Spike generator

[Simon]

I've been tasked with coming up with a spike generator to test our stuff against. We are looking for around 300V. Attached is the specification sheet for the stuff we are trying to produce. Page 14 figure 10 seems to supply a good starting point for a circuit to carry out the test. I'd used rectified mains voltage to get the test voltage.

The only bit I'm a bit unclear on is the recommendation for a reed switch, surely a relay can do this ? if I do need a reed switch how do I operate it ? can you get standard coils to operate reeds with ?

As it is a one off I'll probably look to make it on veroboard, I take it this is ok ? what are the broad rules in terms of track spacing for dealing with 350V ?

[ee.jmlp]

What do you need exactly?

A spark generator? A ESD generator with body model? Other with models?

The easiest choice is to use an step up autotransformer with limited output power. A ferrite, enameled copper wire, a push-button, 2 resistors and a capacitor. If you need an automated one you could use a 555 with a mosfet.

[Simon]

as per the diagram on page 14 we'd be injecting a high voltage of limited power (as it's stored in a 0.1uF capacitor) to see how the circuit can cope with spikes in the power line.

I'm also wondering how much power the human body can cope with to see how deadly the output would be, the spec sheet keeps going on about 15 millijoules

[ee.jmlp]

but a single one? or a pulse train?

[Simon]

The spec sheet is talking about a pulse every 1 second, given that the pulse source capacitor is 0.1uF and the charging resistor is 100K then it will take 50mS for the capacitor to charge back up.

[Andy Watson]

The only bit I'm a bit unclear on is the recommendation for a reed switch, surely a relay can do this ? if I do need a reed switch how do I operate it ? can you get standard coils to operate reeds with ?

Reed switches operate relatively quickly compared to bog standard relays, however, I think the key point that you've skipped over is that they specify a mercury wetted switch. The mercury wetting helps to reduce/eliminate contact bounce. You operate them by putting one (or both) end in a magnetic field, ... or you could just buy a reed-relay.

[Simon]

oh I see, presumably a solid state device will have too high an inductance and series resistance.

[madshaman]

I think some of the old HP pulse generators will go up to 300V, you might be able to buy what you need.

[Simon]

is this like an ebay job ? how much do they go for ?

[minime72706]

Maybe a high-voltage, low rsd-on mosfet would make a good switch?

[Simon]

that would be my prefered option but I'll have to look at what is available and prices as high voltage starts to become a problem quickly with availability and it needs driving too.

[Rufus]

The only bit I'm a bit unclear on is the recommendation for a reed switch, surely a relay can do this ? if I do need a reed switch how do I operate it ? can you get standard coils to operate reeds with ?

Reed relays are fast and mercury wetting means very low contact resistance no real bounce.

Paranoia about mercury has made it more difficult but you can still get mercury wetted reed relays. Try http://www.pickeringrelays.com

I made a similar tester except it was a couple of hundred pF and up to 4kV. The relay only needs to pulse. I just used a SPDT push button charging a 100uF capacitor from 2 PP3 batteries and switching the capacitor across the coil of a 12v reed relay when you pushed.

The relay contact life will likely be limited, maybe down to a few thousand tests.

Spikes in that specification are pretty trivial to deal with, if you stick about 3uF across the supply input of your equipment the spike ends up being a few volts. There is sod all energy in the spike if you choose to use a transient suppressor diode instead. The killer in that spec is the surge voltages unless your equipment can stand 100v for 10s of ms anyway.

[Neilm]

Do you have to make one? I would have thought you could buy / hire something that would do the job and that you know will meet the test standard. If you are trying to qualify it a "home made" setup will not be accepted.

[fcb]

use an IGBT, I love them and there is nothing to be afraid of.

[minime72706]

I recently grabbed a couple of 800V FETs for a high-voltage DC power supply I'm throwing together. They weren't expensive.

[Rufus]

use an IGBT, I love them and there is nothing to be afraid of.

Can they switch a couple of kV in 200pS like a mercury wetted contact?

[Tinkerer]

as per the diagram on page 14 we'd be injecting a high voltage of limited power (as it's stored in a 0.1uF capacitor) to see how the circuit can cope with spikes in the power line.

I'm also wondering how much power the human body can cope with to see how deadly the output would be, the spec sheet keeps going on about 15 millijoules

Off the top of my head, I think around even only 35mA is enough to cause problems and even be deadly, as long as there is enough voltage. So if you are dealing with high voltage of even only mA, you need to be careful. I think the lower threshold amount for causing pain/discomfort starts at about 5mA. Just one thing of note, body size does matter here too. A large person can withstand more power than a smaller person.

[Simon]

We are just trying to verify the product before shipping, we know that our customers do not meet the spec so want to try and get ahead of the game. It's not meant for any certification just to prove the suppliers design.

I think i'll try and find the reed, not sure why it will have a short life ? we are talking under 1uF here and mains-ish voltages.

[Rufus]

We are just trying to verify the product before shipping, we know that our customers do not meet the spec so want to try and get ahead of the game. It's not meant for any certification just to prove the suppliers design.

I think i'll try and find the reed, not sure why it will have a short life ? we are talking under 1uF here and mains-ish voltages.

The contacts are typically rated to switch 0.5A and depending on what load your equipment presents shorting out a capacitor into it could produce hundreds of amps. Mercury boils at 350C so a bit will evaporate each time and there isn't much to start with on the tiny reed contacts.

I have no real idea what the life would be but I would be prepared for it to be short. If you get something from Pickering ask them they are the experts. 

[ee.jmlp]

The spec sheet is talking about a pulse every 1 second, given that the pulse source capacitor is 0.1uF and the charging resistor is 100K then it will take 50mS for the capacitor to charge back up.

I like to use microcontrollers because of the flexibility they give you after design. You can use arduino, microchip, atmel, psoc, or whatever you want



If you don't like the relay output stage because of the erosion of the contacs, you can use a mosfet moved with bootstrapping and with the discharge signal. (I wouldn't care)

[madshaman]

is this like an ebay job ? how much do they go for ?

Yeah, at least that's what I was thinking.  I remember seeing some of the older models go at least to 100V.  I'm pretty sure I have one sitting on a shelf that goes higher, but not to 300V (not at home atm so can't check); I didn't pay more than $100 iirc.

I did a quick search on eBay and nothing in that range at the moment.

However, high voltage pulse generators seem abundant, e.g. http://www.avtechpulse.com/medium/

Might be a used one you guys can buy cheap.

[Simon]

as per the diagram on page 14 we'd be injecting a high voltage of limited power (as it's stored in a 0.1uF capacitor) to see how the circuit can cope with spikes in the power line.

I'm also wondering how much power the human body can cope with to see how deadly the output would be, the spec sheet keeps going on about 15 millijoules

Off the top of my head, I think around even only 35mA is enough to cause problems and even be deadly, as long as there is enough voltage. So if you are dealing with high voltage of even only mA, you need to be careful. I think the lower threshold amount for causing pain/discomfort starts at about 5mA. Just one thing of note, body size does matter here too. A large person can withstand more power than a smaller person.

Well yes i know about the 35mA figure but here I'm storing it up in a capacitor and being told by the spec sheet that it holds 15mJ, so how do I equate mJ into mA for a given voltage ?

[Simon]

is this like an ebay job ? how much do they go for ?

Yeah, at least that's what I was thinking.  I remember seeing some of the older models go at least to 100V.  I'm pretty sure I have one sitting on a shelf that goes higher, but not to 300V (not at home atm so can't check); I didn't pay more than $100 iirc.

I did a quick search on eBay and nothing in that range at the moment.

However, high voltage pulse generators seem abundant, e.g. http://www.avtechpulse.com/medium/

Might be a used one you guys can buy cheap.

Ah yes i did come across them, with my limited knowledge what we require costs around 2K, as usual we are cheapscates.

I'll make a micro driven design so that I can press a button and have the series of negative and positive spikes delivered and then an LED come on to tell when the test is over.

[Simon]

how is an electrolytic capacitor going to react to 50mS reversed polarity surges ?

[Rufus]

Well yes i know about the 35mA figure but here I'm storing it up in a capacitor and being told by the spec sheet that it holds 15mJ, so how do I equate mJ into mA for a given voltage ?

The whole point of the spike test is to simulate what is likely to be encountered on vehicle 24vdc supplies (OK a bit worse than what is likely). Are vehicle 24vdc supplies considered hazardous? Defibrillators put 200-300J into the subject.

how is an electrolytic capacitor going to react to 50mS reversed polarity surges ?

The specification doesn't have any -ve surges.

[Simon]

on page 15 there is a mention of  a "+/-40V total amplitude" to be applied to the test subject. Do they mean peak to peak or that it needs to be negative and positive pulses.

I'm not too fussed about negative pulses as a reversed polarity diode on the input will handle this.

How will the electrolytic cap cope with high voltage positive spikes even or should i make it a 400V one to be sure.

[Rufus]

on page 15 there is a mention of  a "+/-40V total amplitude" to be applied to the test subject. Do they mean peak to peak or that it needs to be negative and positive pulses.

Don't know what they mean with +/-. The same section states figure 11 is an acceptable test circuit and that circuit has no ability to apply -ve voltages.

How will the electrolytic cap cope with high voltage positive spikes even or should i make it a 400V one to be sure.

100nF at 250v has stored energy of 3.1mJ. Dumping an extra 3.1mJ into a 10uF capacitor charged to 28v will increase its voltage to 37v. If the capacitor is straight across the input it needs to stand 100v for the surge tests anyway.

[Simon]

I'll get a 450V cap then they are readily available and not too expensive.

3.1mJ at 250 V ? the spec sheet bangs on about 15mJ so I'd actually need 0.5uF

[Rufus]

I'll get a 450V cap then they are readily available and not too expensive.

3.1mJ at 250 V ? the spec sheet bangs on about 15mJ so I'd actually need 0.5uF

The spec says "an energy content of not less than 15 millijoules" and a circuit which can not deliver that much within the 250v constraint.

In Rev D of the same specification those words are changed to "a maximum energy content of 15 millijoules"

Rev D also changes that +/-40V to +40V.

[Simon]

well that simplifys matters, but 3mJ is a far cry from 15mJ I think i'll up the capacitance a bit and maybe have the ability to switch in more if required.

[madshaman]

is this like an ebay job ? how much do they go for ?

...

However, high voltage pulse generators seem abundant, e.g. http://www.avtechpulse.com/medium/

Might be a used one you guys can buy cheap.

Ah yes i did come across them, with my limited knowledge what we require costs around 2K, as usual we are cheapscates.

I'll make a micro driven design so that I can press a button and have the series of negative and positive spikes delivered and then an LED come on to tell when the test is over.

Understandable.  For some time I've been meaning to built a high voltage high energy pulser for the N2 laser project that will probably go undone for a few years.

An N2 laser needs a very fast rise-time as the lasing action is self-terminating within a few tens of nanoseconds.

After lots of research I decided to use a marx generator-like configuration with BJTs in avalanche mode for my design.  There are lots of IEEE articles on the subject (some of which I've noticed border on plagairising older articles).

You might consider using a single high voltage BJT in avalanche mode to give your pulses a fast edge.  Obviously the leads delivering the pulses will soften the edge due to inductance, there's also the problem of group delay distorting your pulses as they travel down the wire; dealing with this is a whole discussion in itself and I'm no expert; this is all stuff I'm in the process of learning myself.

As aside, although you probably already knew, the formula for energy storage in a capacitor is:

E = 0.5 * C * V^2 (everything in SI units)

[Simon]

so I take it E is energy in jouls ?

I'd rather stick to the spec sheet recomendations and use a mercury wetted relay, these seem to be designed for just this sort of thing and handling well over 1KV is quite normal. Fortunately pickering are a Uk company.

[madshaman]

so I take it E is energy in jouls ?

Exactly.

I'd rather stick to the spec sheet recomendations and use a mercury wetted relay, these seem to be designed for just this sort of thing and handling well over 1KV is quite normal. Fortunately pickering are a Uk company.

If the spec sheet has an app-note specifically for this kind of circuit, or something similar, then hell yeah, definitely follow their recommendations.

[Simon]

I'm a bit uncertain as to whether i need any particular type of 5uH inductor. My instinct tells me an air core should be used as it is meant to replecate the cables but i can only find iron core ones readily available.

Will this be ok ?

[madshaman]

I'm a bit uncertain as to whether i need any particular type of 5uH inductor. My instinct tells me an air core should be used as it is meant to replecate the cables but i can only find iron core ones readily available.

Will this be ok ?

Hi, I'm only answering as I see no reply and I've got time on my hands :-).  I think you're right in your thinking wrt to choosing an air core; afaik, the main difference (aside from efficiency and energy losses) is that an air core will not exhibit hysteresis and neither will the cables.  Also, the air core won't saturate.  It will obviously need to be much bigger.

Not sure this helps and you might already be aware of this.

cheers.

[Simon]

Well I have found an inductor that saturates the core at about 3-4A so I'm guessing that a 0.3uF dump at 340V will not be able to saturate it.

[Eviltech]

I don't see why not use a thyristor to discharge the cap, similar to a capacitive discharge ignition on a engine. They just put out what has been put in - the energy from the capacitor. I'm not shure how fast they are, but to run an 4-cylinder engine at 15k RPM you need the ignition to spark 30000 times/s. That's well over the 50 or 60Hz you get from standart AC sockets. You might look at modifying one of those old AC stabilizers for the lamp TV's, you know, the bricks with the 2 balasts inside...

[Simon]

the expected rise time is 50nS, so if you call that 10% of the high cycle of a square wave thats 1-2 MHz type switching speed.

[Eviltech]

the expected rise time is 50nS, so if you call that 10% of the high cycle of a square wave thats 1-2 MHz type switching speed.

You are right than, they will not do. I dont know alot about IGBT's but i don't think any of them will do the trick... The once i have seen in invertors usualy go up to the 10-100Khz range, which will not be sufficient for your pourposes. Maybe use a radio lamp? Some incredibly powerfull triode?

[Simon]

thing is the relays are only £8-9, I'd bet that you can't get a silicone/mos device to perform for the same price and no headaches.

[Eviltech]

thing is the relays are only £8-9, I'd bet that you can't get a silicone/mos device to perform for the same price and no headaches.

In here i use a normal bipolar transistor "13007" for a similar pourpose and i believe it might be enough for your needs. I have to check it's datasheet again, but it's mainly used in switchmode powersupplys.

[SeanB]

MJE13007 will die very violently in this application. If you want to use a SCR you need a 25A unit with a peak current in the 1kA range. Those tend to be a lot more than a mercury wetted relay. If you are worried about relay life put it in a socket.

[lorth]

For implantable defibrillators, they use between 4 and 10 Joules. The higher, more risk of damage to the heart. Same I think, not 100% sure, about portable defibrillators.

Here the review paper: http://europace.oxfordjournals.org/content/3/4/275.full.pdf

[Simon]

Well my 0.3uF with 350V in them seems to work out at 0.05mJ

[Rufus]

Well my 0.3uF with 350V in them seems to work out at 0.05mJ

I think you worked that out wrong.

[Simon]

Well my 0.3uF with 350V in them seems to work out at 0.05mJ

I think you worked that out wrong.

Well I sure hope I did 

But 350V * 0.000'000'3F * 0.5 = 0.000'0525J

What did i do wrong ?

[Simon]

ok got it, forgot to square the voltage, works out at 18mJ

[trackman44]

I know this might not work but what about a flash from an old Nikon SLR camera? Take out the bulb and wire it to a circuit and take a measurement. Just a thought.

Will

[Simon]

The problem is the camera flash specs are unknown to us, we need to replicate the conditions suggested in the spec and so simpler to just follow their guidlines

[Simon]

Well I'm nearly at the end of my design but have noticed that the spec sheet is also saying to do a 100V test. I don't know where to get 100V from, my transformer will peak at 50 so I could use a voltage doubler rectifier to get it. To try and take a short cut and keep a common ground between my 50V line and the 100V line I could use the diode bridge rectifying the 50V as one of my diodes in the voltage doublers and sharing the capacitors, like the attached PDF, will that work or am I deranged ?

[Andy Watson]

Won't work!
Consider the point labelled as "2" on the bridge rectifier, it can never get above one diode drop of "3", which is the other side of the capacitor that you're trying to charge.

Do you need full-wave rectification?

[Simon]

I need all the power I can get at 50V and less on the 100V. Should I just use a separate 2 diode voltage doubler or will it still not work ?

[Andy Watson]

Should I just use a separate 2 diode voltage doubler or will it still not work ?

Two separate diodes and an extra cap should work.

... and less on the 100V.

How much less? You could go to a whole bridge and two extra caps to get a bit more power, but even then, you'll be wanting reasonably large valued extra caps to realise any significant power output. You would effectively be A.C. coupling the output of the transformer through twice the impedance of the extra cap.

[Simon]

Well the 50V is dumping directly into the load. The 100V is dumping via a 0.5 ohm "source" resistance so less power will go through.

So you say make a completely separate voltage doubler, i still need to have a common negative though

[Simon]

my idea was that one of the diodes in the bridge would in effect be in parallel with one of the voltage doubler ones so why no commonize it and use the capacitors as 50V and as the lower half of the 100V

[Simon]

what is I just take the bridge out and use a voltage doubler instead, then I can take 50 or 100 volts at will, but I'd be inclined to have a larger capacitance on one side and what happens when take significant power from one side only, will the other be affected or will it affect the side I'm taking power from ?

[Andy Watson]

Do you have a figure for the amount of power? If all you want to do is charge a capacitor once in a blue moon, then a voltage doubler is the way to go. On the other hand, further up this thread, you seem to be suggesting that you want to drive 100V into 0.5R continuously! That would be entirely another level of engineering and not one that I would consider practical for a voltage doubler.

[Simon]

No i just need to dump the 100V for 50mS through the 0.5R resistor into the load under test. The dumps need to happen once every 1 second.

[Andy Watson]

No i just need to dump the 100V for 50mS through the 0.5R resistor into the load under test. The dumps need to happen once every 1 second.

An average of 10A, i.e. 1kW!  Do-able, with a voltage doubler, but if it were my problem I'd be looking at using another transformer - probably cheaper than the cost of the extra caps - and more reliable.

[Simon]

I don't think the idea is to supply constant 100V through 0.5R but just dump a load so it will peak at 100V and decay. details in the spec sheet are sketchy.

My concern now is switching that 100V, mosfet and BJT's are scare and expensive at those voltages and the BJT's have very poor gain, I'll end up needing more base current than the load is having dumped on it.

[Rufus]

I don't think the idea is to supply constant 100V through 0.5R but just dump a load so it will peak at 100V and decay. details in the spec sheet are sketchy.

It isn't sketchy. It is trying to simulate a load dump on a 28v dc system when the battery has also been disconnected and as I said on the first page of this thread

"The killer in that spec is the surge voltages unless your equipment can stand 100v for 10s of ms anyway."

The test set up in another similar standard is 100v of large lead-acid batteries and a contactor.

What your test set up needs to be depends on what you are testing - what current it is going to pull with 100v on the supply. If for example you have a 33v TVS diode on the input the theoretical supply current would be 134A or 13.4kW from the supply and 4.4kW in the TVS diode both for 50mS.

[Simon]

hm, I'll try not to point that out to the boss  He might not appreciate the requirement for something huge on a trolley. I think the main concern is the high voltage at 250V but as i already said to them it's more about the power.

The current power capabilites I have designed (680uF @ 100V) seems to work out at about 3.4J i have no idea what that is in the grand scheme of things.