Multi layer PCB (more than 2)

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

bluebird

Well-known member
Joined
Jun 11, 2004
Messages
1,070
Location
Los Angeles
Just wondering how many people here use more than 2 layers. Audio usually fits in a 1 or 2U box without having to downsize too much, we're not building cell phones here. But often I lay out all my parts, and then realize I've cramped everything to close and either have to enlarge the board, or thin out my traces more than I would like.  I'm not yet comfortable with SMD enough to experiment and prototype with it yet. I also find it boring and a bummer to rout power supply traces. They always seem to get in the way of my orderly logical signals...

So I've been thinking about 4 layers and if that makes a big difference for people. Anyone "tried it and never going back"?
 
I'd be curious to hear about this as well. During mixer channel assemblies I opted to connect critical signal runs with free wire to avoid power traces. Of course it came at an enormous cost of labor and time.
 
I have done 4 layers. It can be a great space saving tool and a design marvel but:

Layout is imperative and so is watching your ground plane. I have made the mistake of having to scarp a whole design or add surface jumpers because of missing what is called islands and surrounding a ground plane with a trace thus of course resulting in an open circuit.

There is a possibility of also increasing the parasitic capacitance and inductance issues in both digital and analog designs.

Failures can result in total PCB scrapping as it becomes really hard if not impossible to fix a blown trace and a catastrophic injury can take out multiple circuit paths. All of which is obvious if you think about it, more is more!

https://resources.pcb.cadence.com/blog/2019-how-parasitic-capacitance-and-inductance-affect-your-signals

https://server.ibfriedrich.com/wiki/ibfwikien/index.php?title=Remove_groundplane_islands#:~:text=Islands%20in%20ground%20planes%2C%20orphan,in%20respect%20to%20electromagnetic%20compatibility.
 
I only worked on one project with a 4 layer PCB  (a dsp with a bunch going on) and it was easier to route with the extra layers, but a little harder to debug if traces are hidden.

JR


 
I've done multilayer RF and mixed signal,  8+ layers is not uncommon,  but it's more a necessity  for some things.

With audio so far it's been 2 only.  It's generally feasible with good pre-planning,  and you can always use zero ohm jumpers if you back yourself into a corner.

I agree with what others have said.  4 layers add cost, complexity,  and debug issues to consider.  But if your board is really dense and you need it,  then go for it.
 
Let me bring up another thing, via's. Somewhere along the line, I was made to believe that they should be avoided. I've never had a problem with a via not connecting in any PCB I've created with the current company I use. I feel they are large enough to at least have as much copper as the average trace I use. That's just a rational guess though.

If I was to let go of this superstition, of them being bad, I could route a lot more direct. Or at least jump around top to bottom with power and ground traces without stressing about it.

Is there anything wrong with a 2 sided PCB having lots of via's if the manufacturing is done well? The only draw back I can think of is creating little spots of capacitance everywhere, but if aware of it, could be mitigated.

Thoughts?

P.S.
I might have posed this question before... Suppose a second time around would still be educational for people.
 
I have only done more than 2 layers if there is some mixed signals, high frequency stuff needs ground as close as possible.

 
Somewhere along the line, I was made to believe that [vias] should be avoided
I recall asking about it in the drawing board a few years back, and the general reply for 2-layer was to use as many as constraints dictate.

Found it: https://groupdiy.com/index.php?topic=68629.0

I did read IBM recommends wicking them with solder, but I've never had an unwetted via fail (fingers crossed).

Edit: Extra rules likely apply for multilayer, blind vias, etc.

By the way this quote from Zam reminded me why I stopped caring about the via and learned to love the BOM:
but when you use SMD you have no choices except via,
So, think of all the working PCB's out in the world where they needed to place a component on a pad with traces on the opposite side.
Via's in the zillions.


 
I never ever had a via fail me with my current suppliers, I'm not worrying about it

Gustav went full-in on four-layer some years ago, and never looked back. He says it helps audio layout tremendously to have a ground (and a power) plane, and makes it possible to do more aesthetic decisions..

/Jakob E.
 
I wouldn't distrust modern vias,  PCB technology seems to get better every year. Some modern SMD ICs have heatsinks on a bottom pad, and vias are an effective way to move heat from top mounted IC to a larger heat spreading pad on the bottom (at least it worked for me, but I didn't do measurements, and the tiny class D audio chip probably doesn't make that much heat).

I am a fan of KISS but 4 layers, or more, provides the capability to add extra shielding for sensitive internal runs. Keep in mind that you are potentially adding capacitance and other unintended variables that may not be easy to tweak in development.

JR
 
Keep in mind a plated through hole is essentially a large via that a component leg happens to pass through and is free from solder mask.  So there's nothing to be concerned with about using them.

I've yet to come across a layout (at least in audio) that dictated 4 layers.  Nowadays the cost adder has shrunk to essentially nothing (especially in volume) so I wouldn't be opposed to trying it out.
 
bluebird said:
Let me bring up another thing, via's. Somewhere along the line, I was made to believe that they should be avoided. 
That is wisdom of the past. Due to the smaller diameter of vias, electroless plating (the phase before electro plating, that could be considered as the "seed") bid not give a good yield, so there was significant reject. QC was done "manually", which would be very time-consuming. It was common to have quotations based on the number of holes and vias.
Progress in chemistry, process control and automated QC have made that obsolete. Continuity and isolation control of PCB's is done in seconds now.
BTW, I"ve never had to design an analog board for which I regretted not having more than 2 layers.
The only 4 layer board I had to collaborate was a digital GEQ.
 
Continuity and isolation control of PCB's is done in seconds now

Oh yeah! fwiw (aside) watched a modern chinese fab house walkthrough and they had a high speed 'e-testing' cam machine poking and zooming around way faster than humanly possible. It was neat. That and the optical cam for pick and place QC, which this house considered to be a bit of a trade secret.  The automation and size of these places are shocking.

E-Testing (flying probes) https://youtu.be/ljOoGyCso8s?t=1359

 
Hello Bluebird,

More than 2-layers....... well, it depends on what the end-target is.

Over the past 15 years or so, most of my designs have been 4-layer, with a handfull of 6-layer boards and a couple of noteworthy 8-layer boards, and yes, they are all used in high-end analog audio.... no digital clocks :)
I do have the advantage of having in-house pick & place, so most designs are SMD, however, putting that aside for the moment, the advantage of multilayer are many if you seeking high-performance in regard to crosstalk control and power distribution.

Don't be scared of multilayer; the advantages are many.

This particular project required 4 and 6 layer PCBs:
http://www.cmlaudio.com/nm1249-v-mini-master-sidecar-control-panel/
The reasons were:
a) to achieve a particular set of facilities in a pre-defined physical space.
b) to achieve high levels of performance (mimimal crosstalk) between a number of signal paths that are in close proximity.
Both requirements were met and would not have been possible with double sided PCBs.

There are always traps that can be fallen into; others contributing to the thread have noted some of them such as parasitic capacitance. This is more of a problem where circuitry is at medium to high impedance (DI input stages, equalizers, signal processing circuitry) and less of a problem in amplifier stage outputs and switching matrix designs, however, an awareness of component and trace proximities, signal levels, impedances and their effects on each other will go a long way to designing-out the "banana skins" that can cause you to slip-up and create a less than optimal design. I have been doing PCB design for over 40 years and still slip-up at times..... usually when trying to get a design done in a hurry, or squeezing in that accidentally omitted connection.

A decent quality CAD package is also a high priority.  You don't need to go as high-end as Altium (you are very fortunate if you can) but it is not a necessity. What is a necessity is that the CAD can really do multilayer well and that you can control the copper-flood / copper-pour / power-plane / ground-plane parameters.  A rock-solid forward-annotation and back-annotation ability is another essential facility. Also, totally trustable dimensional error checking that can be set for custom trace-trace, trace-pad and pad-pad gap checks is an essential. CAD is a really personal choice, my own preference being Seetrax Designer XL. Trust and reliance on your CAD to tie the schematic to the PCB and always keep the two in snyc are essential. JR noted that inner-layer traces are impossible to debug and you have to know that the CAD has done what you have expected it to do.

One really obvious advantage of hiding traces on the inner layers is that it is really hard to reverse engineer a design. :)
Multilayer is a good way to add a level of protection to your design IP.

Always, and I really do mean ALWAYS use thermal relief on any hole that connects to a solid copper area. If you dont, you will break plating on through-holes when trying to de-solder a component. Also, by making holes bigger than necessary, you can often suck out solder around all of a component leg, therefore reducing hole plating damage when removing components.

Where multilayer really scores in analog audio is the ground-plane. The number of designs I have seen over the years where performance has been trashed or compromised by poor ground connections (usually scrawny-thin traces) is depressingly high. A multilayer with a groundplane would have solved many problems.

I have never needed or wanted to put blind or buried vias into a design. Too expensive, special manufacturing processes required, specific layer pairs required, definitely need to discuss the manufacturing parameters and requirements with the PCB manufacturers...... in short: blind & buried vias - just say no!
Through-vias are not a problem at all. Matador describes them perfectly.

Choosing what to do with your layer stack is something that some designers fret over and over-analyse.
This, in overview, is what works for me as a 4-layer stack and has been proven to be highly successful:
Layer 1 (component side): inter-component signal traces & any other short point-to-point traces (decoupling caps to ICs).
Layer 2: Groundplane.... all of it.... no breaks.
Layer 3: Signals & power, mainly signals.  Flood open spaces with power or ground.
Layer 4: Signals & power. Ground flood of open spaces.

Using L3 & L4 for power allows wide low-impedance power traces. Use of multiple parallel vias to link power traces between layers is a neat trick to adopt. The same is true when linking the L2 groundplane to other ground-flooded areas of other layers.

In a 6-layer board, my layer assignment is typically:
L1: Top components & interlinking signals & short point-to-point traces.
L2: Groundplane
L3: Signals & power & screening ground
L4: Groundplane
L5: Signals & power & screening ground
L6: Bottom components, Signals & power & screening ground

An 8-layer board is initially as-per the 6-layer, then,
L6: Groundplane
L7: Signals & power & screening ground
L8: Bottom components, Signals & power & screening ground

Some design-rules for adjacent layers with signals on them include:
Avoid trace-crossings, except at 90 degrees,
Never run traces parallel above each other,
Use copper floods for screening of adjacent layer signals.

Most of the PCBs in this console design are 4-layer, with 6 & 8 layer where needed.
https://www.soundtechniques.com/our-gear-test/
The heart of the console is the centre-section card cage. This has 26 plug-in cards (all multilayer) that deal with balanced inputs, balanced outputs, mix amps, console logic, monitor selection, group signal paths, etc, etc. The backplane system for the card cage comprises of a pair of 8-layer boards, one deals with inter-card connection in the card-cage and the other handles the cabling to the centre section modules, the console channel buckets and the console external connections.

Cheers,
 
Great post Gareth.  As I said some time ago, you did a magnificent job on the Sound Techniques desk,  and it's fantastic to get your insights on here.

Stay well
 
Winston O'Boogie said:
Great post Gareth.  As I said some time ago, you did a magnificent job on the Sound Techniques desk,  and it's fantastic to get your insights on here.

Stay well


Thank you  for your kind words; greatly appreciated :)
There are other guys involved in the design work, but the PCBs are all down to me with peer review by colleagues.
 
Wow! Gareth, thank you so much for an in depth answer. This will benefit many people. All contributions will, I'm sure.

Regarding the CAD program, I use Eagle. I have the old version before they moved to the subscription based model. I have considered buying into that but after playing with the newer demo version, decided for what I do, I don't need the new features. There have been many threads here about CAD programs so we need not go there.

The debugging problem is the biggest downside, as many pointed out, and probably why I will stick with 2 layer for now as most of the projects I do now go very slowly, and rarely turn into a final product that will be duplicated.

This should probably be in its own thread, but I will bring it up anyhow...

boji! great video, that guy needs to chill on the Red Bull, Lol...

Pick and place service, my current PCB company only offers a template for solder paste. Can anyone explain quickly how it works when you get SMD components soldered to the final board? Do you submit a BOM and the company has standard parts already? Do you send an order of parts to the PCB company? How exactly does that work?

 
Excellent post Gareth!

IMHO, the PCB accommodates the design, not the other way around.  good engineering practice is to start with your limitations:  the board outline, components that must be in specific places (like jacks, pots, through holes for mounting screws, etc).  You can then move around the components until the ratsnest doesn't look so 'ratsnesty', then move components into their trial positions and then attempt to route, scooting things around as necessary to clear paths for traces.  If the routing complexity is high enough that you just can't manage a solution after half-a-dozen tries then (and only then), move to more layers.

Of course, the nice thing about 4 layers is that, assuming you are completely PTH components, that all of the power and ground nets immediately 'disappear' from your ratsnest once you place the pours.  Very often a congested route cleans up considerably after this simple step and reveals the solution.

It's always been a point of pride for me to route a 2-layer board with no 'extra' vias (because again, evey PTH lead is a via), and I'll spend a bit of extra time moving stuff around to get all extra vias removed.  But for audio, there's nothing to be gained by doing this, and done is always better than perfect. :)

bluebird said:
Pick and place service, my current PCB company only offers a template for solder paste. Can anyone explain quickly how it works when you get SMD components soldered to the final board? Do you submit a BOM and the company has standard parts already? Do you send an order of parts to the PCB company? How exactly does that work?
Hard to answer as different companies require different things.  Some just want a list of part numbers you want, and they'll buy for you (for a fee).  Others want the parts sent to them.  Even others will provide a list of parts they have, so if your design accommodates their parts (especially for simple stuff like resistors and bypass caps) then you'll save some cost.

Common stuff most board houses have on hand in every value, and typically they'll source your BOM (or compatible parts) for assembly.  PTH costs a fortune to hand assemble (even in China), so a good compromise is to have the board house do all of the easily-sourced stuff (like 0805 resistors, MLCC bypass caps, SMD inductors, etc), then you just have to add the PTH stuff.
 
Matador said:
IMHO, the PCB accommodates the design, not the other way around.  good engineering practice is to start with your limitations:  the board outline, components that must be in specific places (like jacks, pots, through holes for mounting screws, etc).  You can then move around the components until the ratsnest doesn't look so 'ratsnesty', then move components into their trial positions and then attempt to route, scooting things around as necessary to clear paths for traces. 

Yes, this is loosely my process as well. I was just doing a particularly large project, that I kept adding features too... "Oh wouldn't be cool if the EQ had a mid/side function?" So I was starting to "jam" parts into the same space....


Matador said:
Hard to answer as different companies require different things.  Some just want a list of part numbers you want, and they'll buy for you (for a fee).  Others want the parts sent to them.  Even others will provide a list of parts they have, so if your design accommodates their parts (especially for simple stuff like resistors and bypass caps) then you'll save some cost.

Common stuff most board houses have on hand in every value, and typically they'll source your BOM (or compatible parts) for assembly.  PTH costs a fortune to hand assemble (even in China), so a good compromise is to have the board house do all of the easily-sourced stuff (like 0805 resistors, MLCC bypass caps, SMD inductors, etc), then you just have to add the PTH stuff.

I will do my homework, thanks!
 

Latest posts

Back
Top