EEVblog #678 - What is a PCB Spark Gap?

[EEVblog]

In this tutorial Dave explains what a PCB spark gap is and how it can be a useful zero cost addition to your PCB layout to help prevent ESD damage.
He shows how to easily design them into your board and calculate the approximate voltage rating.
And of course has some fun applying 5kV to some gaps to show how them at work.
Plugpack teardown:

Support the EEVblog through Patreon!
http://www.patreon.com/eevblog

[bktemp]

Can somebody explain, why they have put the spark gap under the common mode input filter to bypass it?
It makes no sense to me, because the only thing it protects is the common mode filter. Instead all the energy can pass directly into the circuit.

[Kohanbash]

It looks like with 1000V (from 8:12 to 8:14) there is a very small change that occurs near the top of the circle.

[EEVblog]

It looks like with 1000V (from 8:12 to 8:14) there is a very small change that occurs near the top of the circle.

I didn't see a spark, but did see it deform the board material or some such?

[coppice]

Dave uses a strange  formula for breakdown. It has no factor for humidity, which is the biggest factor affecting discharges with open air spark gaps.

[RobertoLG]

Dave uses a strange  formula for breakdown. It has no factor for humidity, which is the biggest factor affecting discharges with open air spark gaps.

ya, I was thinking about that too

[EEVblog]

Dave uses a strange  formula for breakdown. It has no factor for humidity, which is the biggest factor affecting discharges with open air spark gaps.

It's a ballpark, and as I said a formula that is thrown around in a few places. It's the best one I could fine, and the only one I have seen before. If you have a better formula backed with some research then please share.

[coppice]

Dave uses a strange  formula for breakdown. It has no factor for humidity, which is the biggest factor affecting discharges with open air spark gaps.

It's a ballpark, and as I said a formula that is thrown around in a few places. It's the best one I could fine, and the only one I have seen before. If you have a better formula backed with some research then please share.

The variation in air pressure, even at the top of Everest, isn't huge. Try the effect of a seaside town in the especially humid season.  The 3000 in the formula shown is the usual 3000V/mm breakdown figure for dry air at STP. Damp air could divide that by several. If the board is of any significant age, and has started to accumulate a film of pollutants from the air, moisture absorbed into that film will greatly reduce the breakdown across the board surface itself, which is why slotting the board is such a big help. It really stabilises the breakdown point. It also avoid the carbon track build up when breakdowns are happening, although narrow slots aren't always easy to achieve. If the board assembly process doesn't include a really rigorous surface cleaning step, slotting increases its usefulness further.

[coppice]

Most equipment is sold with a spec that says its for use in humidities up to 70% or maybe 80%. If your equipment won't actually work up to 98% you are effectively saying "not for sale in Delhi, Mumbai, Singapore, Hong Kong, Houston, .............".

[Smokey]

Some environmental specs are such a joke.  My favorite ones are the "Max Altitude" spec, which is usually around 2000meters, or 6500 feet.

6300 feet.... ok....
6400 feet.... ok....
6500 feet.... ??.....

Silly.

[T3sl4co1l]

Some environmental specs are such a joke.  My favorite ones are the "Max Altitude" spec, which is usually around 2000meters, or 6500 feet.

6300 feet.... ok....
6400 feet.... ok....
6500 feet.... ??.....

Silly.

That's probably more like,

6500 feet, 1% chance of breakdown
8500 feet = 10%
10000 feet = 50%
18000 feet = 90%
etc.

Tim

[coppice]

i went to take a look at GDT

http://www.farnell.com/datasheets/1640723.pdf

lets say this device, the 2045-07, 75v DC or 500v AC

i dont quite understand the specification when they put it as V/s (for DC) and V/us (for AC). i can understand the physics that a gap is a gap and if designed for 75V it will spark over (im sure there is some special gas mix inside that facilitates easier sparking at lower voltages, but does the speed of rise of voltage affect ability to spark over?), so how do you use the V/s specs to size for a design? any input appreciated TYVM

I am not sure of all the factors which come into play, but arcs take time to form. If you hit a discharge tube with a high but very narrow pulse it doesn't discharge at all.

[Wilksey]

Interesting video!
I found it particularly useful having the PCB demonstration, you should do that more often!

I would quite like to see a video on various input protection methods, I tend to use TVS diodes on mine, but it's always good to see what other people are doing and how they are achieving it.

[T3sl4co1l]

i went to take a look at GDT

http://www.farnell.com/datasheets/1640723.pdf

lets say this device, the 2045-07, 75v DC or 500v AC

i dont quite understand the specification when they put it as V/s (for DC) and V/us (for AC). i can understand the physics that a gap is a gap and if designed for 75V it will spark over (im sure there is some special gas mix inside that facilitates easier sparking at lower voltages, but does the speed of rise of voltage affect ability to spark over?), so how do you use the V/s specs to size for a design? any input appreciated TYVM (or maybe not, its just a measurement thing?)

A spark is a very stochastic (= random...because big words are big) process, where the probability of discharge increases with applied voltage and duration.

You can think of the mechanism like this:

Normally, there's a very small amount of free charge (and thus conductivity) available in the gas.  Usually due to radiation: cosmic rays, background radiation, radioactive isotopes in the construction materials, and radioisotopes intentionally added to the electrode coatings.

Now, free charges will move about, and be accelerated by electric field.  If they gain enough energy to ionize some gas atoms, they'll probably strike one, which frees more charges, and so on.  In a weak electric field, not much energy is gained between collisions, so the conductivity doesn't increase much.

In a stronger field, more and more cascades will occur, and you get avalanche breakdown.  Assuming enough collisions occurred, and there were enough charges around to start the event.

So, for an arbitrarily low voltage, you will always get discharge events, they'll just be imperceptible as leakage current.  It's only above a certain point that you get enough average current (and big enough charge dumps) that it's a problem (what we call a spark or arc discharge).  Conversely, very sudden voltages might pass unimpeded until, after some nanoseconds, enough cascades have occurred to build a spark proper.

So, it's constantly a very random thing, and doing some averages you can come up with characteristics.  There's no operational guarantee that it should work at all at almost any voltage -- it's just that it's 99.9999% likely to discharge if you put, say, a thousand volts across it.  And accordingly, don't expect it to break down at e.g. 75.0V -- you'll only be able to use a 75V device at maybe 50V nominal, and still have to protect your circuit against rapid pulses in the 500V+, couple ns range.  Fortunately, that does mean beefy 8/20 or 10/1000 surges are truncated to ESD-level energy, which is a *LOT* easier to deal with!

Tim

[EEVblog]

The variation in air pressure, even at the top of Everest, isn't huge. Try the effect of a seaside town in the especially humid season.  The 3000 in the formula shown is the usual 3000V/mm breakdown figure for dry air at STP. Damp air could divide that by several. If the board is of any significant age, and has started to accumulate a film of pollutants from the air, moisture absorbed into that film will greatly reduce the breakdown across the board surface itself

I totally agree.
There are two points though:
a) Where is a researched formula for including such humidity? (that's why I didn't include it)
b) In this case of PCB sparkgaps, a reduction in the arc voltage is generally better, not worse. So you could argue that it's maybe not a good thing to include humidity in the calculation, because it is reducing the worst case figure.

[coppice]

The variation in air pressure, even at the top of Everest, isn't huge. Try the effect of a seaside town in the especially humid season.  The 3000 in the formula shown is the usual 3000V/mm breakdown figure for dry air at STP. Damp air could divide that by several. If the board is of any significant age, and has started to accumulate a film of pollutants from the air, moisture absorbed into that film will greatly reduce the breakdown across the board surface itself

I totally agree.
There are two points though:
a) Where is a researched formula for including such humidity? (that's why I didn't include it)
b) In this case of PCB sparkgaps, a reduction in the arc voltage is generally better, not worse. So you could argue that it's maybe not a good thing to include humidity in the calculation, because it is reducing the worst case figure.

You need to have a realistic figure for the gap, or it will be discharging at all the wrong times when the humidity is high. Ever tried selling to Delhi or Hong Kong? :-)

Your practical minimum breakdown voltage might be quite high. For example, how will your product get through the fast pulse transient tests needed for IEC 61000 compliance? If you allow the 4.5kV pulses to arc across your spark gap it will probably erode quite badly before the tests are complete. You probably only want your spark gap to catch the really extreme conditions, and let something better defined protect you from the IEC test cases.

[RobertoLG]

a) Where is a researched formula for including such humidity? (that's why I didn't include it)
b) In this case of PCB sparkgaps, a reduction in the arc voltage is generally better, not worse. So you could argue that it's maybe not a good thing to include humidity in the calculation, because it is reducing the worst case figure.
[/quote]

Clear, thanks 

[EEVblog]

You probably only want your spark gap to catch the really extreme conditions, and let something better defined protect you from the IEC test cases.

Of course. When did I suggest that this spark gap approach should in any way be used for any sort of defined protection scenario? I said the opposite several times I thought.
As I said, this a very poor-mans protection device.

[coppice]

As I said, this a very poor-mans protection device.

You make it sound like a technique that would only be used in lousy equipment, which nobody really cares too much about. Its actually a mainstream way of achieving a number of protection goals in equipment of all quality and price levels. Any equipment from a reputable maker needs to pass at least IEC61000 (and frequently something way beyond that in Australia, because of its desert) and needs to apply its protection in an appropriate manner to do that. Spark gaps are a really key part of the overall protection mix.

[EEVblog]

You make it sound like a technique that would only be used in lousy equipment

That was (clearly I thought) not my intention.

Any equipment from a reputable maker needs to pass at least IEC61000 (and frequently something way beyond that in Australia, because of its desert) and needs to apply its protection in an appropriate manner to do that. Spark gaps are a really key part of the overall protection mix.

Sure, in which case you use a specified and repeatable GDT.

[SeanB]

Went hunting yesterday for a spark gap on a pcb, but only found a CRT base panel out of a dead monitor. took it apart and opened it up to show the tiny gap and tested it with a little voltage.



Then tried the rest of the board. It did not work afterwards. As I was at work I was not going to go and let out too much smoke, there would have been complaints.

[coppice]

The expression 3000pd + 1350 is quoted very often, but does anyone know where the 1350 is derived from? The expression is obviously a rather rough and ready one, as it suggests a zero width gap will not break down until 1350V. I have some evidence which appears to contradict this.

[StefanHamminga]

Dave, very interesting video! I noticed your example recommending the rounded trace sticking in the ground fill, but isn't the triangular design better, because a sharp edge promotes spark forming? It would erode quicker, sure, but the initial protection would be better.

[coppice]

Dave, very interesting video! I noticed your example recommending the rounded trace sticking in the ground fill, but isn't the triangular design better, because a sharp edge promotes spark forming? It would erode quicker, sure, but the initial protection would be better.

Dave covered this. The point works great the first time, but over the life of the product its performance is highly inconsistent as it wears. The rounded end traces with ground fill around them will be far more consistent. If you look around a lot of people show a similar idea, but they do one key thing in a poorer way than Dave. The use a square ended trace surrounded by ground fill. The corners of that squared off end will not give the long term consistent behaviour that Dave's rounded end, with its resulting even spacing and lack of pointy bits, will.

I guess if you only need your product to survive once, because its something disposable, a pointy approach might be superior.

[alter Ratz]

I liked this video very much, since input protection is something which is IMHO very important, yet it is difficult to find good literature about it. Just recently I had to protect an input using a MOV+PTC and it took ages to find information about part selections, and even then I was not 100% sure that I have done it right (but I must have since it passed TÜV).

So a video about the basics of input protection (methods, part selection, combination + pros/cons of methods) would be very welcome.

best regards,
Bernhard