JLC PCB Prototype Assembly - Incredibly Cheap

[cyberfish]

I also asked them, specifically about USB connectors.

They replied asking me for LCSC part numbers. I gave them a few for USB micro-B. They said they will ask the assembly people to add them, but no guarantees.

That was about a week ago. Doesn't look like they have been added.

I'm holding off my trial run because of that - I can't use a stencil to reflow the missing connectors myself if there are already components on the PCB.

[ebclr]

Do they have a list of those  689 Basic components

Where can I find that fist?

[yombo]

Here is the component database search page link:

https://jlcpcb.com/client/index.html#/parts

[pmnxis]

Wrote some python script for rotate parts in your Pick and Place csv file.

https://github.com/pmnxis/RotatorPnP

It will be useful if you use own footprint library.

I rotated my new revision of PCB with this script.

And checked it's working well in my case.

and you need to check and modify rulebook csv file for your footprint library and components

[stefanh]

I am curious to get the complete list of Basic parts. 

is there a way to display more than 10 items per page?

A spreadsheet with a list of all the Basic parts would be ideal, to assist with making a specific LSCS basic parts library.

[teksturi]

A spreadsheet with a list of all the Basic parts would be ideal, to assist with making a specific LSCS basic parts library.

Here https://www.eevblog.com/forum/manufacture/jlcpcb-smt-assembly-service-208764/msg2728288/#msg2728288

You cannot trust this list blindly. Always check with tools that JLC gives and updates.

[rrinker]

 I guess I really need to start checking in sections I otherwise wouldn't have much interest in - I don't do SMD at home, but I could make my boards a lot smaller using at least SMD passives and using their assembly service to load them. I use EasyEDA and have my boards made by JLC anyway, the last one I also used the BOM and transferred it to an LCSC order and ordered the parts I needed as well, since I didn't have enough on hand to build them all anyway. Might as well integrate one more step. I've used commercial boards which I suspect used a service such as this to offer a "kit" version in which the micro and all the passives were SMD already soldered on and the only actual assembly for the kit was to add a pair of switches and a pair of leaded 3mm LEDs - saving $2.50 per unit by spending a coupe of minutes soldering.

[DBecker]

I just got back my first run of boards through JLC assembly.

My biggest challenge during ordering was the always-mentioned part rotation.  I believe that they use EIA orientation, which is ad hoc, while Kicad consistently puts pin 1 in the upper left quadrant (or on the left in the degenerate two pin case).  I spent many hours working with Kicad scripting and their use of Python 2 (!) in an attempt to automate or document the correction, but in the end I submitted a hand-edited .CPL file.

Their assembly layout visualization tool was a huge help with this.  It clearly marks the polarity or pin 1 of asymmetric devices with a red dot.  Combined with tweaking the silk screen on a few of my footprints it was easy to verify the rotation.

I didn't experience any case of offset components that others have reported.

I had two issues with the BOM matching.  I added LCSC part numbers to everything I wanted populated, but their system always failed to match one or two components.  I changed the component description, but then some other part wouldn't be matched.  I concluded that this matching failure was related to the position in the CSV file, not some error in the file I was generating.

Because I was trying to automate file generation, I went through the ordering process many times.  I quickly tired of clicking 'Confirm' for each of a few dozen part types.  If I provide the LCSC part number in the BOM CSV, I shouldn't have to click confirm unless there is a suspected mis-match.  And then it should tell me enough to fix it for the next attempt.

Finally, the pricing of some of the part was wrong.  It exactly doubled the price of crystals and regulators between confirming the BOM and presenting the final total.  That only added $2-$3 to the total, not enough for me to delay my order by a few frustrating exchanges with customer service, but worth mentioning in a review.

My boards ended up taking an extra day or two longer than projected, with the delay largely between fabrication and assembly.  That pushed the finish time to Saturday, and the DHL pickup wasn't logged until Monday evening (6:57pm Shenzhen time).  But amazingly DHL delivered the boards to California on Tuesday 1:22pm PST.

[sokoloff]

@DBecker, are any of your tools for automated generation suitable for open source consumption by any chance?

[pmnxis]

I tried to fix rotation automatically by recongnize image on datasheet.
But giving value how many degree to fix myself per part was more effective.

I am glad meet you thought same issue with me and tried to solve this!

[JohnnyBerg]

My boards arrived today. 226 x 40 mm.
I've used EasyEDA, and send the Gerber. BOM and Pick and Place files without any editing. Worked!

[1Ghz]

My boards arrived today, too.

[sokoloff]

Seems like you probably would have wanted to exclude some of those pads from the stencil layer.

[jonroger]

I just submitted an order - it all went fine, just had to rotate one component by editing my CPL file.   

I'm curious - what algorithm might they be using for mapping from my unknown origin to theirs?    Ie, how do they know where   MidX=10,MidY=10 is on the board?

[cyberfish]

I just submitted an order - it all went fine, just had to rotate one component by editing my CPL file.   

I'm curious - what algorithm might they be using for mapping from my unknown origin to theirs?    Ie, how do they know where   MidX=10,MidY=10 is on the board?

Your gerber files use the same origin.

[SMTech]

There's no particular reason for the SMT line to ever see the gerber file. Your co-ordinates give an accurate account of where each component is in relation to each other, add some dimensions and a rough enough location of one Fiducial to get inside the requisite search window and the machine will handle the rest.

[cyberfish]

There's no particular reason for the SMT line to ever see the gerber file. Your co-ordinates give an accurate account of where each component is in relation to each other, add some dimensions and a rough enough location of one Fiducial to get inside the requisite search window and the machine will handle the rest.

Sure, but they don't have fiducials.

[SMTech]

There's no particular reason for the SMT line to ever see the gerber file. Your co-ordinates give an accurate account of where each component is in relation to each other, add some dimensions and a rough enough location of one Fiducial to get inside the requisite search window and the machine will handle the rest.

Sure, but they don't have fiducials.

So they use some other feature as a fiducial and curse you or not using any, its what I'd do...

[cyberfish]

There's no particular reason for the SMT line to ever see the gerber file. Your co-ordinates give an accurate account of where each component is in relation to each other, add some dimensions and a rough enough location of one Fiducial to get inside the requisite search window and the machine will handle the rest.

Sure, but they don't have fiducials.

So they use some other feature as a fiducial and curse you or not using any, its what I'd do...

They panelize designs from different customers, so presumably they add fiducials on the panel? They still need a way to get things to line up within each design, and the only way they can do that is through gerber -> CPL origin mapping.

[nigelwright7557]

I used PCBWAY to start with before I found JLCPCB.
I found PCBWAY a bit more expensive especially on postage.
A couple of GBP on the price might not be much but can push you over the edge into paying import/VAT charges.

[BlackICE]

I also just noticed that they do assembly. Just ordered my 1st board with a through hole design because SMT is a PITA for old hands and eyes. Now with assembly at a reasonable cost SMT is a good option.

I'm just learned DIP Trace and would like to know is there is a easy way to get a DIP trace compatible, basic and extended parts library to easy the pain of parts selection and use.

I guess I can learn to use EASYEDA but I prefer not to. DIP trace is the easiest one that I have tried. Previously played with KiCAD and Eagle.

[up8051]

I am preparing for the first assembly in JLC PCB.
I have small PCB (14mmx21.6mm), normally I have V-Grove panel 6x4 but on JLCPCB SMT Assembly FAQ I found information:
Please panelize your boards with stamp holes, V-cut panels are not supported currently.
But I don't found any rules for stamp holes panelization.
Has anyone used JLC PCB to mount the  PCB panel.

Regards
up8051 (JarekC.Diy)

[essele]

I'm getting close to trying this service, so I've written a KiCad plugin that will generate the correct BOM and placement files with a single click, and without having to edit any of the standard outputs.

https://github.com/essele/kicad_jlcpcba

You need to add an LCSC attribute to the components in the schematic with the LCSC part number, then everything else should be taken care of. Extra rotations needed are in the rotations.cf config file ... I blatantly stole the ones in there from somewhere else and have then updated with specifics from the board I'm experimenting with ... these are footprint rotations rather than component rotations, which seems to work fine. They are far from complete.

CAVEAT LIST:

1. I'm not a python programmer (which will probably be evident from the code), this is a rough translation from a perl version, but seems to work ok.
2. I've never written a KiCad plugin before, so this is probably horrible.
3. There isn't a lot of error checking, it seems to work fine for my test board, but I'm sure there will be lots of failure scenarios.
4. I've only tested it on OSX .. don't know if it will work ok on Linux or Windows.
5. I haven't actually ordered any boards with this ... please examine the placement view on JLCPCB very closely before committing.

It it's of use, then I'll put some more effort into tidying etc.

[DBecker]

I'm getting close to trying this service, so I've written a KiCad plugin that will generate the correct BOM and placement files with a single click, and without having to edit any of the standard outputs.

https://github.com/essele/kicad_jlcpcba

You need to add an LCSC attribute to the components in the schematic with the LCSC part number, then everything else should be taken care of. Extra rotations needed are in the rotations.cf config file ... I blatantly stole the ones in there from somewhere else and have then updated with specifics from the board I'm experimenting with ... these are footprint rotations rather than component rotations, which seems to work fine. They are far from complete.

CAVEAT LIST:

1. I'm not a python programmer (which will probably be evident from the code), this is a rough translation from a perl version, but seems to work ok.
2. I've never written a KiCad plugin before, so this is probably horrible.
3. There isn't a lot of error checking, it seems to work fine for my test board, but I'm sure there will be lots of failure scenarios.
4. I've only tested it on OSX .. don't know if it will work ok on Linux or Windows.
5. I haven't actually ordered any boards with this ... please examine the placement view on JLCPCB very closely before committing.

It it's of use, then I'll put some more effort into tidying etc.

Don't promote this until you have tried out a range of parts.  You don't need to order the boards, but at least check their board visualization tool.

Basing rotation on the footprint alone is *not* correct.  It needs to be based on the part type as well, with the possibility of over-ride when that does not work.  Compare the 0-degree orientation of a tantalum caps, electrolytic caps, and diodes.

[essele]

Thanks DBecker.

I did post it with the expectation that other people could experiment, feedback, and help improve it -- it's far from perfect (as per my long list of caveats)

I was definitely a little hasty, as in my efforts to check what you've suggested I have discovered that it doesn't work for the top side of the board, I'm sure it won't be a complex fix.

In any case, I've just checked a few polarised caps and diodes and (so far) they all seem absolutely fine with no rotation [not correct, see below]. Do you have any examples to hand that you know are wrong?

Also on the footprint vs part ... I probably don't understand enough about how this works, do you have some examples that I could test against?  The code to add per device rotation would be really simple, but I'd like to see some real world examples to test against.

EDIT: top fixed, but you are correct about the caps ... diodes all seem fine with no rotation,  but both electrolytics and tants (at least the ones I've looked at) do need 180 rotation -- I will do some experimenting.