If by "both ways," you mean with small and large output caps, yes. The small cap gave higher noise and occasional stability issues. So I went to a larger C and have never had an issue since.
Maida's original circuit had some flaws (notably regulator dropout before the rated current is achieved), but that's to be expected- an apps engineer gets the circuit working, writes it up, then moves on. It's up to others to adapt it and improve it. I had the benefit of Errol Dietz's research showing why large caps were desirable there- Maida didn't.
Maida's original circuit had some flaws (notably regulator dropout before the rated current is achieved), but that's to be expected- an apps engineer gets the circuit working, writes it up, then moves on. It's up to others to adapt it and improve it. I had the benefit of Errol Dietz's research showing why large caps were desirable there- Maida didn't.
It's all a matter of SOA of the cascode device. If you are regulating to safe and sane voltages (say, 3~400 V) you'll have a much lesser chance of blowing the cascode device. But I'm starting with 600 V (worst case) and regulating to 470 V. With a 120 mA current limit, the 47 uF cap on the regulator output will need some 250~300 ms to fully charge. Granted, the cascode doesn't see the full voltage the whole time, but it still needs to handle a rather large amount of power without blowing up. I have yet to find a MOS or BJT in a package size up to TO-247 capable of handling the power. Especially, if you should commit the deadly sin of power cycling the amp while the regulator is hot.
I would also prefer a better current limiter circuit than the one in the Maida regulator. Basically, it relies on the voltage drop across the emitter/source resistor of the cascode to rise to the point where the LM317 looses regulation. That's fine for low voltages, low currents, but for 100's of mA and higher voltages (>500 V) this is not good enough to protect the cascode during a short circuit.
~Tom
Yes, current limit is not great. I have had some survive arcs on the output and some die and never found the fault.
I have been using FQAF11N90C as upper pass device, making 500V out from 600V in. TOP-3 isolated package. I don't know how it fares compared to those you have looked at. My circuit is not as demanding as yours, no downstream capacitance, but they have worked great for me so far.
The Builder we are are trying to help specified 565V supply and 400mA current limit.
If the limit circuit worked OK (which it probably doesn't, BTW), at turn-ON the circuit reduces to a capacitor and a 400mA current source. we don't need to warm up SPICE to see that the output voltage will be a linear ramp at V = 0.4/C V/s, or 8V/ms. This means the series transistor will see 300V or more and 400mA for the first 32ms, which is nearly dc-level of stress in SOA terms. Tom has already shown that even TO247 parts (which are expensive) can't handle this sort of startup! With bipolar transistors you are so deep into second breakdown that even 200W huskys like the BUH1215 will fail without even power cycling a warm device.
Naturally, the performance of the circuit should be honed to the requirements, but if you use the basic Maida circuit, the 1u/2.7ohm is the place to start. Larger caps risk degrading the dynamics of the regulator, which are inferior to a prpoperly designed dscrete circuit already.
With respect to the SOA issue, and short duration withstand of high V and I, it may be worth bypassing the FET/darlington with a simple MOV (or MOV-R combinations) designed to repetitively handle the energy and significantly reduce the peak current through the semi device. MOV tolerance and capacitance don't appear to be a detraction in this application.
The transient loadline of the semi would probably depend very much on the parasitic resistances of the source, MOV and load capacitance
Ciao, Tim
The transient loadline of the semi would probably depend very much on the parasitic resistances of the source, MOV and load capacitance
Ciao, Tim
Bingo! I have had no issues whatsoever starting up into an open circuit or resistive load. Large cap load ==> silicon slag.
Yeah... I used STW12NK95Z (950 V, 10 A, 230 W @ Tc = 25 C) for the cascode device. Assuming Tc stays at 25 C during turn-on (obviously not the case), the device should survive in excess of 380 mA with 600 V (my worst case input voltage) across it. So I set the source resistor to start the current limiting at 166 mA -- Vzener = 12 V, Vgs = 5 V, LM317 = 2.5 V ==> 27 ohm source resistor. The silicon still went poof on start-up. I could ramp it up slowly on the variac just fine, but an abrupt turn-on turned the silicon into slag.
~Tom
Adding protection parts to a power circuit is a reasonable design practise, even if they address low risk issues such as hot Tj (lower SOA), low levels of added/parasitic output capacitance, varying source voltage characteristics and impedance, start-up sequencing of various circuit sections, and whatever else may move devices in to their stress area. Choosing a minimal and elegant amount of protective components is an appropriate aim. The use of FET Vgd clamping fits in to that category, due to parasitic Cgd.
The use of a MOV across the FET, and say a zener across the 317, may sim up to show a reasonable reduction of FET stress in unfortunate circumstances, and a suitable choice of simple protective parts. A given design could then place an upper limit to, say, output capacitance that is well below any stress region and any operation of protecive devices.
Ciao, Tim
The use of a MOV across the FET, and say a zener across the 317, may sim up to show a reasonable reduction of FET stress in unfortunate circumstances, and a suitable choice of simple protective parts. A given design could then place an upper limit to, say, output capacitance that is well below any stress region and any operation of protecive devices.
Ciao, Tim
You can use protection devices across the cascode and the regulator as long as you can guarantee that the output voltage won't go above its design target. Now, of course, your protection devices will need to be able to handle the over-stress associated with the short circuit or start-up into cap load condition.
~Tom
~Tom
Some simulation may be the simpler way to gauge the benefit from added protective parts. Devices such as MOV and zener will have fairly low resistance when conducting, so actual circuit voltages and currents will depend significantly on the supply equivalent circuit values chosen, and the voltage setpoints and on resistances of the protective devices - and then that all has to correlate back to dynamic loadline of FET to gauge improvement in SOA, and to gauge energy dumped in to protective devices.
I've come across this scenario too..... repetive failures in PSU passes suddenly shut up with a large chip die.
Somewhat of a thread hijack,
BTW. I find it a somewhat disturbing trend that electronics inexperience gives way having to resort to spice and other simulation programs in order to get an idea how the circuits are behaving...fine if ones brain suddenly clicks and solution solved....instead of the previous generations of maths,physics & experience of being able to sort serious problems out. My experience with simulation is that it isn't Gods tool, not every scenario can be simulated and I've seen some pretty bad/dangerous mistakes where people have made in HV circuits where cognitive experience could have quickly dealt with problems.
I'm not all against simulation programs but so much these days is worked adhock without a complete understanding of what's going on and this is bad.
So unfortunately we now have so-called engineers who can use simulation programs but if the computer power goes down, cannot be asked to work out the data sheets, interpret graphs or be asked to solve even the basic problems. This now is inevitable.
Once a physics teacher, I regulary come across true engineers on the Franco/Germanic continent who have a good understanding of maths as a tool, whereas it seems other countries education has gotton for the worst for uncognitive thinking and that's bad. Yes, UK singled out.
Slam and criticize me if you want to, but it's shows up...becareful using those simulation tools.
richy
I totally agree richy - sh*t in, sh*t out. But stereotyping the tool is an issue to.
I guess many of us have our radar up when first assessing any measurement/result/comment/assertion. I started simulating on a card reader dec computor and trying to interpret line printer plots. I enjoy seeing keen interested people simulating circuits and running in to all the pitfalls of not understanding real life. But same thing happened with oscilloscopes - what ever was displayed was believed to be actually there in the circuit. Measurement techniques - simulation techniques - analysis techniques - they all have to be learnt.
With SOA stuff and transients for this application (given that the basic circuit has been simmed already), I think doing a sim followed by a controlled cro measurement is a very fast way to hone in a better appreciation of protection/stress performance - rather than just conjecturing (unless the conjecture is based on a closely comparable situation).
Ciao, Tim
I guess many of us have our radar up when first assessing any measurement/result/comment/assertion. I started simulating on a card reader dec computor and trying to interpret line printer plots. I enjoy seeing keen interested people simulating circuits and running in to all the pitfalls of not understanding real life. But same thing happened with oscilloscopes - what ever was displayed was believed to be actually there in the circuit. Measurement techniques - simulation techniques - analysis techniques - they all have to be learnt.
With SOA stuff and transients for this application (given that the basic circuit has been simmed already), I think doing a sim followed by a controlled cro measurement is a very fast way to hone in a better appreciation of protection/stress performance - rather than just conjecturing (unless the conjecture is based on a closely comparable situation).
Ciao, Tim
A simulation is only as good as the device models and test benches used. No argument there.
However, simulation, when done properly, is an invaluable tool for circuit design and debugging. It is far easier to experiment with different circuit topologies and component values in the simulator than it would be to build the circuits in the lab. Not to mention, much, much less expensive. In addition, simulation allows for prediction of circuit behavior versus temperature, component variation, etc. in ways that would be insanely cumbersome to implement in the lab. There's a good reason IC design is done by circuit simulation followed by extensive lab evaluation (and correlation with simulation) these days. Design-by-silicon-iteration is just not an efficient - or inexpensive process.
That said, simulations should not be used as a substitute for good engineering. Practical experience still comes from "smoke time" in the lab. I generally find that the best circuit design engineers are the ones that master the practical aspects, simulation, as well as the theories behind it all.
Many "math people" tend to forget that math has its limitations too. 99.999% of the math used in circuit design is based on the assumption that the circuit is operating in its linear region. You try to calculate how an op-amp behaves as it enters and recovers from saturation. I wish you the best of luck.
~Tom
However, simulation, when done properly, is an invaluable tool for circuit design and debugging. It is far easier to experiment with different circuit topologies and component values in the simulator than it would be to build the circuits in the lab. Not to mention, much, much less expensive. In addition, simulation allows for prediction of circuit behavior versus temperature, component variation, etc. in ways that would be insanely cumbersome to implement in the lab. There's a good reason IC design is done by circuit simulation followed by extensive lab evaluation (and correlation with simulation) these days. Design-by-silicon-iteration is just not an efficient - or inexpensive process.
That said, simulations should not be used as a substitute for good engineering. Practical experience still comes from "smoke time" in the lab. I generally find that the best circuit design engineers are the ones that master the practical aspects, simulation, as well as the theories behind it all.
Many "math people" tend to forget that math has its limitations too. 99.999% of the math used in circuit design is based on the assumption that the circuit is operating in its linear region. You try to calculate how an op-amp behaves as it enters and recovers from saturation. I wish you the best of luck.
~Tom
Hey Guys,
I redesigned my 500V regulator using MOSFETS instead of BJT's as suggested by SY.
I moved the current sensor circuit before the pass MOSFET transistor. When the current gets to 0.4A (0.7V across R5), R5 turns on transistor Q1 and pulls the voltage at the gate of Q3 to about 15V. Q3 then shorts out R10 which drops the voltage of the regulator. I put a 3.3 Ohm resistor (R16) on the output of the regulator so that when the regulator is at its minimum voltage of 1.25V there will be 0.4A of current supplied.
Since there will be a huge voltage drop from input to output during a short, there is going to be a lot of heat dissipated. This is why I put 4 MOSFETS in parallel that are heavily heat sinked. Furthermore, I put three 10W resistors in series which feed the 13V zener. During a short there will be approximately 30W dissipated by R1, R2 and R3 combined and just under a watt for D1.
I added a soft start using an RC circuit and Q4. When the amp is turned on, 12V is seen on the gate of Q4 thus Q4 is on. I put a 10k resistor in series with the Q4 so that the regulator starts at 100V. As the capacitor charges, the voltage at the gate drops and then the slowly rises to 500V. I set the RC time constant to 10s.
I built this regulator and it works really well. However, I have not tested the current limiting circuit since I have Q1 on order. The soft start works great too. At 0.333A load I was getting 60dB of ripple rejection so I think R15 and C3 will suffice.
Is there any comments or concerns? I came up with the current limiting circuit myself so I am curious if anyone has seen this method before. I hope the current limiting works. I should be able to test it tomorrow. Thanks everyone for your input. This is my first tube amp build and Maida regulator; so far everything seems to be going smooth.
I redesigned my 500V regulator using MOSFETS instead of BJT's as suggested by SY.
I moved the current sensor circuit before the pass MOSFET transistor. When the current gets to 0.4A (0.7V across R5), R5 turns on transistor Q1 and pulls the voltage at the gate of Q3 to about 15V. Q3 then shorts out R10 which drops the voltage of the regulator. I put a 3.3 Ohm resistor (R16) on the output of the regulator so that when the regulator is at its minimum voltage of 1.25V there will be 0.4A of current supplied.
Since there will be a huge voltage drop from input to output during a short, there is going to be a lot of heat dissipated. This is why I put 4 MOSFETS in parallel that are heavily heat sinked. Furthermore, I put three 10W resistors in series which feed the 13V zener. During a short there will be approximately 30W dissipated by R1, R2 and R3 combined and just under a watt for D1.
I added a soft start using an RC circuit and Q4. When the amp is turned on, 12V is seen on the gate of Q4 thus Q4 is on. I put a 10k resistor in series with the Q4 so that the regulator starts at 100V. As the capacitor charges, the voltage at the gate drops and then the slowly rises to 500V. I set the RC time constant to 10s.
I built this regulator and it works really well. However, I have not tested the current limiting circuit since I have Q1 on order. The soft start works great too. At 0.333A load I was getting 60dB of ripple rejection so I think R15 and C3 will suffice.
Is there any comments or concerns? I came up with the current limiting circuit myself so I am curious if anyone has seen this method before. I hope the current limiting works. I should be able to test it tomorrow. Thanks everyone for your input. This is my first tube amp build and Maida regulator; so far everything seems to be going smooth.
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.