Does this look ok as an inrush current limiter?

[Greg Robinson]

Hi everyone.

I'm just wondering if anyone has any thoughts on this basic circuit topology I put together for inrush current limiting of a high voltage power supply (for a tube amplifier).
The charge time is set by the time constant of R1 and C1, and diode D6 stops the gate voltage from discharging from peak value. Of course there will be some voltage drop from drain-to-source of the mosfet, but that's ok. When power is removed from the circuit, C2 and C3 will discharge and then the gate protection zener diode D1 will become reverse biased and allow C1 to discharge into the load and bleeder resistors. I know I may need to use a gate stopper resistor for stability depending on the load, but have omitted it for now.

I'd prefer to use this type of active inrush current limiting than using an NTC thermistor to prevent nuisance blowing of fuses if the amp is turned off and on again quickly, without having to use a higher current slow-blow fuse that will provide less protection for the amp.

I cannot foresee any problems with this circuit, but wonder if anyone can spot anything I haven't thought of, or if they have any better circuit suggestions they could point me towards.

Thanks for your time!

[David Hess]

Another way to do it is to use a power resistor in parallel with the MOSFET and only turn the MOSFET on once the charge current drops below a threshold.  This way the MOSFET never operates at high power.  Check out the early Tektronix 2213/2215 power supply for an example.

[Greg Robinson]

Thanks for the pointer David, I'll look it up.

[Zero999]

The downside to using D6 is it stops C1 from discharging, when the power is turned off. It will take a long time for C1 to discharge via the parasitic resistances, so the M1 will remain on for awhile after the power is disconnected. If the circuit is switched on before C1 is discharged, then the inrush limiter won't work.

Remove D6 and make the RC time constant long enough to filter the ripple on the rectifier.

Another option is to connect the capacitor between the gate and source and use the Miller effect to make the capacitor look larger, as the MOSFET turns on.

The circuit below was designed for voltages under 15V, so you'd need to add a zener or gate-source resistor (to make a potential divider) to limit the gate voltage to a safe level.

[Greg Robinson]

I've played around with the circuit a little more, and feel I've made some minor improvements. I've added a PMOS in parallel with the main pass fet's gate capacitor, so that when power is disconnected from the circuit, that capacitor is rapidly discharged regardless of whether the full load is connected to the power supply - it is common for the majority of the load to be removed from a tube amp power supply when the power tubes are removed during service, so it would be impractical if the soft start circuit stopped operating correctly under this condition. Instead there is a second, shorter time constant formed by C4 and R5 that will discharge quickly on turn-off which then turns the PMOS on and discharges C1. The zener diode D9 keeps the gate at higher potential than the source to make sure it stays hard off during normal operation - its value would need to be selected according to the chosen FET.



David, the service manuals I was able to find for the Tektronix 2213/2215 oscilloscopes did not include the mosfet and parallel power resistor in the power supply that you were describing, do you have a link to a version that does?

Thanks again for looking everyone.

[Greg Robinson]

Thanks Hero999, you posted while I was typing. Yes, I was dealing with that problem you mention, have come up with a fairly convoluted solution.
I'll have a think about what you've said and try re-working things.

[David Hess]

David, the service manuals I was able to find for the Tektronix 2213/2215 oscilloscopes did not include the mosfet and parallel power resistor in the power supply that you were describing, do you have a link to a version that does?

Did you check the 2213A/2215A by mistake?  They are a different model which replaced the 2213/2215 and lack the current limiter.

http://w140.com/tekwiki/wiki/2213
http://w140.com/tekwiki/wiki/2215

I am including the relevant section of the schematic.  There is a thyristor preregulator off the left side of the schematic so the current limit board operates at 42.5 volts but the idea works at higher voltages.  I am not saying that this is the way to do it but just presenting it as an alternative design to consider.  Power resistors are tougher and more reliable than power transistors.

[Greg Robinson]

Ok, I've completely re-thought my approach as there were some issues with my plan of secondary-side soft-start. As this is intended for a tube amp, that wouldn't help at all with the heater/filament inrush current, and would need to be duplicated on auxillary rails too. So I'm back to primary side soft-start.
So my choices are either series resistor or NTC. I think an NTC will be an ok choice so long as I can switch it out off circuit after a few seconds.
I'm thinking that a capacitive dropper circuit would be the best option, it should work from 100-240VAC and turn on a relay after a delay. I know the principles of capacitive droppers, but I've never actually designed one before, so I'd appreciate if anyone can point out if I've made any bone-headed mistakes.
Thanks.

[T3sl4co1l]

Inrush rate limiter perhaps, but a current limiter it ain't.  Try a short circuit load...

Delay relays are simple enough, but you may want a tighter threshold than a BJT's Vbe.  Some hysteresis would be good, too.

Also, you don't need such massive components for just a half watt relay.

Tim

[Greg Robinson]

Hi Tim, thanks for your reply.

Inrush rate limiter perhaps, but a current limiter it ain't.  Try a short circuit load...

Accepted, but I felt it implied that it was only for limiting inrush current, not for an actual current limit.

Delay relays are simple enough, but you may want a tighter threshold than a BJT's Vbe.  Some hysteresis would be good, too.

I honestly don't know what purpose a tighter threshold or added hysteresis would serve. I may have been gilding the lily with my previous secondary side inrush limiter attempts, but I think this is about as good as it needs to be. Perhaps you could elaborate on what benefits you think that could reap?

Also, you don't need such massive components for just a half watt relay.

I think I've rated everything appropriately, did you have any criticism in particular?
At 240V+10%, the dissipation in R1 should be just under 3W. I always make sure to derate resistors by at least 50%, and keep in mind that the intended application is a tube amp which is basically a space heater that also happens to make noise.
R2 dissipates about 200mW, could probably get away with a 500mW resistor there - it could be omitted if the primary of the transformer could be guaranteed as a load, but if a thermal fuse blows then it's needed to discharge the X2 cap C1.
C2 I've listed as 400V in case the zener fails.
The zener dissipates about 1.4W.
All this seems quite reasonable to me but please let me know where I'm wrong.

[julian1]

I've thought about capacitor droppers as a voltage source for powering inrush circuits. I've read suggestions there may be issues with PFC (or maybe EMI) and compliance - although with the low current it's hard to imagine it's a problem.

Rather than a 5watt zener (do they make them?), maybe use a lower rated one - but use it to control a transistor in order to shunt the current to drop the voltage to 24VDC.

I experimented with a similar capacitor charge circuit and single bjt to drive the relay - but the relay switch on was unsatisfactorily slow/weak, and I ended up using a comparator to get a fast/hard coil close. It would be interesting to know how well the Darlington works.

[David Hess]

Tektronix did not bother with inrush current limiting on their big tube oscilloscopes but they did use a 6N045T delay relay to drive a multipole relay which switches the various B+ supplies on.  I assume this was to prevent cathode stripping.  Do the tube audio amplifier people worry about this?  I have not noticed.

I would probably use a power resistor or NTC on the primary side with a relay or AC semiconductor switch on a time delay.  That is essentially what the 2213/2215 schematic I posted shows but there it is on the rectified DC source before the filter capacitor.  Linear inrush current limiting is fun but demands a lot of the semiconductor.

For a stage applications, it would be really annoying if a momentary power sag caused the time delay to reset if that resulted in the amplifier failing to operate correctly so maybe tie the delay to the heater temperature or require a minimum off time to reset.

[Greg Robinson]

Rather than a 5watt zener (do they make them?), maybe use a lower rated one - but use it to control a transistor in order to shunt the current to drop the voltage to 24VDC.

Hi Julian, thanks for your reply. Zeners are commonly available up to 50W. Yes, I could use a lower power amplified zener, but 5W parts are just about as cheap in small quantities as 0.5W, having to add an extra resistor and transistor would negate any cost benefit, and it should work fine as shown anyway. But, I've taken another pass at the circuit and done some optimisation anyway, I think a 2W zener will suit me fine.

Tektronix did not bother with inrush current limiting on their big tube oscilloscopes but they did use a 6N045T delay relay to drive a multipole relay which switches the various B+ supplies on.  I assume this was to prevent cathode stripping.  Do the tube audio amplifier people worry about this?  I have not noticed.

Oh, don't get a guitarist with enough knowledge to be dangerous started on cathode stripping! Cathode stripping simply doesn't happen in receiving tubes, the voltages just aren't high enough. You need around 4x10^6V/m to begin cathode stripping. The spacings of the elements and (relative to transmitting tubes) low voltages in receiving tubes don't even approach that. However, Fender began putting standby switches in some of their amps - some of them because of the mercury rectifiers that they were using (to allow the mercury to be boiled off after moving the amp), and others because they cheaped out on the voltage rating of the reservoir capacitors, before the tubes cathodes had heated up and conduction started, the transformer wasn't loaded enough to stay within the voltage rating of the capacitors. Allowing them to heat up first before applying the high voltage allowed marginal capacitors to be used. The most notable amp that they did this on was the Fender Bassman, which is the amp that was first copied by Marshall and became the JTM45. Marshall copied it verbatim including the standby switch.
I'm not sure exactly where the idea that the reason for standby switches was cathode stripping entered the conciousness of guitar players, but it's a tough myth to dissuade people from. It's gotten to the point were other large amp brands are bowing to pressure from consumers and including standby switches on reissues of classic designs that never originally featured it. Notably, Vox implemented a terrible standby switch on their reissue AC30CC - they put the standby switch between the tube rectifier and the main reservoir, which lead to many amps failing due to arcing in the rectifier (hot switching of capacitive loads on tube rectifiers is a BIG no-no).
Standby switches also lead to other problems like cathode poisoning, if the cathode is kept hot without current flowing for an extended period, interference resistance can increase the noise in the tube and lower its gain, which is particularly important with tubes because they are almost always run open-loop, so any reduction in gain is usually immediately obvious.

I would probably use a power resistor or NTC on the primary side with a relay or AC semiconductor switch on a time delay.  That is essentially what the 2213/2215 schematic I posted shows but there it is on the rectified DC source before the filter capacitor.  Linear inrush current limiting is fun but demands a lot of the semiconductor.

I'd come to the same conclusion - primary inrush management side also deals with heater/filament inrush, and any other rails too, without having to replicate everything for each rail. Below I've got my updated thought-process.

For a stage applications, it would be really annoying if a momentary power sag caused the time delay to reset if that resulted in the amplifier failing to operate correctly so maybe tie the delay to the heater temperature or require a minimum off time to reset.

That's something I've been thinking about, I'll tackle that on my next revision.

Ok, and finally an updated and somewhat more detailed and optimised version. I selected C1 as 820n for a Xc of ~3k3 at 60Hz. With R2 470R in series that's 3705R. At 100VAC -10%, max current is 24mA, just enough headroom for the 17.5mA needed by the relay.
Rather than the brute-force smoothing I was using before, I've added a CRC pi filter. C2 is rated for 400V in case R3 goes open circuit somehow. C3 is rated for 150V in case the zener goes open circuit. They'd both be small and inexpensive enough with those rating that I think thats reasonable.
These optimisations have also allowed me to downgrade the power ratings on some of the resistors over my previous work. The ratings are still quite conservative though, that's just how I prefer to do things.
I've also detailed some of the rest of the primary side circuitry - the transformer voltage selection switching is simplified, but it should convey the intent. It also shows that the value of NTC thermistor would be doubled for 240V over 120V, and how that would be accomplished.
Any thoughts welcome.

[T3sl4co1l]

I honestly don't know what purpose a tighter threshold or added hysteresis would serve. I may have been gilding the lily with my previous secondary side inrush limiter attempts, but I think this is about as good as it needs to be. Perhaps you could elaborate on what benefits you think that could reap?

Relying on a Vbe threshold means your circuit will start up at different rates depending on temperature, and the relay will be turned on very slowly when it does.

The latter may not be a problem, because the relay itself exhibits hysteresis.  I suppose there could be a chance of chattering, though.

There's also not much to discharge the circuit when power is turned off, though I'd have to run the numbers to see if that's good enough or not.  (It should have just enough ride-through to deal with intermittent power sources, but just short enough that it gives a proper full time delay beyond that.)

So if it's inside a "space heater", and adding some heat of its own, you may find it only delays 5 seconds instead of 10 or even 15, once it's been on, and heating up, for a while.

Easily fixed with a comparator (e.g. LM311), or another transistor to provide some snap action, and the threshold can be stabilized with a zener perhaps -- you've got plenty of supply (~24V) to work with, so using a voltage divider on that, or a second zener (at, say, 6.2V), would be a fine way to go.

I think I've rated everything appropriately, did you have any criticism in particular?
At 240V+10%, the dissipation in R1 should be just under 3W. I always make sure to derate resistors by at least 50%, and keep in mind that the intended application is a tube amp which is basically a space heater that also happens to make noise.
R2 dissipates about 200mW, could probably get away with a 500mW resistor there - it could be omitted if the primary of the transformer could be guaranteed as a load, but if a thermal fuse blows then it's needed to discharge the X2 cap C1.
C2 I've listed as 400V in case the zener fails.
The zener dissipates about 1.4W.
All this seems quite reasonable to me but please let me know where I'm wrong.

I don't use triple ratings, only double.  That should be enough overrating for resistors, even in a space heater like this.

I suppose the other 1/2W resistors are just reusing what you have on hand, so that's fair enough.

The zener's not going to fail, and if it does, it'll fail shorted (FYI: semiconductors mostly fail as N-way shorts, and usually only blow open if the short draws enough fault current to physically blow it up!).  And if it opens up, you've got much bigger problems to worry about than a tiny electrolytic capacitor going poof!

Calculating backwards from the relay (24V, 0.42W, 17.5mA, 1.37kohm), and the input voltage range, I see the input current and zener power are about what they need to be, so a 3 or 5W zener is about right, after all.  (A "boosted zener", using a BJT, is also a common method, but probably won't help you much in cost, size, etc.)

You can save on other power sources by simply reducing their value: R1 (now R2 in the above schematic) is there for transient limiting, and doesn't need to be 10W sized.  It does need to handle surge energy, not just from C1 but also from mains surge.  On the upside, it can be a fusible type resistor, and then you don't need the extra fuse (F1).

For the same reason, you don't want R3, and don't need a pair of filter capacitors.

FYI, TVSs are NOT to be used across mains loads.  If you need transient suppression, use MOVs.  Mind that the worst-case peak voltage across a MOV (under surge conditions) is about triple the rating (while for TVSs it's more like +50%).

HTH,
Tim

[Greg Robinson]

Relying on a Vbe threshold means your circuit will start up at different rates depending on temperature, and the relay will be turned on very slowly when it does.

The latter may not be a problem, because the relay itself exhibits hysteresis.  I suppose there could be a chance of chattering, though.

There's also not much to discharge the circuit when power is turned off, though I'd have to run the numbers to see if that's good enough or not.  (It should have just enough ride-through to deal with intermittent power sources, but just short enough that it gives a proper full time delay beyond that.)

So if it's inside a "space heater", and adding some heat of its own, you may find it only delays 5 seconds instead of 10 or even 15, once it's been on, and heating up, for a while.

Easily fixed with a comparator (e.g. LM311), or another transistor to provide some snap action, and the threshold can be stabilized with a zener perhaps -- you've got plenty of supply (~24V) to work with, so using a voltage divider on that, or a second zener (at, say, 6.2V), would be a fine way to go.

I'd be pretty impressed if I managed to get the relay to chatter, given that turn-off voltage is typically about 10% of nominal coil voltage.
As to the timing accuracy - it really kinda doesn't matter. The cathodes of the tubes need time to heat up before the amp is useful, and the exact warm up delay varies between different tube manufacturers. Making it arbitrarily long should suffice.

The zener's not going to fail, and if it does, it'll fail shorted (FYI: semiconductors mostly fail as N-way shorts, and usually only blow open if the short draws enough fault current to physically blow it up!).  And if it opens up, you've got much bigger problems to worry about than a tiny electrolytic capacitor going poof!

You're probably right, Im just most used to seeing capacitive droppers used in things like LED lighting supplies where there's a very real chance of the load becoming disconnected (either from LED's burning out or physically disconnected if its a particularly deadly design), where best practice is to use a high voltage cap after the rectifier. But yes, I'd have much bigger problems to worry about if the load had become disconnected or the zener had gone open circuit.

Calculating backwards from the relay (24V, 0.42W, 17.5mA, 1.37kohm), and the input voltage range, I see the input current and zener power are about what they need to be, so a 3 or 5W zener is about right, after all.  (A "boosted zener", using a BJT, is also a common method, but probably won't help you much in cost, size, etc.)

Yep, could use an amplified zener, but as you'll note in my revision I've lowered the power rating of the zener to 2W - I calculate the dissipation at 240V+10% to be just below 0.9W.

You can save on other power sources by simply reducing their value: R1 (now R2 in the above schematic) is there for transient limiting, and doesn't need to be 10W sized.  It does need to handle surge energy, not just from C1 but also from mains surge.  On the upside, it can be a fusible type resistor, and then you don't need the extra fuse (F1).

Oops, sorry, should have kept the component labelling more consistent. Note I've lowered R2's power rating from 10W to 5W.
Fuse F1 is also for transformer protection. It's a nominal value for 100V/120V operation. F2 is a higher rated non user-accessable fuse because musicians are idiots who will do anything to finish a gig, including wrapping blown fuses in chewing gum wrappers or aluminium foil, or replacing them with iron nails. F3 is only switched in when the transformer coils are put in series, and is the nominal value for 220V/240V operation.

For the same reason, you don't want R3, and don't need a pair of filter capacitors.

Not sure I follow why you say that, the CRC pi filter just allows the use of smaller value caps.

FYI, TVSs are NOT to be used across mains loads.  If you need transient suppression, use MOVs.  Mind that the worst-case peak voltage across a MOV (under surge conditions) is about triple the rating (while for TVSs it's more like +50%).

I often see that thrown about, but noone was ever able to point me to any reference as to why. I realise that up until recently manufacturers weren't making any TVS diodes they recommended for AC power lines, but that hasn't been true for a while now. Bourns Littelfuse.
Given that MOV's progressively degrade, and often catch fire or explode leaving char and ash all over the inside of equipment (I'm a repair technician, I see this a LOT, I've even seen a number of times equipment that had a proper mechanical switch, and yet the MOV was connected across the line BEFORE the switch, so was subject to all transients 24 hours a day if it was left plugged in but unpowered!), whereas TVS diodes have an effectively indefinite lifetime so long as they have adequate power rating and are appropriately fused (and are not subject to acts of god), and they also have lower peak voltages and faster response times, both of which provide improved protection for the load. These all seem like advantages to me.

[David Hess]

Tektronix did not bother with inrush current limiting on their big tube oscilloscopes but they did use a 6N045T delay relay to drive a multipole relay which switches the various B+ supplies on.  I assume this was to prevent cathode stripping.  Do the tube audio amplifier people worry about this?  I have not noticed.

Oh, don't get a guitarist with enough knowledge to be dangerous started on cathode stripping! Cathode stripping simply doesn't happen in receiving tubes, the voltages just aren't high enough. You need around 4x10^6V/m to begin cathode stripping. The spacings of the elements and (relative to transmitting tubes) low voltages in receiving tubes don't even approach that. However, Fender began putting standby switches in some of their amps - some of them because of the mercury rectifiers that they were using (to allow the mercury to be boiled off after moving the amp), and others because they cheaped out on the voltage rating of the reservoir capacitors, before the tubes cathodes had heated up and conduction started, the transformer wasn't loaded enough to stay within the voltage rating of the capacitors. Allowing them to heat up first before applying the high voltage allowed marginal capacitors to be used. The most notable amp that they did this on was the Fender Bassman, which is the amp that was first copied by Marshall and became the JTM45. Marshall copied it verbatim including the standby switch.
I'm not sure exactly where the idea that the reason for standby switches was cathode stripping entered the consciousness of guitar players, but it's a tough myth to dissuade people from. It's gotten to the point were other large amp brands are bowing to pressure from consumers and including standby switches on reissues of classic designs that never originally featured it. Notably, Vox implemented a terrible standby switch on their reissue AC30CC - they put the standby switch between the tube rectifier and the main reservoir, which lead to many amps failing due to arcing in the rectifier (hot switching of capacitive loads on tube rectifiers is a BIG no-no).

I had to refresh my memory but the book "Getting the Most Out of Vacuum Tubes" by Robert B. Tomer makes a case starting on page 18 for cathode striping in all tubes based on ion contamination if high current is drawn before the cathode reaches operating temperature.  Rectifiers are just particularly susceptible to this because of how they are applied.

The ability of the cathode to supply electrons, and hence to replenish the space charge, depends on temperature.  When the cathode is warming up from a cold start, its ability to supply electrons is fairly limited.  This is what is known as its region of temperature-limited emission.  If very heavy plate current demands are made on a tube while it is temperature limited, leaving the cathode exposed to bombardment by the heavy negative gas ions that are always present.  These negative gas ions are repulsed by the negative space charge when it exists, but when it has been drawn away, there is nothing to stop the ions from plunging violently into the cathode coating where they erupt the surface like miniature volcanoes.

The erupted cathode coating becomes vaporized in the electron stream, and more gas ions are formed.  In far less time that it takes to describe, a gas arc has built up and serious pitting or stripping of the cathode has occurred.  Quite frequently, the arc is sufficient hot to burn a hole through the cathode sleeve and extend on into one of the heater folds where the arc current then finds a path to ground, opening the heater int he process.

As described, this ion bombardment could result in increasing levels of ion contamination eventually leading to arcing and cathode stripping.

Standby switches also lead to other problems like cathode poisoning, if the cathode is kept hot without current flowing for an extended period, interference resistance can increase the noise in the tube and lower its gain, which is particularly important with tubes because they are almost always run open-loop, so any reduction in gain is usually immediately obvious.

It takes a while for interface resistance to manifest.  The brief standby time during warmup is not significant.

This comes up in connection with CRTs when a user mistakenly turns the intensity off to preserve CRT life when leaving an oscilloscope or similar test instrument operating for long times which might happen for data acquisition or to use some feature which does not require the CRT.  It is better to lower the intensity and defocus the CRT.

Tube oscilloscopes have some logic circuits which should eventually suffer from interface resistance but this never seems to happen.  Maybe Tektronix was aware of the problem and used tubes rated to operate for extended time when cutoff.

I would probably use a power resistor or NTC on the primary side with a relay or AC semiconductor switch on a time delay.  That is essentially what the 2213/2215 schematic I posted shows but there it is on the rectified DC source before the filter capacitor.  Linear inrush current limiting is fun but demands a lot of the semiconductor.

I'd come to the same conclusion - primary inrush management side also deals with heater/filament inrush, and any other rails too, without having to replicate everything for each rail. Below I've got my updated thought-process.

The same book mentions inrush current limiting helping to extend heater life significantly.

Any thoughts welcome.

Add a small capacitor across the relay contacts to prevent arcing?

[T3sl4co1l]

I'd be pretty impressed if I managed to get the relay to chatter, given that turn-off voltage is typically about 10% of nominal coil voltage.
As to the timing accuracy - it really kinda doesn't matter. The cathodes of the tubes need time to heat up before the amp is useful, and the exact warm up delay varies between different tube manufacturers. Making it arbitrarily long should suffice.

Yeah, but then that digs into your usability range -- how come everyone else's rig is ready to go in ten seconds and this one takes 40 seconds after pulling it out of the trunk on a cold day?  It's so inconsistent and...

Obviously you know your customers better than me, so it's up to you (and them) if that's any bother.



My mind also wanders to ever-more-efficient methods, like... you could put a 120VAC coil in series with the FWB, solving several problems at once (less voltage to drop across R or C, transient protection thanks to substantial coil inductance), then use the stuff after the bridge as a modulated load.  When power is applied, it starts at high impedance (100kohm+?), and the relay stays off.  After the time delay, its impedance drops (0 to ~kohms), supplying current to the relay.  But what about the voltage range?  It's a constant current sink, so it doesn't overvoltage the relay.  Power dissipation can be lower, on account of the relay coil voltage being closer to mains voltage.  (Well, maybe; AC coils may need more current than their DC counterparts, which wouldn't be so friendly to the current limiting circuit.)

But whatever... just an amusement, nothing worth looking at unless you were making 100k of them.

Oops, sorry, should have kept the component labelling more consistent. Note I've lowered R2's power rating from 10W to 5W.
Fuse F1 is also for transformer protection. It's a nominal value for 100V/120V operation. F2 is a higher rated non user-accessable fuse because musicians are idiots who will do anything to finish a gig, including wrapping blown fuses in chewing gum wrappers or aluminium foil, or replacing them with iron nails.

Yeah, I get that need.   A separate (and much smaller) fuse for the delay circuit would be the trick.

Not sure I follow why you say that, the CRC pi filter just allows the use of smaller value caps.

Hmm, but you've still got 100uF in there, so you've got 110uF total..?

Remember the power isn't delivered in spikes like it is for a C-input filter; the series dropping C, with the FWB, makes the voltage nearly square-wave, and the current comes in humps.  (This is all obvious on the simulation of course, and you can see just how much filtering is actually needed, whether it's 100uF, or more or less.)

I often see that thrown about, but noone was ever able to point me to any reference as to why. I realise that up until recently manufacturers weren't making any TVS diodes they recommended for AC power lines, but that hasn't been true for a while now. Bourns Littelfuse.

Hmm, ~70V devices?  What good is that...

Lovely ratings on those, though!  It takes a metric shitload of silicon to stomach 15 kiloamperes.

Speaking of which...
http://www.digikey.com/product-detail/en/bourns-inc/PTVS10-380C-TH/PTVS10-380C-TH-ND/5053079
So if you get one, that by the appearance, is probably a literal stack of the single unit parts (which are, themselves, probably anti-series pairs of dies) -- it's expensive enough to buy a matched quad of 6L6s!

Also, on the other hand, something a little worrying... no energy rating. There's a temperature derating, but no way to guess what it can offer for waveforms other than the 8/20us surge.  What if I need to test 10/1000us surge?  What if I have a high voltage application that occasionally generates wicked ESD (like, accidental strikes from a 50kV+ source).

Given that MOV's progressively degrade, and often catch fire or explode leaving char and ash all over the inside of equipment (I'm a repair technician, I see this a LOT, I've even seen a number of times equipment that had a proper mechanical switch, and yet the MOV was connected across the line BEFORE the switch, so was subject to all transients 24 hours a day if it was left plugged in but unpowered!)

Yeah, so realize that you've seen all the shitty designs: too-small MOVs, in the wrong places, no protection, no graceful failures, just smoke and shit.

Suppose, then, that it is possible to use MOVs, responsibly, and to get long life.

Much longer life than the $100 TVS will provide, and at a tiny percentage of the cost!

An MOV with the same energy rating as that TVS is $0.60 or less.  It won't take /that/ big of a surge forever, no -- but a larger one (a buck or two) will last decades.

Tim

[julian1]

In case it's of interest here's my admittedly rather over-engineered inrush limiter,
 


I avoided the cap-dropper because of the weird voltage potentials and possible compliance issues. Power is a mini-transformer, and zener controlling a series-pass transistor. switch-on is with a comparator and the timing is adjustable with a pot.

For a dedicated design - I'd just wind another coil onto the toroid.

The resistor in fact never gets warm. The only reason it's an NTC is so that if the relay fails mechanically, it won't burn.

[David Hess]

For my own edification, I found two different designer notes from Tektronix about why they used the time delay.  On page 4-15 of the 545A service manual:

A Time-Delay relay K600 delays the application of dc voltages to the amplifier tubes int he instrument for about 25 seconds.  This delay is to allow the tube heaters time to bring the cathodes up to emission temperatures before operating potentials are applied.

This makes it seem like they were worried about cathode damage.  Maybe they were just paranoid or there is some circuit arrangement in these oscilloscopes which is particularly susceptible to damage.

And the 547 designer notes mention for practically the same circuit:

The relay is connected so the 100v and 350v filter capacitors charge when the scope turns on and the -150v and the 225v capacitors charge when the relay closes.  Delaying part of this current surge prevents a single current surge that could blow the primary fuse or damage the power switch.