Clearance between PCB tracks with 300V of potential difference?

[ocset]

Hello,
We have a  230VAC  offline,  non isolated LED lamp with LEDs mounted on MCPCB which sits on an earthed heatsink. Due to space constraints on the MCPCB, we have PCB tracks which  have a potential difference of 300V between them, with just 0.3mm of separation. The tracks obviously are covered in solder resist.
Table 6-1 of page 43 of IPC2221 appears to state that this is too  little spacing, but is not clear.
However, will we be likely to see flashover, or dielectric breakdown of the solder resist material? Also, why does  electrical spacing need to be greater at higher altitudes?…eg, the wider regulatory spacings seen for  >3000 metres?

IPC2221 (page 43.....comes up as 53rd page)
http://www.edaboard.com/showthread.php?t=372473&p=1595141#post1595141

[ocset]

Thanks, i wish i could find that UL60950 on the web..we cant afford it.
I wonder what perils will happen if we ship the circuit with this 0.3mm of spacing between the tracks which are 300V apart in voltage?

Will we see dielectric breakdown of the  solder resist?
Or will we see the conductors flashing over and arcing?.....even though they are both insulated from each other with the solder resist which covers them.
What kids me is that  the  PCB tracks are on MCPCB...and the MCPCB sits on an earthed heatsink...so therefore all the tracks are only effectively separated form the earthed heatsink by the 60um of insulator which covers the MCPCB.  Do you think we need a thermally conductive/electrically insulating pad between the MCPCB and the Earthed heatsink?

[ocset]

Thanks very much , and i dont think there will be fire, the MCPCB is enclosed in a enclosure full of non flammable materials.
I think we agree that no problem will result which is directly caused by the 0.3mm of spacing between tracks with 300V of difference?...since they are covered in solder resist.

I appreciate regulations....but they are to at least a degree written as a factor of protectionism of whichever market by influence of whichever large engineering corporation is acting with self-interest.

The following website shows how the flashover distance for conductors which are 1000V apart is 0.13mm....

https://www.cirris.com/learning-center/calculators/50-high-voltage-arc-gap-calculator

....so us having 0.3mm of distance with 300V of potential difference will surely bring about no problem whatsoever?

I once went for a job at a missile detonator company. The Detonator voltage was some several kV, and they simply couldn't achieve anywhere near regulatory spacings between conductors...but none of their products suffered problems due to the non-regulatory spacings.

It is simply impossible for anyone to be physically harmed by us having a spacing of 0.3mm between our traces which are 300V apart...surely you agree?
 
I use the following example to support my claim about the righteousness of using PCB spacings considerably less than what the standards require (eg IPC2221 and IEC 60950)……in this case, the situation of inadequate spacings seen in offline SMPS……

It is interesting to note that many major power supply manufacturers in their low-power off-line designs are widely using 500-800V MOSFETs in TO220 package operating at 400V and higher. With this package you can get about 30 mils spacing between the pads, while the documents would require at least 100 mils. Even if you spread the leads on the PWB, you can't do anything with 50-mil spacing between the TO220 leads along the surface of the package.

The above quote is taken from the following….
http://www.smps.us/pcbtracespacing.html

[vzoole]

Solder mask or solder resist is resist against soldering as its name shown and nothig else. So it isn't mean anything in isolation.
Most standards require minimum 1.5mm clearance between traces in main circut at 300V.
If you do not keep this value and your device will fail it will be your responsibility.
If it is a demage of property around 1000$ thats not a big deal, but it will be a product recall and that's will hurt.
If a building burning down and somebody die you risk jail.

ps.: and one of your competitor can check your design as well and it will be a product recall that cost you a lot.

[ahbushnell]

Thanks, i wish i could find that UL60950 on the web..we cant afford it.

Google "ul60950-1 pdf" and choose the first result (on my computer, from a cdn14.21dianyuan.com domain).

I wonder what perils will happen if we ship the circuit with this 0.3mm of spacing between the tracks which are 300V apart in voltage?

Will we see dielectric breakdown of the  solder resist?
Or will we see the conductors flashing over and arcing?.....even though they are both insulated from each other with the solder resist which covers them.
What kids me is that  the  PCB tracks are on MCPCB...and the MCPCB sits on an earthed heatsink...so therefore all the tracks are only effectively separated form the earthed heatsink by the 60um of insulator which cover the MCPCB.  Do you think we need a thermally conductive/electrically insulating pad between the MCPCB and the Earthed heatsink?

I've tried 0.15mm for 1.1kV for experiment boards with no breakdowns at all, but I won't trust my company reputation, money and freedom from jail on that.
As the saying goes, no one gets fired for buying IBM. Sometimes standard conforming is a shield from jail time.

Or a shield from killing someone.

[ocset]

Thanks, is it even OK to use nonisolated   mains on a MCPCB with LEDs on it?
We have a 300V LED string on MCPCB, and the dielectric is just 60um thick.
Is this wise?
The product is non isolated and runs off 230VAC.
The MCPCB sits on an earthed heatsink.

[logictom]

Assuming this is a product and not just some project, you must meet safety standards to obtain the CE marking required to sell in the EU. The low voltage directive now requires compliance with 61010 3rd edition which requires, best case, 1.5mm clearance.

If you are your team don't understand the safety requirements for such a product, let alone the regulatory requirements, you need to pay someone to get you up to speed quick or drop the project.
To make a one off in your own home without the knowledge is bad enough but to try and sell this as a product is something else entirely.

[capt bullshot]

You can get away with lesser spacing than the standard for such "internal" voltages. You'll have to do some component failure tests on the board to achieve CE or whatever. The testing is to prove that your product isn't dangerous in in case of failure (of a component or in this case the 0.3mm distance). If you adhere to the minimum clearances according to the standards, it is believed they won't fail and testing isn't necessary for this particular part of the product.

A layer of 60um isolation may be thick enough to provide a safe isolation from the live circuit parts to safety ground. Depends on the material, thermal stress, ... You'll have to check this with the PCB manufacturer against the applicable regulations.

Wider clearance is required at higher altitudes due to the lower atmospheric pressure. The same voltage sparks at wider distances at lower atmospheric pressure. This can be really nasty in particular if you have electrical wiring inside housings operating at low pressure (like pre-vacuum).

Solder resist isn't an isolation means. It can stand impressively high voltages, but it is not guaranteed to do so. You can't rely on that.

If your 300V sparks across the 0.3mm gap (which is totally possible to happen, maybe due to some transients or surges on the power line), this spark can turn into an arc and cause heavy destruction. Depends on the upstream fuse. So if you do breakdown testing, it is alway necessary to specify and use a fuse, either part of your product or specified to be installed by the end user (for business customers, not recommended for consumer customer).

Conclusion: Better go for a wider distance ...

[Ice-Tea]

I use the following example to support my claim about the righteousness of using PCB spacings considerably less than what the standards require (eg IPC2221 and IEC 60950)……in this case, the situation of inadequate spacings seen in offline SMPS……

It is interesting to note that many major power supply manufacturers in their low-power off-line designs are widely using 500-800V MOSFETs in TO220 package operating at 400V and higher. With this package you can get about 30 mils spacing between the pads, while the documents would require at least 100 mils. Even if you spread the leads on the PWB, you can't do anything with 50-mil spacing between the TO220 leads along the surface of the package.

As others mentioned: please respect the 'rules'. They are there for a reason. You also need to understand better what you are doing, even the example that seems to be the basis of your scepticism more or less complies to them: 30 mil works out to almost 0.8mm (which is what IPC2221 calls for). This is less than 1.5mm but I would assume this is because they are no longer directly on the line but rather a bit further, after a MOV per example that guarantees surges to be blunted.

[dmills]

Are you sure that 300V is just 300V even when something 'interesting' happens on the mains input?
1,000V transients on the mains are so common as to be basically uninteresting and 5,000V is not particularly remarkable.
 
Now add a film of hydroscopic dirt and some badly applied solder mask, there is a reason the clearance distances are what they are, what pollution class are you designing to anyway? 

I have seen a (Granted Paxolin) board with proper creapage burn where an arc tracked between two mains connections and the heat then carbonised enough board to form a relatively low resistance path which then drew amps of current. The numbers in the standards are minimums, try to design for more, it pays in reliability.

300V with a gap like that might be ok if the energy was severely limited (Think GM tube bias supply), but I would not want to design it in if there was any way for real energy to end up there (Including due to failure of any single component).

One thing to watch on MC pcbs is tracking around mounting holes and board edges, the epoxy layer is actually quite robust, and earthing the metal side via the heatsink is usually sufficient if you are using a decent fab. There are standards for PCB voltage robustness (CTI index is the magic search term for surface tracking issues).

Regards, Dan.

[ocset]

Thanks all, we do have a littelfuse TMOV module in the AC line. Also, we have a TVS downstream of a current clamp.....so we have transient protection.

If your 300V sparks across the 0.3mm gap (which is totally possible to happen, maybe due to some transients or surges on the power line)

Thanks, though the flashover distance for 1100V is  110um. So  its hard to see it happen.
As for the chances of gettign the creepage related shorts burned in to the  baord......we have a fuse, and also, to make the board bigger just because of the possibility of such a rare occurrence, wouldnt be financially viable.

I agree though that we wont be able to export our product....competitors woudl kill us on regs.
With our transient protection, fusing etc, lack of product flammability , its hard not to be  convinced that  no one will ever be harmed by our 300V with 0.3mm clearance.

[dmills]

It is not just unable to export, but unable to bring to market at all.

Someone points it out to trading standards, and they will have you pulling it off the market and recalling product, very expensive.

Trading standards are under funded, but they will react if someone points something out.

I would worry about any design where a few mm of extra board area made or broke the product BOM cost, it just does not seem like a market that would be remotely interesting to play in.

Who is going to sign the declaration of conformity for this thing (They end up with personal liability as I understand it).

Regards, Dan.

[ocset]

Thanks, there are multiple places where theres only 0.3mm clearance between the 300v_apart_tracks.
To change all these places to be 0.8mm apart would need a much bigger product.
I cant see why Our country's standards would get us done over....after all, we only ship within our country...and we pay tax to our country's government, so its not worth them doing us over.

We cannot see how our product can possibly harm any person.
If there was the slightest chance of this we would pull it now.

[IanB]

We cannot see how our product can possibly harm any person.

Can a failure produce smoke, fumes, fire? For instance, what happens if there were to be a low level short with not enough current to trip the protection, but enough current to produce charring, smoke or flame?

[ocset]

Can a failure produce smoke, fumes, fire? For instance, what happens if there were to be a low level short with not enough current to trip the protection, but enough current to produce charring, smoke or flame?

Thanks, thats a good point, but no i would be amazed if the 0.3mm of clearance between the 300v_apart tracks could catch fire. The tracks are covered in solder resist. The flashover distance for 1000V is 113um, so us having 3000um between tracks 300v apart will surely stop any fire from happening. Its a 220VAC product with a 3.15A fuse, and the product has a power level max of 150W.

30 mil works out to almost 0.8mm (which is what IPC2221 calls for)

Thanks, but we are category B4, of IPC2221A, page 43 (53rd page), and that calls for 0.4mm for the category of   251-300V......

IPC2221A
http://www.sphere.bc.ca/class/downloads/ipc_2221a-pcb%20standards.pdf

[ocset]

1,000V transients on the mains are so common as to be basically uninteresting and 5,000V is not particularly remarkable.

Thanks, so if we place a 1.5SMC440A, 440v TVS  in series with a 100mA quick-blow fuse across the mains,   do you believe that 200mA,  fuse will blow within a month? (if not a month, how long?)

1.5SMC440A TVS
https://www.vishay.com/docs/88303/15smc.pdf

[capt bullshot]

So is this 300V voltage difference isolated from the mains and regulated - then the 0.4mm clearance shall be OK.
If it is rectified mains, or more or less derived from the mains, with 0.3mm clearance, sooner or later you'll have defective products due to this. Your 3.15A fuse will protect the end user from dangerous situations given your housing is as good as you claim, but you'll see destruction on the PCB happened.

Better not use a MOV across the mains input, these blow up after long enough time and some overvoltage events happened. Sometimes they are required by regulations, but a "normal" consumer device's mains input isn't the right place to put a MOV onto.

Do you have a rectifier followed by a reasonable bulk capacitor (100uF and up) at your input? The bulk capacitor will absorb quite a few of the line transients - but you should place a NTC in series with your mains input to limit the inrush current. Be careful if you use a PFC stage, sometimes the diode there can't stand the inrush or transient current caused by a line surge.

The TVS device you mentioned is quite fail-safe. When overloaded, they fail short and blow the upstream fuse. And even better: they do not have the long term degradation mechanism of your typical MOV. So if you don't overload it, the fuse won't fail.  Overloading the TVS is a matter of pulse energy (of single transient / surge events) or too high powér loss due to constantly applied overvoltage.

[T3sl4co1l]

Solder mask or solder resist is resist against soldering as its name shown and nothig else. So it isn't mean anything in isolation.

Based on my reading of IEC 60950-1, I conclude that soldermask or conformal coating is acceptable insulation, if an inspection process is used.

I asked one major PCB manufacturer about this safety rating and they were relatively clueless, so I doubt it is used very often.  Base your calculations on assuming the soldermask isn't there.

Note that, for traces in an off-line circuit, internal clearances only need meet functional ratings.  It is not clear where basic or reinforced insulation (mains transients) becomes functional (transient limited), but it's probably somewhere after the input filter or rectifier, since semiconductor components cannot function at transient voltages, regardless of whether they survive or not.  (Safety doesn't care if the circuit survives, only that it fails in a safe condition.  Semiconductors failing shorted and blowing the fuse is an acceptable mode of operation.)  Including an MOV provides a clear demarcation of this, but it's not a necessary component for average reliability products.

Tim

[Ice-Tea]

30 mil works out to almost 0.8mm (which is what IPC2221 calls for)

Thanks, but we are category B4, of IPC2221A, page 43 (53rd page), and that calls for 0.4mm for the category of   251-300V......

No, it doesn't. Firstly, you mention you have a 230Vac product and at another point, you have 220Vac product. If you are in the UK, it should be 230Vac. There's a tolerance on that: 10%, so 253Vac. IPC2221A (the most recent one is B, by the way) has categories for "Voltage Between Conductors (DC or AC Peaks)". Your AC peaks are 357V. Well above 300. In addition, the only reason why you can rely on it not being higher at that point is your TVS/MOV/Whatever. The one you selected will fire, worst case, only at 462V. And it will have a clamping voltage of up to 602V. So, double of what you are considering.

[DerekG]

after all, we only ship within our country...

Yep, this makes it very easy to catch up with you.

and we pay tax to our country's government, so its not worth them doing us over.

A bit like a criminal arguing that he/she should not be sent to jail as they pay their income tax on time.

[ocset]

Thanks,
It is indeed non-isolated mains...but the tracks i speak of are within the multiple led banks, and the voltage across the led bank (total) is clamped to 300V max. ...so even in a mains transient, the voltage should stay below 300V between the  discussed tracks.

[capt bullshot]

Thanks,
It is indeed non-isolated mains...but the tracks i speak of are within the multiple led banks, and the voltage across the led bank (total) is clamped to 300V max. ...so even in a mains transient, the voltage should stay below 300V between the  discussed tracks.

So there's some electronics in between the rectified mains and your tracks? Depends on the nature of the components, but in general it is now quite possible to have 0.4mm (or whatever the applicable standard says) clearance and be safe with that. Can't tell this remotely without ever having seen your circuit, even then I only would be able to have an opinion but no statement that might save your boss or you from liability. if you're in doubt go to some kind of authority (in Germany e.g. VDE, or TÜV) that can give you a statement.

[T3sl4co1l]

60950-1 allows a given potential, and its associated clearance, to be broken up into N+1 gaps of width (1/N or minimum gap, whichever is greater).  That is, the total gap always meets the minimum, even in a single failure event.

This is most applicable to chains of resistors, but the same logic applies for LEDs, yes.

It would be rather stupid to not be able to get 7 mil clearance between traces that are 10V apart, just because they're both floating at 300V.  This standard, at least, allows for such.

Writing accurate design rules, to reflect these considerations, can be very difficult.  It is probably better within the OP's capabilities to simply make a board as he pleases, submit it to the local approval group (the UK's equivalent of UL) for review, and revise until correct.  (Again, this is a very expensive procedure, but it seems to be no more expensive than other processes the OP has revealed, and is more likely to be correct than free information over the internet here.)

Tim

[ocset]

Thanks, What would be very helpful to us, and i believe other companies too, is details of court cases where companies have been fined or liquidated due to having made electrical standards infringements.....but you never see this, it seems like the standards are some "Holy order", only trasnparent to those "important" enough to have the right to know about them.

Has any company ever been fined, or forced to do a  total product recall, due to a standards infringement.

If these standards really are written in the name of safety, and indeed saving peoples lives, then i am amazed at how opaque they are. If its for saving peoples lives, then each country should summarize them and circulate the info, so as to prevent death of citizens.

Of course, im not rabbitting on at, or blaming in any way, you kind and helpful respondents here.

[T3sl4co1l]

At least in the US, I don't think it's a court level thing.  I don't actually know the ultimate legal basis under which you might be fined or otherwise liable if your products aren't UL recognized here -- but it's also highly unlikely that a scrupulous retailer would ever risk selling your non-UL-recognized products here, for that reason.  And I imagine the basis is similar in UK law and practice.

Does it ever occur to you that asking about these things is analogous to asking why gambling is illegal?  Like, "what is the actual basis for that statement"?  Because that's what you're doing here, gambling with your customers' safety.  You're, at best, an incompetent engineer, and at worst, an unethical and irresponsible one.

Tim

[Ice-Tea]

To further the analogy: you're now the criminal that's asking what the chances are of getting caught.

Standards are drafted by experts. They see the iceberg, where you only see the tip. As a result, they may seem opaque but mostly it's just good and knowledgeable People trying to be precies. Granted, it's a bit of a jungle though.

[capt bullshot]

Has any company ever been fined, or forced to do a  total product recall, due to a standards infringement.

Never heard of such a case ... Wait, that's not true, I remotely remember reading of companies beeing forced to recall products (I don't think of the burning Samsung Smartphone now), but this never happens "out of the blue", there's always been an incident or someone volunteering to find flaw on a products. A competitor doing so comes to my mind ...

But there's no such thing as a mandatory and preventive check of your product before you can put it to the market. You declare it conforms to the rules. You can do so by just cheating, no one will notice.

By doing so, you take all the responsibilty and liability on your own risk. And now for the special part: If you made your product according to applicable standards and you have proof of this, then in case of damage, it is believed the it's not your fault by default. Otherwise you'll have to prove the particular damage isn't your liability due to bad product design. Disclaimer: IANAL, just an EE.

What is well known: customs rejects importing of goods that aren't declared to be conform. But they do not check if the conformity is real or cheat.

[capt bullshot]

Does it ever occur to you that asking about these things is analogous to asking why gambling is illegal?  Like, "what is the actual basis for that statement"?  Because that's what you're doing here, gambling with your customers' safety.  You're, at best, an incompetent engineer, and at worst, an unethical and irresponsible one.

Don't be so harsh :; It's the way things work in a small company (at least ten years ago when I worked for a quite small company). It wasn't even considered to research all the applicable standards, too much effort ... So why question the rules if you don't even know them?

Now I'm working in a different context, nobody wants to take responsibility, so obeying applicable standards is the way to go.

[ocset]

Thankyou , I must now tell of the story of a fantastic little company called Cyden. 

They pioneered hair removal by high intensity light application. They developed an original  portfolio of fantastic hair removal products which were brilliant, absolutely perfectly safe,  and had started to gain export sales.   

Then one day, a huge  engineering corporation phoned them up and asked them if their product conformed to all of hundreds upon hundreds  of quoted  regulatory standards.
The huge corporation then told Cyden that they were going to bring Cyden’s products into their standards lab…and rip them apart to  see if they passed  absolutely all sub-clauses of every single regulatory  standard.            
The huge corporation told Cyden that if its products did not conform to absolutely  all  standards from top to bottom, then they would be forced to do a total product recall from  several of the  countrys  in which this huge corporation resided…as well as receiving a big fine.

However, the huge corporation gave Cyden an alternative……   

If Cyden agreed to sell itself  off to the huge corporation concerned then  none of the standards checking would happen.
And indeed the company sold itself off………and still makes great products, -but it  is simply now a little employee of the big corporation, and gets payed a pittance of a wage. They are certainly now not allowed to put their own name on their own products…the name of the huge corporation  goes on there instead.
Their products were entirely safe and reliable, but   the huge corporation has an enormous department full of standards  checking employees….and there was no way that Cyden could guarantee being able to conform to the letter of all that  enormous sea of standards.
The same thing happened with uvintegration.com…..now sold off to an overseas company, totally.
..and scores of others have gone the same way.
Very often, the gaffer of the small company being bought out gets given a bit of a “nugget”, to sweeten the disappointment, shall we say.
After having the "sweetener" forced upon him/her, the gaffer is probably less inclined to speak about what happened.
Certainly, speaking out about it  might be viewed rather dimly by the buyout company....the gaffer's position in the newly acquired "subsidiary" mght come under "review", shall we say.

Then you end up with a situation like in the UK where 66% of all industry (>500 staff) is foreign owned. Quoted from Baroness Neville-Rolfe, Uk minister of business in 2016

 

[ocset]

1,000V transients on the mains are so common as to be basically uninteresting and 5,000V is not particularly remarkable.

Thankyou , please consider the following capacitor/diode being placed across the output of a mains rectifier. (no other circuitry involved)
How often would you expect this capacitor’s voltage to be taken above 450V? (I  am speaking about connection to the mains in UK or Europe or USA or Australia or New Zealand)

[T3sl4co1l]

Note that RFN5TF8S avalanches around 900V:
http://rohmfs.rohm.com/en/products/databook/datasheet/discrete/diode/fast_recovery/rfn5tf8s.pdf
(Typically, breakdown is a modest percentage above Vrrm.)

Assuming no series resistance beyond the surge network, the excess voltage can be calculated.

IEC 61000-4-5 surge is based on a 20uF capacitor.

Assuming charge equality at the waveform peak, a 1.5kV (line-line) surge will divide into 33uF giving 933V.  The diodes will feel it.

Tim

[ovnr]

Honestly, I'd never put mains on a MCPCB no matter what. Not saying it can't be done, but it's iffy and outside my comfort zone. I always go for full reinforced insulation specs (>= 4mm clearance) in any event.

Filtered, clamped DC at 300V on a MCPCB? Sure, not a big deal.

I don't think it was entirely clear if your product is single-board, or if you have a MCPCB for the LEDs and a separate board for the power supply?

[ocset]

I don't think it was entirely clear if your product is single-board, or if you have a MCPCB for the LEDs and a separate board for the power supply?

The power supply is linear led current regulators feeding in to the multiple led banks on the pcb.

IEC 61000-4-5 surge is based on a 20uF capacitor.

How common are these type of transients?
How often do they occur?

Assuming charge equality at the waveform peak, a 1.5kV (line-line) surge will divide into 33uF giving 933V.  The diodes will feel it.

Thanks, but i just cant believe that the vast majority if mains transients would be able to take the capacitor of Reply #31 above 450V.

[Gyro]

IEC 61000-4-5 surge is based on a 20uF capacitor.

How common are these type of transients?
How often do they occur?

You're approaching that question from completely the wrong direction. From the perspective of a cautious product designer, the question should be 'How often can I afford these transients to occur?' It sounds as if your current answer may be 'less than once'.

[Ice-Tea]

IEC 61000-4-5 surge is based on a 20uF capacitor.

How common are these type of transients?
How often do they occur?

[/quote]

1) Once is enough
2) 1 minute google in attachment.

Assuming charge equality at the waveform peak, a 1.5kV (line-line) surge will divide into 33uF giving 933V.  The diodes will feel it.

Thanks, but i just cant believe that the vast majority if mains transients would be able to take the capacitor of Reply #31 above 450V.

Belief doesn't really hold a lot of weight. And once is enough.

[ocset]

Thanks Iced Tea.....thats great data on transients.   

However, it doesnt allure to how much energy is in each of those transients.

There is no way of knowing how many of those transients would cause the 33uF capacitor of post #31 to go above 450V.

Its presumbaly possible that none of the transinets of that graph could take the capacitor of Reply #31  above 450V? 

 

[ovnr]

I think something you're forgetting is that the cap isn't perfect. If you get a nice, fat transient down the line, you're going to have pretty short rise times. The capacitor ESR will prevent it from absorbing the whole thing; you can still get higher voltages on the board, even though the cap isn't charged above 450V.

Also, keep in mind that it's not the arc-over from the transient that's going to burn your house down. It's the newly ionized (... and possibly carbonized) channel that allows the line voltage to flow across your board, causing much excitement.


I can't help but feel like your attitude to the problem is a bit wrong too. "Can't we cross our fingers and hope that nothing goes wrong?" isn't the safety-minded approach; you should instead say "What can we do to make sure this thing doesn't kill someone even under adverse conditions?".

The standards are there for a reason. If you follow them, no slimy lawyer can go "blah blah negligence" later on. If you skimp on things... well.

[Ice-Tea]

As Tesla stated: the standard is modelling a surge with a 20uF. Again: there's a reason they do this, in this case probably because it approximates real life. So it's easy enough to deduce the occurence from the graph I found. Which is, by the way, annecdotical. Could be a lot worse. Or not. To eliminate that kind of guesswork there are... standards!

[ocset]

Thanks,  incidentally  I dont think there is danger  of fire in an enclosure full of non-flammables.
FR4 PCB isnt flammable.

I once accidentally plugged a three phase input into the output connector of a prototype 10kW electric drive.
The explosion was quite earth shattering.
It was like being next to a cannon
The PCB was melted and contorted beyond belief, but there was no flame, and no fire.

[Gyro]

You are frighteningly complacent! 

[ocset]

Thanks, thing is, look inside any offtheshelf offline PSU you like, you'll see just a mov transient protector, -nothing more..give that a good few transients...and that MOV fails......and then your product has no transient protection.

And the failed mov won't blow the fuse, because manufacturers these days are putting movs upstream of the input fuse.

  

[T3sl4co1l]

Most things don't have a MOV at all, and survive just fine.

A keen observer would take this and run:

IEC 61000-4-5 surge is based on a 20uF capacitor.

Assuming charge equality at the waveform peak, [the] surge will divide into [capacitance] giving ...

33uF is insufficient.

What about a value much greater than 20uF?  What about 100uF, or 1000uF?

Indeed, check this out: https://zerosurge.com
How's that work?  Simple: https://www.google.com/patents/US4870528

Looks like they use a stage or two of 220uF to deal with it.  Series inductance to help even more.

It won't work for unimaginably large surges, or bursts of several surges together, that charge the capacitors too far.  But it will help with the most common events, which are also comparable to, or smaller than, the standard it's tested to.

Put another way: to a certain extent, it's gaming the system.  Why should the protection device withstand any more than what the standard specifies?

Now consider what "any offtheshelf [sic] offline PSU" has inside it -- most I see don't have MOVs at all.  They survive just fine!

I am now a bit curious about the smaller supplies.  Probably, resistance in the CMC goes a long way, reducing the impact.  Or current inrush, for which some don't even use an NTC at all, but a fusible resistor (usually 4.7 ohm).  Make that resistor surge-proof as well, and there you go.

Tim

[Zero999]

When at outer layers without encapsulation, you need 1.6mm creepage and 1.5mm clearance (for functional insulation)

1.5mm of clearance and 1.6mm of creepage sounds overkill for functional insulation.

Why did they pick such impractically large figures? The creepage distance between the pins of many power semiconductors is much less than that. It's not possible to use a 400V TO-220 TRIAC and maintain 1.5mm creepage on the PCB, without bending the leads.

I can understand having a large creepage distance, on the PCB, before the fuse, as failure can result in fire, but afterwards seems unnecessary!

[capt bullshot]

When at outer layers without encapsulation, you need 1.6mm creepage and 1.5mm clearance (for functional insulation)

1.5mm of clearance and 1.6mm of creepage sounds overkill for functional insulation.

Why did they pick such impractically large figures? The creepage distance between the pins of many power semiconductors is much less than that. It's not possible to use a 400V TO-220 TRIAC and maintain 1.5mm creepage on the PCB, without bending the leads.

I can understand having a large creepage distance, on the PCB, before the fuse, as failure can result in fire, but afterwards seems unnecessary!

You can use smaller clearances and get e.g. CE compliance certification from a notified body. It just involves more effort like testing what happens if these components fail. Test setups etc. are well-defined. You'll have to consider other failures too, e.g. what happens if the reduced clearance at your triac fails and a heat source that is intended to work on low duty cycle gets full power then? This can easily start a fire if no other countermeasures are provided.

[Zero999]

When at outer layers without encapsulation, you need 1.6mm creepage and 1.5mm clearance (for functional insulation)

1.5mm of clearance and 1.6mm of creepage sounds overkill for functional insulation.

Why did they pick such impractically large figures? The creepage distance between the pins of many power semiconductors is much less than that. It's not possible to use a 400V TO-220 TRIAC and maintain 1.5mm creepage on the PCB, without bending the leads.

I can understand having a large creepage distance, on the PCB, before the fuse, as failure can result in fire, but afterwards seems unnecessary!

You can use smaller clearances and get e.g. CE compliance certification from a notified body. It just involves more effort like testing what happens if these components fail. Test setups etc. are well-defined. You'll have to consider other failures too, e.g. what happens if the reduced clearance at your triac fails and a heat source that is intended to work on low duty cycle gets full power then? This can easily start a fire if no other countermeasures are provided.

That sounds fairly sensible but is it really necessary to perform tests?

Most potentially dangerous scenarios can be eliminated by good design. For example, an appropriately rated and approved thermal fuse, can be used to prevent a heater from overheating, if the TRIAC's junction fails short circuit, which is more likely than the functional insulation between the pins failing.

[capt bullshot]

That sounds fairly sensible but is it really necessary to perform tests?

Most potentially dangerous scenarios can be eliminated by good design. For example, an appropriately rated and approved thermal fuse, can be used to prevent a heater from overheating, if the TRIAC's junction fails short circuit, which is more likely than the functional insulation between the pins failing.

Well, that depends ...
Testing: You'd discuss this with your notified body or whoever certifies your product, differs with applicable standards.
Example: Yes, that's right, the TRIAC is more likely to fail than the clearance between its pins. Especially if you use TO220 cases, the wireframe has less distance than required for your PCB - that's kind of irritating and often solved by testing (TRIAC fails short includes isolation failure of the PCB).