Top Fifteen Mistakes People Make When Designing Prototype PCBs

This is a pretty technical video, not related to replica props, but very related to electronics design. I am not an EE so everything I've learned has been through experience and making mistakes which is an interesting way to learn!

Did I miss the mark or do you agree with my top 15? I've got some more ideas in mind for other things people miss in their first designs but would like to hear from you!

Download KiCad:

----------------------------------------------------------------------

1. Stale Breadboards:
• Breadboards are great for learning and testing but can lead to frustration in complex projects.
• Loose wires and potential failures increase with time.

2. Making Hardboards DIY:
• Soldering hardboards or perf boards can have similar issues to breadboards.
• DIY PCB methods like CNC milling or acid etching may be messy and have limitations.

3. Designing for Production:
• Design the first PCB expecting it to fail, focusing on functionality testing.
• Size and shape considerations can come later; prioritize testing various features.

4. No Test Points:
• Lack of test points hinders debugging and fixing mistakes.
• Test pads for common functionalities reduce the risk of blocking progress.

5. No Power or Diagnostic LEDs:
• Diagnostic lights for voltage levels and operations save time in identifying simple mistakes.

6. Overcrowding Components:
• Avoid packing components tightly during prototyping; leave space for adjustments.
• Keep passives relatively large for easier removal during testing.

7. Underutilizing Silk Screen:
• Clearly label components on the silk screen for easy assembly and orientation.
• Ensure markings are readable on the smallest boards.

8. Not Using Isolation Jumpers:
• Incorporate zero-ohm resistors or cutable jumpers for easy isolation during testing.
• Facilitates methodical bring-ups and simplifies troubleshooting.

9. Not Breaking Out Unused GPIOs:
• Break out additional GPIOs for testing and fixing mistakes without ordering a new PCB.
• Adds flexibility for rewiring components or integrating external modules.

10. UART Mixups:
• Ensure correct pairing of transmit and receive pins in UART components.
• Use jumpers or specific designs to easily correct mistakes.

11. Locking Into I2C Addresses:
• Provide options to change I2C addresses using resistors for flexibility.
• Prevents the need for a new PCB revision due to address conflicts.

12. Separate Power PCB:
• Consider splitting the design into multiple boards, especially separating power.
• Enables testing power solutions independently without scrapping the entire PCB.

13. Choosing Labeled Surface Mount Resistors:
• Opt for labeled surface mount resistors for easier visual inspection and testing.

14. Verify Footprints:
• Check dimensions on the data sheet against PCB footprints in your design software.
• Prevents ordering the wrong footprint for components.

15. Check Parts Availability:
• Consider part availability before designing the circuit.
• Speculatively order critical parts before PCB production to mitigate shortages.

Music:
• Kunal Shingade - Betelgeuse
• New Day - Lakey Inspired
• Silent Partner - Bet on It
• Kevin MacLeod - Lobby Time
• Topher Mohr and Alex Elena - Where I Am From

[Olaftr]

I like to print a 1:1 layout of my board before ordering, then place the parts on the printout, checks: footprint, pin 1, connector clearance, etc.

[Factory]

I have made all those mistakes.... some of them numerous times.
My only addition is to 'sleep on it'. When I finish a design and feel I have done everything necessary to send it to the stop for the day and re-evaluate in the AM with a fresh look.
I still manage to catch a few dumb things and happy I waited.

Great video.

[pauldavis]

It is true that the first PCB will never be 100% correct. I used to work at a electronics company and we only once got a PCB perfect first time. We celebrated and then went to mount the only mechanical bracket on the board and found the mounting hole was in the wrong place.

[henrychan]

Also put in mounting holes for your PCB first, not last.

[____________________________.x]

8:24 switchable TX RX pads. Thank you for this tip 👍

[MSM]

AS a manufacturer of bed of nails PCBA test fixtures I gotta note that quite often PCB designers make silly mistakes when placing test points such as placing multiple test pads in one spot on a one side of the PCB right below the components mounted on the other side. This leads to leaving no chance to provide an adequate counteracting support such as a pushfinger that in turn introduces a risk of damaging both the PCB and the components during testing in the fixture. Mechanical damages caused to ceramic capacitors is one of the biggest problem in the industry.

[kensmith]

A few points:
Don't buy the super cheap solderless breadboards. A little extra money on them can save you time later.
Sticking the solderless down with "3M command" sticky things is a very good idea. You can do this with very little stress on the part but still have it stay in place well enough until you need to move it.
The technology of wire-wrap is still useful sometimes. A wrapping tool is a good thing to have when that is needed.
Generally all PCB stuff is inch based measures. It is best to just call the "mil" your spacial increment and get over the idea that it is anything but the "PCB spacial increment". Converting between inches and metric gives way too many places to make a mistake. If your layout software natively talks in mils, then get out your spreadsheet and convert everything before you start.
I find I can do 0603 parts well enough.
On your power supply section, it is really a good idea to bring the power to the rest of the PCB via an RF bead. This helps on your EMI in the final product but on the prototype gives you a thing you can install only after you have checked that the +3.3 isn't making 12V.
Putting a test point on the power after the RF bead gives you a way to bring in power if all else fails.
If you have a laser printer with scaling you can trust, print out the pads and try your components of fit if you can.
On the parts with a big fat thermal pad on the underside, you can use a great big hole on the prototype. This gives you a way to solder the pad. It is unlikely you will be doing a heat test so the heat sinking doesn't need to be as good as in production.
Yes to the LED on the power line but be careful not to put too much current through it. When working on the PCB you will look directly at the LED from a close range.
On test points, use through hole ones with a hole big enough that you can solder a wire.
On your power input, do something to protect against reverse power connections. One early morning mistake can cost you a whole week.
I use ExpressPCB <--- Not an ad for them. They can do 6 mil traces. This is about the finest I've needed.
When you layout a fine pitch part fan your traces out to a large spacing as quick as you can. If you have to cut a trace a 25 mil grid is easier to do it on.
Try to stay with using any internal layers as just a solid pour. Internal traces on prototypes are not needed usually.
Add several ground test points. You will need to hook your scopes's ground and you want it out of the way.
For easier debug avoid parts on the back of the PCB other than bypass capacitors.
If you are going to had solder the board, check the footprints a 2nd time to see that they are not too small to do by hand. They are often the IR reflow type of footprint and they are hard to solder by hand.
Try to find parts in TQFP if the other option is the QFN package. TQFP parts can be done by hand and you can lift on leg in a pinch.

[somejoe]

Thank you 1000x for calling out KingBright's non-standard LED markings. I like their LEDs, but I end up having to quadruple-check the footprint, datasheet, and marking of every one of their LEDs every time I put one on a board.

[Graham_Wideman]

Couple of thoughts to add:
1. Don't be shy about using SMD resistors and capacitors on prototyping perf board (the kind that has plated through holes in 0.1" grid pattern). 0805 components solder nicely between adjacent hole pads. But I do agree, move to PCB as soon as possible
2. Make a checklists of things to check before submitting a job to PCB manufacturer. This includes a number of things in this video: test points, disconnect points, communicative silk screen labels, especially on connectors, component polarity, a board id, and also consider including mounting holes so you can mount your board to a fixture conveniently. Such a checklist cuts down on annoying forgotten features during last minute rush before order.
3. Make a checklist of all the options you normally select when placing the PCB order.. This cuts down on stress while placing the order, and gets you consistent results, or at least provides a consistent baseline that you can customize for special requirements.

[Nik]

I've switched RX and TX so many times, that not every time i use a UART somewhere in a design, i have a resistor "bridge" of 4 resistors, so i can flip RX and TX.

[SkylineIn]

sThis video is GOLD for new comer. I learnt all these mistake in a hard way during work. My tips is Slap as much test point as you can in first version, leave the track as thick as you can for future possible cut and reroute. Sleep and check again next day before submit. And you are good to go.

[Mike-H_UK]

Excellent video. Two things that I would add from my own experience. 1. Be consistent with the position of labels on the silk screen. I was sloppy on my first PCB (some labels above resistors, some below, some on the side) and this created untold confusion! 2. Using dual footprints where possible creates flexibility. For example both a TO92/TO39 footprint and a SOT-23 footprint. This allows alternative components to be tried without adapter boards.

[Fg-youtube]

I am just now dealing with a PCB prototype I made 4 years ago. I made almost every one of those errors, I wish I had seen your video back then. Thank you very much for helping others to prevent that.

[portblock]

100% on everything said!
Here is the way I have been doing some things:
1: I make sure I have all parts in stock in front of me before ordering PCB
2: I print out my PCB design and place new components I have not used before on the print to check footprints
3: I have a small pcb with reset and LED's that sits between my programmer and my device, so I have one header for programmer and test LED's, serial, reset etc.
4: if I am on the fence about a sub circuit section, I design it in the board, and add a header to bypass and use external version
5: I break out all unused pins to headers (I dont solder unless needed)
6: I have a header with all voltages at are used (usually vin, 5v, 3.3v) so I can power board before populating power supply components
7: I populate and test at intervals, first the uController, xtal, decoupling, then test make sure programmer and see it and program it, then I build up in stages with power supply components last

EDIT: 8: I forgot to mention, if I am able to use DIP package, I will use a DIP socket - a current project I am working on, I was able to move 3 DIP chips across 4 versions of the board. its not that I am cheap, its just I dont like wasting parts on bad prototypes

As a note, on a number of occasions I have build up half the board and did a bunch of testing and firmware changes only to realize I need to revise the board in one way or another, and got lucky because I never completely populated the board.

[atsleeprepeat]

i dont have much to add other than - layout the PCB is sections so it looks neat and is organized (kind of plays into having those jumpers to cut off parts), standardise your parts and sections as soon as possible (helps when you have multiple designs) so you can reuse them and you dont need to redesign the PCB every single time and you know what components work (because you have some idea about their MTTF). Always use the shortest trace width possible.

You will fail, you have to accept that and not let that demotivate you at all. Make mistakes and make them fast. thank you for this great video and thanks to everyone in the comment section.

[jimmy]

Awesome tips. I discovered all of these the hard way; the tricks with 0 ohm resistors are genius.

[julianbrown]

Going to add one more in - autofill of zones/areas. It is really, really easy to use plugins in your eda to generate submissions to those common PCB prototype fabricators and even easier to make a change to a trace and hit the "send" button without refilling those zones resulting in a board with a dead short. DRC should refill and check for you but without forming a habit to always run DRC after making changes, it is all too easy to forget

[repatch]

Ha! I worked for a company where a full ASIC came back with the RX and TX on some high speed serial stuff reversed. Yes, an actual ASIC! That was a costly fix.

[____________________________.x]

“Designing for Test” is a good practice too. In the good old days you’d have circuits added just to run digital signatures through chips to check they were working. If you have spare logic gates, get them to combine something elsewhere in the design which shouldn’t be active in normal operation. Spare inputs could be used by a test boot sequence to run diagnostics, as feedback to check an output is functioning.
Adding a low value resistor feeding each ICs power pin will highlight a short circuit condition with that chip

[AnonyDave]

I will definitely say don't get metric and imperial smd caps/resistors mixed up. Nothing worse than thinking you're putting in a big 0805 smd resistor only to find out you've picked the metric footprint and it's super small

But also if you design for a larger footprint on those same devices, you'll still easily be able to get a size or two smaller to solder in. At least if it so happens that 0805 becomes unpleasantly expensive, 0603 will fit on the same footprint just fine. For that time you finally finish off the reel of 10k 0805 resistors you picked up a few years ago for a couple dollars