Dave should mirror the greenscreen Video from his head. This would allow him to look to the right side of his screen, wich would than match the PCBs right side. Now it’s „wrong“ he looks to the right but explains something on the left.
Young engineers will understand this and older ones might not but mentoring is so unbelievably important that I cannot stress it enough. This type of "Everyone grab a cup of coffee and head to the conference room in 5min" video is unbelievably helpful. Its literally like the boss told the group of junior engineers to head to the conference room to go over a random layout one of the new guys is doing to go over as a working example. This is crash course learning at its finest. Everything is boiled down to an essence. Dave, you are doing a service. The floating talking head BTW is perfect. Anything else for a video like this just wouldn't have worked. Everything on this video was 100. Thank you!
Yeah, can only agree and upvote. I'm a logistic-programmer, the learning for my job where mostly some word/powerpoint-documents i found in the company-network and a few small courses. And i came right when they changed the software and team-assignments, so when i had a problem nobody even knew who i could ask for help. When the next 2 guys came into our team i had saved those documents that took me a long time finding (and that since have been deleted !!! even) and made sure to be there when ever they had any problems, constantly looked over what had produced and also gave them tips of helpful things (for example we have some small helper-libraries for stuff you need to do over and over again - but there exists no documentation for that).
Reviews and mentoring is worth its time in pure gold, if it could be monetized. More eyes,different skill sets and (prior) experience is the best way to debug 'whoooopsies' I have a question though, why wouldn't older engineers understand the need for mentoring? - They for one should know how inexperienced new designers are, but I guess some people will give off the impression that they know and can do it all - and hence don't need mentoring and reviewing. Anyways, those conference room meetings which lasts only a few hours imho usually saves days and countless hours of fustrating work down the line.
@@EgonSorensen *" I have a question though, why wouldn't older engineers understand the need for mentoring?"* Cause to them (or you) it might be "obvious". had this a couple of times happen to me already despite already being very careful to clarify what i mean. One example is the terminology used in warehouses to describe the packages/cartons/whatever used to transport goods around the warehouse or out of it - there are several different abbreviations for that, some that are discouraged, some that have been used 15 years ago but supposedly no longer used and so on. I have worked with them, i know them, i don't think about them being anything worth mentioning. But then it happens that for a new person you give them something like "this has been done for XYZ already - you can look it up there and try to replicate it here" and then they wonder what the hell those 3-letter-thingies are. There is a reason why becoming a teacher takes quite some time. Once you have learned something you know how to do it - but that also means you are no longer in the position of not knowing even what to ask. Being a good mentor is not easy (and i certainly would not be a good one - i can only try to help for the few things i know they will struggle, but not nearly enough to get them started on my own)
@@ABaumstumpf "I guess some people will give off the impression that they know and can do it all - and hence don't need mentoring and reviewing." - this goes especially for both new and experienced people New engineers have their heads full of stuff they don't yet understand, making it hard to cram more stuff in there - so mistakes are (likely) made. It takes time to settle in. Old engineers have had their seats for long, starting to take things for granted and what they have learned has become 'practically understood' - and they might have forgot how it is to be new at a task, so they take things for given - I agree. I also have had the 3-letter word experiences, and it must have been a pain for some to explain what I asked about. It is a 'giving game' to walk the line, it's far from easy and not always fun and games - but then, that's engineering in a nut shell to me.
Older engineers will understand the value of this video. (Well, I do,.) It may actually be used in the conference room in some places. It should definitely be required viewing for any new engineers.
Absolutely agree. This is literally the boiled down essence. Its the boss grabbing all the new engineers to the conference room to go over a design. Its great.
Seriously, the way you look down to the PCB from this little floating head is so hilarious! I think it's great :D And cool content, thanks for that! I guess I kind of knew most of what you said in this video already, but it's still a very neat overview with a lot of reminders to the little details. Will probably help me with my next designs ;)
Thank you for this video, as an engineer i will highly appreciate more design videos and design's review as it gives more information to all of us. I will also highly appreciate more advanced and complicated things for (non beginner people). Thanks for all the information you gave us all these years.
The designer probably had them right the first time but then decided to turn the chip 90° for signal routing "optimisation"?, but did not move the caps with it. Just guessing ofcourse...
These are definitely my favourite EEV videos. Great for two reasons; reinforces the things I do know, and it's rare I'll watch one and not pick up 2 or 3 new things. Love it. Keep up the great work.
I'll often connect bypass caps to an ICs power pins but not connect the pins to power until after an initial placement of components on the pcb, that way the rast nets clearly show what pin an cap belongs together
This kind of video was about 50% of the reason I subscribed. Thanks, it's been a while. I especially like your ideology for routing traces. I will soon be laying out the first 4 layer PCB I've done in years and this type of video helps with remembering how to make the decisions with prioritization and overall layout.
Green-screen looks great. 👍🏽 Can it be mirrored so your head moves with the point-of-interest? 9:35 you noticed it, my learning style really appreciates you using this.
These types of videos I really enjoy - It's almost like having Dave as a mentor explaining what to do. Regarding the decoupling caps, I always like to think of it as knocking out any inductance that's introduced from the long trace from the power supply to the chip, placing the cap as close to the IC as possible means the trace post cap is super short (and therefore very low inductance!). But, the discussion on things like layer stack up and WHY you should have it that way is priceless, I'll definitely be watching this video again when I need to lay out another board!
I have graduated a few months ago, but I've known you for years now. you are my role model Dave, please keep doing these. I have learned stuff that would've taken years of mistakes from these videos. thank you.
It looked like he tried to do the layout for a 2 layer PCB, then had to later change the design. Those 2-layer power traces definitely need to go in favor of vias to the power planes.
This was great, I'm working on a design now and you gave me heaps of tips I may not have thought of. Informative videos that help people with their own work are my favourite kind.
Your PCB layout videos gave me the confidence to design and build a PCB for a small project a few months ago, so thank you for the excellent content! Now that I've been through that and have a grasp on more of the basics, I'm finding that I understand and even predict more of your advice. I'm looking forward to trying my hand at more complex boards in the future!
This video is perfect! I always find myself learning the most by looking at existing electronics and studying how the traces are connected. Love the content, hope to see more!
This is absolute gold. Fantastic content. I'm a newbie at doing anything with this stuff, and all these points that Dave makes provide a direction for how I can get better.
I like how you’re talking a lot about having a best practice mindset. If you consistently do something even when it’s not required it becomes a habit. I’m still new in designing my own boards so this helps a lot. It’s often the simple things that get you until someone points it out. Oh and the green screen is a definite winner! I feel it all looked better to follow along this way. Winner winner, chicken dinner 😜
The USB-to-Serial chip is CH340, and its CH340C version is exactly the same SO-16 as others, but doesn't need an external crystal, it has an internal one, the XTAL pins just stay NC! The chip is available on LCSC
Internal oscillators have fairly wide frequency swings over temperature. I'd prefer to use the crystal for the much tighter frequency tolerance. USB doesn't like varying signal rates.
I've done several hundred boards with the 340C and generally it's been good for me. However I did have this one customer who absolutely could not get the products to operate on one of his PC. It could have been many reasons - including this.
@@Dave5281968 Wouldn't you rather use the host's time base in the device? Guaranteed to be ±0 variance from the host's time base, once the x48000 PLL locks.
I find these types of videos to be incredibly helpful and informative. I studied EE over 20 years ago and eventually ended up being a programmer for the last 2 decades, but I've always been a hobbyist in electronics. It's easy to build all kinds of circuits today with resources like Adafruit and Sparkfun modules. Just stitch them together and add a bit of code, but sometimes you have something that you'd really like to turn into your own board and this practical examination of a real world project is awesome. I'd like to see a whole series of videos aimed at the amateur that covered duplicating the functionality of these modules into a custom board (within copyright and design practices).
As a maker I am just about to plan my first PCB. Some days ago i did not have a clue how do that. So watched ca. 30 videos here on youtube... They all turned out to be just reportages. By that i mean they did not succeed to teach anything useful, i literally just could watch other people using some SW to draw a board, but none of them explained why and how they make design decisions, rendering all that stuff useless for educational purposes. Then i stumbled upon your videos about capacitors and then this one. I just can say thank you for these! Watching just some of your content made me understand the very basics of PCB design in a way that i think i have good starting point for now.
I really liked this one, and as others have said, covers off some basics that are common mistakes, even by people that have been producing boards for years now!
Thank you so much I'm a young electronical engineer and this will be sooo helpfull. PCB design is not well teached in school I think, all thesee littles tricks are new for me. Thank you Dave
I'm working on my first real complex PCB and this video will probably keep giving me good info along the way, as well as a good chuckle - floating Dave is great!
Dave, the green screen is a great upgrade...much more flexibility and makes room for activities. I'm a shade-tree electronics hobbyist and I'm starting do build my first board in Kicad. It doesn't require any signaling, but I will definitely use your suggestions if I ever get to that point. Thanks.
I never run tracks between pins, like on that lower left, unless absolutely necessary. It's easy to end up with whiskers to be left when etching the board. It's also not ideal to be soldering either side of the through trace because you can get a solder bridge if the solder mask isn't perfect. There's no need to have done that since you can route the power to avoid it. This board uses very few hole sizes which is good. However, make the vias as big as you can, because small drills, say 0.4mm or less, have much shorter flute lengths than bigger ones. PCB manufacturers typically want to drill through 3 x 1.6mm boards in a stack because drilling takes a lot of time. As soon as you drop below 0.4mm, you're pretty much forcing them to drill them 2 or even one high. Ask your PCB manufacturer about this sort of thing, you can save a lot of money. Decades ago, all PCBs were drilled and routed 4 PCBs high, and via holes were 0.7mm diameter to allow that. If you have lots of through holes, try to rationalise them to use the same sizes wherever possible. Again, PCB drilling machines have to stop the spindle, swap tools and then restart which all takes time. It's also another possible source of error, because drills can look the same size. Another thing we used to do was to drill the first and last hole of every size in the scrap outside of the circuit frame. That way a quick visual check can spot if a drill broke. If there are thousands of holes, you won't spot this until much later in the process. Again, ask your PCB manufacturer for their best design practices, and better still, go and visit them and sit down with the people who are the victims of your poor design decisions. Use wide tracks rather than unnecessarily narrow ones, and don't route them up tight against things when they don't need to be. Run a gap check, and increase the clearances it flags up. Increase the allowable gap clearance and do it again. That way you can make the board easier to make and have less chance of shorts or interference.
I have a renewed respect for computer motherboards now. Routing traces seems to be quite complicated, with a lot to consider. I learnt a lot from this video - thank you!
I can't even fathom the amount of work in routing multiple BGA devices with hundreds to thousands of connections on each. I'm sure it takes large teams to design PC motherboards. I know I will never even attempt it or anything even close to it. Too confusing. Too many wires to route.
I would really love to see you layout this circuit. Not that you need a complete stranger assigning you work but a before-and-after on this thing would be really informative. Thanks for the explanation.
I think the talking head is great Dave, Worked well. Nice video which is now making me revisit some of the stack-up in a 4 layer PCB I am currently working on so thanks.
Another bit of advice for newbies, if you can't figure out how to lay it out just start(call it a draft). Like Dave said don't be afraid to move components to make routing easier. Don't be afraid to toss it all out and re do the layout/ remove all the traces and try again often after you have move components around to fix issues.
A lot to unpack here. Upfront, I worked at Agilent in Mulgrave for 27 years, the last 12 as a PCB designer (ECAD Engineer) for our EE engineers. The last 5 years as their only dedicated pcb designer. We used Protel/Altium exclusively up until I left 6 years ago, now retired. They have now mistakenly gone corporate with Mentor Graphics. That’s another story. First off, top layer should be BLUE, not red. Red is the Altium default and is wrong. Blue top stems from the old red and blue Bishop tape days whereby the positives and negatives used the colours to determine which was the top and bottom layer. This was of course for 2 layer THT (Through Hole Technology) pcbs. Yes, it’s only a colour but standards should have been kept. Next, use metric, NOT imperial. All Agilent pcbs were designed in mm, not mils. The world is metric, not American standard imperial. All through this video I had trouble trying to convert mils to mm. The only imperial that we used at Agilent was oz for copper weight but even that we usually specified as 18 or 37um. I.e. thickness rather than a silly oz per sq foot or whatever it is. Now to the video… • The pcb should have a ~5mm edge around the pcb for smt (Surface Mount Technology) placement machines to grip the pcb securely. Only reason not to is if the pcb is part of a panel with routed breakoffs or defined routing afterwards. • I agree the cap locations around the image chip are terrible. • A major rule to follow in design is “follow the current, follow the current, follow the current”. • Traces at acute angles are the single worst error in pcb design. I guess the creator used auto-routing , terrible idea. • THT crystals are quite common for reliability and getting the right type of crystal, they’re ok. • 4 Layer is fine for this. It could have been done with 2 layer but with EMC in mind and the fact that 4L is only 30-50% higher in cost than 2L then it’s a good idea. Even in very complex pcbs like PC motherboards you very rarely need to go over 4L. I designed over 100 pcbs I reckon with many large designs and I only used a 6L or higher about a half dozen times, and these are A4 size pcbs with many many chips. • Altium footprints are not optimum, even though they say they are PbF and IPC “standard”. They aren’t truly PbF pads and courtyards don’t allow for height when heat flowing solder • For the large FPGA, yes they are usually designed to be at 45deg if using lots of pins such as in a PCI card. They’re designed that way. • The usual chip bypass is 100nF but I agree it is negotiable. Also remember that vias to the gnd and power plane will provide HF bypass which might negate the requirement for 10nF caps. Another method is vias inside a chip’s smt footprint but that’s a whole new ballgame. • I agree that a large cap used as bulk capacitance and the local power plane capacitance might be enough, maybe not. • Don’t use auto-routing, it’s more trouble than it’s worth unless you are dealing with a really large design that is time constrained, but even then you need to do most of the placing and power planning beforehand. Auto-routing loves 6+ layers. It’s getting better but just cannot easily do analogue stuff properly. • I did not see any gnd or +v vias around the chips, but I suppose it was early in the design. Design is a compromise between signal trace coming out and close placement of power cap and its vias. • I personally could see no value in wiggly trace lengths of these signals. I may be wrong. Remember 150mm is 1ns. I think about 30mm was the approx. length which equates to about 200ps. Is that a big deal in this circuit, I doubt it. I may be wrong. • I agree the signal traces should have been the priority on the top layers. Vias are impedance interruptions which cause reflections etc and cause more trouble than trace length IMO. • This is a simple design and in my day I would have finished this design, done the BOM (Bill of Materials) and schematic cleaned up and exported ready for the pcb manufacturer to start work in half a day. My work was always using an EE engineer’s schematic design with which we collaborated in the pcb design phase. • As mentioned by Dave I would possibly have added a GND fill on bottom and top layer. We NEVER relied on a GND fill to ensure proper grounding of SMT pads. Always use GND vias to power caps. • Yes, top layer signals allow modifications of signal tracks if required. • There are many choices of grounding here. Star, gnd fill and power planes. We must look at the current flows and “follow the current”. BTW, Follow the current saying comes from a very famous HP designer many moons ago who quoted this to all of his pupils. • Vias to pwr planes is maybe more important than external smt caps. • Less vias is less cost. 0.6mm via holes use up a lot of drill bits if you have lots of vias
I use the Saturn PCB toolkit too! It's such a great tool. Other observations: - length matching isn't what's important; it's time matching. Different layers will have different propagation speeds, so if possible, all the signals should go on the same layers (e.g. all go L1 then L4 then L1) and ideally the on a layer the lengths should be matched (e.g. each trace has 500mils on L1 and 400 mils on L2). Also remember that vias add length, so either take it into account or have the same amount of vias. - when changing layers on a high speed signal, there should be a stitching capacitor (between the two reference layers) near the via, to give an AC return path - DC/DC converter routing is poor, but this might be a work in progress with temporary traces to show that it routes ok and eventually will add regions/polygon pours - looks like the xtal case might touch the usb shield pad. Probably not an issue. - use thicker traces for bypass/decoupling caps. 12-20 mils - traces are really close to your mounting holes - hard to tell but solder mask sliver looks too thin on the MCU - very often the thermal pads are supposed to be connected to ground; check the datasheet. - you've got tons of space at the bottoms, consider adding surface testpoints at the bottom); it's free real estate - high speed signals going to middle layers cause stubs at the via (e.g. if you go from L1 to L2, you have a stub between L2 and L4) - looks like all your edges are going to be routed, so you'll need mousebites for your arrays, but I don't see many places when you'll be able to put them because your traces run so close to the edges - don't forget to add fiducials. They can be on your array, but you probably want to have one near the BGA. Preferably at the top right of the BGA so it's also close to the MCU.
I recently had to design a very compact PCB with multiple ground and power areas and eventually, out of frustration, decided to go to 4-layer with grounds and power in the inner planes. It was absolutely worth it. This kind of design in 2-layer is an absolute nightmare. It would be helpful if you could talk about how best to do multiple digital/analogue/power grounds.
7:45 Hey Dave! The buck converter layout is NOT GOOD! The switching node and the trace to the inductor is too long. Also, the output for the feedback is taped at the wrong place. Furthermore, there are no power planes and the input and output capacitors are not close to each other. The only thing which is good, is that all power traces are on the top layer.
@@MetalheadAndNerd Hi! Please check out this video: ua-cam.com/video/gq-0ZpcGm8E/v-deo.html There it is explained in detail. There are two current loops, one if the switch is cloased and one if the switch is open. You want to have the DIFFERENCE as small as possible and his will be achived, if the negative terminals of Cin and Cout are close to each other.
Haven't been at the channel for months. This video on its own is a quintessential distilled practical knowledge of PCB design. Learned a lot. Anyway, may I ask someone to suggest a good, up-to-date book (or course) for PCB design, please?
Maybe you could do a video where you actually fix the design, would be nice for the guy and would be nice for us watching as we would see the design progressively improve
That would require a heck of a lot more time, you know - many hours. For we visual learners, it isn't necessary. I can see that for others it would be valuable. Also, it would require Dave needing the source files as well as be familiar with and have all the software.
The DC/DC could also be improved. The input caps are far away, they should be priority one. Place them as close as possible to the IC and even connect their GND-pad directly to the IC without any vias. Also do not connect the feedback line to the inductor, connect it to the output cap instead. Additionally, I would rotate the chip to make the switching node as short as possible (prio 2 directly after the input cap).
From a video production perspective, and for teaching purposes seeing, you head really helps. Great job! Ideally flip the head video left-right (mirror), so that your movement matches the actions in the demonstration area. Ie. When you look after camera, and then bring in a data sheet overlay, they come in from the same side.. When you point at components in the corners, your head/eyes are tracking to the same corner.
Most of the issues with this design come from routing power first. Which for whatever reason is a common thing told to beginners to do in a lot of online courses (Udemy etc.). Have seen a couple of newbie engineers I work with do this.
Maybe it's a throwback from one- or two-side PCB times when one didn't have the luxury to just drop to a power plane on the spot wherever needed. After all, pesky chips with 4-6-8 pins needing all connected to a non-plane power trace cutting off everything else as they wind around the chip can really ruin your day if you start by laying out your signal traces first only to realize there's no way to connect all the required power at the end without redoing absolutely ALL the other traces from scratch. This has even become worse recently ever since THT (where as Dave mentioned you get a free via on each pin which will get soldered in anyway) got replaced by SMD, assuming you'd rather stay on a single side which is much simpler to produce at home without the headache that DIY vias incur (once you're ok with having vias and a two-side PCB it stops mattering as much).
Beautiful example of why FPGA bypass caps are often placed on the bottom layer. Signal lines first, then drop vias to the bypass caps, and Bob's your auntie.
Great video, I learned a lot about multi layer boards. Floating Dave is awesome, but don't forget to mirror the cam image to match the screen content :)
Great video as always! One thing that struck me though that I was kind of expecting to be mentioned is the placement of the FPGA and converter each placed on opposite side of the image sensor Just switching place between the uC and FPGA would help shift the ground-plane return currents (assuming there is a solid ground plane ofc) of the FPGA from passing directly under the image sensor But thats just my own humble opinion :) Keep up the great work!
Great video as always. Its a good idea to consider how your going to distribute the power during placement but I always route most to least critical IO then I circle back to the pdn.
Dave. Great video, particularly when taken with other ones that you've made on PCB layouts. I too like the floating head. It would be educational to see your routing solution, even if incomplete, for this project. Of course I realise that this would be counterproductive for soFPG and for your time. Perhaps soFPG will put a revised layout up for further review because a followup would definitely be of interest
I love this as a video segment :D It's a great teaching aid, plus people can submit their own designs! I know I got some PCB's I'd love you to have a look at :) Also there's more immediacy to it, unlike a mailbag where stuff can get backlogged for months cause of delivery times and so on
I used I2C IO extender for arduino project, it was around 1,5m away, coupled with standart cable, going power for it in same cable. Just needed to to drive data lanes bit harder (lower termination resistor values), working just fine, no need to worry about I2C lines on the PCB, they will never be so long. Just imagine how big a PCB must be for lines to be over 1,5m xD
Problems with I2C will usually show up quickly if you go do your proper EMC testing. It's easy to get your communication completely messed up when you apply some RF fields to your designs. Sure, it might work in your lab and on a bench, but if you have some radio emitters (like cell phones) nearby, your device might suddenly act all crazy... Seen that happen quite a lot on I2C, even on relatively small 4 layer PCBs with proper circuit and layout designs and ground planes.
Designing my first board layout since uni for a project, the first iteration on hobby perf board suffers from cross talk between the output and feedback sections as well as the inputs. I have a lot to learn still. “incoming message from the big giant head!” vibes with the floating head, I like it.
The Top-Power-Ground-Bottom stackup is terrible for EMI, as it places the power planes far away from one another. This acts a lot like a slot antenna and radiates noise out the sides of the PCB. This can be disastrous for high-speed designs. A much better 4-layer stackup would be Top/Power-Ground-Ground-Bottom/Power
I have to say Dave, the green-screen was worth every penny in this vid, excellent placement!
I thought so! I'm enjoying the green screen, I think it works pretty well.
The placement just made sense here as there were no tracks in that spot.
@@EEVblog You should have flipped only horizontally, to mach your movements ;)
@Aidan Macgregor ROFL this is to prepare us to the red dot when Dave will be a transhuman.
@@b_force4079 Yep, agreed.
Dave should mirror the greenscreen Video from his head. This would allow him to look to the right side of his screen, wich would than match the PCBs right side.
Now it’s „wrong“ he looks to the right but explains something on the left.
Young engineers will understand this and older ones might not but mentoring is so unbelievably important that I cannot stress it enough. This type of "Everyone grab a cup of coffee and head to the conference room in 5min" video is unbelievably helpful. Its literally like the boss told the group of junior engineers to head to the conference room to go over a random layout one of the new guys is doing to go over as a working example. This is crash course learning at its finest. Everything is boiled down to an essence.
Dave, you are doing a service. The floating talking head BTW is perfect. Anything else for a video like this just wouldn't have worked. Everything on this video was 100.
Thank you!
Yeah, can only agree and upvote.
I'm a logistic-programmer, the learning for my job where mostly some word/powerpoint-documents i found in the company-network and a few small courses. And i came right when they changed the software and team-assignments, so when i had a problem nobody even knew who i could ask for help. When the next 2 guys came into our team i had saved those documents that took me a long time finding (and that since have been deleted !!! even) and made sure to be there when ever they had any problems, constantly looked over what had produced and also gave them tips of helpful things (for example we have some small helper-libraries for stuff you need to do over and over again - but there exists no documentation for that).
Reviews and mentoring is worth its time in pure gold, if it could be monetized. More eyes,different skill sets and (prior) experience is the best way to debug 'whoooopsies'
I have a question though, why wouldn't older engineers understand the need for mentoring?
- They for one should know how inexperienced new designers are, but I guess some people will give off the impression that they know and can do it all - and hence don't need mentoring and reviewing. Anyways, those conference room meetings which lasts only a few hours imho usually saves days and countless hours of fustrating work down the line.
@@EgonSorensen *"
I have a question though, why wouldn't older engineers understand the need for mentoring?"*
Cause to them (or you) it might be "obvious".
had this a couple of times happen to me already despite already being very careful to clarify what i mean. One example is the terminology used in warehouses to describe the packages/cartons/whatever used to transport goods around the warehouse or out of it - there are several different abbreviations for that, some that are discouraged, some that have been used 15 years ago but supposedly no longer used and so on. I have worked with them, i know them, i don't think about them being anything worth mentioning.
But then it happens that for a new person you give them something like "this has been done for XYZ already - you can look it up there and try to replicate it here" and then they wonder what the hell those 3-letter-thingies are.
There is a reason why becoming a teacher takes quite some time. Once you have learned something you know how to do it - but that also means you are no longer in the position of not knowing even what to ask. Being a good mentor is not easy (and i certainly would not be a good one - i can only try to help for the few things i know they will struggle, but not nearly enough to get them started on my own)
@@ABaumstumpf "I guess some people will give off the impression that they know and can do it all - and hence don't need mentoring and reviewing."
- this goes especially for both new and experienced people
New engineers have their heads full of stuff they don't yet understand, making it hard to cram more stuff in there - so mistakes are (likely) made. It takes time to settle in.
Old engineers have had their seats for long, starting to take things for granted and what they have learned has become 'practically understood' - and they might have forgot how it is to be new at a task, so they take things for given - I agree.
I also have had the 3-letter word experiences, and it must have been a pain for some to explain what I asked about. It is a 'giving game' to walk the line, it's far from easy and not always fun and games - but then, that's engineering in a nut shell to me.
Older engineers will understand the value of this video. (Well, I do,.) It may actually be used in the conference room in some places. It should definitely be required viewing for any new engineers.
I feel like analyzing designs & workflows is a really underrated kind of learning. Great video & thank you for sharing
Yes, almost nothing more valuable than competent peer review and critique!
Absolutely agree. This is literally the boiled down essence. Its the boss grabbing all the new engineers to the conference room to go over a design. Its great.
Especially workflows...
So true, would have killed for a class dedicated to this stuff in undergrad.
Seriously, the way you look down to the PCB from this little floating head is so hilarious! I think it's great :D
And cool content, thanks for that! I guess I kind of knew most of what you said in this video already, but it's still a very neat overview with a lot of reminders to the little details. Will probably help me with my next designs ;)
I like it too!
Some have said they find it weird and don't like it.
@@EEVblog
I love it, don't listen to everybody else!
@@EEVblog We can ignore those people, they are wrong.
@@EEVblog dave you gotta make the floating head into its own series, hilarious and educational
@EEVblog Hi Dave, would you tell us your capture software/camera setup specifically for this type of video? It’s quite an interesting format
Thank you for this video, as an engineer i will highly appreciate more design videos and design's review as it gives more information to all of us.
I will also highly appreciate more advanced and complicated things for (non beginner people). Thanks for all the information you gave us all these years.
This series could happen more often, this is great learning from Dave.
Thanks, Dave! I'm just in the curious stage with wanting to start designing a PCB sometime in the future, and stuff like this is really fascinating.
I have plenty of other PCB layout tutorials as well.
@@EEVblog Thanks! I'll check them out.
This is the kind of video that keeps me coming back. Awesome Dave!
Those decoupling caps around the image sensor looks like they are playing musical chairs.
Last one down gets the inductance!
@@EEVblog I think all of them got the inductance in this case!
@@EEVblog Is inductance contagious? In the new normal, it needs to be asked.
I was reminded of a May-pole dance. But where are the zeroes?
The designer probably had them right the first time but then decided to turn the chip 90° for signal routing "optimisation"?, but did not move the caps with it. Just guessing ofcourse...
it's 4 am at my local timezone and this video drops. nothing better to watch at the moment. i have no life.
Nerd.
I've just watched it at 3am haha
Same. Its 3:32am UK time when typing this comment
Great video! 10 most important PCB design rules by showing how not to do it. And after you explain it it's all so obvious. Thanks Dave!
Dave Jones is a great teacher and mentor. Very well spoken.
These are definitely my favourite EEV videos. Great for two reasons; reinforces the things I do know, and it's rare I'll watch one and not pick up 2 or 3 new things. Love it. Keep up the great work.
Always look forward to these Dave as you do point out the real nuts and bolts of good PCB layout Thanks
I'll often connect bypass caps to an ICs power pins but not connect the pins to power until after an initial
placement of components on the pcb, that way the rast nets clearly show what pin an cap belongs together
This kind of video was about 50% of the reason I subscribed. Thanks, it's been a while.
I especially like your ideology for routing traces. I will soon be laying out the first 4 layer PCB I've done in years and this type of video helps with remembering how to make the decisions with prioritization and overall layout.
My boss is giving me my first PCB design project starting next week, this is perfectly timed!
How did it go?
this is the sort of stuff that made me watch Dave all these years
This is an excellent piratical tutorial. I learned a lot in something I haven't had to think of in decades.
Floating Dave ist ABSOLUTELY GREAT!!! Makes a very clean image and much room for the content.
Green-screen looks great. 👍🏽 Can it be mirrored so your head moves with the point-of-interest? 9:35 you noticed it, my learning style really appreciates you using this.
Good point. Yes it can be mirrored easily.
Floating head looks awesome! :D Thanks for fantastic video! Lots of knowledge for newbies. Waiting for more such content.
These types of videos I really enjoy - It's almost like having Dave as a mentor explaining what to do. Regarding the decoupling caps, I always like to think of it as knocking out any inductance that's introduced from the long trace from the power supply to the chip, placing the cap as close to the IC as possible means the trace post cap is super short (and therefore very low inductance!). But, the discussion on things like layer stack up and WHY you should have it that way is priceless, I'll definitely be watching this video again when I need to lay out another board!
I have graduated a few months ago, but I've known you for years now. you are my role model Dave, please keep doing these. I have learned stuff that would've taken years of mistakes from these videos. thank you.
As someone who is only getting started, this is absolutely priceless. Thank you so much for making these.
Loved it!
And the casual ".. and Bob's your uncle." LOL 🤣
I'd really like to see more of this.
Cheers!
All good advice Dave. The power wandering around on a 4 layer board instantly threw me for a loop (pun intended). 🙂
It looked like he tried to do the layout for a 2 layer PCB, then had to later change the design. Those 2-layer power traces definitely need to go in favor of vias to the power planes.
This was great, I'm working on a design now and you gave me heaps of tips I may not have thought of. Informative videos that help people with their own work are my favourite kind.
thank you for the video, we really need more video like that. maybe some schematic review.
I did a schematic recently, but it was more BOM related.
Yes, I agree. A schematic design video would be great with some "best practices" for larger multi-page organization.
Your PCB layout videos gave me the confidence to design and build a PCB for a small project a few months ago, so thank you for the excellent content!
Now that I've been through that and have a grasp on more of the basics, I'm finding that I understand and even predict more of your advice. I'm looking forward to trying my hand at more complex boards in the future!
As many other comments have said - great video! Love the content, very educational and informative. Also love the green screen!
This video is perfect! I always find myself learning the most by looking at existing electronics and studying how the traces are connected. Love the content, hope to see more!
This is absolute gold. Fantastic content.
I'm a newbie at doing anything with this stuff, and all these points that Dave makes provide a direction for how I can get better.
Please do more of this....very interesting.
I like how you’re talking a lot about having a best practice mindset. If you consistently do something even when it’s not required it becomes a habit.
I’m still new in designing my own boards so this helps a lot. It’s often the simple things that get you until someone points it out.
Oh and the green screen is a definite winner! I feel it all looked better to follow along this way.
Winner winner, chicken dinner 😜
I like the Floating Head overlay on the PCB graphics...Great PCB basics tutorial.
The USB-to-Serial chip is CH340, and its CH340C version is exactly the same SO-16 as others, but doesn't need an external crystal, it has an internal one, the XTAL pins just stay NC! The chip is available on LCSC
Hmm I heard people say it's very unstable though, don't know why. That's why I just used the G version in a board a while ago
Internal oscillators have fairly wide frequency swings over temperature. I'd prefer to use the crystal for the much tighter frequency tolerance. USB doesn't like varying signal rates.
I've done several hundred boards with the 340C and generally it's been good for me. However I did have this one customer who absolutely could not get the products to operate on one of his PC. It could have been many reasons - including this.
@@Dave5281968 Wouldn't you rather use the host's time base in the device? Guaranteed to be ±0 variance from the host's time base, once the x48000 PLL locks.
@@AirzonesBlasters Yikes
dave thank you so much. you are so good person for productivity.
I find these types of videos to be incredibly helpful and informative. I studied EE over 20 years ago and eventually ended up being a programmer for the last 2 decades, but I've always been a hobbyist in electronics. It's easy to build all kinds of circuits today with resources like Adafruit and Sparkfun modules. Just stitch them together and add a bit of code, but sometimes you have something that you'd really like to turn into your own board and this practical examination of a real world project is awesome.
I'd like to see a whole series of videos aimed at the amateur that covered duplicating the functionality of these modules into a custom board (within copyright and design practices).
Same here but 30 years ago.
Great, engaging content! Like the layout analysis and how it can be improved. Thank you for sharing!
As a maker I am just about to plan my first PCB. Some days ago i did not have a clue how do that. So watched ca. 30 videos here on youtube... They all turned out to be just reportages. By that i mean they did not succeed to teach anything useful, i literally just could watch other people using some SW to draw a board, but none of them explained why and how they make design decisions, rendering all that stuff useless for educational purposes. Then i stumbled upon your videos about capacitors and then this one. I just can say thank you for these! Watching just some of your content made me understand the very basics of PCB design in a way that i think i have good starting point for now.
I really liked this one, and as others have said, covers off some basics that are common mistakes, even by people that have been producing boards for years now!
Green screen talking head rocks! Noticed it earlier in the video. Another great down under video! Thanks!
I love the green screen and floating Dave! always learn a lot from these, thanks!
Thank you so much I'm a young electronical engineer and this will be sooo helpfull. PCB design is not well teached in school I think, all thesee littles tricks are new for me. Thank you Dave
I'm working on my first real complex PCB and this video will probably keep giving me good info along the way, as well as a good chuckle - floating Dave is great!
Dave has finally assimilated himself into a PCB!
Great video! Right to the point. Signal, GND, PWR, Bottom. Yep. And think of SMD as single layer. Right to the point!
Needs a green shirt for best floating head effect. :D
Turtleneck please
i really appreciate this content as a layout engineer.
WOW! \Some proper electronics on EEVB - been a while ;-)
Green Screen Excellent , Video Content Excellent. More of these Please
This is great Dave. I just finished a 4 layer board and I’ve picked up some good tips here for my next revision.
this content earns you my subscription, thank you
Dave, the green screen is a great upgrade...much more flexibility and makes room for activities.
I'm a shade-tree electronics hobbyist and I'm starting do build my first board in Kicad. It doesn't require any signaling, but I will definitely use your suggestions if I ever get to that point. Thanks.
I never run tracks between pins, like on that lower left, unless absolutely necessary. It's easy to end up with whiskers to be left when etching the board. It's also not ideal to be soldering either side of the through trace because you can get a solder bridge if the solder mask isn't perfect. There's no need to have done that since you can route the power to avoid it.
This board uses very few hole sizes which is good. However, make the vias as big as you can, because small drills, say 0.4mm or less, have much shorter flute lengths than bigger ones. PCB manufacturers typically want to drill through 3 x 1.6mm boards in a stack because drilling takes a lot of time. As soon as you drop below 0.4mm, you're pretty much forcing them to drill them 2 or even one high. Ask your PCB manufacturer about this sort of thing, you can save a lot of money. Decades ago, all PCBs were drilled and routed 4 PCBs high, and via holes were 0.7mm diameter to allow that.
If you have lots of through holes, try to rationalise them to use the same sizes wherever possible. Again, PCB drilling machines have to stop the spindle, swap tools and then restart which all takes time. It's also another possible source of error, because drills can look the same size.
Another thing we used to do was to drill the first and last hole of every size in the scrap outside of the circuit frame. That way a quick visual check can spot if a drill broke. If there are thousands of holes, you won't spot this until much later in the process.
Again, ask your PCB manufacturer for their best design practices, and better still, go and visit them and sit down with the people who are the victims of your poor design decisions.
Use wide tracks rather than unnecessarily narrow ones, and don't route them up tight against things when they don't need to be.
Run a gap check, and increase the clearances it flags up. Increase the allowable gap clearance and do it again. That way you can make the board easier to make and have less chance of shorts or interference.
I have a renewed respect for computer motherboards now. Routing traces seems to be quite complicated, with a lot to consider. I learnt a lot from this video - thank you!
I can't even fathom the amount of work in routing multiple BGA devices with hundreds to thousands of connections on each. I'm sure it takes large teams to design PC motherboards. I know I will never even attempt it or anything even close to it. Too confusing. Too many wires to route.
i just love your videos, you have such a nice way to explain things even complicated topics
I would really love to see you layout this circuit. Not that you need a complete stranger assigning you work but a before-and-after on this thing would be really informative. Thanks for the explanation.
I think the talking head is great Dave, Worked well. Nice video which is now making me revisit some of the stack-up in a 4 layer PCB I am currently working on so thanks.
Helpful video. I liked it
Great video. I would love to see the final revision of the board after your advise.
Great video. I've noted so many crucial points that you have highlighted. Definitely learnt something from this video. Many thanks :)
This was so insightful! Thank you so much for your videos sir.
Another bit of advice for newbies, if you can't figure out how to lay it out just start(call it a draft).
Like Dave said don't be afraid to move components to make routing easier.
Don't be afraid to toss it all out and re do the layout/ remove all the traces and try again often after you have move components around to fix issues.
A lot to unpack here.
Upfront, I worked at Agilent in Mulgrave for 27 years, the last 12 as a PCB designer (ECAD Engineer) for our EE engineers. The last 5 years as their only dedicated pcb designer. We used Protel/Altium exclusively up until I left 6 years ago, now retired. They have now mistakenly gone corporate with Mentor Graphics. That’s another story.
First off, top layer should be BLUE, not red. Red is the Altium default and is wrong. Blue top stems from the old red and blue Bishop tape days whereby the positives and negatives used the colours to determine which was the top and bottom layer. This was of course for 2 layer THT (Through Hole Technology) pcbs. Yes, it’s only a colour but standards should have been kept.
Next, use metric, NOT imperial. All Agilent pcbs were designed in mm, not mils. The world is metric, not American standard imperial. All through this video I had trouble trying to convert mils to mm. The only imperial that we used at Agilent was oz for copper weight but even that we usually specified as 18 or 37um. I.e. thickness rather than a silly oz per sq foot or whatever it is.
Now to the video…
• The pcb should have a ~5mm edge around the pcb for smt (Surface Mount Technology) placement machines to grip the pcb securely. Only reason not to is if the pcb is part of a panel with routed breakoffs or defined routing afterwards.
• I agree the cap locations around the image chip are terrible.
• A major rule to follow in design is “follow the current, follow the current, follow the current”.
• Traces at acute angles are the single worst error in pcb design. I guess the creator used auto-routing , terrible idea.
• THT crystals are quite common for reliability and getting the right type of crystal, they’re ok.
• 4 Layer is fine for this. It could have been done with 2 layer but with EMC in mind and the fact that 4L is only 30-50% higher in cost than 2L then it’s a good idea. Even in very complex pcbs like PC motherboards you very rarely need to go over 4L. I designed over 100 pcbs I reckon with many large designs and I only used a 6L or higher about a half dozen times, and these are A4 size pcbs with many many chips.
• Altium footprints are not optimum, even though they say they are PbF and IPC “standard”. They aren’t truly PbF pads and courtyards don’t allow for height when heat flowing solder
• For the large FPGA, yes they are usually designed to be at 45deg if using lots of pins such as in a PCI card. They’re designed that way.
• The usual chip bypass is 100nF but I agree it is negotiable. Also remember that vias to the gnd and power plane will provide HF bypass which might negate the requirement for 10nF caps. Another method is vias inside a chip’s smt footprint but that’s a whole new ballgame.
• I agree that a large cap used as bulk capacitance and the local power plane capacitance might be enough, maybe not.
• Don’t use auto-routing, it’s more trouble than it’s worth unless you are dealing with a really large design that is time constrained, but even then you need to do most of the placing and power planning beforehand. Auto-routing loves 6+ layers. It’s getting better but just cannot easily do analogue stuff properly.
• I did not see any gnd or +v vias around the chips, but I suppose it was early in the design. Design is a compromise between signal trace coming out and close placement of power cap and its vias.
• I personally could see no value in wiggly trace lengths of these signals. I may be wrong. Remember 150mm is 1ns. I think about 30mm was the approx. length which equates to about 200ps. Is that a big deal in this circuit, I doubt it. I may be wrong.
• I agree the signal traces should have been the priority on the top layers. Vias are impedance interruptions which cause reflections etc and cause more trouble than trace length IMO.
• This is a simple design and in my day I would have finished this design, done the BOM (Bill of Materials) and schematic cleaned up and exported ready for the pcb manufacturer to start work in half a day. My work was always using an EE engineer’s schematic design with which we collaborated in the pcb design phase.
• As mentioned by Dave I would possibly have added a GND fill on bottom and top layer. We NEVER relied on a GND fill to ensure proper grounding of SMT pads. Always use GND vias to power caps.
• Yes, top layer signals allow modifications of signal tracks if required.
• There are many choices of grounding here. Star, gnd fill and power planes. We must look at the current flows and “follow the current”. BTW, Follow the current saying comes from a very famous HP designer many moons ago who quoted this to all of his pupils.
• Vias to pwr planes is maybe more important than external smt caps.
• Less vias is less cost. 0.6mm via holes use up a lot of drill bits if you have lots of vias
I use the Saturn PCB toolkit too! It's such a great tool.
Other observations:
- length matching isn't what's important; it's time matching. Different layers will have different propagation speeds, so if possible, all the signals should go on the same layers (e.g. all go L1 then L4 then L1) and ideally the on a layer the lengths should be matched (e.g. each trace has 500mils on L1 and 400 mils on L2). Also remember that vias add length, so either take it into account or have the same amount of vias.
- when changing layers on a high speed signal, there should be a stitching capacitor (between the two reference layers) near the via, to give an AC return path
- DC/DC converter routing is poor, but this might be a work in progress with temporary traces to show that it routes ok and eventually will add regions/polygon pours
- looks like the xtal case might touch the usb shield pad. Probably not an issue.
- use thicker traces for bypass/decoupling caps. 12-20 mils
- traces are really close to your mounting holes
- hard to tell but solder mask sliver looks too thin on the MCU
- very often the thermal pads are supposed to be connected to ground; check the datasheet.
- you've got tons of space at the bottoms, consider adding surface testpoints at the bottom); it's free real estate
- high speed signals going to middle layers cause stubs at the via (e.g. if you go from L1 to L2, you have a stub between L2 and L4)
- looks like all your edges are going to be routed, so you'll need mousebites for your arrays, but I don't see many places when you'll be able to put them because your traces run so close to the edges
- don't forget to add fiducials. They can be on your array, but you probably want to have one near the BGA. Preferably at the top right of the BGA so it's also close to the MCU.
Thanks Dave. I've learned so much from you!
This was incredibly helpful and insightful. Thanks so much for the great advice!!
Enjoyed the review... Thumbs up Dave :)
I recently had to design a very compact PCB with multiple ground and power areas and eventually, out of frustration, decided to go to 4-layer with grounds and power in the inner planes. It was absolutely worth it. This kind of design in 2-layer is an absolute nightmare. It would be helpful if you could talk about how best to do multiple digital/analogue/power grounds.
Thanks Dave! Just like the old days.
love this kind of content, thanks!
7:45 Hey Dave! The buck converter layout is NOT GOOD! The switching node and the trace to the inductor is too long. Also, the output for the feedback is taped at the wrong place. Furthermore, there are no power planes and the input and output capacitors are not close to each other. The only thing which is good, is that all power traces are on the top layer.
I know, but I specially said I wouldn't get into that.
Why should the input and output capacitors be close to each other?
@@MetalheadAndNerd Hi! Please check out this video: ua-cam.com/video/gq-0ZpcGm8E/v-deo.html There it is explained in detail. There are two current loops, one if the switch is cloased and one if the switch is open. You want to have the DIFFERENCE as small as possible and his will be achived, if the negative terminals of Cin and Cout are close to each other.
Woow, great tutorial, thanks!
It would be interesting to see you re-route some of the bad parts, I suppose that would take way too long
I didn't have the Altium file otherwise I would have.
EEVblog so we need part 2 with the Altium files?
Very interesting, more of this please! :)
Haven't been at the channel for months. This video on its own is a quintessential distilled practical knowledge of PCB design.
Learned a lot.
Anyway, may I ask someone to suggest a good, up-to-date book (or course) for PCB design, please?
Thanks for the video Dave. I learned many things today. I'd like to see more videos like this. I like the green screen too.
Excellent review. Learned so much! Thanks.👍
Maybe you could do a video where you actually fix the design, would be nice for the guy and would be nice for us watching as we would see the design progressively improve
I thought the same thing. Also a side by side before-after comparison at the end would be really cool
Maybe the guy could submit the altium file instead of a screenshot.
That would have helped alot.
That would require a heck of a lot more time, you know - many hours. For we visual learners, it isn't necessary. I can see that for others it would be valuable. Also, it would require Dave needing the source files as well as be familiar with and have all the software.
I dig the talking head green screen, Dave. Keep up the great content.
The DC/DC could also be improved. The input caps are far away, they should be priority one. Place them as close as possible to the IC and even connect their GND-pad directly to the IC without any vias. Also do not connect the feedback line to the inductor, connect it to the output cap instead. Additionally, I would rotate the chip to make the switching node as short as possible (prio 2 directly after the input cap).
From a video production perspective, and for teaching purposes seeing, you head really helps. Great job!
Ideally flip the head video left-right (mirror), so that your movement matches the actions in the demonstration area.
Ie. When you look after camera, and then bring in a data sheet overlay, they come in from the same side..
When you point at components in the corners, your head/eyes are tracking to the same corner.
Yep will flip next time
Most of the issues with this design come from routing power first. Which for whatever reason is a common thing told to beginners to do in a lot of online courses (Udemy etc.). Have seen a couple of newbie engineers I work with do this.
Maybe it's a throwback from one- or two-side PCB times when one didn't have the luxury to just drop to a power plane on the spot wherever needed. After all, pesky chips with 4-6-8 pins needing all connected to a non-plane power trace cutting off everything else as they wind around the chip can really ruin your day if you start by laying out your signal traces first only to realize there's no way to connect all the required power at the end without redoing absolutely ALL the other traces from scratch. This has even become worse recently ever since THT (where as Dave mentioned you get a free via on each pin which will get soldered in anyway) got replaced by SMD, assuming you'd rather stay on a single side which is much simpler to produce at home without the headache that DIY vias incur (once you're ok with having vias and a two-side PCB it stops mattering as much).
I loved this video, and the green screen. I especially love those Lattice FPGAs.
Beautiful example of why FPGA bypass caps are often placed on the bottom layer. Signal lines first, then drop vias to the bypass caps, and Bob's your auntie.
Great video, I learned a lot about multi layer boards. Floating Dave is awesome, but don't forget to mirror the cam image to match the screen content :)
Great video as always! One thing that struck me though that I was kind of expecting to be mentioned is the placement of the FPGA and converter each placed on opposite side of the image sensor
Just switching place between the uC and FPGA would help shift the ground-plane return currents (assuming there is a solid ground plane ofc) of the FPGA from passing directly under the image sensor
But thats just my own humble opinion :)
Keep up the great work!
Great video as always. Its a good idea to consider how your going to distribute the power during placement but I always route most to least critical IO then I circle back to the pdn.
Dave. Great video, particularly when taken with other ones that you've made on PCB layouts. I too like the floating head.
It would be educational to see your routing solution, even if incomplete, for this project. Of course I realise that this would be counterproductive for soFPG and for your time. Perhaps soFPG will put a revised layout up for further review because a followup would definitely be of interest
I love this as a video segment :D It's a great teaching aid, plus people can submit their own designs! I know I got some PCB's I'd love you to have a look at :)
Also there's more immediacy to it, unlike a mailbag where stuff can get backlogged for months cause of delivery times and so on
I used I2C IO extender for arduino project, it was around 1,5m away, coupled with standart cable, going power for it in same cable. Just needed to to drive data lanes bit harder (lower termination resistor values), working just fine, no need to worry about I2C lines on the PCB, they will never be so long. Just imagine how big a PCB must be for lines to be over 1,5m xD
Problems with I2C will usually show up quickly if you go do your proper EMC testing. It's easy to get your communication completely messed up when you apply some RF fields to your designs. Sure, it might work in your lab and on a bench, but if you have some radio emitters (like cell phones) nearby, your device might suddenly act all crazy... Seen that happen quite a lot on I2C, even on relatively small 4 layer PCBs with proper circuit and layout designs and ground planes.
1: signals + power pour
2: ground
Core (center part is the thickest)
3: ground
4: signals + power pour
Isn’t that the stack up Rick Hartley recommends?
thanks for the tips :-). more such videos please :-)
Thank you, more videos like that please
Thank you very much sharing this. Just great
Designing my first board layout since uni for a project, the first iteration on hobby perf board suffers from cross talk between the output and feedback sections as well as the inputs. I have a lot to learn still.
“incoming message from the big giant head!” vibes with the floating head, I like it.
Love this video
The Top-Power-Ground-Bottom stackup is terrible for EMI, as it places the power planes far away from one another. This acts a lot like a slot antenna and radiates noise out the sides of the PCB. This can be disastrous for high-speed designs. A much better 4-layer stackup would be Top/Power-Ground-Ground-Bottom/Power