I actually always round up by default anyway to at least +0.05mm and sometimes even +0.1mm, just to be safe. If I REALLY need that tolerance I always contact the manufacturer first. I also stopped working in inches/imperial. To many companies and people screw these numbers up. Excellent video Dave!!!!!
@@outsideworld76 Yes, I guess, working in imperial units will give you a lot of other issues, like talking about ±5% or something is all of a sudden a much bigger step. Which really works weird when you have something like non perfect fraction of inches. I actually always wondered how they do that with Fahrenheit, since it's a non-linear scale. ±10% at 20 degrees F will give you a different tolerance compared to ±10% at 100 degrees F.
@@p_mouse8676 I really wonder how people in Celsius calculate melting and boiling points, as the underlying physics is based around absolute zero. Tin at 100 Deg. C is actually pretty close to melting and even 25 Deg. C is actually for metal migration still pretty warm. Yeah it takes time for the tin to eat into chopper - but hey: You car radio will be there for 10...15 years and it does get toasty and/or chilly in there. What I want to say: using any degree system is non-sense when talking about temperature. Yes, Celsius is nice for boiling and freezing water and we happen to be salty water bags (close enough anyway), but for metals the scale is pretty much of no use.
@@sarowie That's not the point I was making. If you want to use a proper scale, use Kelvin. The point is, that these scales are linear. It really doesn't matter what a scale is based on. I don't understand why people keep saying that. It's totally not relevant, because the number of a scale just simply shifts. Properties of materials also don't change with different scales. Otherwise you are just to much stuck in absolute numbers and clearly don't understand what a scale is all about. So people calculate melting points and boiling points the same way, just the numbers look different. So I would really advice to read a bit more on the subject
The floating head is great, imagine if you had a Microsoft Clippy-like Dave head in your PCB design software, "Uh-oh, it looks like that trace is too close to the pad!"
I have never seen anyone green screening and making a youtube vid as a floating head! This is genius, it lets more background be visible! And head is the only important thing anyway.
Their viewer is the best I have seen. A few more features would be nice. 1) More layers (Yes I ran out) 2) Remember the file name for a layer and put it in the hint when you hover over the "visible". 3) A "there's your problem" layer that you can draw on
Love the floating head. I recently ordered some PCBs from JLC as a complete and utter noob. The service was amazing. I messed up the gerber multiple times, and had issues with clearances etc; they were super helpful in getting it sorted (I felt pretty bad since it was only 5 boards!!). 1wk from order to delivery to UK so pretty stoked in the end.
First time watching this channel. Very funny presentation. I've been designing PCB's and PWB's, yes there is a difference, for 42 years. I started using tape an mylar. There is no type, or technology of board that I haven't done. 25+ years for DOD and government and all types of commercial. Some important things for all of you to consider. Never design to minimum. Bigger is always better. Gerber is only good for 7 decimal places( microns). Always output at max resolution to avoid truncation and roundoff. Consult with your fabricator before you start and as needed throughout your design.
You were fancy-pancy with your Bishop Graphics tape on mylar! Dalo pen on copper clad for the win. Followed by stick-on resist footprints and tape on copper clad. Then comes the fancy-pancy stuff!
This rounding error is definitely JLCPCB's fault. They've stated that is minimum trace width, so the minimum of the values should be sufficient. This also implies they should take care to round up, not down, if they're going to convert it. The row above implies either that mil is the source unit (as 0.127 is exact) or that they have 3 significant digits, so it should've said 3.55. Messy. Of course we can take the engineer's route and always pick the pessimistic value instead, but having to will cause cumulative margin errors. The pad to track tolerance probably has more to do with guaranteeing the solder mask won't leave the track exposed or interfere with the pad soldering than any drill; drills are involved in the PTH and NPTH specs. This means the tolerances of three liquid processes combine (copper etching, soldermask, and soldering itself). Track to track is covered by solder mask so only the etching affects it, pad to pad may require masking off.
Like Dave said, you shouldn't be pushing the limits with a prototype PCB manufacturer anyway. That will simply leave you paying for a lot of defective boards with broken and/or shorted traces. I do my best to stay away from the minimums, and will only come close when necessary. If what you are designing needs the bare minimums listed by the manufacturer you are probably better off selecting a different (likely also more expensive) manufacturer to send your gerbers to.
I can see it both ways - on the one hand, if they say 3.5mil, they should do 3.5mil, otherwise it's false advertising and that's a very slippery slope. On the other hand, when you're pushing the envelope, it's your job to get the complete specification - at those widths, you need more than one number to describe their accuracy - how wavy is the trace, how parallel are it's edges, how wide is it's top compared to it's bottom... PCB manufacturer is not a black box operation - good designers need to know how things work inside.
@@johnalexander2349 "they should do 3.5mil, otherwise it's false advertising and that's a very slippery slope" They do not, in any way, guarantee that 3.5 mil works - they specifically say it is the minimum they accept.
I know there are cultural issues involved and such but frankly I think it's just not a sensible decision to work to deprecated measurement regimes in any field of science or industry in 2020. Why would you introduce such a source of potential error or conflict into your workstream when you have the simple choice not to?
I'm in the U.S., but do most of my mechanical design with metric dimensions and hardware as those parts are easy enough to source. But when you're putting together a board for people to add through-hole parts and solder themselves, well again it comes down to parts availability and the ubiquity of 0.1" (2.54mm) spacing. It does get weird when I have both JST 2.50mm connectors and 2.54mm pin headers, but overall it seems to work out best when the grid is in thou instead of mm.
It is very simple to do and costs are low, allow both today - no real problem, select a distant swap over date 10-100 years from now when only metric will be used, and all new stuff sold must use either both or metric only from a different future cut-off date.
Soviet Union simply used a 2.5mm pitch instead of 2.54mm (0.1inch) when copying stuff. It was pretty much only noticeable with chips like 40pin CPU's where the error was enough to cause misalign. So we could start from such a "minor" tweak, I guess some machines could be adjusted for the small difference.
The reason for the issue with 0.9mm is not that it automatically rejects 0.0889mm, but they round it up, and if that causes you to break clearance, that’s the rejection. The reason they put a larger clearance for pads to trace is because they want room for solder mask variance so they don’t expose your trace based on the offset tolerance levels. That’s why pad to pad don’t have that because they are both exposed, and trace to trace are both covered. They should have called it soldermask clearance though, but that’s probably what it’s for. Lastly, as you pointed out later, most manufacturers consider smaller BGA chips to be “advanced” so yeah, different capabilities for different levels of manufacturer.
They seem to treat their specifications in two ways - for some they'll reject a board and not try to manufacture it, but for others, I guess they just make it and if there are any manufacturing errors they can say "well, we told you you needed 0.2mm clearance".
Dave, you are amazing, I love the way you speak, your accent, and how deep you go into the subjects you talk about. As a just graduated electrical engineer it's amazing to watch your videos, thank you very much, your contribution to electrical engineering is priceless !!
This is good info to have. Some fab houses like the one OshPark uses will take your board with whatever specs you have and try to fabricate it. They have minimums listed on the site, but if you go below that they will warn you that your yield could suffer and they will not attempt to remedy the situation for you. You're stuck with faulty boards. I'm sure all fab houses are different, so knowing that some will just kick it back without even attempting it is very good to know.
It's amazing, the floating head Dave. Please let it randomly jump around the screen. Now also get some green sleeves, so a floating head and 2 floating hands.
Excellent information, clearly presented as always. Could listen to Dave waffle on about PCB design for hours and learn a lot. There are that many considerations when you push the limits. Unfortunately.
Thank you for all your detailed explanations that me as a dyslexic can understand. Wish coding could be explained so nicely. I struggle and it’s touch and go for me.
Could you Imagine if there was an Altium plugin like Mr Clippy for MS Word XD Every time you do something dumb his head pops up "oi! whatya doin ya bloody bludger!" "that's not 0.9mm!"
Fun fact: Some board houses have different annular ring specifications for outer layers versus inner layers and sadly KiCAD doesn't let you specify these different tolerances for different (e.g. inner vs. outer) layers.
This will be available in Kicad v6. If you are adventurous enough you can already test it in the nightly builds (be aware that the graphical interface for entering the new DRC rules isn't done yet).
@@Eletronicafg I think a plain text ascii configuration file and editing it is good enough. Beating on the code that implements the feature seems a worthy thing to do next time I have nothing to do.
13:20 There are a few different reasons to have different spacing specs for plated and non-plated through holes, drilled vs. non drilled pads, and features on the same or on different nets and they are all etching and plating related. PCBs in the 10/10 micron lines/space range are often produced in pattern plating mode. I.e. the pattern is not etched from a full thickness copper foil. Instead a thin seed layer is plated in electroless mode, then the photoresist is applied and the pattern is filled into the openings of the resist. After stripping the resist, the seed layer must be etched away between the traces and pads. This will always lead to some wedge-etching under the traces and my leave bridges, if the gaps are too small. This is what limits the line and space width capability of the process, and you can see, how shorts within the same net are not that critical. With pads around holes, the problem is usually adhesion-related. During the PTH process, chemicals can creep between the pad and base material. That weakens adhesion of the whole pad, and causing air or water caught in these pockets that can explode in the reflow oven. And there are lots and lots more things that can go wrong during muli-layer PCB manufacture...
Things that may explain stuff. The pads may be done on a different pass than the traces and this may lead to a repeatability issue. It certainly did in the past. The soldermask edge to copper accuracy is not as good as copper to copper. This can make trouble of exposing a trace that was supposed to be under the mask. Last I checked, Kicad stores X,Y as integer numbers of 1E-4 inch increments. You can suffer from two roundings as the dimensions are converted. This is done because the numbers for almost all packages come out more even in tiny fractions of an inch. History rules destiny on this.
Yes, and I think there is supposed to be a fine line of soldermask between those two exposed pads, and the soldermask is going to have a minimum width. There's going to be some small registration error between the soldermask and the pads.
If JLC is using metric, they should really write that spec as "0.09mm (3.6mil)". The way it's written, I'd *totally* assume the actual value is US customary and they're providing a metric approximation for convenience. Providing US customary below the actual spec (rather than rounding up) helps no-one.
TL:DR: 1) Rounding errors between Imperial and metric (3.5 mil vs 0.09 mm) 2) Different clearance distances among different type of elements, such as Plated drilled through hole, pads, BGA pads.3) Formatting resolution of exported Gerber coordinates (mil/mm; decimal precision)
Floating severed Dave head is best! I agree with the turtleneck comments too. And extra bonus points if your ethereal head moves around and your eyes look at whatever you're talking about.
Dave, floating head is great. It looks great and it leaves more space in the video. Regarding the pads vs trace clearance, pad clearance will be limited by the solder mask tolerances, which often times is lower. About the BGAs, I think that in-pad vias are the way to go, if supported buy your board house.
@EEVblog in kicad you can go in and set the minimums in the board setup but all you have to do is make sure that you check with the board maker company and get what their minimums are.
love the floating dave head XD. Though when you have your shoulders in and your body is cut off it's kind of neat as you can position yourself to be coming out of text boxes or whatever you are displaying at the time. Have fun with it, why not?
Thanks for the video. I use JLCPCB myself, and it doesn't seem that they touch the gerbers at all. However, I haven't done any fine pitch (nothing finer than TSSOP-2, and no traces finer than 6mil) PCB's and so have not had any need to push the limits of their process.
It’s airhead Dave! Hilarious and creepy at the same time. I like this format actually, Dave, with these kind of videos. Wonder how the green screen would fare with Mailbag.
I tend to set my defaults to 6mil/6mil unless i need something special. I do this because after comparing most the common players for prototype PCBs, that is the largest spacing and width for their prototype boards. And that is another trap that I have noticed. Some places will list 3mil/3mil as their capabilities, but when you read the fine print, it is larger for their low volume prototype service.
I have switched many pcb production fabs and made a rule. Make all tolerances +1 LSD. For example 0.127 => 0.128, 0.2 => 0.21. With only execption is holes because it is makes no sence. Also, keep in mind rounding errors in checking and drawing software.
Love the floating head Dave. I can hear it now: "Your design is stupid and you should feel bad!" - after reviewing a 2 layer breakout board for my 128 pin MIL circular connector
Just wanted to add something regarding JLCPCB minimum trace widths @3:50. After reading the comments, it seems most people got away with 3.5 mil wide traces. Therefore, I just wanted to note that the minimum spacing of 0.090 mm seems to be a stricter requirement because I was told to resubmit or cancel my board because I also had a dense set of parallel traces spaced at 3.5 mil (0.0889 mm) as well. Anyway, just wanted to state I wish I had seen this video earlier because redoing those traces was a pain.
The pad to pad distance is about the solder mask, there are actually many companies that mess with that layer. For example my main supplier requires the solder mask in gerbers to have the same dimensions as pads have. But they adjust the actual dimensions in the tooling process (+2mil for small pads, +4mil for bigger). My experience with those prothotype PCB manufacturers (jlcpcb and allPCB) is that they are actually very helpfull and they will go over some issues with you (I don't know if that's a common thing for a $5 board or if they do it for more expensive PCBs, mine was about $80). For example I had an issue where they weren't able to put solder mask bridge between PADs and after a discussion with their engineer, I ended up making them a bit smaller.
Small drills do wander, but you also must align the the holes and vias among the various copper layers. When many layers are assembled, the probability of having all of the wanderings be small or in the same direction decreases.
I actually like the floating head because you can see more of the computer screen, whilst retaining expressions, assistance from lip reading, and hand expressions
Pad to track clearance is about the accuracy of the placement of the solder mask. If the track is too close to the pad it might be accidentally exposed.
Pad to track tolerance could be linked to accuracy of the solder mask. If the solder mask can't reliably cover the track within .2mm next to a pad you risk being unable to solder it without risking a bridge
In my experience JLC will adjust your gerbers to make it work. For example www.eevblog.com/forum/manufacture/anyone-used-these-through-hole-usb-c-2-0-sockets/msg3098449/#msg3098449 they chopped off the edges of round pads to increase the clearance to the traces running between them. I also think JLC specifically won't "reject your board" they'll adjust if possible, just leave it if not and send it to be produced - I've accidentally made several boards with traces that are too thin, too close to vias and via rings that are too small and they've made them all - I had one that had a bridge due to me violating the limits but the others worked fine.
If you're going to do that level of floating head, make sure you include a long sleeve tshirt so your hands are cut off at the wrists and shoot from low enough that there isn't the bottom edge.
The first point in the video literally just happened to me. BGA and all. To be fair, I started my design before JLCPCB updated their site to include the mm spec, but it sure was annoying. They still manufactured it though, and everything works so far! 😆
I use 0.199mm minimum spacing for all PCBs of 1oz-20z copper, manufactured hundreds from 2 layers to 8 layers, with several different manufactures and never a single issue. 0.199 is usually well within the capabilities of most manufactures. JLCPCB has delivered some of the worst quality boards ive ever seen but still worked with those limits even on 2 oz copper.
I've never heard that version before. In my part of the US it's 'you don't get upset.' Naturally I'm biased but I prefer this, maybe because 'get' and 'upset' rhyme.
I still remember the first time I heard it in preschool. "You don't throw a fit" is what I heard, so that's been engraved in my mind ever since. I like it better as well.
@@brocktechnology Maybe we have different pronunciations. The way I say it get and upset both end with '-ette' while fit ends like the word it. Do you pronounce any of the words differently?
20:00 Oh come on Dave, one doesn't muck around with the caliper tool in the Gerber viewer if all one wants to know is the size of a trace or pad - there's a reason they are marked with the "Dxx" code of the tool they are getting painted with; you note the number, and select it from that drop-down list at the top called "DCode", which lists the size of the tool right next to its code (and also highlights all elements painted with that tool).
13:40 would that have to do with the soldermask? They seem to made their masks a slight bit larger, if your two pads are too close together, maybe that could break the separation? [edit] oh, you actually mention this later in the video, haha, I was too quick to comment there, sorry.
I read English / Metric conversion rounding caused issues with the Soviet IC industry. They rounded the 0.1" spacing. If boards/sockets and chips were from mixed systems, the more pins the harder to insert,
I add 16mil isolate on inner layers since I've had boards come back with the holes too far off. I also have had boards with spots that are not completely etched. I also had a board house not remove Silk over pads.
For an excellent Gerber viewer suggest you have a look at FAB3000 (paid version) or DFM Now (free version) In above you would have been able to double click on a track to see it's width. Also with the paid version Gerber manipulation becomes crazy easy compared to say CAMtastic ;^)
If the defining spec is in mm, then that's what should have been written first. The number in parenthesis should be the conversion. Also, since this is a minimum the conversion should have been rounded up (or else omitted altogether). The way this boardmaker has written it is erroneous on their part and leads to unnecessary confusion. If this has been pointed out and they've refused to fix it then they're just being stubborn. Definitely a good point to pay close attention to this. I never considered it before, and I think I have my track spacing design rule set in mils so I should change it. Many people say "never design to the minimum" and I agree with that if you have the flexibility to be relaxed. Personally I'm not in a situation where I need to push the limits so my design rules are set very conservatively. But the manufacturer is being asked for this information for a reason. If they say they can build something then they should be able to build it because some people are in a situation where they have less flexibility and these specs actually matter. I don't expect them to argue and negotiate over a $5 job, but if the same issue comes up repeatedly from different people then it should be taken as a sign that the published specs need correction or clarification. But I won't lose sleep over the flaws of some China industrial business' web site, it's more useful to just know how to deal with them so that point is well taken.
I think jlc do that board with 3,5mils, even with 3,4mils ... But also care of that these are the values for 4layer boards, a 2 layerboard needs 5mils (other machine is working on that) They just do that fine. The bigger "problem" is, that i recently feel like the coworkers get more and more closer looks to the boards and recognize things like that, before giving it to production. That might be the problem, that they text you they wont do that for you. If they trespass the board to production, everything will be fine, without problems.
If you spend a whole lot of time designing a board that got rejected for some tolerance issue, I'd just look at the other prototype PCB companies before spending another week redesigning.. maybe for another buck a board they can manufacture it It would probably be good practice to look at a variety of companies and use the WORST SPEC of any company for each design feature, then you have the most flexibility later if you want to get it built by someone else
Good to know, I typically work in mils when setting up my design rules. Lowest I've had was 4 mils with jlcpbc, so guess I was right on the edge of I needed to do a bit lower.
You can assign a track to a net in KiCad (eeschema) by editing the track and picking the net at the top of the dialog, then it should let you connect them.
I haven't been designing PCBs for long, but it kinda looks like everyone is leaning towards metric now, most surface mount components I use have package dimensions reported in mm.
I don't go below 8/8 thou (0.2 mm) with these Chinesium board houses, not that I've had reason to. Printed silk screen registration is usually the worst followed by solder mask.
it is not just drill wander. when you laminate a 4 layer board, things will not line up perfectly. Top and bottom are slightly off, inner layers are slightly off, the whole thing is slightly bend and warped, ... (let alone: vias are "texas sharp shoter", as the hole is drilled before the chopper is etched away) As I like to say: Yes, we are splitting hairs here - because we literally are in physical dimensions, where a hair is a relevant unit of measurement.
I actually always round up by default anyway to at least +0.05mm and sometimes even +0.1mm, just to be safe.
If I REALLY need that tolerance I always contact the manufacturer first.
I also stopped working in inches/imperial. To many companies and people screw these numbers up.
Excellent video Dave!!!!!
I guess we European circuit board designers where always aware of the imperial/metric issue.
@@outsideworld76 Yes, I guess, working in imperial units will give you a lot of other issues, like talking about ±5% or something is all of a sudden a much bigger step.
Which really works weird when you have something like non perfect fraction of inches.
I actually always wondered how they do that with Fahrenheit, since it's a non-linear scale. ±10% at 20 degrees F will give you a different tolerance compared to ±10% at 100 degrees F.
That's exactly the way to go. Never rely on the minimums of the manufacturer.
@@p_mouse8676 I really wonder how people in Celsius calculate melting and boiling points, as the underlying physics is based around absolute zero. Tin at 100 Deg. C is actually pretty close to melting and even 25 Deg. C is actually for metal migration still pretty warm. Yeah it takes time for the tin to eat into chopper - but hey: You car radio will be there for 10...15 years and it does get toasty and/or chilly in there.
What I want to say: using any degree system is non-sense when talking about temperature. Yes, Celsius is nice for boiling and freezing water and we happen to be salty water bags (close enough anyway), but for metals the scale is pretty much of no use.
@@sarowie That's not the point I was making. If you want to use a proper scale, use Kelvin.
The point is, that these scales are linear.
It really doesn't matter what a scale is based on. I don't understand why people keep saying that.
It's totally not relevant, because the number of a scale just simply shifts. Properties of materials also don't change with different scales.
Otherwise you are just to much stuck in absolute numbers and clearly don't understand what a scale is all about.
So people calculate melting points and boiling points the same way, just the numbers look different.
So I would really advice to read a bit more on the subject
Dave is finally in his final form
Brett Smith just need a jar for his head....
Lool
Yeeeeesssss... he will soon build his eternal body
Floating head? Looks like you're just 'flapping in the breeze' Dave 😂
He should also wear a green neck cover :D
@@antnk9040 Why not a green balaclava while we're at it. :)
yeah. a li'l bit of hot snot should fix it though.
lol I was like "WHATT??". :)
@@antnk9040 Yes, that Holly's (Red Dwarf) look would be great.
The floating head is great, imagine if you had a Microsoft Clippy-like Dave head in your PCB design software, "Uh-oh, it looks like that trace is too close to the pad!"
plus wah wah wahhh as the error sound
Sounds like a kicad plugin waiting to happen
The floating head is brilliant. I love it
Just imagine what a green balaclava would be like
I have never seen anyone green screening and making a youtube vid as a floating head! This is genius, it lets more background be visible!
And head is the only important thing anyway.
I can't be the first, surely?
@@EEVblog Dave, don't underestimate your brilliance in the youtube space. A true trendsetter.
@@EEVblog A professor on Twitch does the same. Most videos are deleted, but here is a clip: www.twitch.tv/jhkrueger/clip/JollyShinyDotterelSpicyBoy
@@EEVblog Two years ago: ruclips.net/video/M76g81G3bi8/видео.html - I like how he switches during the different parts of the video
We always lie that the head is the most important thing.
A shirt with just non green collar could avoid being strange
Change the t-shirt to a turtleneck so you have a floating head.
But Australia man.
It gets hot... Guess air conditioning could help.
Dave using KiCAD's Gerber Viewer instead of Altium's one. Getting there, slowly, slowly... :)
Their viewer is the best I have seen. A few more features would be nice.
1) More layers (Yes I ran out)
2) Remember the file name for a layer and put it in the hint when you hover over the "visible".
3) A "there's your problem" layer that you can draw on
@@kensmith5694 you can go to the github project and propose as a feature
Come to the FOSS side, Dave! Bwahahahaha!
I guess after more than a decade of The Amp Hour, Chris's love of KiCad is finally rubbing off on Dave
Love the floating head. I recently ordered some PCBs from JLC as a complete and utter noob. The service was amazing. I messed up the gerber multiple times, and had issues with clearances etc; they were super helpful in getting it sorted (I felt pretty bad since it was only 5 boards!!). 1wk from order to delivery to UK so pretty stoked in the end.
First time watching this channel. Very funny presentation. I've been designing PCB's and PWB's, yes there is a difference, for 42 years. I started using tape an mylar. There is no type, or technology of board that I haven't done. 25+ years for DOD and government and all types of commercial. Some important things for all of you to consider. Never design to minimum. Bigger is always better.
Gerber is only good for 7 decimal places( microns). Always output at max resolution to avoid truncation and roundoff. Consult with your fabricator before you start and as needed throughout your design.
You had tape on mylar, luxury!
ruclips.net/video/ue7wM0QC5LE/видео.html
My first design was with "ink on copper" method.
You were fancy-pancy with your Bishop Graphics tape on mylar!
Dalo pen on copper clad for the win. Followed by stick-on resist footprints and tape on copper clad. Then comes the fancy-pancy stuff!
@@EEVblog Jeez. I might still have my original circa '79 Dalo pen in one of my junk boxes.
I've got flashbacks of vertically-hanging mylar sheets with tape in a pile on the bottom of the cabinet!
@@theantipope4354 I still have my exacto kit. Lost my slide rule.
This rounding error is definitely JLCPCB's fault. They've stated that is minimum trace width, so the minimum of the values should be sufficient. This also implies they should take care to round up, not down, if they're going to convert it. The row above implies either that mil is the source unit (as 0.127 is exact) or that they have 3 significant digits, so it should've said 3.55. Messy. Of course we can take the engineer's route and always pick the pessimistic value instead, but having to will cause cumulative margin errors.
The pad to track tolerance probably has more to do with guaranteeing the solder mask won't leave the track exposed or interfere with the pad soldering than any drill; drills are involved in the PTH and NPTH specs. This means the tolerances of three liquid processes combine (copper etching, soldermask, and soldering itself). Track to track is covered by solder mask so only the etching affects it, pad to pad may require masking off.
Like Dave said, you shouldn't be pushing the limits with a prototype PCB manufacturer anyway. That will simply leave you paying for a lot of defective boards with broken and/or shorted traces. I do my best to stay away from the minimums, and will only come close when necessary. If what you are designing needs the bare minimums listed by the manufacturer you are probably better off selecting a different (likely also more expensive) manufacturer to send your gerbers to.
I can see it both ways - on the one hand, if they say 3.5mil, they should do 3.5mil, otherwise it's false advertising and that's a very slippery slope. On the other hand, when you're pushing the envelope, it's your job to get the complete specification - at those widths, you need more than one number to describe their accuracy - how wavy is the trace, how parallel are it's edges, how wide is it's top compared to it's bottom...
PCB manufacturer is not a black box operation - good designers need to know how things work inside.
@@johnalexander2349 "they should do 3.5mil, otherwise it's false advertising and that's a very slippery slope"
They do not, in any way, guarantee that 3.5 mil works - they specifically say it is the minimum they accept.
@@ABaumstumpf No, in some countries' laws, it has to be functional for at least some applications.
I know there are cultural issues involved and such but frankly I think it's just not a sensible decision to work to deprecated measurement regimes in any field of science or industry in 2020. Why would you introduce such a source of potential error or conflict into your workstream when you have the simple choice not to?
I like the floating head, takes up way less space this way.
Agreed, it works well if the webcam view is at the top of the screen.
"Hi, it looks like you're trying to design a PCB!" unexpected flashback :D
*dink dink*
*shudder*
The resistance to adopt the metric system boggles me
I'm in the U.S., but do most of my mechanical design with metric dimensions and hardware as those parts are easy enough to source. But when you're putting together a board for people to add through-hole parts and solder themselves, well again it comes down to parts availability and the ubiquity of 0.1" (2.54mm) spacing. It does get weird when I have both JST 2.50mm connectors and 2.54mm pin headers, but overall it seems to work out best when the grid is in thou instead of mm.
It is very simple to do and costs are low, allow both today - no real problem, select a distant swap over date 10-100 years from now when only metric will be used, and all new stuff sold must use either both or metric only from a different future cut-off date.
As they say in the UK, we go metric inch by inch!
The machines were often designed "in the era of the inch". Notice how so many connectors are 2.54mm increments.
Soviet Union simply used a 2.5mm pitch instead of 2.54mm (0.1inch) when copying stuff. It was pretty much only noticeable with chips like 40pin CPU's where the error was enough to cause misalign. So we could start from such a "minor" tweak, I guess some machines could be adjusted for the small difference.
I like the floating head! Though it did remind me of Clippy.
"Hey! You're trying to design a PCB. Do you want......."
Love the floating head.
The reason for the issue with 0.9mm is not that it automatically rejects 0.0889mm, but they round it up, and if that causes you to break clearance, that’s the rejection. The reason they put a larger clearance for pads to trace is because they want room for solder mask variance so they don’t expose your trace based on the offset tolerance levels. That’s why pad to pad don’t have that because they are both exposed, and trace to trace are both covered. They should have called it soldermask clearance though, but that’s probably what it’s for. Lastly, as you pointed out later, most manufacturers consider smaller BGA chips to be “advanced” so yeah, different capabilities for different levels of manufacturer.
They seem to treat their specifications in two ways - for some they'll reject a board and not try to manufacture it, but for others, I guess they just make it and if there are any manufacturing errors they can say "well, we told you you needed 0.2mm clearance".
Dave, you are amazing, I love the way you speak, your accent, and how deep you go into the subjects you talk about. As a just graduated electrical engineer it's amazing to watch your videos, thank you very much, your contribution to electrical engineering is priceless !!
This is good info to have. Some fab houses like the one OshPark uses will take your board with whatever specs you have and try to fabricate it. They have minimums listed on the site, but if you go below that they will warn you that your yield could suffer and they will not attempt to remedy the situation for you. You're stuck with faulty boards.
I'm sure all fab houses are different, so knowing that some will just kick it back without even attempting it is very good to know.
It's amazing, the floating head Dave. Please let it randomly jump around the screen.
Now also get some green sleeves, so a floating head and 2 floating hands.
Sounds like a cool screensaver.
Can we get an ice cream van to go with the Green sleeves?
Excellent information, clearly presented as always. Could listen to Dave waffle on about PCB design for hours and learn a lot. There are that many considerations when you push the limits. Unfortunately.
Thank you for all your detailed explanations that me as a dyslexic can understand. Wish coding could be explained so nicely. I struggle and it’s touch and go for me.
Could you Imagine if there was an Altium plugin like Mr Clippy for MS Word XD
Every time you do something dumb his head pops up
"oi! whatya doin ya bloody bludger!"
"that's not 0.9mm!"
"oi mate, yer DRC failed!"
Fun fact: Some board houses have different annular ring specifications for outer layers versus inner layers and sadly KiCAD doesn't let you specify these different tolerances for different (e.g. inner vs. outer) layers.
It is a feature that could be added. The nice thing about open source is improvements can be driven by what people actually want.
This will be available in Kicad v6. If you are adventurous enough you can already test it in the nightly builds (be aware that the graphical interface for entering the new DRC rules isn't done yet).
@@Eletronicafg
I think a plain text ascii configuration file and editing it is good enough.
Beating on the code that implements the feature seems a worthy thing to do next time I have nothing to do.
13:20 There are a few different reasons to have different spacing specs for plated and non-plated through holes, drilled vs. non drilled pads, and features on the same or on different nets and they are all etching and plating related.
PCBs in the 10/10 micron lines/space range are often produced in pattern plating mode. I.e. the pattern is not etched from a full thickness copper foil. Instead a thin seed layer is plated in electroless mode, then the photoresist is applied and the pattern is filled into the openings of the resist. After stripping the resist, the seed layer must be etched away between the traces and pads. This will always lead to some wedge-etching under the traces and my leave bridges, if the gaps are too small. This is what limits the line and space width capability of the process, and you can see, how shorts within the same net are not that critical.
With pads around holes, the problem is usually adhesion-related. During the PTH process, chemicals can creep between the pad and base material. That weakens adhesion of the whole pad, and causing air or water caught in these pockets that can explode in the reflow oven.
And there are lots and lots more things that can go wrong during muli-layer PCB manufacture...
Things that may explain stuff.
The pads may be done on a different pass than the traces and this may lead to a repeatability issue. It certainly did in the past.
The soldermask edge to copper accuracy is not as good as copper to copper. This can make trouble of exposing a trace that was supposed to be under the mask.
Last I checked, Kicad stores X,Y as integer numbers of 1E-4 inch increments. You can suffer from two roundings as the dimensions are converted. This is done because the numbers for almost all packages come out more even in tiny fractions of an inch. History rules destiny on this.
Yes, and I think there is supposed to be a fine line of soldermask between those two exposed pads, and the soldermask is going to have a minimum width. There's going to be some small registration error between the soldermask and the pads.
Love the floating head....
Somebody please send him a green balaclava
If JLC is using metric, they should really write that spec as "0.09mm (3.6mil)". The way it's written, I'd *totally* assume the actual value is US customary and they're providing a metric approximation for convenience. Providing US customary below the actual spec (rather than rounding up) helps no-one.
TL:DR:
1) Rounding errors between Imperial and metric (3.5 mil vs 0.09 mm)
2) Different clearance distances among different type of elements, such as Plated drilled through hole, pads, BGA pads.3) Formatting resolution of exported Gerber coordinates (mil/mm; decimal precision)
I think the floating heads lends a unique character to the videos. It's a winner!
Floating severed Dave head is best! I agree with the turtleneck comments too. And extra bonus points if your ethereal head moves around and your eyes look at whatever you're talking about.
Dave, floating head is great. It looks great and it leaves more space in the video. Regarding the pads vs trace clearance, pad clearance will be limited by the solder mask tolerances, which often times is lower. About the BGAs, I think that in-pad vias are the way to go, if supported buy your board house.
@EEVblog in kicad you can go in and set the minimums in the board setup but all you have to do is make sure that you check with the board maker company and get what their minimums are.
Thanks. You just saved me a hassle with the silk screening on a couple device patterns on a board that I am designing now.
love the floating dave head XD. Though when you have your shoulders in and your body is cut off it's kind of neat as you can position yourself to be coming out of text boxes or whatever you are displaying at the time. Have fun with it, why not?
I don't know, on one hand its awesome to see you, on the other hand its a bit creepy.
Why not try a green scarf next time? :D
or green turtleneck
Why stop there - green facepaint - disembodied eyes, glasses and mouth, anyone?
Thanks for the video. I use JLCPCB myself, and it doesn't seem that they touch the gerbers at all. However, I haven't done any fine pitch (nothing finer than TSSOP-2, and no traces finer than 6mil) PCB's and so have not had any need to push the limits of their process.
It’s airhead Dave! Hilarious and creepy at the same time.
I like this format actually, Dave, with these kind of videos.
Wonder how the green screen would fare with Mailbag.
I tend to set my defaults to 6mil/6mil unless i need something special. I do this because after comparing most the common players for prototype PCBs, that is the largest spacing and width for their prototype boards. And that is another trap that I have noticed. Some places will list 3mil/3mil as their capabilities, but when you read the fine print, it is larger for their low volume prototype service.
I have switched many pcb production fabs and made a rule. Make all tolerances +1 LSD. For example 0.127 => 0.128, 0.2 => 0.21. With only execption is holes because it is makes no sence. Also, keep in mind rounding errors in checking and drawing software.
I just placed an order with JLCPCB that has 3.5mil (0.0889mm) traces. It was approved. We'll see how it turns out in a couple weeks.
Love the floating head Dave.
I can hear it now:
"Your design is stupid and you should feel bad!" - after reviewing a 2 layer breakout board for my 128 pin MIL circular connector
LOVE head-on-a-plate-Dave!
Just wanted to add something regarding JLCPCB minimum trace widths @3:50. After reading the comments, it seems most people got away with 3.5 mil wide traces. Therefore, I just wanted to note that the minimum spacing of 0.090 mm seems to be a stricter requirement because I was told to resubmit or cancel my board because I also had a dense set of parallel traces spaced at 3.5 mil (0.0889 mm) as well.
Anyway, just wanted to state I wish I had seen this video earlier because redoing those traces was a pain.
The pad to pad distance is about the solder mask, there are actually many companies that mess with that layer. For example my main supplier requires the solder mask in gerbers to have the same dimensions as pads have. But they adjust the actual dimensions in the tooling process (+2mil for small pads, +4mil for bigger). My experience with those prothotype PCB manufacturers (jlcpcb and allPCB) is that they are actually very helpfull and they will go over some issues with you (I don't know if that's a common thing for a $5 board or if they do it for more expensive PCBs, mine was about $80). For example I had an issue where they weren't able to put solder mask bridge between PADs and after a discussion with their engineer, I ended up making them a bit smaller.
I love Dave's head just flying everywhere
Small drills do wander, but you also must align the the holes and vias among the various copper layers. When many layers are assembled, the probability of having all of the wanderings be small or in the same direction decreases.
I do love anything you post on PCB design...more please!
I actually like the floating head because you can see more of the computer screen, whilst retaining expressions, assistance from lip reading, and hand expressions
Pad to track clearance is about the accuracy of the placement of the solder mask. If the track is too close to the pad it might be accidentally exposed.
It sometimes appears your head is appearing from a via. With your energy level, make sure you add some thermal relief!
Pad to track tolerance could be linked to accuracy of the solder mask. If the solder mask can't reliably cover the track within .2mm next to a pad you risk being unable to solder it without risking a bridge
Actually I've ordered PCB at JLC with two 3.5mil tracks between BGA pads (at 0.6mm raster). They just made it without any complains.
For the floating head bit, make it bounce around like the old DVD screensaver logos, will he ever hit the corner of the screen perfectly? who knows!
In my experience JLC will adjust your gerbers to make it work. For example www.eevblog.com/forum/manufacture/anyone-used-these-through-hole-usb-c-2-0-sockets/msg3098449/#msg3098449 they chopped off the edges of round pads to increase the clearance to the traces running between them. I also think JLC specifically won't "reject your board" they'll adjust if possible, just leave it if not and send it to be produced - I've accidentally made several boards with traces that are too thin, too close to vias and via rings that are too small and they've made them all - I had one that had a bridge due to me violating the limits but the others worked fine.
Loving the severed floating head.
I've never laughed this hard on an educational video before, thanks Dave for the bobbing head :)
Ive always wanted a little floating Dave on my desktop
0:40 missed opportunity to do the pac-man WAKAWAKAWAKAWAKA
If you're going to do that level of floating head, make sure you include a long sleeve tshirt so your hands are cut off at the wrists and shoot from low enough that there isn't the bottom edge.
The first point in the video literally just happened to me. BGA and all. To be fair, I started my design before JLCPCB updated their site to include the mm spec, but it sure was annoying. They still manufactured it though, and everything works so far! 😆
Thanks for putting up this video, I'm just about to take the plunge and have my first board manufactured!
I use 0.199mm minimum spacing for all PCBs of 1oz-20z copper, manufactured hundreds from 2 layers to 8 layers, with several different manufactures and never a single issue. 0.199 is usually well within the capabilities of most manufactures. JLCPCB has delivered some of the worst quality boards ive ever seen but still worked with those limits even on 2 oz copper.
Here (Michigan) it is "You get what you get, and you don't throw a fit".
Flows better to my ears.
I've never heard that version before. In my part of the US it's 'you don't get upset.' Naturally I'm biased but I prefer this, maybe because 'get' and 'upset' rhyme.
I still remember the first time I heard it in preschool. "You don't throw a fit" is what I heard, so that's been engraved in my mind ever since. I like it better as well.
@@eDoc2020 Um,... get and upset don't rhyme. Get and fit do though.
@@brocktechnology Maybe we have different pronunciations. The way I say it get and upset both end with '-ette' while fit ends like the word it. Do you pronounce any of the words differently?
Weird-floating-dave-head is awesome! Some spider or octopus legs would make it even better!
Floating head is weird, but I like it. As another user suggested, make it bounce around like the DVD logo!
20:00 Oh come on Dave, one doesn't muck around with the caliper tool in the Gerber viewer if all one wants to know is the size of a trace or pad - there's a reason they are marked with the "Dxx" code of the tool they are getting painted with; you note the number, and select it from that drop-down list at the top called "DCode", which lists the size of the tool right next to its code (and also highlights all elements painted with that tool).
13:40 would that have to do with the soldermask? They seem to made their masks a slight bit larger, if your two pads are too close together, maybe that could break the separation?
[edit] oh, you actually mention this later in the video, haha, I was too quick to comment there, sorry.
I'm loving the floating head thing!
Love the floating head! I cannot avoid to think how floating hands also would feel
I read English / Metric conversion rounding caused issues with the Soviet IC industry. They rounded the 0.1" spacing. If boards/sockets and chips were from mixed systems, the more pins the harder to insert,
Big thumbs up for the floating head!
I add 16mil isolate on inner layers since I've had boards come back with the holes too far off. I also have had boards with spots that are not completely etched. I also had a board house not remove Silk over pads.
For an excellent Gerber viewer suggest you have a look at FAB3000 (paid version) or DFM Now (free version)
In above you would have been able to double click on a track to see it's width. Also with the paid version Gerber manipulation becomes crazy easy compared to say CAMtastic ;^)
Even before you announced it I fell in love with your new shirt
If the defining spec is in mm, then that's what should have been written first. The number in parenthesis should be the conversion. Also, since this is a minimum the conversion should have been rounded up (or else omitted altogether). The way this boardmaker has written it is erroneous on their part and leads to unnecessary confusion. If this has been pointed out and they've refused to fix it then they're just being stubborn.
Definitely a good point to pay close attention to this. I never considered it before, and I think I have my track spacing design rule set in mils so I should change it.
Many people say "never design to the minimum" and I agree with that if you have the flexibility to be relaxed. Personally I'm not in a situation where I need to push the limits so my design rules are set very conservatively. But the manufacturer is being asked for this information for a reason. If they say they can build something then they should be able to build it because some people are in a situation where they have less flexibility and these specs actually matter.
I don't expect them to argue and negotiate over a $5 job, but if the same issue comes up repeatedly from different people then it should be taken as a sign that the published specs need correction or clarification. But I won't lose sleep over the flaws of some China industrial business' web site, it's more useful to just know how to deal with them so that point is well taken.
Floating Dave finally!!! Awesome :-D
I love the floating Dave head!
Floating Dave head is probably one of the creepiest things I've seen in awhile, but hilarious
The difference in clearances between traces and pads must have to do with solder paste or solder masks -- something other than the copper.
Floating dave head is amazing, you should most definitely do it again!
The gerber viewer has grid options for 0.1 thou. It lists imperial first and then metric.
Any news of the custom LCD µsuppy ?
It's on the to-do list.
@@EEVblog Item #7,421 :D
The head thing reminds me of RIchard III from the first Blackadder episode. "Have you got transport Dave?"
I think jlc do that board with 3,5mils, even with 3,4mils ... But also care of that these are the values for 4layer boards, a 2 layerboard needs 5mils (other machine is working on that)
They just do that fine. The bigger "problem" is, that i recently feel like the coworkers get more and more closer looks to the boards and recognize things like that, before giving it to production. That might be the problem, that they text you they wont do that for you. If they trespass the board to production, everything will be fine, without problems.
Absolutely love the floating Dave
If you spend a whole lot of time designing a board that got rejected for some tolerance issue, I'd just look at the other prototype PCB companies before spending another week redesigning.. maybe for another buck a board they can manufacture it
It would probably be good practice to look at a variety of companies and use the WORST SPEC of any company for each design feature, then you have the most flexibility later if you want to get it built by someone else
Good video, thanks! Looks like old good Dave is back with us. :)
Good to know, I typically work in mils when setting up my design rules. Lowest I've had was 4 mils with jlcpbc, so guess I was right on the edge of I needed to do a bit lower.
LOL floating head was the best. Great video.
You can assign a track to a net in KiCad (eeschema) by editing the track and picking the net at the top of the dialog, then it should let you connect them.
I haven't been designing PCBs for long, but it kinda looks like everyone is leaning towards metric now, most surface mount components I use have package dimensions reported in mm.
The severed talking head format is great Dave! It's hilarious, keep it!
i like the dave head without shoulders, saves screen space.
I don't go below 8/8 thou (0.2 mm) with these Chinesium board houses, not that I've had reason to. Printed silk screen registration is usually the worst followed by solder mask.
LOVE floating Dave :D. it is better than pull up or pull down Dave .
I agree with you Dave, it did look better with the shoulders aswell.
it is not just drill wander. when you laminate a 4 layer board, things will not line up perfectly.
Top and bottom are slightly off, inner layers are slightly off, the whole thing is slightly bend and warped, ...
(let alone: vias are "texas sharp shoter", as the hole is drilled before the chopper is etched away)
As I like to say: Yes, we are splitting hairs here - because we literally are in physical dimensions, where a hair is a relevant unit of measurement.
This is the exact problem I've faced with jlcpcb. If you have bgas is quite hard to route out signals. Due to those tolerances of pads and vias!