It’s hard to say what the underlying problem may be from just a few screenshots.
There are lots of dialects around of the Gerber format (which started 40+ years ago) and it seems quite common that PCB manufacturers stick to old buggy software because it’s “working” in their machines.
It is a bit suspicious that apparently two different manufacturers see the same circles, but it does not mean much because they may very well use the same buggy software. It does appear though that they have a different amount of circles and on different layers.
Some time ago there was a thread here about a bug with circles in the gerber software of JLCpcb (especially the online viewer). It was about radiusses in pads not being rendered correctly, and even someone from JLC participated in that thread to get to a solution.
KiCad’s gerber files are generally quite good but sometimes uses newer features of the Gerber format that not all manufacturers can handle properly.
I have a vague memory that it sometimes helps to uncheck Pcbnew / File / Plot / Plot format: Gerber / Gerber Options / [ ] Use extended X2 format to gain more compatibility with PCB fab’s that use older or buggy software.
Another unrelated issue I see on your PCB is that you (apparently) have lots of Silk Screen text on top of pads. This is not done.
There are a few different ways that PCB fab houses handle this.
Some just refuse your PCB.
Others print the Silk screen directly on the pads, which make them unsolderable, but it is what you specified.
Others point you to this, and ask you if you want to change this, or want to make the PCB as is.
Yet others simply clip the text to keep the pads clean.
Another unrelated issue I see on your PCB is that you put a bunch of via’s inside pads. This does not matter much for hand soldering, but if you use a solder stencil, then a fixed amount of solder is available for each joint, and most of that solder will wick into the via’s and this leads to bad solder connections and rework. You have plenty of available space to put the via’s elsewhere.
These are pretty common mistakes beginners make.
It probably is a good idea to do a more thorough review before you attempt to manufacture this PCB.
Hi,
thanks for you feedback on the pcb. regarding the silk on pads, I thought that the solder mask had preference over the silk, so there wont be silk and no problems so far with this. I guess most of them went for option 4. I put them there to get them out of the way. Regarding the via on pad, fab has confirmed that they can do it and so far its the 4 spin of the pcb with no issues related to vias on pads. I don’t do it unless Its necessary, there is not that much space as it may seem. I will take into consideration your advice for future designs, thanks a lot.
I did not read the tread (again), but I did see the screenshots with the wrongly parsed rounded corners in the pads.
I did not imply you have the same issue though, I just meant it as an example of PCB manufacturers using buggy software.
I have vague memories of some popular program as interface between the Gerber files and the actual plotters that apparently gets pirated frequently and has known bugs / issues or otherwise does not work correctly.
What you can do is find some online viewer that does show these circles, then simplify your PCB and upload again. Iterate till you have a small file that still shows the error, and then look at the gerber file what is the root cause of these discrepancies.
You can also determine the center point of those circles, and then search for those coordinates in the gerber file. You may get lucky and find the right lines quickly.
I do have indeed circles in the silk. And in the latest screenshot I got from the fab they are using CAM 350 v9.5. Thanks a lot @paulvdh@davidsrsb and @der.ule for the threads. Really interesting reads for sure. I wish kiCAD had cam output presets like eagle used to have. I read through the threads shared and try the posibilities described.
You can have KiCad clip the silkscreen when you generate the gerbers. Look at the screenshot of the plot dialog provided by @paulvdh and you can see that he has Exclude pads from silkscreen selected. I suspect that you don’t have that selected, thus the silkscreen over pads in your gerber rendering from Ucamco.
Try ticking “subtract solder mask from silkscreen” this should at least remove silk from the pads. Whether it is still too close for your fabricator based upon how fine they can print is a different question
One would expect fabs to read files that comply the Gerber specification. Point.
Of course, bugs happen. That is OK, but something must be done about them.
This bug is there forever, and is in a feature that is at least 30 year old, but those fabs simply make no attempt to have it fixed. I results in confusion, waste of time and scrap over and over again. JLCPCB, for one, is a repeat offender, as attested by numerous posts in this forum.
Is it too much to ask that fabs handle 30 year old features correctly?
yes as it is “good enough” and to be compliant to the spec costs money. This is one reason I always tick the X2 box because if a fab house preview or their CAM engineer comes back with oddities I do not use them because if they are unable to parse X2 in a setup that only does X1 they have some serious problems and thus what else do they not do?
Nope. It’s not “good enough”.
I find it quite astounding that there is still no standard way of specifying simple things such as for example V-grooving in Gerber files.
Many man-hours are wasted with double and triple checking Gerber files, and ambiguity in standards cause many delays in PCB manufacturing. For some kind of reason I do not understand most people seem to accept this as “normal” and the cost of human hours wasted seems to be not a part of the equation.
[Edit:]
Accidentally bumped into this post from a few weeks ago again today, and did some reading about Gerber in the time in between.
There is a possibility to specify V-grooving and a lot of other features in the latest Gerber standard. Real pads can be defined for flying probe tests for example and lots more, and those may have been there for some time too. So now it’s already time to implement those things in the software and start using them
This whole discussion demonstrates it is not ‘good enough’.
As a trade, we want to become better and more efficient, one supposes. By sticking to 20 year old, maybe cracked, software JLC etc block progress.
I could not agree more.
Complying to existing standards, as they are, would be a good first step.
As long as JLC etc refuse to use anything else than 20 year old, maybe cracked, software we are going nowhere, however brilliant the specification.
