Top Fifteen Mistakes People Make When Designing Prototype PCBs
ฝัง
- เผยแพร่เมื่อ 26 เม.ย. 2024
- 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:
www.kicad.org/download/
----------------------------------------------------------------------
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
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.
I second this. If you have the components already this is a fantastic sanity check - one which has saved me from many a mistake
Exactly ... I was just about to make the same comment. I always print my layout and place the parts to make sure they fit. This is after spending a good deal of time studying footprints and measuring parts, to make sure they're correct in the first place.
YES! same here, solving both points 14 and 15
Yup. Helps a lot. It does require that you have the components on hand already.
As you get more experience you'll move to what is said in the video: Double check the sizes in the datasheet with the footprints.
Normally I rely on that, but recently I printed a finished design again. I had not done that in quite a while, but the technique is still available and I pulled it out of the closet for that project.
Sadly, this doesn’t work when you have hundreds of components on board))
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 fab......I 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.
I recommend developing and maintaining a checklist of "everything to check before submitting a PCB", and another checklist of what options to select when ordering the PCB.
'Sleep on it' helps me too. It takes away almost all hesitation, and sometimes gives new better ideas tomorrow.
Also put in mounting holes for your PCB first, not last.
Amen!!
For a horror example of mounting holes see Raspberry Pi 1. Only 2 mounting holes and in random locations.
@@Foersom_ Or the arduino uno where the headers prevent you from putting a screw into one of the holes...
Exactly. Probably the first unsuspected PCB builder horror with heatsinks coming as second.
8:24 switchable TX RX pads. Thank you for this tip 👍
I've run into that quite a few times. Nothing worse than wrong signal or voltage polarity on part of your board/s
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.
Interesting issue!
so you recommend that wherever test points are places, there should be enough space to apply "pressure" on the PCB ? Would it make more sense to provide 4 "push points" on all 4 edges of the PCB as a standard practice ? How are ceramic capacitors getting damaged ? due to uneven loading on the PCB ?
@@3atsleeprepeatPCBs are very flexible, ideally you’d want an opposing support for every test pogo pin rather than just the corners. And yes, if you are fixing a ceramic component to a board, that’s the first thing to break when you bend it. Leaded Resistors have a bit of give but they’ll break too
@@3atsleeprepeat _"so you recommend..."_
I recommend to use a proper engineering approach to designing PCBs which involves FEA. If mechanical parameters aren't going to change from one design to another then some manufacturers like BOSCH might come up with an empirically proven method when the PCB is divided with a grid into regions within each of which 1 push finger is applied per every certain number of test pads. Min. distances from PF to a ceramic capacitors as well as other components are also specified for that. Designing and making test fixtures isn't that simple. There are many hundred different test probes available with different shapes of heads, different sizes, different compression force, different materials they're made. All these parameters are taken into consideration when designing a test fixture for a particular PCBA as every new PCBA is like a bran new story. If there is a need to do ICT(in-circuit test) at the component level where the vast components have direct access via TP (test points) then the test fixture might easily have a few hundreds(!) of probes. If let's say there are 400 probes 2 Newton each all up in the fixture then the entire DUT (device under test) experiences 800 N or 81.5 Kilogram-Force, So for in-line PCBA testing either pneumatic press or vacuum interface are used which are capable to develop quite huge amount of force in a fraction of a second. With this method the entire board gets tested within just 6 - 10 seconds.
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.
It makes me wonder if they did a production run and then realized the mistake so just changed the datasheet instead of scrap the parts? Who knows. Thanks for watching! Pass it on to a friend.
@@cosplaylightandsound4143 Apparently the issue is the manufacturing process for SMD LEDs combined with how Kingbright wants to arrange their manufacturing equipment. There is a substrate that is doped one way (P or N), and then a top silicon layer that is doped the other way (N or P), making a diode. To change colors, they change the type and location of the dopant, so in some LEDs the cathode ends up on the top silicon layer, and in others it ends up on the substrate. Most LED manufacturers then will mark the package with the cathode, no matter which lead it is. However, Kingbright marks the substrate always, calls that "pin 1", and then expects you to refer to the datasheet to determine whether pin 1 is the cathode or the anode. It's maddening because every other diode on the planet has the mark on the cathode.
@@somejoe7777 Truly fascinating. Thanks for the clear explanation. I still avoid their LEDs though! :)
Agilent (previously HP) actually have LEDs of both persuasion, if you choose HE red you get one pin out and if you choose normal red you get the other one. Basically with LEDs always check.
Hahaha -- LED polarity is often vague, including some that are only marked on the back, which of course you can't see when you place them. I end up taking a microscope photo of the top view of the actual LED component, then testing it with a battery and resistor to be 125% sure what the polarity is, and add that info onto the photo which then goes in the assembly doc for later reference
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
I agree with all your points except "Generally all PCB stuff is inch based measures." In my experience, in the US,, this has not been true for quite a number of years. Virtually all components these days have their dimensions documented first in metric, specifically mm. We're rarely using DIP components, so 0.1" grid is not very important. And all the outside constraints on board size and interface to other boards or case are measured in mm in CAD. The one exception is 0.1"-pitch pin headers, but those can be placed just fine using mm, and don't cause any particular problem when everything else is in mm. And ECAD software can generally switch easily from one system to the other. But, if you're more comfortable in inches (or mils), go for it. I just wanted to provide a counterpoint to the idea that PCBs are still intrinsically based on the once-upon-a-time ubiquitous 0.1" grid.
Yes, Enig is better, but a bit more expensive.
Smol addition: I've heard from multiple sources that pen wiring beats wire-wrapping. I forget the details, but it's quicker and easier and works very well.
@@eekee6034 Those sources are evil and trying to lead you down a dark path. Wire wrap is the one true way.
This is a really good and thorough video! From the analog and RF world, a couple more critical mistakes I've seen are:
-Add non-populated footprints to solder passive electronics on. Maybe put it in a position for a simple filter (0R resistor in series, unpopulated spot to ground). A good practice is also to add a space to put a capacitor (or a spot for an entire passive network) in a control system such that there's negative feedback to prevent unwanted oscillations (like a small capacitor close to the output and inverting input of an opamp, or between the collector and base of a transistor, etc.). A voltage divider configuration, but not populating the resistor to the ground, can be very handy sometimes as well. This is ALSO super handy when dealing with microstrips and coplanar waveguides, it allows you to add a couple components here and there if there are any impedance matching issues :-). More often than not like 80% of these spots remain unused, but the other 20% prevent you from molesting your boards with a scalpel in order to untidily slap a component somewhere...
- Poor ground planes, it's super easy to rely on software to handle grounding for you, but this can result in weird voltage drops everywhere, ground loops, etc. You'll be scratching your noggin slapping capacitors and jumpers across everything to suppress noise and oscillations because this is a really hard problem to fix. It's good practice to draw in a trace for ground of the appropriate width, so you are very well aware of what path the current will take in its worst-case scenario. Then you generate the rest of your ground plane to properly take care of shielding. (Obviously, a 4-layer PCB takes care of most of these issues, but a 2-layer PCB can be just as good and cheaper as long as you do this correctly)
- Forgetting decoupling capacitors also happens every now and then, it's good practice to double up by putting a big and a small one in parallel. Don't be shy about adding them to other active devices as well rather than only chips.
- Signal traces parallel to high current ones such as power supply traces causing crosstalk is also something almost unfixable, the closer to perpendicular these crossings are, and the further parallel lines are apart, the better.
- As short as possible trace lengths for high impedance lines with very low voltages! The shorter the trace, the less susceptible to noise.
- Prevent traces from switching layers as much as possible, once is fine, twice is acceptable, but more often not only makes troubleshooting horrible, but sometimes the added impedance of the line can cause voltage drops or higher susceptibility to noise as well...
For RF, I'm obviously not going in-depth, simply because half the components of your designed circuit exist of simply the copper, the board material itself, platings, and solder masks also having a big impact. But one mistake I've seen multiple times with GPS or GSM-related projects for things like drones:
- Too small of a clearance between generating a ground plane and your signal traces for RF boards. You can make the most beautifully spaced coplanar waveguide from a GPS module to the connector to which you want to mount your antenna (according to the datasheet), but your software doesn't care about that. Make sure to specify for the software not to mess around with generating the ground plane around such a strip, as usually the ground plane clearance with regular traces is much smaller than what you need for your stripline. From the modules/antennas perspective, this is a short and it's unfixable.
I mean... there's many many more to this, but just to name a few 🙂
As for perf board and etching... It's the only part I don't really agree with. Learning to properly make circuits on a perf board takes skill and a lot of practice, but a well-made perf board is just as reliable as an etched PCB. For a one-off complicated board, it still has my preference. It takes a bit more time than designing a board and ordering one, but it's cheap, and I won't end up with 5 PCBs, four of which can be binned right after receiving the package. I see a really rapid decline in the soldering-iron-handling capabilities of new electrical engineers where even soldering a wire is done in such a way it breaks off after moving the board around on the desk twice... Adding this to trying to solder an oxidized perf board with lead-free solder and a crusty tip of a soldering iron that has never been cleaned, I agree the result is as reliable as a breadboard... Then again, at my university, I've seen quite high-stakes equipment being torn down where entire parts were built (very cleanly) on perf boards that were in service for literal decades before the entire machine got decommissioned due to old age...
As for etching them manually... When you're strapped for time it's nice to have the materials, especially for high-current/voltage stuff with big components... In the past I've done RF prototyping, using PCBs with a photoresist layer. That stuff is super tight on tolerances and it worked perfectly. Again, it is more of a skill issue rather than the technology that's the problem. For through-hole stuff, you can even buy stickers and an aerosol-based solder mask to make them look cool (or you use a tinning solution to keep the copper clean). Thankfully the etching solutions are a lot more tolerable nowadays as well! I know engineers who used to use iron(III)-chloride and wash their boards in benzene :-). Nowadays there's stuff like sodium persulfate, which bleaches your clothes but it's not really a health hazard anymore... As for solvents, to get rid of photoresist and stuff acetone works well, but if you hand-paint your traces, draw them on with a paint marker, or use the technique to transfer ink from a printer to the board, your partner's makeup remover is already strong enough to clean them up! Don't tell her though... before you know it she'll become suspicious and hide it like mine did :-P).
But again, these are my personal findings, and we all use different methods for different kinds of technologies! I deal with digital chips for my job every now and then, but thankfully it's all I2C and it's really convenient to deal with... I'm thankfully not the person programming them hahaha, writing code is already a pain to get going, but to make it run reliably and efficiently... that stuff is absolute witchcraftery to me
What a fantastic, detailed list thank you!!!!
If someone makes a perfboard *instead* of a PCB and really puts in time and thought into it, then they can be works of art. But I think a lot of people make perfboards as a quick-and-dirty solution to avoid the fiddlyness of breadboards and because they don't want to wait a week for a PCB. So they make mistakes and just rush to lay it out. If they stopped and thought about the layout, they might as well make a PCB. I readily admit I'm biased against putting a ton of effort into something you only get to use once. It's a personal choice everyone has to make.
@@cosplaylightandsound4143 I can absolutely see where you're coming from, but I think it highly depends on the workflow and the type of project itself. Whenever I design a circuit where I'm 90% sure it's gonna work (trivial stuff like a powersupply, a temperature controller, a servo system or whatever) I definitely agree with you. It's quick enough to draw up a PCB, wait for a week, and accept the cost of manufacturing. And especially when you may need multiple boards in the future it's lucrative. Obviously with RF, especially when you go do stuff above 100MHz, there's no alternative anyway. I find this is also often the case when you're working with digital technology. You have the benefit of working with premade modules rather than individual parts that were purposefully designed to click together with ease (maybe you have to add some level-shifting circuitry, but it's really robust to what nature tries to throw at it). This makes the process quite linear, you preselect your parts, you draw up your schematic with them, you design your PCB, and then mayhem breaks loose because now you have to tie the nuances down in code.
In the analog domain, this doesn't work this way at all. Personally, I often need to do designs where there are almost no (tested) schematics (or very dated ones) to draw inspiration from. For every part on the way, you're not really sure if it's gonna do what you want. Your mathematical analysis says this, your intuition says that, your simulation says something else, your colleague from the time transistors weren't even around also has an opinion, the documentation doesn't have anything useful at all, and in real life, neither of them all really predicted the outcome correctly anyways. So you either fix it, scrap that part, find/develop a new topology, find a new (more suitable) device with probably a completely different footprint, redo your research, discuss your findings with other knowledgeable people, try a completely different technology altogether, or restart completely from the beginning. At least, whatever went beforehand, you already tested it and you're convinced it works, you only need to tear down a small part of the board, check your math, get a second opinion from someone else, and try again. You don't have to wait a whole week for confirmation if the next section is gonna work, so you know it right away so you can continue your prototype. This means that you don't start with a ready-made schematic, but a general idea (like a block diagram with a set of signal processing operations or something), and then the schematic slowly rolls out, together with your PCB design and testing simultaneously. Of course, nowadays, there are some building block type components so you don't need to construct a high impedance input stage of an opamp yourself, or a highly stable voltage reference, or a phase locked loop, etc, but even these parts usually behave un linearly with unreasonable tolerances making them a bit unpredictable.
On an unrelated note, this is also why RF design is so incredibly time-consuming and expensive. It's even less predictable than things like analog control/measurement systems, and if something doesn't work, the only thing you can do is slap a capacitor or a resistor somewhere for your impedance matching and cross your fingers. You first try all sections out on separate PCBs, once you get them working after 3 iterations, you try to add things together, get an aluminium case milled as simply the shielding of the ground planes in the PCB material isn't enough and you go through another 3 iterations, FR-4 isn't always suitable for this either so you end up buying Rogers PCB material, where suddenly board size becomes an incredible constraint as that stuff costs like 10 euros per square centimeter just for the material alone... And finally, you can measure your system for phase stability, jitter, temperature dependency, etc., to realize it's actually not good enough and you can try again if you're unlucky hahaha. When it comes to designing filters and cavities etc, software options are incredibly limited and obscenely expensive and limited to big institutions, so you sometimes have to write your own simulation software which is super time-consuming.
I really enjoy the flexibility of perf board that would otherwise make prototyping hell if I'd be constrained with only premade PCBs, and if you do it well, you also end up with a piece that will be highly reliable. In case you need more, all you have to do now is digitize your schematic and quickly draw up a PCB and you have a system that is almost guaranteed to work right off the bet with little to no bodges needed. And if you need only one PCB (since I mostly build test equipment, this is frequently the case) you're already ready to go on to the next device :-).
I think it's important in this case to keep an open mind to perf boards really depending on the type of project you're doing rather than completely disregarding it entirely. A perf board is a great alternative to a breadboard for prototyping that takes the "long" and intensive route. It takes a bit of extra time to work with it, but it's flexible. You can leave it for a month, and you can continue cobbling your circuit together, if you take a lot of extra care and really learn to work with them, you can leave it for longer than your own existence and it will remain working reliably (if you avoid electrolytic caps and keep the board out of the rain :-).
As for the surface mount stuff... I mean for a lot of analog jobbies, manufacturers often still produce versions in DIP packages, every now and then you have to write them an email and they can provide them as a sample. But ever since surface mount technology exists, breakout boards exist as well :-).
In the end, it is purely preference and highly dependent on the job. I know I'm in what's nowadays quite an unpopular and small branch in electrical engineering, but I just wanted to give an example where it's still a critical part of the process.
@@m.e.8273 I wouldn't dream of trying to design an analog/RF circuit at my current level of knowledge. Reading your comments I feel like an amoeba trying to comprehend the universe. If I ever get stuck, you know you're my first call now, right? 😆
@@cosplaylightandsound4143 Mail's in the description for a reason hahaha, I feel exactly the same way when people start discussing digital stuff honestly XD. Different technologies tackle different problems with different solutions 😀. Your board designs are absolute works of art!
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.
...and never start with 0204 components and minimal PCB size...
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.
ALso add MISO and MOSI pins to this list!
Awesome tips. I discovered all of these the hard way; the tricks with 0 ohm resistors are genius.
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.
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.
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.
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.
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.
Great video! And good tips. Don’t have much too add since I’ve started with PCBs, but a lot to learn from your video. One that has helped me a lot in identifying mistakes in footprints is to print the PCB and do an inspection there, is much easier than measuring everything in Kicad. This has been the best YT recommendation I’ve had.
Printing the PCB design is a brilliant suggestion!
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
Fortunately for the most part, the common imperial size numbers don't match any metric size numbers and vice versa. So you can mostly tell which are which. I do agree that it's a pitfall. When specifying one or the other, I try to adopt a scheme where a letter identifies the system, like U0603 vs M1608 (U=US, M=Metric), or a combo size number like U0603M1608. While working, it's often helpful to be reminded of the actual dimensions in both schemes (0.06x0.03in, 1.6x0.8mm). The prefix letter has the side benefit that when listing such data in a spreadsheet it avoids the string being turned into a number, and losing the leftmost zero.
@@Graham_Wideman it's easy enough to get them confused in the kicad base libraries, I've done it once 😅
Its one of those things that can happen if you just go a little too quick
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
This is a great compilation, will recommend it to all my juniors from now!
I've learnt the hard way for half of those mistakes but you've saved me from making the other half, thanks!
A LED hooked to the micro controller can be used to output diagnostic info. Like it blinks 3 times to indicate the I2C bus can't find the IO expander chip.
If you're going that far, might as well add a serial line and use it to emit some more detailed info, no?
I usually make sure I can do serial but using the status light for things like this is really cool too.
@@cosplaylightandsound4143 It was just an example. You can encode a lot of information in a blinking LED.
Good idea, just like the error beeps that computer motherboards make
It is really useful to have a status led for any project. It allows you a quick debug of basic problems especially if you are having to debug remotely. Programmed right you can see the status of the hardware. Like maybe it's in an unconfigured state or stuck doing something.
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.
Glad to see that as an amateur, I'm doing most things correctly, main takeaway for me was the diagnostic LEDs, great tip!
Just getting into the whole designing of PCBs this year so this was a very well timed and valuable video!
Great video, I really appreciate that you listed all your points in the description of your video.
So I could add some to my already 300+ pcb-WatchOut list 😊
Very good ideas here, especially about test points, test hooks solder pads that can be cut to isolate key sections for additional testing and troubleshooting then soldered back together. Those are great tips for production designs as well.
“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
This is an excellent video. These types of mistakes are easy to make for beginners but also for more seasoned designers. Thanks.
Thanks for these tips! What a great video. I've watched so many of these PCB tips videos, and this is one of the most useful ones I've seen. I'm a programmer, dabbling in hardware, and making some PCB's, and I will be using some of these tips.
A lot of great tips here. Nice work!
We had a working prototype and I had to draw the PCB for production, back in those day getting it done on a CAD system would cost a fortune. Yes, it was last Century. I had to draw it by hand, luckily I drew all connections of all components and every connection that went to something, even if there was no connection made to it. That was my luck, because, my mistake was that I did not draw one line between the connector and an IC. When the PCBs were made non of the soldered board were working, every board needed one small wire that I didn't draw. One lesson learned.
This is excellent content. Please make more sharing your experiences with electronics.
Thanks! I've got a second list in mind.
This is an awesome video, I'm designing my first (two) pcb's for my project. An Rc controller and rc car.
Been over a month, because I keep redesigning it.
A few of these I'll definitely keep in mind though.
Fantastic hints!
Wonderful and very useful video, thanks!
Awesome instructions! As part of the silkscreen a date/version costs nothing but can help in the future ✊🏿 thanks for the video 😃
Great video, these are all very helpful tips!
I like to add a large solder mask cutout in my ground planes at the edge of the PCB so I can stick an aligator clip on for easy probing with a DMM and oscilloscope. For RF designs, I have started adding 0-Ohm 0402 resistors in-line of the RF path with a solder mask cut-out over a ground plane nearby so that I can stick a coax cable facing either direction for cheap testing. I know there are switch connectors out there but the counter-part test probes tend to be a bit expensive, and the connectors are typically uni-directional so you would have to rotate them anyway.
I have literally forgot to power a MAX232 chip one time, also did the RX TX mix up in a last minute break neck project for a trade show on the other side of the country. found out after I arrived and the booth display didn't work. Butane soldering iron, exacto knife, individual stranded wire strand, and the patience of Job fixed that. Great video, you have earned my sub!
I'm defiantly going to use some of these designee tips in my next PCB build.
Excellent advice!
There is one additional quite subtle problem with configurable devices, e.g. via I2C:
Make sure the circuit/board works not only for the configured, but also for the unconfigured device.
Picture a designated GPIO pin which is intially a Reset pin. Bummer.
This list is so true! Some of the things I have done and failed myself and some others I am still guilty of (e.g. premature minitiaturization). I like to add a additional thing: Not reading layout requirements in datasheets correctly. These things can really ruin your day.
Great advice, thanks!
LOL, I almost fell down my chair when #1 rolled up.... nice one...
That footprint for UART mix ups - in going to use that everywhere! Thanks!
Thank you so much for this. I had been unable to sleep for 3 days and this immediately knocked me out
Thank you for this Excellent compilation of many precautions that needs to be taken at concept verification and prototyping stage. Provision for appropriate heatsink and mounting (mechanical fouling) is not foreseen many times. In some specific cases EMI precautions, balanced tracks etc could be considered.Also mains isolation - (regulation if any to be followed) ...... if needed...
I learned electronics with a breadboard and ferric chloride made pcbs in a club, so might be biased to them (also some friends had a lot of trouble with customs after ordering a pcb in China). My general advice would be :
1) If you use perfboard, plan your layout beforehand to avoid soldering wires as much as possible, a 5mm grid sheet of paper is really handy for that.
2) also when using perfboard, don't bridge pads with solder only, use a single wire AWG26 or AWG24 and solder it at each pad on the way after bending it cleanly with plyers, it doesn't seem much but it make the circuit a lot more clean, and less prone to accidental shorts, and also adds rigidity to the board
3) always use single sided perfboard, with non-metalised holes, you never know, you might need to replace a component, and plated holes are a nightmare after more than 3 aligned pins, dip supports is a great opinion, but don't go to the cheapest, you will have reliability issues.
4) avoid metalised holes when working with through-hole components, it's just way simpler to replace a component on a single side-board
For #14, wrong footprint, a tip I often use is to print on paper the PCB layout and visually check sizes with my components.
Saved me wrong footprints.
Breadboards are nice only for small subcircuits that you want to test. Perf boards are nice if you want those small subcircuits to be permanent. I am finding I also like perf boards if I am combining numerous pcb modules together as well, but you should not be attaching a microcontrollers pinouts directly to a perfboard unless you have absolutely no other way of making the circuit (and if so, plug it into female headers, your future self will appreciate you). I find it strange that people will automatically go to making a pcb without testing at all. One of the biggest recommendations I have is to try and buy modules and connect with jumpers instead of relying on a breadboard or perfboard. For instance, I know a few newer guys that didnt know they couldnt control a stepper without a stepper motor driver, they got an H bridge, then made a pcb out of it for their project not knowing it would not work at all. If they did their research, got an individual module off of amazon, and then made the appropriate code, you would know that your circuit would work and that the code works before having a 3 week lead time. The time is the biggest killer.
Agree 100% with these observations. At this point, I buy modules for everything and just start doing isolated tests of each function and then start to put together a speculative PCB with extra headers, extra test pads, and places to patch things in assuming I'll screw things up. I don't get excited about enclosures and board shape until rev 2. To me the first PCB is just to eliminate loose wires and bad soldering from the pool of mistakes.
@@cosplaylightandsound4143 If you're going to wire together a bunch of modules into a configuration that you'll be testing or writing code for, it can be super handy to 3D print a frame to attach the modules to so the assembly is a bit more robust.
I dont know how did i found you but i am right before designing a pcb for a project! Thats useful for sure!
Thank you!
More GND clips!
Another good tip for the footprint saga. A lot of EDA's will allow you to print a 1:1 version of the board to a piece of paper. If you have connectors or some non-standard components already at hand you can print out both sides of the PCB on paper and check the components on those Printer printed PCB's :)
Great tips!
A great video! Thanks 👍
I don’t know how to design PCB’s or anything but still loved this video
The NXP i2c call out was cathartic
These are so true. Even while I know I should not, I often make these mistakes.
As of the space constraints and optimizing, it is not always possible. I..e in some cases the requirement is very much about exact layout and dimensions, when i.e. making a replacement part, mod to some other product, making a portable device (still in more complex setups, one should make enlarged prototype and optimize later), or working with some existing size standard (i.e. DIN Rail enclosure). I have often not too complex designs, that if they cannot fit in a specific enclosure, then they are dead, and whetevever circuit works or not, does not matter, if I cannot fit it. And if it fits, will often depends a lot of layout, usage of specific parts, and some other design decisions. So if I make a bigger version, or with connectors, mounting holes, size, different, sure I can make it work, and make it easier to debug, but that does not help me, when I need to shrink it to fit, which might require a complete relayout, different parts, different micros, different ICs, differnet packages, connectors, etc, etc. So sometimes I will do spend way more than I should on first layout, because I want to make sure it is even possible to fit it. Also noise or thermals might be extremely depended on layout, and component proximity, so having bigger design doesn't fully allow me to test functionality. I agree, it is prudent, to test one aspect at the time, and not be distracted by interdependent problems.
Still, sometimes, when I make sure that layout can fit, and in principle I have some layout, I will quickly a simplified version, for testing, and make it slightly bigger, manufacture quickly and test.
Even on minimal design, I will make sure to include things like test points, isolation jumpers, spare IOs, etc. And for production run remove any if they are impeding size, due usage of this last 1% of space. Because it helps a lot. If I can fit everything, and test / debug features, then it is pretty obvious I can remove them later.
But sure, most of the time, it is not the case.
Few small things to add: 1) Using too thin traces everywhere. The fact that you can, does not mean you should. Thin traces introduce resistance, and also makes reworking board harder. 2) Doing mindless copper fills everywhere. Fortunately some software will remove islands, but you still should make sure you do not have too many holes in your power planes, i.e. due to dense vias placement.
Great video!
I'm just getting back into designing my own PCBs again after having taken a break for a couple of years. This video's been a nice refresher and reminder that I've been on the right track since I came back to it, heh. Another tip on checking the size (especially if you've already bought the components) is to just print it out on paper and lay them on the printout. It's 100% scale so will be exactly the same size as the final PCB. Makes component size discrepancies real obvious. :)
Great video! I go straight from KICAD design to PCB. If I check it enough times, and the datasheets are accurate then it should work first time. Except the datasheets are not always accurate, are ambiguous, or fail to mention something important. Biggest problem is with connectors. Had one once with internal shorts between some of the pins (some sort of signal isolation idea). Not a mention of that on the datasheet. This sort of defensive approach to design is essential.
A lot of these were pretty good. I personally always breakout unused pins on ICs and MCs during the first phase along with spare PWR and GND pins.
I basically always plan my first board around plan Bs to what what could possibly go wrong. Back up pull Up/Down resistor pads, bridges (0 resistors as you call them here) in spots I don't feel 100% on and indicator LEDs are my latest feature I will never do without now.
Cool video.
One thing about I2C and other such functionalities that may need reconfiguration - instead of providing placement for resistors (0Ohm), you can add headers to allow using jumpers. That way you don't need to resolder anything later on. Mind that multiple changes would require multiple resolderings which increases the chances for deterioration of the copper traces and may even damage surrounding components.
Good information = Thank you! Final proofing tip = Draw a 100mm line on the Comments Layer (etc.) of your PCB design page. Print the final PCB layout on paper at 100% scale. Include the Comment Layer in the print settings, so that the 100mm line will print. Do not trust the scale setting (more to follow.) Verify the line's length with a proven measuring stick. Why? Because some printers (particularly inkjets) are known to not print to scale, even when set accordingly. If yours is not to scale, do the math and make a correction with the custom print settings function. Once you have proven the scale is correct, proof each component on the printed copy with the actual components. Example: My inkjet prints 100mm as 98mm. So, I have to increase my print scale to 102% I hope that helps.
You just need to stretch your printer by 2% so it prints properly!
Thats what he wrote in the last sentence ...@@Graham_Wideman
For all components that use holes , it is important that the wires are NOT an interference fit prior to soldering .... this is because any rework will rip out the plated through connection .... a good practice is to fully load the board ... turn it upside down and EVERYTHING falls out ! ... this design method uses slightly more solder , but a faulty component can be replaced easily saving the cost of the whole ( expensive ? ) PCB .... ( tried - n - tested ) ........... DAVE™🛑
Nice video. I'd love to see a beginners guide to PCB design or something like that.
Great vid!
Great video. Cheers.
NXP spun off from Philips who invented I2C so I guess they can argue that technically everyone else is doing it wrong. One mistake that got me recently is that the ADC2 inputs of the original ESP32 cannot be used while WiFi is enabled. WiFi is the ESP32's raison d'etre so the second ADC is essentially useless and therefore apparently not a good way to monitor your battery voltage. So make sure you read the fine print as well as the pin-outs. Like others have mentioned I also print my PCBs 1:1 (which almost always isn't the same as 100% in the printer settings) and drop components on to the printout (provided I have them) to check any novel footprints. Make sure connectors aren't mirrored incorrectly too. When you're sure the design is right it can be a good idea to wait a day and take a step back before sending it off. There's been a few times when I've realized I've missed something critical after stepping away. It's not the PCB cost but the shipping and time that hurts.
Great video, thank you! 20-Year EE and I approve of this list. I've suffered through wire-wrapping, solderless breadboards, etching, PCB Artist, Pulsonix, Advanced Circuits, and finally FINALLY migrated to KiCAD / JLCPCB workflow. My mind is blown at how I can have a framed stencil and PCBs for perhaps under $500, when I used to quote customers a minimum of $1200 materials cost for the simplest PCB due the workflow I used. It's an exciting time to be an electronics designer. Do you have a website or way to get in touch to discuss projects? Always looking to improve my network of colleagues.
Thank you very much. A lot of good advices 🙂👍🏻🇧🇻
These are interesting ideas to think about for new people who wish to develop circuits like me, this video was useful
I've made a few of these, KiCAD 8 is just to say out now as of this writing, and it feels more mature and almost quietly better which is a good thing.
Not saying I have made this mistake on a big project 😟but sending the design files out with all the polygon pours shelved :( It was a 4 layer PCB 300mm x 200mm and there was no ground connections as I made them all with a topside fill :( Learnt really fast on that one to check the Gerbers before sending them ;)
Fortunately KiCad now prompts you to refill pours before generating Gerbers. 300x200 4 layer ouch.
Thank you from a home hobbyist! Lots of good tips; some I’ll use right now (805 caps!!!)
Having designed hundreds of PCBs and providing designs for over a hundred more, I can totally agree with every single item mentioned here (when possible at least). Adding test points and bringing out unused microcontroller and FPGA pins has been a life saver countless times during initial testing. The global component shortage caused me to redesign several boards to a multi board design where we would swap out the say a microcontroller block for one using a different controller easily. It also makes testing for new solutions and upgrades easier. Probably the last thing I learned to do maybe a decade or so ago was to spend more time on the silk screen, not only does it help in troubleshooting (proto and production), it is very satisfying to have a great looking board. Kudos to an excellent video.
One more point, I often use SMD inductors or ferrite beads to isolate different parts of the power. Initially it was to keep noise from going to different sections, but it has been way helpful in isolating short circuits to specific sections, especially when a BGA is involved. We now place extra just for troubleshooting.
THANK YOU~~~~
Good video. I stay away from really tiny parts being a hobbyist... for complete beginners, you should mention all north/south wires go on one side of the PCB, and all east/west wires go on the other side... use vias to jump from one side to the other, when necessary... this keeps unused pathways open and available for future paths. Also provide open holes at a few places on the board, in case you need to run a jumper from one PCB side to the other. I also identify pin 1 on chips with a square pad, same for anodes of diodes and LEDs, and positive term of 'lytic caps. If you are using thru-hole components, make the drill sizes on the pads for the component leads as large as practical so it is easy to wick out all the solder, and replace a part without tearing out the thru plating on the board. To remove thru hole ICs, ( failed chips ) clip all the leads off and remove them one at a time.
I always place a text somewhere that identifies the exact filename of the PCB CAD file for that board, so I can get the source info for it. Also my ham radio call sign, or email address so others can find me with any questions. If there is any spare board space, I put extra footprints there for components and ICs, so I can add them "temporary" for circuit changes, using point-to-point wiring. If wires run off the board, try to make them all enter/leave along one edge of the board, so you can just turn the board over to get to the backside for troubleshooting... like turning the pages of a book. If wires run off 2 or more edges, that can't be done... and leave adequate slack in the wires that are installed, to permit this. Unless I know exactly where and how a PCB will be installed in an enclosure, I avoid using parts that will dictate how the board is mounted... like displays, switches, controls or other things a person must operate... because they MUST be mounted on an external panel or surface.
Use your imagination... put some forethought into it... don't just do the obvious stuff...
Give yourself some room for design mistakes and changes down the road.
Temporary thing end up being permanent, and permanent things.... aren't.
Excellent.
Extra GPIO are ideal for bringing out to a header so you can hook them to an inexpensive Salae logic analyzer or clone. There are many ways this is useful but my favourite is to use it to directly measure performance of code by toggling the GPIO at various points in the code flow! It's a terrific way to optimize or just verify operation in timing sensitive areas of operation.
that #3 hit hard
Great list. I would also add a few things. 1) Ignoring the DC bias of MLCC (learned that the hard way) 2) not paying attention to how rare is the part you are going to use: it's better to pick a more popular part that is present in large quantities in the warehouse instead of a 10% cheaper alternative that will disappear while you are designing PCB layout and you need to re-do that thing with a new part... 3) order parts FIRST and PCB - SECOND, same reason
Great video
Oh this is good advice.
Great set of design checks. LED's for power is a must, 0603 with 0.7mm height /0.18mm width silk is ok under a 10X microscope. I would add, if you make your own footprints, check its top down or bottom bottom up. Also Ive imported connector footprints into Altium where a tab was imported as a hole not a slot as the length was ignored. and pinouts for the same part can change between SC70 and sot-23.
I need to do bottom view in the next video.
Back in the 1980's I used wirewrap to build projects. Doing PCBs is so much cheaper these days. I made my first one in 2017 since the early 1990's. Many good point for beginners. I always use plenty if silk screen info to make things much easier for troubleshooting. KiCAD is a great program.
Thank you sir
I’ll add another one - print your board on paper with a simple printer. You’ll get the feel of the physical size of the thing and you can test fit components.
Back in the dark ages when I did PC board repair, I would routinely CUT up various failed PCB products. The idea was, WHY REINVENT the wheel?
If you need a power supply, just CUT it out of a FAILED product. If you needed a timer circuit, display circuit, or perhaps an I/O circuit, JUST find the needed circuitry on a failed PCB and CUT IT OUT with Tin Snips. I would then overlay the CUT UP CIRCUIT boards onto a piece of wood or perhaps a blank PC board and often use double sided tape to insulate and MOUNT the various pc boards into a working prototype with jumper wires.
Yep. FRANKENSTEIN DESIGNS ! But it was FAST, and for the most part - quite reliable. Best of all, since I got the parts at ZERO COST, the designs were very effective training aids. OK, they were FUN to build too. 🤪
Personal preference is to aim for prototypes (as opposed to development/have a play boards) to be as close to end target as possible. This is because it will help prove the practicalities of fitting the design in the target format and it will also give the closest representation of real world performance. Okay so it's probably not a big deal if it's a simple LED flasher, but with mixed mode and high speed circuits it can make all the difference. I also don't want to have to go through a major design revision just for the sake of making revision 1 quick and easy.
I think you could make a mockup project that includes the smaller footprints of every IC and part you intend to use just to keep an idea of the overall real estate of your design so you can know if it will fit within the enclosure. But I still affirm that trying to produce your first PCB to fit the final enclosure is going to guarantee multiple respins and end up taking longer than making a board that's easy to test. But we all have opinions. :)
I wrote up something similar on Reddit before Christmas. I've never seen anyone else discuss this before.
I am doing my first pcb right now and while most of these I didn't think of, I can proudly say I've got headers for all gpio's for testing and potential expansion.
Thanks! Recommend a good video guide on post-soldering diagnosis of what component wasn’t soldered properly and so on? Have assembled an SMT board that is known to be a functional design, but it wasn’t functioning. (Audio filter) Thanks in advance!👍🏻
Well glad to see that "i am marked safe from these mistakes" ofc i did learn the hard way though.
Awesome video! Was a amazing explanation only 10 years later 😂😂😂😂😂 dang!
If you have money for them (they can be a bit pricey), development boards are also really helpful for initial prototyping of specific functionality!
Always , always , always check your gerber file output throughly . I use a separate gerber viewer (gerbv) and check for things like spacing / crowding / obvious line widths / drill hole or connector mechanical violations / and silk screen conflicts .
I want to add onto #7 - use silkscreen as much as possible, but don't sacrifice board compactness when size matters just to fit in the silkscreen.
1:10 Hey I made that perfboard! It was when I just started out with electronics!
imagine scrolling through youtube so hard that you find yourself as stock image on a tutorial
I hope it was not taken as a slam or sleight but only as a "perfboards can get confusing and unwieldy quickly" advisory.
@@cosplaylightandsound4143 ah no worries!
Number 14 got me. But I was able to still bodge it and test the PCB, and my redesign will allow more compatibility with different component suppliers, so that's a plus.
My first major mistake on my first PCB design was not understanding that the Mask layer was a negative layer, though.
Yep, that board had no solder mask on it.
I've made the UART mistake before, but was saved because I had them on jumper resistors. A little fib wiring and it was fixed!