I had a sneaky suspicion that the soft start circuit wouldn't necessarily prevent SOA failures on the cascode device. So I ran a quick sim.
On start-up, the base of the soft-start PNP is grounded by the cap, but the transistor doesn't start turning on until the emitter voltage is at least 0.5~0.6 V. This leaves some time where the regulator is providing the full output current - limited by the Vgs of the cascode, the minimum drop of the regulator, and the voltage across the resistor on the regulator input. As I've described earlier, this current limit is actually rather ill-defined as none of the voltages except that across the resistor are guaranteed by the manufacturer of those components.
Using a MOS for the soft-start transistor makes matters worse as the MOS doesn't turn on until a Vgs of some 4~5 V is reached.
That said, the start-up current is much better controlled on the soft-start version of the regulator. It does peak at the value set by the current limiter - as predicted - but within 80 us it returns to a few mA (charging current for the external cap).
This is looking pretty encouraging. I think I'll build myself a prototype... If it works well, I'll follow up with some boards.
At 200 V, life is relatively easy. I'm using upwards of 600 V in at the worst case corner. Much more energy and rather more difficult to ensure reliable operation.
~Tom
On start-up, the base of the soft-start PNP is grounded by the cap, but the transistor doesn't start turning on until the emitter voltage is at least 0.5~0.6 V. This leaves some time where the regulator is providing the full output current - limited by the Vgs of the cascode, the minimum drop of the regulator, and the voltage across the resistor on the regulator input. As I've described earlier, this current limit is actually rather ill-defined as none of the voltages except that across the resistor are guaranteed by the manufacturer of those components.
Using a MOS for the soft-start transistor makes matters worse as the MOS doesn't turn on until a Vgs of some 4~5 V is reached.
That said, the start-up current is much better controlled on the soft-start version of the regulator. It does peak at the value set by the current limiter - as predicted - but within 80 us it returns to a few mA (charging current for the external cap).
This is looking pretty encouraging. I think I'll build myself a prototype... If it works well, I'll follow up with some boards.
At 200 V, life is relatively easy. I'm using upwards of 600 V in at the worst case corner. Much more energy and rather more difficult to ensure reliable operation.
~Tom
Attachments
Hi, Tom. Good to hear from you. You're right, it is quiet here.
I tried shorting the prototype output with a 47uF capacitor added in parallel with the 1uF. The model predicts a modest increase to 22A during the short, but for a longer duration.
For the benefit of others who may be reading, here's what happens;
On switch on, the output gradually rises over a period of about 10 seconds, being regulated all the way up, until it reaches 144V in this example, and there it settles. When the output is shorted, the capacitors quickly discharge, releasing a burst of energy in the form of a spark, with some current going back through the protection diodes. The choke's high reactance at very high frequencies limits the initial discharge current and prevents high current surges through the protection devices.
Once in the shorted state, with output voltage around zero, the current is limited by the voltage developed across the resistor, I_LIMIT, and the MOSFET dissipates all the heat. The limiting current = (V_gate-output - LM317 voltage drop – MOSFET VGS(TH) ) / I_LIMIT or ~ (15 – 3 – 3)/270 = 33mA in this case. As Tom says, this is very approximate. The voltage across the MOSFET in this shorted state is nearly the supply voltage, so dissipation is around 6W for this design. More heat is generated during the short than it is at the beginning of a normal start up where current is limited by R_load plus that required to charge the capacitors.
On releasing the short, the cycle immediately repeats from zero output volts. If shorted before the final voltage is reached, the ramp is immediately terminated and, on release, the cycle then resumes from zero volts as before.
The spark was much more invigorating with 48uF and, yes, it survived the 10 or so cycles I put it through.
Regards,
Bob
I tried shorting the prototype output with a 47uF capacitor added in parallel with the 1uF. The model predicts a modest increase to 22A during the short, but for a longer duration.
For the benefit of others who may be reading, here's what happens;
On switch on, the output gradually rises over a period of about 10 seconds, being regulated all the way up, until it reaches 144V in this example, and there it settles. When the output is shorted, the capacitors quickly discharge, releasing a burst of energy in the form of a spark, with some current going back through the protection diodes. The choke's high reactance at very high frequencies limits the initial discharge current and prevents high current surges through the protection devices.
Once in the shorted state, with output voltage around zero, the current is limited by the voltage developed across the resistor, I_LIMIT, and the MOSFET dissipates all the heat. The limiting current = (V_gate-output - LM317 voltage drop – MOSFET VGS(TH) ) / I_LIMIT or ~ (15 – 3 – 3)/270 = 33mA in this case. As Tom says, this is very approximate. The voltage across the MOSFET in this shorted state is nearly the supply voltage, so dissipation is around 6W for this design. More heat is generated during the short than it is at the beginning of a normal start up where current is limited by R_load plus that required to charge the capacitors.
On releasing the short, the cycle immediately repeats from zero output volts. If shorted before the final voltage is reached, the ramp is immediately terminated and, on release, the cycle then resumes from zero volts as before.
The spark was much more invigorating with 48uF and, yes, it survived the 10 or so cycles I put it through.
Regards,
Bob
Bob,
For low currents and low input voltages, such as the 30 mA, 200 V in your example, you're not likely to run into SOA violations if you're using a reasonably beefy MOSFET - or even better BJT - for the cascode. I have no doubt that your circuit works.
My design target was 200 mA, 600 V and the regulator was designed to current limit around 250 mA as I recall. I turned several beefy (10 A, 950 V, 230 W rated) MOSFETs into silicon slag before I realized that they couldn't handle the in-rush current during start-up. However, that was without soft start. I think the soft-start should make it possible for the device to survive start-up.
Another option would be to use a tube rectifier. 5AR4 with its slow warm-up characteristic would be a good choice.
With the lower voltages, currents in your case we're not comparing apples to apples. Good for you that your circuit works, though.
~Tom
For low currents and low input voltages, such as the 30 mA, 200 V in your example, you're not likely to run into SOA violations if you're using a reasonably beefy MOSFET - or even better BJT - for the cascode. I have no doubt that your circuit works.
My design target was 200 mA, 600 V and the regulator was designed to current limit around 250 mA as I recall. I turned several beefy (10 A, 950 V, 230 W rated) MOSFETs into silicon slag before I realized that they couldn't handle the in-rush current during start-up. However, that was without soft start. I think the soft-start should make it possible for the device to survive start-up.
Another option would be to use a tube rectifier. 5AR4 with its slow warm-up characteristic would be a good choice.
With the lower voltages, currents in your case we're not comparing apples to apples. Good for you that your circuit works, though.
~Tom
Did some more simming. Found another potential pitfall. Again, this applies to my 600 V in, 470 V out, 200 mA design. Significant power is dissipated in the soft-start transistor during start-up. One will have to use a fairly beefy MOS here as well. High voltage, high power PMOS and PNP are becoming scarce. Hmmm...
~Tom
~Tom
Yeah.... 600 V would be OK, but wouldn't survive in a regulator blow-up with Vin > 600 V. Not that that's necessarily a requirement, but just something to keep in mind.
I did see some beefy IXYS PMOS parts in TO-247 packages somewhere. Probably Mouser or Digikey. They were expensive, though. $10/each.... That's a bit much for a soft-start transistor. I would also hate to get the design done and have part availability go to zero because IXYS decided they weren't making enough money on high-voltage PMOSes.
~Tom
I did see some beefy IXYS PMOS parts in TO-247 packages somewhere. Probably Mouser or Digikey. They were expensive, though. $10/each.... That's a bit much for a soft-start transistor. I would also hate to get the design done and have part availability go to zero because IXYS decided they weren't making enough money on high-voltage PMOSes.
~Tom
I appreciate that we're designing for different purposes, Tom, and that your design is much more demanding on safe operating area of power devices, but that isn't the only demanding criterion or the most critical application.
I'm trying to design the equivalent of a valve rectifier preamp power supply on a simple PCB which will perform better, last longer, be just as tolerant to abuse and, most importantly, have a similar soft start and be stable at DC – hence the Maida interest.
I specified my reasons earlier in that I've designed a two stage feedback DAC valve amplifier with subsidiary DC feedback, and this requires a DC supply for stability. Further, because of the relatively high (0.47uF) output coupling capacitors, it will undoubtedly present high voltage spikes (>100V) to the input of the power amplifier at switch-on without a soft start and with cold valves with infinite rp. When valves are hot, voltage spikes are irrelevant at switch-off, given reasonable smoothing capacitor values and decay time constants.
The simulations I've shown may be unrepresentative of real life with parasitic inductances and capacitances and the component models may be wrong but they led me to believe that I'd identified the reason why these SS circuits fail (usually by abuse such as shorts) in HT supplies – a chain reaction beginning with the failure of protection zeners through high current surges, and I hope my posts have identified a solution to this problem as well as offering a soft start which ought to be seriously considered for every valve preamp design with output coupling capacitors => 0.1uF feeding into a power amp, especially if it's SS or already switched on.
Life isn't that simple with a preamp power supply either.
I'm trying to design the equivalent of a valve rectifier preamp power supply on a simple PCB which will perform better, last longer, be just as tolerant to abuse and, most importantly, have a similar soft start and be stable at DC – hence the Maida interest.
I specified my reasons earlier in that I've designed a two stage feedback DAC valve amplifier with subsidiary DC feedback, and this requires a DC supply for stability. Further, because of the relatively high (0.47uF) output coupling capacitors, it will undoubtedly present high voltage spikes (>100V) to the input of the power amplifier at switch-on without a soft start and with cold valves with infinite rp. When valves are hot, voltage spikes are irrelevant at switch-off, given reasonable smoothing capacitor values and decay time constants.
The simulations I've shown may be unrepresentative of real life with parasitic inductances and capacitances and the component models may be wrong but they led me to believe that I'd identified the reason why these SS circuits fail (usually by abuse such as shorts) in HT supplies – a chain reaction beginning with the failure of protection zeners through high current surges, and I hope my posts have identified a solution to this problem as well as offering a soft start which ought to be seriously considered for every valve preamp design with output coupling capacitors => 0.1uF feeding into a power amp, especially if it's SS or already switched on.
Life isn't that simple with a preamp power supply either.
Soft start can be made to mimic the slow turn-on of a valve rectifier. I think you have that down. I'm also pretty impressed by your short circuit abuse.... Careful not to electrocute yourself there!
But for the HV spikes on the amp outputs; have you considered adding a beefy zener diode to clamp the output voltage? Depending on the plate load on the output stage of your preamp, that may be an option. Or run a normal 1N4007 diode from the power amp input to its positive supply.
You could of course also just use an output relay... Or a time delay on the B+ supply.
I'd probably add 100 kOhm or so to ground on the output of your amp. At least try that to see if it helps any on the output voltage spikes.
Thanks for lighting the Maida fire under me again. Good stuff!
Oh, and by the way. On the lower voltage Maida regulator I used a while back, I had a 5AR4 rectifier running the supply. The slow turn-on of the tube rectifier prevented any SOA failures.
~Tom
But for the HV spikes on the amp outputs; have you considered adding a beefy zener diode to clamp the output voltage? Depending on the plate load on the output stage of your preamp, that may be an option. Or run a normal 1N4007 diode from the power amp input to its positive supply.
You could of course also just use an output relay... Or a time delay on the B+ supply.
I'd probably add 100 kOhm or so to ground on the output of your amp. At least try that to see if it helps any on the output voltage spikes.
Thanks for lighting the Maida fire under me again. Good stuff!
Oh, and by the way. On the lower voltage Maida regulator I used a while back, I had a 5AR4 rectifier running the supply. The slow turn-on of the tube rectifier prevented any SOA failures.
~Tom
Last edited:
Yes, I considered and dismissed all of these options, Tom.
Preamp output loads are design dependant.
Protection devices such as zeners which are permanently in place across the preamp output have significant resistances and capacitances which are a complex function of output voltage. The capacitance can be minimised by placing a diode, i.e. 1N914, in series with the zener. Even so, it's a compromised solution and there'll still be a pop through the speakers.
Opening of NC relay contacts across the channel outputs after a timed delay period is a better solution but it gets overly complex ensuring that the timing circuit is fool proof with repeated switch-on / switch-off cycles.
A delay on the B+ supply will create a smaller spike when the output valve is hot and it's rp is low but it can still be very significant and, again, making fool proof timing circuits gets overly complex. Closing of NO relay contacts across a high value resistor in the B+ supply line after a timed delay has similar problems.
Using a valve rectifier can also cause a significant pop through the speakers with repeated switch-on / switch-off cycles if this is done when the rectifier cathode is still hot.
It was the dismissal of all of these usual options which prompted me to look into the soft start regulator design. It's poor design practice to create a voltage spike and then try to eliminate it where its creation can be avoided. In this design, there can never be a voltage spike at the regulator output (except if it's shorted), even with repeated switch-on / switch-off cycles, which means that there can never be a voltage spike from the preamp output (unless output coupling capacitor values are ridiculously high offering almost DC to the power amp). It's totally fool proof in this respect. During repeated switch-off / switch-on cycles, the output voltage simply stays put until the smoothing capacitor voltage approaches the regulated voltage, at which point the (always regulated) output voltage droops, following the smoothing capacitor's time constant, and then steadily resumes the ramp from that point on switch-on.
I've satisfied all of the design goals which I initially set myself (soft start, no output voltage surges, regulation to DC, always regulating at all voltages i.e. no start-up / shut-down transient preamp instability, tolerant to abuse from shorts, cheap, good longevity).
It doesn't need any further development so I'll leave it there and wish you every success with your own high energy project.
Oh, and thanks for the pointer to LT's improved LM317. I'll have to buy a bunch.
Regards,
Bob
Preamp output loads are design dependant.
Protection devices such as zeners which are permanently in place across the preamp output have significant resistances and capacitances which are a complex function of output voltage. The capacitance can be minimised by placing a diode, i.e. 1N914, in series with the zener. Even so, it's a compromised solution and there'll still be a pop through the speakers.
Opening of NC relay contacts across the channel outputs after a timed delay period is a better solution but it gets overly complex ensuring that the timing circuit is fool proof with repeated switch-on / switch-off cycles.
A delay on the B+ supply will create a smaller spike when the output valve is hot and it's rp is low but it can still be very significant and, again, making fool proof timing circuits gets overly complex. Closing of NO relay contacts across a high value resistor in the B+ supply line after a timed delay has similar problems.
Using a valve rectifier can also cause a significant pop through the speakers with repeated switch-on / switch-off cycles if this is done when the rectifier cathode is still hot.
It was the dismissal of all of these usual options which prompted me to look into the soft start regulator design. It's poor design practice to create a voltage spike and then try to eliminate it where its creation can be avoided. In this design, there can never be a voltage spike at the regulator output (except if it's shorted), even with repeated switch-on / switch-off cycles, which means that there can never be a voltage spike from the preamp output (unless output coupling capacitor values are ridiculously high offering almost DC to the power amp). It's totally fool proof in this respect. During repeated switch-off / switch-on cycles, the output voltage simply stays put until the smoothing capacitor voltage approaches the regulated voltage, at which point the (always regulated) output voltage droops, following the smoothing capacitor's time constant, and then steadily resumes the ramp from that point on switch-on.
I've satisfied all of the design goals which I initially set myself (soft start, no output voltage surges, regulation to DC, always regulating at all voltages i.e. no start-up / shut-down transient preamp instability, tolerant to abuse from shorts, cheap, good longevity).
It doesn't need any further development so I'll leave it there and wish you every success with your own high energy project.
Oh, and thanks for the pointer to LT's improved LM317. I'll have to buy a bunch.
Regards,
Bob
I agree 100 %.
Awesome. And thank you.
I grabbed the LT1086 because it was available in LT Spice. You'll probably find that National Semiconductor's version of it - LM1086 - is less expensive. It wouldn't surprise me if Fairchild has a version of it as well.
~Tom
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.