GND plane vs. ESD protection choice

[Chancy99]

Hi -

I'm in the final stages of the Datalogger design for cars. I had originally wanted to include clamping diodes on all the analog inputs, using the PACDN017 smd array from CalMicro.

http://www.calmicro.com/products/data/pdf/pac_dn017.pdf

After much pushing and shoving in the PCB layout, there is room, finally. Unfortunately, it would only be on the bottom layer - simply no room on the top layer.

Given that all the other components are SMD, the GND plane below the analog section is nice and almost unbroken. Good for analog signals If I add in this ESD array on the bottom, it chops that GND plane up but good.

So the question is, which is more important for a vehicle application ? Nice smooth GND plane, or ESD clamping diodes for each input ?

Dean.

 

[melone]

You want the ESD clamps as close to the source as possible, i.e. at your connector. Actually, you want caps there as well. Therefore, you should not be mucking up your ground plane too much if you place your components there. Also, you might want to shy away from using an IC, since that takes up a large surface area. Try individual surface mount caps(100pF), and 3 terminal diode clamps (look in Digikey) located immediately following the connector. Sure you might take up more board real-estate, but it should be easier to fit on your board.

If your ground plane gets segmented up, esp under your micro, you are asking for trouble. Ground return current is probably the most neglected current. Allow for the current in the ground to follow the same path as the supply current. Often, this is exactly why segmented ground planes (good in theory but VERY difficult in practice) fail. Digital should be seperate from analog, but IC's often contain both digital and analog. What ground do you connect too?

Anyway, take a look at increasing the number of components, to help alleviate board real-estate problems, and keep your ground plane solid! If your ground starts bouncing or sits at a higher potential (.25, .3, .35V, etc.) you are robbing the voltage from your circuit, and asking for trouble. If your IC is designed to run at 5V, would you want to supply it at 4.75V?

Good luck and keep us posted on your progress! (Great project!!!)

 

[Lewish]

melone, very well said.

Chancy, how many of these are you planning to build? If it is a 100 or more, I can give you a source for 4 layers boards, that will cost about the same, or less, than 2 layers boards built in the U.S. That way you can have a solid ground plane, except for vias, under all your parts.

 

[Chancy99]

Melone :

Thanks for the quick information. See below for a slightly older version of the layout. The connector and analog section are still the same. Look to the left side of the board.

http://www.avrfreaks.net/phorum/download.php?f=3&file=MegaLogger-pcb-V5b.pdf

The thread is at

http://www.avrfreaks.net/phorum/read.php?f=3&i=23140&t=23140#23140

Note the resistor arrays just inside of the connector, they feed into the dg333a muxes. The pacdn017 ESD array chip would go right under these on the bottom layer. The traces would go directly to the input pins on the connector.

There are a couple of bottom layer traces near there, but not much. When I do the copper pour, it fills in nicely. With that chip in place though, that whole section gets chopped up.

The analog and digital GND planes split right under the Mega128 cpu. The cpu is well covered by GND plane except where the decoupling caps are.

CalMicro also makes those arrays in a couple other sizes. The pacdn-016 is a 6 channel array, and they also have a 2 channel one. They will fit there as well, but it doesn't really help the chopping up of that bottom layer much. The single pacdn-17 is way easier ...

Dean.

 

[Chancy99]

Lewish :

Cool ! Thanks for the offer. I really have no idea how many of these will get built.

It was originally a project for myself and a few friends. I've had lots of request for it though, and can easily see a good 20-30. How much would that count be for the 4-layer ?

Dean.

 

[melone]

2 suggestions:

1) Completely breadboard up your circuit and verify full functionality (esp. in vehicle). Nothing worse than finding a stupid logic error after you have spent several hundred (or possibly thousand) dollars. Might be a pain in the butt, but it will only help you! You can use the breadboarded version to develop your software, algorithms, timing verification, power assumptions, spike survival, anticipated waveform verification, etc.. Also, with a breadboarded version, you are not restricted by the number of components or size. Make it big and clunky, but put everything in it, even the kitchen sink! Put in headers to measure hard to get to signals, power, clocks, etc.. Once again, lots of work that can save you tons of time, money, and frustration after you get your PCB's.

2) Go 4 layers!!! Why kill yourself trying to fit it into 2 layers, when for a little bit more, you can have a solid power and ground plane? Especially if you have a fully functional breadboarded solution, then you can easily justify the couple of extra bucks for the 4 layer. Don't get fancy with trace size, min. 8 mil for digital signals, and min. 15 for power (obviously the bigger the better). Don't get crazy with vias either! Add multiple vias on power or high energy traces (via are inductive & resistive and putting them in parrallel decreases the inductance/resistance). Also, opt for the 100% electrical test. Nothing is more annoying that troubleshooting a "bad circuit" and finding out there is a short / open in the PCB.

Congratulations, it sounds like you are well on your way to completing a pretty interesting project!

 

[carnage]

a few things
In response to the question itsself I say the esd protection is more important.

For noise cancelation it is ok to cut up the ground plane as long as it is done right. The key is to have the ground plane look like a net with holes and lots of interlocking as opposed to fingers that just radiate away from the plane. That said it is still nicer to keep the ground plane as solid as possible.

I am assuming that you will mount this in an enclosure that will also help with noise cancelation.
BTW Nice Design, I can't wait to hear the results

 

[Chancy99]

Thanks guys. This has been waaay more work than I ever anticpated Then again, it's done an awful lot of feature creep and evolving/mutating ... I appreaciate the comments, I really do.

I'm a breadboard manic. There are always several on the desk here, with various "ooh, let me try this" things on them. Unfortunately, this particular project is using quite a bit more SMD devices than I normally (if ever) play with. So breadboarding it is a little tough.

I am using the Atmel STK500/501 kits, and interfacing with all the thru-hole stuff now. That's fine and dandy, doing one bit at a time. The analog inputs I'll do with a few generic opamps (lm386 or something) and discretes. Use a 4016 or something as a switch instead of the DG333a. Proof of concept type deal.

So for some stuff, the pcb will be the breadboard The more eyes that spot glitches/gotchas before that happens, the better.

As for the GND plane, the ESD array chip will be way over on the left, right next to the 44pin connector. So it's really only over there just at signal entry that the plane will be broken to shreds. The rest of it should be OK. I think I'll keep it in.

It would be great to go with a 4layer board. But this is a project, not (yet anyway) a product So cost is definitely an issue. It was originally for myself and a couple of others. Though there has been enough interest to indicate it may later be worth doing a run of boards and making up kits, or perhaps building them myself. We'll see how it goes.

For 2layer boards, Olimex will do $26/ea, 4pcb will do $33/ea, and e-teknet is $25/ea. 4layer is basically double all that. I'm already at around $180 in parts alone, not including pcb ...

Oh, the enclosure is a Hammond 1455k1601 extruded alum ...

http://www.hammondmfg.com/1455.htm

Dean.

 

[melone]

Be careful, diode are not the good ESD protection devices, they are good clamping devices. ESD is high energy for a VERY short duration. Unfortunately, diodes take some time to turn on, and cannot be counted on for ESD protection. Even if they do turn on, you will be dumping a lot of energy into your supply at a high rate. I don't think that you want your micro too see a sudden power surge during operation.

An ESD cap, will shunt the energy to ground, and only raise the ground voltage. Given the choice of exceeding the voltage ratings on a component, and not meeting the ratings, I would much rather save my components (and money), and cause the component to potentially stop operations due to insufficient power.

Anyway, I would highly recommend putting in the caps instead of the diodes. I design enginer controllers for one of the big 3 automakers, and am familiar with what ESD protection is included on engine controllers (VERY HARSH ENVIORNMENT). Plus, these caps should be cheaper than the diode array.

Did you get the SMD->Thru-pin sockets for all of your components? Some of them might be difficult to find, but let us know where you get stuck. I know for a fact that there are sockets that accept BGAs, PGAs, Gullwing, J-lead, etc..

Finally, why the aluminum case? Isn't it kinda expensive, when a sturdy plastic one will sufice? Are you thermally tying the PCB to the case for heat sinking?

 

[Chancy99]

melone :

I figured that given that these (the CalMicro arrays) are touted as ESD protection, the switchon time would be very quick. It's not noted in the datasheet though. Based on what you say, I think the caps are the way to go then, particularly since I have to make a choice

Can't fit both on the pcb - not enough room. So it's either the diode array or caps. Digikey sells a nice Panasonic capacitor array ...

http://www.digikey.com/scripts/us/dksus.dll?Detail?Ref=63630&Row=240537

They're 220pF, and two of those 6.4x3.1mm chips would fit nicely. Would 220pF be enough here, or do we have to go with individual caps ?

I have a couple of SMD prototyping mini-boards, but not yet for everything. Initial breadboard tests will be with thru-hole equivalents, just to validate things. Then I'll use the real things.

I like the Hammond enclosure Also, the extruded design with the mounting rails and end caps make it easy to handle the connectors etc. on each end of the board. I had originally started with a typical plastic project box, but getting the board down in there with the connectors protruding was problematical.

Dean.

 

[melone]

ESD is a very tricky problem. The amount of energy stored in a pulse, you can jump tracks on a PCB. Therefore, it is ESSENTIAL that your caps be located as close as possible to the connector. Threfore, a cap array might not be the best solution. Investigate individual components (1206, 0805, or even 0603 sizes should easily fit into your design.). Pick voltage rating of at least 100V, preferably around 200 or higher. Also, individual caps allow for the most protection with the least likelihood of indavertant susceptability (already protected trace [after a cap] is run next to a unprotected [before a cap]).

The thing about ESD is that the results of an ESD shock are so unpredictable. You can shock the save device 10 times and not see any immediate circuit degredation, and shock the "same circuit" once and see catastophic failure. But here is the kicker, the worst case scenario is that there is a ESD hit, and there is no apparent failures, just degredation. This means that your circuit will fail at some later time for no apparent reason. This can cause huge amounts of effort in troublshooting the failure because the event that started the failure occured some time earlier.

Anyway, try looking at the individual caps. If those won't work in your layout, stick with the cap array, and locate it as close to the connector as possible. Finally, be meticulous in your routing of signals. Keep traces as far apart as possible, and do ground fills everywhere else. (He is working low enough frequencies that the change is track impedances due to change in surrounding ground islands will not really affect him.)

 

[Chancy99]

Hmmm. I guess that Panasonic cap array won't do it then. It's 25V rated. Since this is going to be hand-built, I don't think I want to go below a 1206.

You can see a jpg of the analog input section, sans the GND plane at :

http://photos.yahoo.com/fnatmed

Just dig down into the My Photos section, and choose FullSize. Red is top layer, blue is bottom layer. The really thin lines are nets still to be completed. You'll see a bunch of them leading up off the top of the board. Up there are the footprints of the ESD array and two of the capacitor arrays.

The ESD chip would go right under the four 180k resistor packs, or the two cap-arrays would go there. Or a whole sea of 1206 size 0.1uF caps ...

Oh, what size caps ? Any particular type ? Thanks for the tips/help melone - I really appreciate it.

Dean.

 

[melone]

1206, 200V ceramic caps from Panasonic should work just fine.

About the help, no problem! This is exactly why this forumn is so useful. You have put in the effort, need help, and are not afraid to ask questions. We need more engineers with that type of work ethic!

 

[Chancy99]

Cool, thanks. I'm not looking forward to hand-soldering an extra 16 SMD caps, let me tell you I'll check tonight to make sure I can fit them all in.

Dang, an array would be much easier. Panasonic makes a 50V 1206-size array of 4 caps. Although rated at 50V, the dielectric will withstand 300% of rated voltage for 1-5 seconds. It actually says :

Class 1: rated voltage x 300%
Class 2: rated voltage x 250%

Would these do ?

http://rocky.digikey.com/WebLib/Panasonic/Web data/ECJ Array Type Series.pdf

Dean.

 

[melone]

You can get away with the X7R caps in the array form. Just double check the crosstalk numbers will be acceptable in your app. Try searching the web for "Human Body Model" for ESD. This will accurately define an ESD pulse that your circuit is likely to encounter.

 

[Chancy99]

OK, thanks again. Any particular cap value to use ? Given that the pulses should be of very short duration, I wouldn't think we would want that big a cap. The largest I can find in the array is 10000pF.

Dean.

 

[melone]

Anywhere from 100pF to .01uF. You will have to crunch the numbers and see what value of capacitance will not affect your circuit too badly.

 

[carnage]

Are you using the aluminum end panel version? that combined with some rf shield tape for the edges (to ensure conductivity between the end plates and the enclosure at all edges) would give you plenty of noise cancelation.
Unfortuneately all of that won't help with the ESD that will no doubt come up your signal lines at times.
Myself I have always prefered diodes to caps but the apps I was dealing with were more geared twards having something left to repair after a catastrofic event, So I would say go with melone's concept it seams to be more appropriate for this application.
(I don't care if the design has protection diodes, it was struck by lightning , I can't fix it)
If you can't tell my experiance is in the service deparment.

 

[Chancy99]

carnage :

Yup - it's version 1 of that enclosure with the alum. endcaps.

http://www.hammondmfg.com/1455.htm

Should be able to keep openings in those caps to a minimum.

Left
- DB25 size for the 44pin - typical metal shield
- Small round for locking Kycon power connector

Right
- USB device socket - typical metal shield
- slot for MMC card
- 2.5mm stereo socket for serial connection

For the caps, I think I'll go with slightly lower values, say 100pF. Want to keep the capacitance loading to a minimum.

*Next* version, with all the enhancements, will be on the 6x4 board, and will have room for everything

Dean.

 

[carnage]

I thought as designs were refined they were supposed to get smaller.