0-350V 0.2A Bench Power Supply

[T3sl4co1l]

The design requirements - budget- low cost, low component count, current-limiting, noise/precision, robustness, building one, building 1,000 etc. seems to throw the design from MOSFET, tube, switching PSU, or a hybrid of these. I don't know what "homebrewed polymer films" need for a OP's PSU.

So, I'm hearing SS.

The tube solution is simple, proven, robust but needs an additional filament+grid-bias power transformer (windings), inefficient, older parts and has the stigma of being old technology.

Well, that, and inherently poorer efficiency; and inherently poorer lifetime, as the tube will do 5 or 10k hours if it's a premium quality / industrial NOS part, and used at reduced ratings.  Or 2khr if you're lucky, if it's a new Chinese production type run at "ratings".

Of course you don't expect to run a bench supply at ratings, 24/7, its entire life.  But it's a time*temp thing.  It'll fail eventually, and very likely before the electrolytics will (unless you've carelessly placed them right above the tube, that is ).

SS doesn't suffer from that limitation; with simple consideration of SOA, and the application of MOVs or TVSs, expected lifetime is wholly the capacitor lifetime.  The SS circuit, in and of itself, can last for centuries!

The MOSFET solution is fragile, complicated SOA protection and "trial by fire". Schrapnel from a TO-247 says "no, not quite right" and hopefully the load survives as well. Replace MOSFETs and try again.

I don't get what you're making this straw man from.

You look up the part.

You look for the SOA.

If there's no SOA plot, or no DC curve, or no power-limited DC SOA, discard it and move on to the next part.

You will find many families of parts this way, which are rated for this service, and reasonably priced.

You apply an accurate and fast current limit circuit -- the BJT shunt sense method is fine, while more accurate and adjustable means can be applied additionally to get traditional CC/CV supply operation -- and the SOA is never, ever violated.

You can even put in a thermistor to throttle or fold back the current when the heatsink gets hot, to deal with clogged heatsinks.  You can use enough heatsink to avoid using a fan at all (which is the most common culprit in everything that needs them!).

The SMPS solution is a lot of engineering. The custom HV magnetics are a bear to make, unless you have cores, bobbins, magnet wire, insulating tape etc. and a lot of patience.

Yes, that would be an expert level project, for many more reasons than those alone!

FWIW, that one battery discharger design I made, is still selling well, and I'm not aware of any problems they've had with it.
https://bumblebeebatteries.com/products/hybrid-battery-accessories/prolong-intelligent-discharger/
I think that used a QFET (among other things).

Tim

[Ian.M]

A dedicated hobbyist might put as many as 12 hours a week, 48 weeks a year on their HV PSU, possibly more if they are single and retired.   That's over 8 years for a good NOS tube.   Get a couple of spares for each tube while they are reasonably cheap and it will probably last you a lifetime.

[floobydust]

...
I don't get what you're making this straw man from.

You look up the part.

You look for the SOA.

If there's no SOA plot, or no DC curve, or no power-limited DC SOA, discard it and move on to the next part.

You will find many families of parts this way, which are rated for this service, and reasonably priced.

You apply an accurate and fast current limit circuit -- the BJT shunt sense method is fine, while more accurate and adjustable means can be applied additionally to get traditional CC/CV supply operation -- and the SOA is never, ever violated.

You can even put in a thermistor to throttle or fold back the current when the heatsink gets hot, to deal with clogged heatsinks.  You can use enough heatsink to avoid using a fan at all (which is the most common culprit in everything that needs them!).

I think MOSFETS have two weaknesses here. I know SS is the future, lol.

Looking up the SOA- not so simple, assuming a reactive load will never be seen, needs a derating as running parts near these limits is a bad idea and their datasheets are full of hype.
TI claims:
"As a final guarantee of the reliability of our SOA curve, we de-rate each measured thermal runaway line anywhere from 30-40%, depending on how much part to part variation we see. So when you are comparing our FETs’ datasheets to competitors’, be wary of the fact that they may not be as conservative. We have seen some vendors who are. We have seen others who publish the actual failure points and claim this as their guaranteed SOA. There is no industry standard and the truth is without the underlying data demonstrating where parts actually failed, it is impossible to know which part is more reliable from the datasheet SOA curves alone."

Mainstream MOSFETS are all about switching, and parts built for tough operation in linear-mode is a gray area. You have to wonder why IXYS actually has a lineup specifically for that.

So maybe a solution is to use several of the venerable IRFP460, for example, giving generous SOA derating.

This leads to the next problem.

The reaction time of the current-limit circuit to discharge the MOSFETs' capacitance before the pulse SOA is exceeded... I think it's difficult with >10,000pF Ciss and the HV.
I hesitate to use single-slope SOA and the sense-resistor/transistor but maybe SPICE will say otherwise. The 470uF filter capacitor at 400V will dump some current into the MOSFETs for a usec or two, during a short circuit on the power supply's output. Let's be generous and add 10R series resistance. Down to 40A peak. Still a lot.

If you you feel I'm spewing hot air, perhaps let's decide tube or SS or both and keep rolling 

Application Note AN-1155 Linear Mode Operation of Radiation Hardened MOSFETS
AN-4161 Practical Considerations of Trench MOSFET Stability when Operating  in Linear Mode

edit: forgot attachment

[oldway]

Linear power supplies are no longer a modern solution and even less so when it comes to high voltages.

Initially, there was a fundamental error of 001, that one could easily realize a linear supply adjustable from 0 to 350V with semiconductors.

For obvious reasons of poor efficiency, this solution is no longer acceptable in modern technology.
The right solution would be a SMPS. 

But it is a solution that is not simple and requires a high level of competence, which is not the case for 001.

That's why there are two solutions:
- or 001 buys another power supply
- or he opts for the simple solution which is that of the series regulator with tubes.

But there is another problem: he wants a complete diagram, with all the details of realization, which requires to develop this power supply and to make a prototype ...

Who has the skills and the time to develop this project on a volunteer basis for him?
Who is willing to sacrifice the costs that such a project will cause?

I gave some ideas but I have neither the time nor the money to develop projects for free.

[T3sl4co1l]

Looking up the SOA- not so simple, assuming a reactive load will never be seen, needs a derating as running parts near these limits is a bad idea and their datasheets are full of hype.

Reactive is no problem.  Fit SOA to worst case, i.e., short circuit condition.  Capacitor can't draw more than that for a short time (seconds, for a fucking huge capacitor), inductor can only draw that after some time.

Oh and, inductor can only draw current the same direction as was applied.  So, charging up a big fat 100H inductor then disconnecting it and watching that sexy arc draw off the terminals, makes absolutely no difference to the supply at all.  Current is dropping during that arc, and voltage remains fixed at the setpoint.  No flyback pulse, no crazy spikes, just a clean burn.

Derating: goes without saying.  My earlier hint seems to have gone unnoticed: to use TO-247s at under 100W, and TO-220s under 50W.  Typically, these will be rated upwards of 300W and 100W, respectively (depends on die size, some are less, too).  The derating serves two purposes: being conservative about the actual SOA, and approximate overhead for RthJC, insulator and heatsink.

Well, derating helps reduce the impact of thermal cycling, which is also nice.

TI claims:
"As a final guarantee of the reliability of our SOA curve, we de-rate each measured thermal runaway line anywhere from 30-40%, depending on how much part to part variation we see. So when you are comparing our FETs’ datasheets to competitors’, be wary of the fact that they may not be as conservative. We have seen some vendors who are. We have seen others who publish the actual failure points and claim this as their guaranteed SOA. There is no industry standard and the truth is without the underlying data demonstrating where parts actually failed, it is impossible to know which part is more reliable from the datasheet SOA curves alone."

I tested a smaller QFET to destruction (FQPF6N40C) and it failed very close to the RthJC limit.

I haven't tested any TI transistors to destruction, so can't speak to the validity of their claims.

Anyway, all this talk about SOA works the other way, too: you can buy a handful of different transistors, and test each one.  Once the supply is complete, and tested for operating characteristics, test it under short-circuit load (protip: don't actually short it, use a low ohm fusible resistor instead).  Test different brands and increase the current limit gradually until destruction occurs.  This plots a single point on the SOA curve.

If that point lies inside the DC SOA curve, that manufacturer is full of shit.

If outside, they're being conservative.

Like I said before, Si IRF740 is not specified for DC, but I have tested it for DC and it passed with a whopping 60% or so of headroom.

There's no secret to it.  You hook it up and either it works or it doesn't!  This is literally all the manufacturer does, and all that you need.  (Well, they also use a runaway detection circuit, so a single device can be characterized over many points, rather than destroyed for each.)

Mainstream MOSFETS are all about switching, and parts built for tough operation in linear-mode is a gray area. You have to wonder why IXYS actually has a lineup specifically for that.

Duh, marketing.

IXYS probably has a worse track record than most, on SOA capabilities.  A lot of their HiperFETs don't even have SOA curves at all.  The ones that do, drop off terribly early (1ms curves, if that; certainly not DC).  These are an exemplar of the previous generation: high power density, but high thermal sensitivity, too.

Now that they've licensed Infineon's CoolMOS, that's all in the past.  I mean, not fully, as those product lines will take a decade to close out -- but you don't have to buy them.

And, if you want the design assurance that they're actually intended for (and presumably, tested for) linear operation, you can buy those -- there's nothing wrong with that, you're just paying a premium for it.

So maybe a solution is to use several of the venerable IRFP460, for example, giving generous SOA derating.

Indeed, old MOS (like the above IR HEXFET family) tend to have crap power density, so they didn't reach into the region of thermal runaway, and are suitable for linear use.

They also have massive dies, perhaps unexpectedly for their price -- indeed, those ancient masks have fully depreciated and they cost almost nothing to make.

You can spend the same money on a new (SuperJunction type) part, and get only a third of the ultimate power dissipation (~die area).  Point being, you also get about triple the switching performance (Rds(on) * Qg), or, even better than that, actually.  Which is fantastic for switching, it just doesn't help us much for the immediate problem.

This leads to the next problem.

The reaction time of the current-limit circuit to discharge the MOSFETs' capacitance before the pulse SOA is exceeded... I think it's difficult with >10,000pF Ciss and the HV.

Have you... run the numbers on this?

Because, for a puny 2N3904 with 1k series base resistor, across a source current sense resistor, the response time is under a microsecond.  I'm not kidding!  Don't believe me?

Most of the circuits in this thread haven't even shown a base resistor (which is a rather unsafe proposition, I should add-), and so the delay will be in the hundreds of nanoseconds.  About a quarter again, if it's a damn 2N2222.

Even going from hard saturation (say, Vgs(on) ~ 10V) to soft limiting, even with very pissy gate drive, doesn't take long, and it certainly doesn't violate the SOA.

It doesn't even necessarily violate the DC SOA, let alone the 10us pulse SOA.

Example: PSMN012-80BS with 50mohm source degeneration resistor, Vgs(on) = 9V, 40V step (t_r < 1us), -20A/div.  This is another current limiting circuit I designed and built.



This is a 150W device (D2PAK, not that you'd ever be able to get 150W from it in practice), 100mJ avalanche energy so it's not a very big die (it's a newer generation -- not SuperJunction at this voltage, but nonetheless significantly improved from the IRFP days), and has a DC power limit of just a few watts in application (PCB cooling only).
https://assets.nexperia.com/documents/data-sheet/PSMN012-80BS.pdf
Fig.3 shows 10us SOA at 40V going well over 100A.

The pulse is way under the limit, and that's with inferior gate drive.  Qg(tot) is a lot lower than IRFP460, but a 2N3904 current limiter will pull down much harder than what's in this circuit, and in a linear circuit, Vgs is already near Vgs(th) -- not all the way up at Vgs(on) -- so it only needs to shove the gate a few tenths of a volt to do the job!

Don't cry Miller effect, either -- that's included in the above pulsed test, of course (at the same time current drops from its peak, to the plateau value of 60-80A, drain voltage rises to the full 40V), and anyway, don't argue from ignorance! Measure! Even SPICE will get this right!

I hesitate to use single-slope SOA and the sense-resistor/transistor but maybe SPICE will say otherwise. The 470uF filter capacitor at 400V will dump some current into the MOSFETs for a usec or two, during a short circuit on the power supply's output. Let's be generous and add 10R series resistance. Down to 40A peak. Still a lot.

Where the fuck is 40A coming from, man?  Really?

It's already in the linear range.  It's drawing as much current as it ever will.  Pentode curves at work here!

A step change in Vds will cause a step change in Vgs, yes -- this amount can be calculated or simulated, and the delta I_d found.  It won't be 40A, and it certainly won't be V(B+) / Rds(on)!

And even if it were, 1us of that is perfectly within the SOA of any device that shows a square SOA on that time scale.  For god sakes even IGBTs can handle that!

(Speaking of, and to reiterate: don't use IGBTs for linear operation.  They have even less silicon than top-performance MOSFETs.  They're made for switchin', and switchin's what they'll do.)

Application Note AN-1155 Linear Mode Operation of Radiation Hardened MOSFETS
AN-4161 Practical Considerations of Trench MOSFET Stability when Operating  in Linear Mode

edit: forgot attachment

Incidentally, note that Trench MOS is a slightly newer generation than HEXFETs -- IIRC, HEXFETs are VMOS, and Trench was roughly the next gen after(..?!).  The trench process has nicely served many designs, so you'll see the whole gamut, as far as stability and performance, in devices with that keyword.  YMMV.

Tim

[oldway]

All these endless discussions about the use of MOSFETS in high voltage linear regulation, all the doubts that there is about the adequate Mosfet, all that does only prove one thing: a mosfet used in these conditions is not reliable.

The most important point in a HV bench power supply is safety because 350V 200mA is lethal.

One thing is certain: pass MOSFETs (which usually fail in short circuit), can not be used in a linear HV bench power where the risks of electrocution of the operator are high, it is too unreliable technology and too dangerous !

[BrianHG]

Hmmm, if I were to spend around 100$ (total) making such a supply and be lazy about the design, very low noise regulated output, NO output Mosfets, I would just use PA340 power opamp at 15$, feed a few NPN transistors in emitter follower configuration, obviously series output resistors on emitters and input on base, and a voltage reference zener, however, since 350v with 60ma continuous out is the PA340's max, you would need to per-regulate it's supply to get it right up to 350v without going over using a single additional NPN to regulate the DC power input (the low voltage drop here and only around 50ma needed for the op-amp means less heat dissipation).  Your max output will be only around 345v due to output v+ limit of opamp and voltage drop across the transistors.

Second smaller regular op-amp for current limiter sensor, by muting the + input to the PA340.

The rest, a cheap surplus old heatsink to hold 8 NPN transistors + 1 smaller one for the op-amp if needed + 1 or 2 transformers to make the un-regulated 375vdc source...

https://www.digikey.com/product-detail/en/apex-microtechnology/PA340CC/598-1917-ND/2700701
https://www.apexanalog.com/resources/products/pa340u.pdf

Transistor choice FJA13009TU: https://www.digikey.com/product-detail/en/on-semiconductor/FJA13009TU/FJA13009TU-ND/1056011
You will just squeeze into the safe operating area with 8 in parallel, at full continuous 200ma load, all the way down at 1v output, I figure a 375v drop from the main un-regulated power source.  In linear operation, you cannot do this effectively with mosfets, unless you precisely match all their Vgs down to the same degree of error as in the Vbe drop of multiple BJTs over a wide temperature range.

The DC safe operating area of STMicroelectronics BUF420AW doesn't look trustworthy or complete in the data sheet, but, you would only need this 1 transistor to drive 200ma at over 375v drop and it's a high price at 11$.  Though 10x of the other transistor is also getting a little pricey at 15$.  I would just use at least 10x the FJA13009TU part since the DC soa is absolutely clearly defined in the data sheet.

I would still spend a little more for better case and protected connections.
I don't believe you can build a 350v 200ma adjustable supply that you could trust for safety and longevity for less than 100$.  And when I say longevity & safety, I don't mean it might stop working after a few weeks or months, I mean with 375v inside, it wont unexpectedly blow up or start a fire after a few weeks or months.

Also, thanks to using a standard opamp, I would also skip the analog volume adjustments and use a PIC or equiv MCU with internal dac/adc to drive the power op-amp with a nice readout adding another 25$ in parts, but with some safety features like over-temp shutdown and true voltage and current digital readings of your output + RS232 com control and push button memory controls.

[Zero999]

I've come up with a solution to the LM317+valve current limiter problem: use an NPN+valve cascode. The voltage across Q3 is limited to below 80V, even when shorted at nearly 300mA; finding a BJT which is capable of this should be possible. A Darlington device still may be a good idea, as the base current will not be high enough, at high voltage settings and paralleling two transistors is still a good idea.


I have never attempted to design a linear power supply with such a high voltage output, so rather than just relying on simulators and a bit of guesswork, I decided to do some research on the matter. The biggest challenge is the output pass device, which dissipates most of the power. Let's look at the three different devices being suggested:

1) BJTs. In short forget about it. The safe operating area of all BJTs I've seen, allows only a 20mA or so, at 400V. Try to get, even a high voltage, high power BJT to dissipate a significant amount of power at a few hundred volts and the result will be secondary breakdown causing the emission of magic smoke. The only way would be to cascode several in series, which is a pain.

2) Valves/tubes. Electrically robust and very resistant to overload. The downside is they're, inefficient, relatively short lived, mechanically fragile and require a negative voltage to turn off, like a depletion mode FET.

3) MOSFETs. These appear to be the best option. They're resistant to secondary breakdown. The only limit seems to be the thermal. 80W for a TO-220 case may be pushing it, so two could be connected in parallel and they should current share fairly well.

Tap switching should be implemented, so the output device never sees the full voltage. This can be implemented using a relay and comparator, monitoring the output voltage.

I don't think I would seriously consider building this. I'd probably opt for a switched mode design.

[David Hess]

All these endless discussions about the use of MOSFETS in high voltage linear regulation, all the doubts that there is about the adequate Mosfet, all that does only prove one thing: a mosfet used in these conditions is not reliable.

I would not consider any pass element or control circuit reliable given pot odds which include easy electrocution.  Once was enough to teach me that excessive bulk filtering capacitance is not a good idea.  See below about including an SCR crowbar.

The most important point in a HV bench power supply is safety because 350V 200mA is lethal.

One thing is certain: pass MOSFETs (which usually fail in short circuit), can not be used in a linear HV bench power where the risks of electrocution of the operator are high, it is too unreliable technology and too dangerous !

I would include an SCR crowbar circuit on the output to blow the fuse in the event of a pass element short, control circuit failure, or high output voltage.

You apply an accurate and fast current limit circuit -- the BJT shunt sense method is fine, while more accurate and adjustable means can be applied additionally to get traditional CC/CV supply operation -- and the SOA is never, ever violated.

This is why I like using integrated regulators as pass elements; they include their own built in fast current limit and safe operating area protection.

I have seen power MOSFET designs where a thyristor was used between the gate and source to implement fast protection.  While not suitable for this application, Motorola produced the amusing MDC1000A (datasheet included below) for power MOSFET switching applications; internally it uses an SCR for fast turnoff.  These days there is no shortage of small, fast, high current, and high gain bipolar transistors suitable for driving charge through high capacitance MOSFETs.

[David Hess]

3) MOSFETs. These appear to be the best option. They're resistant to secondary breakdown. The only limit seems to be the thermal. 80W for a TO-220 case may be pushing it, so two could be connected in parallel and they should current share fairly well.

MOSFETs are *not* resistant to secondary breakdown (Do they call it that?  What do they call it?) at high voltages where the temperature coefficient of the threshold voltage reverses.  This is bad enough on a single die but multiple devices in parallel will result in current hogging.  It is not fatal but it needs to be watched.  That graph floobydust posted illustrates the problem.

Update: If I ever find one of my copies, the Siliconix MOSPOWER Applications Handbook has an excellent discussion of this issue.  Back then with low cell density and horizontal power MOSFETs, it was not nearly as much of a problem but modern high density devices are often worse than bipolar transistors except perhaps in magnitude.  3sl4co1l said that this situation has improved and some datasheets seem to indicate this which is a welcome relief.

[BrianHG]

Get a good heatsink, and parallel 4x PA340, the data sheet claims you will get 50ma continuous each device, no secondary breakdown.  This solution would be 4x15$, plus some series resistors.  However, you would want one transistor regulating the +350v supply to maximize the supply for the op-amp.  And, the output will need series resistors to accommodate minute voltage offset between the 4 devices.  Your output voltage will range between 2v and 342v, but, with 2 series diodes on the output, you will go from 0v to 340v.

I would still use another op-amp in front to amplify a MCU's dac output to 350v, then drive the 4 parallel output driving op-amps in voltage-follower mode.  The adc's on the MCU will measure output voltage and current, and, perform the constant current mode voltage limiting.  Your voltage and current limiting output will now be digital accurate and you will have digital readout.

Your output precision and noise will be incredible, but if your MCU firmware crashes and doesn't perform proper current limiting, you'll blow them.  I would add an additional linear current-limiting circuit muting the MCU's dac output to protect the op-amps.

[001]

Thanx a lot to all!

I see many experienced peoples here 

Can You say is long-tail dual triode schematic have any benefits vs classic small pentode driver?

[David Hess]

I would still use another op-amp in front to amplify a MCU's dac output to 350v, then drive the 4 parallel output driving op-amps in voltage-follower mode.

Be careful about exceeding the inverting and non-inverting input voltages of the operational amplifiers under any circumstances; the necessary input protection can get out of hand pretty quickly.  Of course maximum MOSFET gate to source voltage and bipolar base-emitter voltage require care as well.

This issue is why I did not recommend a bootstrapped operational amplifier implementation although it is certainly feasible to do.

The adc's on the MCU will measure output voltage and current, and, perform the constant current mode voltage limiting.  Your voltage and current limiting output will now be digital accurate and you will have digital readout.

The MCU does not need to be part of a critical feedback loop for digital accuracy and it may be undesirable for reasons of reliability.

[BrianHG]

The MCU does not need to be part of a critical feedback loop for digital accuracy and it may be undesirable for reasons of reliability.

No, I would just use the DAC as my voltage ref set to 1 value unless maximum current is exceeded.  I don't want dac transitional noise on my outputs.  Have a linear circuit transistor mute/clamp the DAC's output if the current exceeds 200ma, a reference set by a second dac from the MCU to a small comparator/opamp on a shunt resistor at GND, with a maximum 200ma range making linear current limiter instead of a crucial software protection loop.  (I guess, making a linear supply, try to make everything crucial stay linear...) The inputs of the opamp do have that 16v maximum offset protection, I would use series resistors when feeding the op-amp inputs.

But yes, drive the inputs hard off then end, you will create a nasty problem.

[BrianHG]

Wow, all you need is this 1 PA92 IC for a 375v, full 200ma continuous output current safe to short to GND as it has a built in programmable current limiter.  Also, you will need a heat-sink.  Though it costs.............. 163$

https://www.apexanalog.com/resources/products/pa92u.pdf

At just under 200$, this one will continuously drive 200ma, at 350v shorted to GND, with a case temperature above 70c:
PA93 https://www.apexanalog.com/resources/products/pa93u.pdf

This would be the simplest, safest assured power supply, which wont explode, as long as you follow the usage in the data sheet.
Easy to make an adjustable voltage, 200ma supply with this IC and a volume control knob and zener diode with a fraction the worry.  Nothing other than a HV transformer, diodes and cap rectification with fuses for DC power.

[001]

I see wery intelligent people here 

And I no more understand anything. I`m only homebrewer 

If You can. help me please with simple tube pass project

[Zero999]

All these endless discussions about the use of MOSFETS in high voltage linear regulation, all the doubts that there is about the adequate Mosfet, all that does only prove one thing: a mosfet used in these conditions is not reliable.

The most important point in a HV bench power supply is safety because 350V 200mA is lethal.

One thing is certain: pass MOSFETs (which usually fail in short circuit), can not be used in a linear HV bench power where the risks of electrocution of the operator are high, it is too unreliable technology and too dangerous !

Just about any semiconductor will fail short circuit. I don't know about valves.

Why is that a safety hazard? If the bench power supply is capable of giving a lethal voltage, then the best safety practice is to assume the output voltage has the potential to be lethal, even if it's set to a non-hazardous voltage. Never touch the output of a bench linear power supply which is powered by a lethal voltage on the other side of the pass element!

The same principle applies to a mains circuit switched of with a TRIAC or thyristor.

[oldway]

...Just about any semiconductor will fail short circuit. I don't know about valves.

Why is that a safety hazard? If the bench power supply is capable of giving a lethal voltage, then the best safety practice is to assume the output voltage has the potential to be lethal, even if it's set to a non-hazardous voltage. Never touch the output of a bench linear power supply which is powered by a lethal voltage on the other side of the pass element!

The same principle applies to a mains circuit switched of with a TRIAC or thyristor.

It's exactly the opposite ... The safety rules require that if a power supply is adjustable from 0 to 350V, it can be used with 12V output without any risk for the operator.
A well-designed 0 to 350V power supply meeting basic safety principles can not be more dangerous when set to 12V than a simple 12V power supply with 7812.

[T3sl4co1l]

All these endless discussions about the use of MOSFETS in high voltage linear regulation, all the doubts that there is about the adequate Mosfet, all that does only prove one thing: a mosfet used in these conditions is not reliable.

You're the only one with these imaginary doubts...

The most important point in a HV bench power supply is safety because 350V 200mA is lethal.

One thing is certain: pass MOSFETs (which usually fail in short circuit), can not be used in a linear HV bench power where the risks of electrocution of the operator are high, it is too unreliable technology and too dangerous !

Tubes fail shorted too.  Try melting one some day.

I guess we should just throw up our hands!  It's an impossible assignment!  We cannot make it shock-proof, touch-proof or lick-proof!  They'll always make a smarter fool no matter what protection we add!

Seriously.  Solid state, high voltage power supplies fly on interplanetary probes.  Indeed, this is because tubes still fly on such platforms -- it's hard to beat a TWT for flight reliability record, gain, power and so on.

Tim

[oldway]

@David Hess:

I would include an SCR crowbar circuit on the output to blow the fuse in the event of a pass element short, control circuit failure, or high output voltage.

Crowbar circuit is a good solution for fixed output voltage power supply but pain in the ass for variable output power supply.
Crowbar threshold must be ajusted every time you change the output voltage of your power supply.....not practical at all....

[T3sl4co1l]

@David Hess:

I would include an SCR crowbar circuit on the output to blow the fuse in the event of a pass element short, control circuit failure, or high output voltage.

Crowbar circuit is a good solution for fixed output voltage power supply but pain in the ass for variable output power supply.
Crowbar threshold must be ajusted every time you change the output voltage of your power supply.....not practical at all....

Easy to detect a pass device fault and activate the crowbar.

Tim

[David Hess]

I would include an SCR crowbar circuit on the output to blow the fuse in the event of a pass element short, control circuit failure, or high output voltage.

Crowbar circuit is a good solution for fixed output voltage power supply but pain in the ass for variable output power supply.
Crowbar threshold must be adjusted every time you change the output voltage of your power supply.....not practical at all....

The crowbar circuit protects against a pass element short or control circuit failure where the output voltage rises all the way to the input voltage.

[oldway]

All these endless discussions about the use of MOSFETS in high voltage linear regulation, all the doubts that there is about the adequate Mosfet, all that does only prove one thing: a mosfet used in these conditions is not reliable.

You're the only one with these imaginary doubts...

The most important point in a HV bench power supply is safety because 350V 200mA is lethal.

One thing is certain: pass MOSFETs (which usually fail in short circuit), can not be used in a linear HV bench power where the risks of electrocution of the operator are high, it is too unreliable technology and too dangerous !

Tubes fail shorted too.  Try melting one some day.

I guess we should just throw up our hands!  It's an impossible assignment!  We cannot make it shock-proof, touch-proof or lick-proof!  They'll always make a smarter fool no matter what protection we add!

Seriously.  Solid state, high voltage power supplies fly on interplanetary probes.  Indeed, this is because tubes still fly on such platforms -- it's hard to beat a TWT for flight reliability record, gain, power and so on.

Tim

The tubes are not dead, there are many applications where the tubes could never be replaced .... You use every day a microwave oven with a magnetron tube and you do not complain .

Just because you do not know anything about this technology does not mean you have to neglect it.

A tube failing shorted (NB: between anode and cathode) ? .... it is extremely rare, I have never seen one in my life .... And on the tubes with connection of the anode on the top of the tube it's just impossible.

[oldway]

I would include an SCR crowbar circuit on the output to blow the fuse in the event of a pass element short, control circuit failure, or high output voltage.

Crowbar circuit is a good solution for fixed output voltage power supply but pain in the ass for variable output power supply.
Crowbar threshold must be adjusted every time you change the output voltage of your power supply.....not practical at all....

The crowbar circuit protects against a pass element short or control circuit failure where the output voltage rises all the way to the input voltage.

How will you use you power supply with 350V output if voltage drop on Mosfet pass transistor is only a few volts ? How will you detect a shorted pass transistor ?

[T3sl4co1l]

The tubes are not dead, there are many applications where the tubes could never be replaced .... You use every day a microwave oven with a magnetron tube and you do not complain .

Just because you do not know anything about this technology does not mean you have to neglect it.

I know everything about this technology, from the underlying physics to the electrical characteristics of many standard parts.

I know better than anyone why tubes are nearly forgotten today, and why solid state has surpassed it in all but a few (electron-wave physics) applications.

Magnetrons are one, I gave another example (TWT) which I guess you didn't pick up on.  Particle accelerators use klystrons, and more broadly speaking, any particle accelerator is itself a large, very complicated tube filled with vacuum, a vacuum tube if you will.

They only remain useful in niche applications.  Everything from DC to UHF has been superseded.  I'm not sure if GaN FETs will ever be cheap enough to replace the microwave oven magnetron, but it'll be an interesting day if it comes.

A tube failing shorted (NB: between anode and cathode) ? .... it is extremely rare, I have never seen one in my life .... And on the tubes with connection of the anode on the top of the tube it's just impossible.

Er, well... screen and grid, let's say. Not a healthy situation for any of the circuits in this thread.

Cracked or melted glass also lets in gas, which acts to increase leakage or short out (sparking) electrodes.

I don't get you, man.  Just because you can't imagine it, it doesn't exist?  What kind of thought process is that?  That's an extremely narrow, unimaginative way to go through life.

Tim