I am new to PCB and electronics design and I just finished designing a PCB that includes POE flyback converter, Ethernet Phy, a li-ion battery charger, and a DC/DC converter. My design is based on the reference design and BOM/datasheet recommendation of all the ICs to be on the safe side. I ended up with about 300 total and 120 unique components. Mostly it is different resistor and capacitor values but also different FETs for the various power switching converters. Assembly service is very expensive due to the high component count. How is PCB Schematic simplification and optimization usually done? Where to start? Are there any best practice tips or resources?
-
\$\begingroup\$ Eliminate unessential components. \$\endgroup\$– Voltage Spike ♦Dec 12, 2019 at 17:22
-
\$\begingroup\$ Can you share your schematic? \$\endgroup\$– The PhotonDec 12, 2019 at 17:36
-
\$\begingroup\$ First you need to thoroughly understand the task each component fulfills... If you think you have alternatives try manually substituting them into the existing board. But beware that just because it works on the bench doesn't mean it will reliably in the field - once was asked to diagnose a consumer product that was going to essentially inevitably blow its non-user-replaceable fuse if not the TRIAC itself as the factory had cost reduced away most of the TRIAC's snubber leaving lots of empty footprints on the board. You could use it anywhere from 1 to 100 times before it failed. \$\endgroup\$– Chris StrattonDec 12, 2019 at 17:47
-
1\$\begingroup\$ You might be interested in How do I make sure my electronics design is good from manufacturing perspective? \$\endgroup\$– The PhotonDec 12, 2019 at 17:50
-
\$\begingroup\$ Here is a link to the schematic: drive.google.com/open?id=1AIQbMqL28CSA07JHKXdx_UPSckQJBKnK \$\endgroup\$– F. HeisenbergDec 12, 2019 at 18:06
2 Answers
In case someone is still having the same question, this video pretty much nailed it:
https://www.youtube.com/watch?v=RwVzLOI4cmA
Summary:
- Try to reduce unique component count as each unique component needs a seperate component reel on the pick & place machine.
- Consolidate close values if possible (e.g. if you have a 11kOhm and a 10kOhm resistor, can't you just use the 10kOhm?)
- Put parts in series/parallel instead of using single parts (e.g. for 5kOhm put two 10kOhm in parallel)
- Try to avoid 0201 components as typically a special pick & place head needs to be used on the machine for handling these parts. He also talks about if 0402 are really necessary...
- Try to just omit components if they are not necessary (e.g. some decoupling caps may not be necessary - but test it in prototype first!)
-
1\$\begingroup\$ Could you summarize what is in the video? I like Dave Jones as much as the next guy but it's not always possible to spend time watching a 40 minute video. \$\endgroup\$– regimonMay 22, 2020 at 9:21
-
2
It is not clear if you intend to go to production or just for yourself. PCBA is very expensive for small runs regardless of complexity, generally one sees prices start to fall once you hit 1ku for moderately complex embedded boards
If this is a production board, First prototypes will be expensive but they are good place to start tinkering and eliminating components, and rethinking certain circuit sections But you should at least prove to yourself your basic design works.
There is no real step by step process it's a basically project management and cost optimization a very complex topic .
Typically during design process , requirements are identified and a target BOM or target cost is identified and kept track of roughly. Requirements are important, for example the input voltage range of the SMPS and Max power hugely influence the cost of components needed to roll the regulator.
At this point this may not have happened, but you may still apply a target cost, and go slashing and hacking starting with the highest per $ part and rethink some circuits.
As an example if you have many different resistors of similar values, and you determine the value isn't very critical it's better to just nudge them all to one value so only one unique part is needed instead of several.
Going through a retroactive requirements exercise will help identify low hanging fruit/useless features/overkill in your design.
Keep in mind it isnt very easy, it is quite possible given a set of requirements that no significant optimization can be done outside of mass volume, in that case it may be valuable to rethink certainty requirements or radically rethink your design.
Sometimes there is something silly you are doing that explodes production costs, the manufacturer could provide consultation on any cost savings they can see, e.g. an expensive process step.
As an example
- if you use ten fewer unique parts we do not need to run through pick n place twice because of reel limitations
- these components must be hand soldered, if you make some change it could be automatically placed.
-
\$\begingroup\$ The board is about 45x200mm and assembly costs for 2 prototype boards is about 800€ per board (without PCB and components). The manufacturer just wrote me there are "more than usual" unique components. The intend is clearly to go into production at some point but right now I just started and want the goal is to get a working prototype. However, any optimization that reduces prototype production costs will help my wallet. \$\endgroup\$ Dec 12, 2019 at 17:54
-
1\$\begingroup\$ @F.Heisenberg, $1500-$2000 minimum lot charge isn't out of line in my experience for on-shore (US) assembly. If you increase to 10 pieces, the price might drop to €200 per board, and at 20 pieces it might be €120 (Just w.a.g'ing numbers that would be reasonable in my experience) \$\endgroup\$ Dec 12, 2019 at 17:58
-
1\$\begingroup\$ On the other side of the coin, since you're only doing 2 pieces, even if you fully optimize the design you might still end up paying €500 each. \$\endgroup\$ Dec 12, 2019 at 18:00
-
\$\begingroup\$ Even if it's still €500 it'll help. I'd rather spend a couple of months longer learning and optimizing than paying for a prototype board that might not work in the end. \$\endgroup\$ Dec 12, 2019 at 18:08
-
1\$\begingroup\$ Just a small suggestion, if this is for your hobby that is a reasonable approach, but even with a rough proto board you may likely cut that development time significantly, and the investment in the board now would likely get you going quicker. What I am saying is that there may be opportunity cost associated with over thinking it instead of just spending the money. Because @ThePhoton is spot on that the lot cost for 1-10 is about the same no matter what, they will always pad it to hit that ~1500ish lot charge. \$\endgroup\$– crasicDec 12, 2019 at 20:07