PCB reliability and cost single vs double sided

[SPK627]

I'm working on a new circuit board layout that has so far gone through 3 revisions and a fourth revision is required. The main problem with the layout has been due to ground trace placement. We are forced to use single sided layouts because of the cost, but I am trying to convince my boss that we need to use a 2 sided layout if we really want to fix the problem.
I have found what looks like a possible fix for the ground problem. It involves re-routing the ground trace. If I had a 2 sided layout the change would be fairly simple to implement.
My boss insists that we stick with the single sided layout but we'll need to place additional zero ohm jumpers. We are already up to 8 jumpers and will probably need 4 or 5 more to make the new changes.
The cost increase for going to a 2-sided FR-4 board versus a single-sided CEM-1 board is about 50 cents more per board which is an increase of about 3-4% of the total circuit cost.
My question is if anybody could quantify the reliability differences between zero ohm SMT resistors and plated through-hole vias. I'm also asking one our manufacturing engineers if there's a process cost penalty for having to mount the extra components. The increased component count also should decrease reliability since there is a better chance of problems such as jumpers not soldering properly, etc.
I'm trying in vain to find ways to justify changing the board to a 2-sided layout, but it seems management is too tight-fisted to do the right thing. Nobody seems to factor in that we have already gone 2 months over schedule re-spinning 3 revisions of a single sided board! ARGH!

 

[VEBill]

Sounds like your boss will learn the hard way that cheaper isn't always less expensive. Be careful about who gets blamed for the ground problems and endless revisions.

Good luck.

 

[Skogsgurra]

What?

Can't believe it. Is this for one of those toys that McDonalds put in their Happy Meals?

How do you at all get a decent ground plane on a single-sided board?

One question that pops up is: What router do you use? A good router shouldn't make three revisions necessary. It should tell you that the job cannot be done. If that is the case.

Gunnar Englund

http://www.gke.org

 

[SPK627]

Gunnar,

I wish this were for a happy meal toy! This circuit is for a xenon strobe with a piezoelectric horn. The grounding problem is aggrevated by the discharge of the xenon flash tube. Basically this circuit is a DC/DC boost converter with a 16 kHz switching frequency. We don't have a ground plane, only a ground trace.
We actually have 3 other versions of the circuit that work just fine on single sided boards, but this new version is a repackaged version that requires that some key components get moved around.
The fix that I want to implement involves re-routing the ground trace to change the direction of the current flow. I am fairly confident that this can be best achieved with the two-sided board layout. We might be able to do it on the single-sided board, but it will require additional zero-ohm jumpers.

What I really need here is a way to quantify the advantage of using a two-sided board to justify the increased material cost.

 

[itsmoked]

Generally you would make a table.

Estimate the size of the board; two sided and one sided.

Generate the different BOMs (Give your best guess on those jumpers)

Get the quote on the price of the 2 sided board verse 1 sided.

You can even do a brief failure rate of all the added jumpers verse the standard (existing) board.

Then give the layout time estimates of each method and
the cost difference.

Then put down the advantages/disadvantages of both methods.

All the cost differences too.

Look at this. It may show that single sided is still the most okay-est choice. It may provide a clear (tabular) argument that the mngmnt can actually understand for moving into 2 sided land.

A lot of times mngmnt needs paper work to change its mind. As this shows that it wasn't just some knee jerk that led to a change but that all these "things" were considered.

 

[melone]

It sounds like 1 sided vs. 2 sided isn't the problem..... You need to throw some layers at this problem! You NEED a ground plane!!!

 

[Skogsgurra]

About costs. A jumper or any low-cost component costs at least ten cents to buy, handle, install, solder and inspect. No exact cost can be given here, but your manufacturer should know.

So, if the cost for a double sided board increases by 50 cents, then your 8 jumpers have already cost you more than a two-layer board would cost. The double board will usually get you plated-through vias and that usually means better reliability. A xenon flash with a horn - hmm, sounds like such a device will take some beating. A double sided board is what you need. No doubt about it.



Gunnar Englund

http://www.gke.org

 

[MacGyverS2000]

If the jumpers are simply zero-ohm resistors, figure $0.02 per jumper for a mass-produced board. If it's longer wires, or worse yet wires with bends in them, price can jump significantly. Unless you're using cheapy fiberboard, 2-layer FR-4 boards are marginally more expensive than single layer (in fact, some places will charge a premium for single-layer boards... don't ask me why, since they use the same damn board). Large vias, like 20+ mils, should be roughly the same cost as a zero-ohm resistor being installed, but will generally save you money because now you don't have to buy the resistor, either.

I would say it's 6 of one, half dozen of the other... cost wise it should be about equal, but the 2-layer will give you a better product.


Dan
Owner

http://www.Hi-TecDesigns.com

 

[itsmoked]

macgyvers you are living in the wrong world. Everything you and I build fits in your and my world but in the single sided world you are usually talking China... a different planet! Yes, I bet he's using fiber board. I bet his quantities are expected, or are, very LARGE. Two sided boards bring a second photo resist, exposure, development, solder resist, drilling process, etc, etc, beat dead horse. It really does add cost to an order of a million units or even 100k pieces.

Now if SPK627 is only building 2,000 of something in the USA they truely should be using 2 sided and stop screwing around.

 

[felixc]

TVs and VCR are still built with sigle-sided punched phenolic boards, with wires, and very thin copper prints. They're also full of ECNs and board revisions. It's a matter of volume vs cost, also a matter of design. Here there are problems with ground noise. Is the xenon discharge capacitor properly prefiltered, and is the current path clearly defined?

A zero-ohm SMT resistor is more like a 50 milliohms resistor than a piece of wire. If the current flows are high they must be taken into account.

 

[walker1]

Sending high current pulses through jumpers will create noise!
If nothing else, a jumper vill have at least two bends, making you a coil, which again will give you an EMF.

Spiked EMF can be a real nuisance in a circuit.

Traces on a PCB carrying currents should in general be kept short and direct. Instead let the signal traces make the jumps.

Then again, the cross-talk (at least capacitive wise) between two wires running next to each other on the same side of a PCB is usually smaller, than if they were placed opposite.

 

[VEBill]

World's shortest EMC summary: Make your current paths look more like transmission line pairs and less like antenna loops.

 

[SPK627]

Thanks to everyone for your responses. What I've decided to do is to do two layouts - one single-sided and the other double-sided. That way we'll be able to see which, if either, of the layouts will work. If they both work, then I guess we'll just stick with the single-sided approach, but if only the double-sided board works then there should be enough justification to continue on with the project. I should know in a couple of weeks what happens. I'll try to post an update.

 

[Noway2]

It does sound like you need a ground plane, as was indicated in a few of the earlier replies. One technique I have used is to put a copper pour on the board tied to ground. The traces are then ploughed through it, though you need to be sure to leave enough clearance. This approach may not always work and in some instances can make noise issues worse.

A solid ground plane, even one full of holes, is infinitely better than a serpentine ground path.

 

[itsmoked]

Noway2 I would guess their already trying to cram 10 pounds onto a 5 pound board making "pours" impossible or so island-ized as to be ineffective.