6.2 V zener with a 15 V to give the LM317 more headroom. We'll see where this leads...
~Tom
You're on the right track. D1 should be at least 12V, especially with a Darlington. The output cap can be larger than 10u if you have the series resistor that you show. It seems weird given the commodity nature of the 317, but I've run into fakes- it might be worth double-checking. And you might put a protection diode from the output to the input. Scope the output to make sure you're not oscillating. And for layout, you want that 220R and the pot very tight to the output pin.
You do NOT want that 100uF load on the reg! It will fail 100% of the time. The regulator replaces that big cap.
You do NOT want that 100uF load on the reg! It will fail 100% of the time. The regulator replaces that big cap.
Tom, The circuit looks mostly OK to me, but my suggestions would be:
- Is the problem in the Darlington follower? Since the load seen by the darlington is a regulator, you could get oscillation during transients [of the Real part of the input impedance goes negative kind]. Especially with only 27 ohm in the emitter. I think that adding 120 ohm between the zener and the Darlington base might improve the margin on that.
- During charging of the 100uF, does the SOA of the follower transistor get violated, or does it avalanche? It might see nearly 600V and the peak charging current simultaneously. At high voltages, the peak power handling for a BJT is much reduced.
- Regulators like the 317 are not always happy working into low ESR capacitors. The high ESR of ordinary electrolytics forms an ultrasonic zero that prevents loop gain falling to one too soon and provoking oscillation. 1.0 or 2.2 ohms at the output of the regulator can help, if low ESR caps are used, or if MKP/MKT or other caps are used as a bypass.
Personally, I think that the risk of destruction during transients makes a 40V part somewhat a liability in a 600V circuit.
Note also that the output impedance of a 317 rises quickly with frequency (ie it looks inductive), and the slope of this effect varies with load current in a nonlinear way. All this plays havoc with the supply **when used with impulsive loads**, and can lead to noise whose spectrum bounces around with the music programme.
The Zener power follower presents no real problems in use, but a Power FET is really needed for 600V, for SOA margin. One sure precaution is 120 ohms right on the gate, and a G-S zener of 12V. With 600V, this stopper resistor should be lower max. value than in low voltage applications.
- Is the problem in the Darlington follower? Since the load seen by the darlington is a regulator, you could get oscillation during transients [of the Real part of the input impedance goes negative kind]. Especially with only 27 ohm in the emitter. I think that adding 120 ohm between the zener and the Darlington base might improve the margin on that.
- During charging of the 100uF, does the SOA of the follower transistor get violated, or does it avalanche? It might see nearly 600V and the peak charging current simultaneously. At high voltages, the peak power handling for a BJT is much reduced.
- Regulators like the 317 are not always happy working into low ESR capacitors. The high ESR of ordinary electrolytics forms an ultrasonic zero that prevents loop gain falling to one too soon and provoking oscillation. 1.0 or 2.2 ohms at the output of the regulator can help, if low ESR caps are used, or if MKP/MKT or other caps are used as a bypass.
Personally, I think that the risk of destruction during transients makes a 40V part somewhat a liability in a 600V circuit.
Note also that the output impedance of a 317 rises quickly with frequency (ie it looks inductive), and the slope of this effect varies with load current in a nonlinear way. All this plays havoc with the supply **when used with impulsive loads**, and can lead to noise whose spectrum bounces around with the music programme.
The Zener power follower presents no real problems in use, but a Power FET is really needed for 600V, for SOA margin. One sure precaution is 120 ohms right on the gate, and a G-S zener of 12V. With 600V, this stopper resistor should be lower max. value than in low voltage applications.
FWIW, I've never lost one in day to day use (fooling around in the lab, sure, burned plenty), and I've built close to a hundred of them. Used them in preamps and power amps, 200-600V.
The whole point of the output RC is to null the inductance; with 4-5 ohms and 40-50u, the output impedance is low (for this use) and flat to well above where the output transformer has given up.
A 100u load with no series resistance is the #1 culprit here, guaranteed failure mode. If Tom pulls that out and beefs his 10u cab up to 47u or so, all will be copacetic.
SY: Picture this (or have a quick look at the schematic): SET output stage, Loftin-White(ish), with a 47 uF from the cathode of the tube to B+. Two channels, 47 uF each = 100 uF (give or take). But they do have a 660 ohm resistor in series, so the total impedance seen by the power supply on start-up should be [100 uF + 330 ohm]. 330 ohm at 470 V.... That's over an amp. I can't really do anything about that as I'm not willing to change the Loftin-White so another topology. And the regulator didn't have issues at 375 V so it should be possible to make it work at higher voltages as well.
Rod: It does seem like the BJT has failed. I'm not sure about the LM317. I'll need to wire it up in a low-voltage circuit to verify. The original Maida had a darlington pair in there as well. You could use a single BJT, but at 100~200 mA output currents, you're looking at several mA in base current for a typical power BJT. This will lead to significant power burn in the base-collector resistor (series resistor for the zener). I'm not sure if it has avalanched or not. It seems to be the B-E junction that's fused to a short. I'll need to double-check. The argument that a 40 V device should not be used in a 600 V supply doesn't really hold water, though, I do understand where you're coming from. That's one thing I like about the zener regulator. All devices are rated for operation at Vin(max) or more. But on the other hand.... If low voltage devices shouldn't be used in a high-voltage design, we shouldn't use LEDs for bias either...
I'll give it another whirl tomorrow. We'll see where it goes. I'll keep you posted. Thank you very much for the initial sanity check.
~Tom
Rod: It does seem like the BJT has failed. I'm not sure about the LM317. I'll need to wire it up in a low-voltage circuit to verify. The original Maida had a darlington pair in there as well. You could use a single BJT, but at 100~200 mA output currents, you're looking at several mA in base current for a typical power BJT. This will lead to significant power burn in the base-collector resistor (series resistor for the zener). I'm not sure if it has avalanched or not. It seems to be the B-E junction that's fused to a short. I'll need to double-check. The argument that a 40 V device should not be used in a 600 V supply doesn't really hold water, though, I do understand where you're coming from. That's one thing I like about the zener regulator. All devices are rated for operation at Vin(max) or more. But on the other hand.... If low voltage devices shouldn't be used in a high-voltage design, we shouldn't use LEDs for bias either...
I'll give it another whirl tomorrow. We'll see where it goes. I'll keep you posted. Thank you very much for the initial sanity check.
~Tom
Hmmm..... Typical pinout: B-C-E for power BJT; G-D-S for MOS. One could drop one in there and try it out. No base current to worry about then.
I have a diode connected in reverse across the output but not one that goes from output to input of the entire regulator (anode at the output). I should be able to get away with just one diode from out to in rather than one across the BJT and one across the LM317, correct?
I saw your comments about the layout earlier. It's pretty tight, actually. The total trace length from OUT to ADJ is on the order of 20 mm if that. It's dominated by the passives in the path.
~Tom
Kevin,
Once I get mine debugged I intend to put it on a PCB and have the PCB professionally made. I'll have spares and would be happy to sell you one. Just shoot me a PM if you're interested and can wait a few weeks while I develop this.
My needs are 470 V, 200 mA but it could be tailored to other output voltages as well. The current needs to stay within the SOA of the components used.
~Tom
Oh.... Note to self: The NPN darlington needs to be operating within its SOA during start-up. I knew that... I bet that's the cause of the fried silicon.
The FJPF5027 I'm using only handles about 20 mA at 600 V Vce. Granted, it shouldn't see that much voltage for very long, but it does need to charge the capacitors in the amp...
~Tom
The FJPF5027 I'm using only handles about 20 mA at 600 V Vce. Granted, it shouldn't see that much voltage for very long, but it does need to charge the capacitors in the amp...
~Tom
Yeah..... It's the in-rush that kills the Maida cascode device. My spice sim hows the 47 uF cathode cap in my 300B amp draws 650 mA inrush. The cathode resistor is 725 ohm. 470/725 = 0.65 or 650 mA...
Way above the SOA of the wimpy cascode devices I'm using. I'm currently eying out some 800 V ~ 1 kV, 10+ A MOSFET devices on Digikey.
Of course, one could change from the Loftin-White topology to using a regular grounded cathode cap. That way only 200 mA is needed during start-up. The drawback of this approach would be that the 200 mA is sourced through the 300B tube. In addition, it complicates the signal path. Simplifying - and shortening - the signal path is the whole point of the Loftin-White circuit. The original L-W circuit from 1930 also included some hum-reducing circuit but with DC heated filaments I don't really see a need for that.
~Tom
Way above the SOA of the wimpy cascode devices I'm using. I'm currently eying out some 800 V ~ 1 kV, 10+ A MOSFET devices on Digikey.
Of course, one could change from the Loftin-White topology to using a regular grounded cathode cap. That way only 200 mA is needed during start-up. The drawback of this approach would be that the 200 mA is sourced through the 300B tube. In addition, it complicates the signal path. Simplifying - and shortening - the signal path is the whole point of the Loftin-White circuit. The original L-W circuit from 1930 also included some hum-reducing circuit but with DC heated filaments I don't really see a need for that.
~Tom
Hey Guys,
Has anyone seen these? They have stupid high voltage ratings. I don't know if they would work in a regulator. That kind of engineering is beyond me.
http://www.st.com/internet/analog/subclass/1229.jsp
Kevin
Has anyone seen these? They have stupid high voltage ratings. I don't know if they would work in a regulator. That kind of engineering is beyond me.
http://www.st.com/internet/analog/subclass/1229.jsp
Kevin
Last edited:
A discrete, protected regulator using a MOS ballast will be much more resilient than any BJT or 317 alternative.
Here is an example of such a regulator that could easily be adapated for other voltages/polarity:
http://www.diyaudio.com/forums/powe...-voltage-regulator-circuit-2.html#post2417920
It doesnt compromise on performances either.
Here is an example of such a regulator that could easily be adapated for other voltages/polarity:
http://www.diyaudio.com/forums/powe...-voltage-regulator-circuit-2.html#post2417920
It doesnt compromise on performances either.
I agree - a regulator with a single pass element should be capable of better dynamic performance than a "stacked" IC regulator. And better durability.
Have you built the circuit on the linked page? I am wondering about the aggressive pole formed by R3 and C1 - and the risk of oscillation at 500k .. 5MHz with that response. Does R4 zero it in time?
I prefer non-LDO circuits at high voltage though - the drain-output configuration in LDOs is much harder to tame, due to the open-loop behaviour, and difficulty with capacitive loads.
Thanks for linking the circuit, though - I always enjoy a good discrete design to check out.
I didn't build exactly that version of the circuit, but I tried a number of similar ones.
The schematic is bit misleading, because it upside down, but if you look more closely, you'll see that those components create in fact a zero, not a pole.
Attachments
Side topic for this thread, but the noise canceling in the original L-W was not for the filaments, but for the power supply. They didn't have access to large value caps, so quiet supplies were expensive. You could easily incorporate the cancellation circuit in you design, if you replace the 3V bias supply with a resistor. That resistor becomes part of a divider to divide down a signal taken from the output tube cathode, by the ratio of the u of the input tube. If you include a trim pot as part of the divider, you can null out the power supply noise completely.
In the original design, the input tube and output tube shared a supply and the ripple phase is the same for both, so the noise for both is nulled. In your case, the noise for the output would be nulled, but not for the driver. If you are regulating both, it's probably a moot point. If you are regulating only the driver, then it is worth considering.
Sheldon
Last edited:
I disagree, actually. Run a sim of the loop gain and it's pretty easy to convince yourself of the advantages of the IC regulators. You do have to ensure that the cascode device will operate within its SOA under all operating conditions. If the cascode fries, so will the IC.
It's the same level of reliability for the discrete regulator. If the pass device in the discrete regulator fries due to operation outside its SOA, the regulator won't regulate.
~Tom
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.