@gbulmer I checked in my board (Maple Bacon) during the design process and every so often sent a copy through OSHPark, and feurig built and tested them. I find the iterative approach to be useful, though you do generate some e-waste.
Community rework of standard maple and maple mini.
(181 posts) (14 voices)-
gbulmer
Moderatorsoycamo - I strongly agree that an iterative approach is effective.
If you guys are willing to get boards made at OSHPark, build and test them, that would be brilliant!Low-cost PCB turnaround time in the UK is quite slow. The boards are sufficiently dense that I have routed a few tracks under components. So using Spirit's 'go naked' takes more care to assemble than a board with soldermask, and I only get one shot at each design.
IMHO good processes need to be adjusted to hit the goals. My goals include me understanding how to design boards that are manufacturable by 'unknown' suppliers. So I am trying to work to 'product release' quality.
As part of that mindset, I am from the school of "don't break the build". I think it can waste a lot of cycles.
I believe I have resolved all outstanding schematic and component issues. Schematic is complete, BoM is complete.
I have two sets of errors; they are only 'Stop Mask' caused by the bottom silk screen. However, they are not trivial, and will take a few hours to fix. I hope to get enough time today to complete, and will then futz around releasing to github.
AFAIK, many PCB manufacturers will post-process the silk screen to ensure it doesn't get applied to pads. However, I also know some do not, and will either make the board incorrectly, or will reject it. So I'd like to get it close enough to be a genuine test of manufacture. That way I can learn from defects, feedback teaches me something, and I am not wasting peoples time and effort.
Okay?
Posted 2 years ago # -
@gbulmer thanks for clarifying your process and goals for the build. That's an interesting approach to board design. (I think it would drive me mad but I do feel a certain sense of calm when using Eagle...)
I just don't want you to get overly wound-up over a particular board design before it's been Suspect Devices-tested. We're pretty good at breaking things. :)Posted 2 years ago # -
siy
Member@soycamo, most boards present at https://github.com/gbulmer/openstm32hw/ are tested. For some of them even photos of assembled boards are provided.
Posted 2 years ago # -
gbulmer
ModeratorI am pleased, and somewhat relieved, to say that I have reached a stable design for both the orone-mini, 0.6" row spacing (identical to Maple-mini, except av+ is supplied by the main regulator), and orone-mini-w, 0.8" row spacing (which is 'schematic identical' to orone-mini, but has the additional 'Power-on' LED and 'External Power' Molex connector).
The regulator has been replaced with a 'Tab-pin on ground' SOT-223. Nominally LM2937 as recommended by Siy, but there are several alternatives if cost is an issue, or voltages above 6V aren't needed.
I took yesterday evening 'off'. (One fix required the STM32F to move South 4mil and East 4mil, so it required every signal to have some ripup and re-route).
It 'passes' all of Seeedstudio's 'Fusion' PCB service design rules, except for one 'bridge'. The 'bridge' is two small pieces of overlapping copper, which connects two signals. It generate clearance and overlap errors, but will not make any difference to the copper. The 'bridge' is there to break 'ground' into two. It makes it a little bit clearer what is happening when routing, stops copper pours connecting where I don't want them, but is actually just continuous copper. It could be replaced by top and bottom 'Restrict' wires, but I find it easier to use while the board is changing.
I might get to github this evening, but more likely over the weekend.
soycamo - Thanks for understanding, and helpful remarks. I consult with a few friends, who are in or near my time zone, while the design is 'in flight', so it isn't as odd as it might sound. I agree, I find a sense of calm using Eagle, and also delight when a board is 'done'.
I know there will be defects, but I don't know what they are (though I'm 'not delighted by' the ground path on C6:-).
However I hope every defect will teach me something, and how much fun is learning?-)(I can have a bacon sandwich now as a reward, and another one when github is updated, Hmm, bacon :-)
(Edit: BTW, when I say 'bacon', I mean something yummier than USA bacon, which we call 'streaky bacon'. http://en.wikipedia.org/wiki/Bacon#United_Kingdom_and_Ireland I hope this helps cross-cultural understanding :-)Posted 2 years ago # -
siy
Member@ala42, that should not be a problem in the real life. This is a guaranteed minimum, vast majority of these LDO's will work from lower voltage. Also, the problem can be completely avoided by using P-channel mosfets for automatic power source selection. They have very low voltage drop.
Posted 2 years ago # -
gbulmer
Moderatorala42 - Very helpful 'catch'. Thank you very much.
This Micrel MIC5209 was one of the alternatives I spotted, it has a 500mV dropout voltage:
http://uk.farnell.com/micrel-semiconductor/mic5209-3-3ys-tr/ldo-reg-16vin-0-5a-3-3v-1-3sot223/dp/2100322
So I will default to that in the ECAD.Siy - I realise the spec is worst case. However someone might get unlucky, then worse might have an intermittent problem. This could be very hard to diagnose, which might be frustrating and disappointing. So I'll change the default part.
"Also, the problem can be completely avoided by using P-channel mosfets for automatic power source selection. They have very low voltage drop."
I understand how to use P-MOSFETs as 'low forward voltage drop diodes' to protect from reverse voltages. How are they used for voltage selection? I can understand selecting between two voltages using two comparators to drive one gate or the other. Is that what you had in mind? That might take quite a lot of board space.Posted 2 years ago # -
siy
Member@gbulmer, just use two "low dropout diodes" and connect sources of mosfets as output (connected to LDO). Drains connect to voltage inputs - one to VIN and one to VBUS. Number of components is exactly the same, but two P-channel MOSFETs instead of two diodes. It is also possible to use devices like IRF5850 and have only one component.
Posted 2 years ago # -
gbulmer
ModeratorSiy - "connect sources of mosfets as output (connected to LDO)"
Yes, understood."Drains connect to voltage inputs - one to VIN and one to VBUS"
Yes, understood.Where do the P-MOSFET gates connect?
If both P-MOSFETs gates are connected to the boards 'ground', then when both Vin and Vusb are connected, both P-MOSFETs will be switched on, and the host PC USB will be subjected to whatever voltage is on Vin, or the other way around. So that isn't the same effect as two diodes.
Further, if the voltage on Vin or Vusb is a 'diode forward voltage drop' larger than the other voltage, then that MOSFET will conduct regardless of the state of the gate. This has behaviour similar to a ordinary, non-Schottky, diode.
Is that how this works?Posted 2 years ago # -
gbulmer
Moderatorsiy - If you come up something please tell us!-)
I understand using P-MOSFETs for reverse voltage protection alone. It can have a tiny voltage drop, e.g. 0.1R at 0.5A = 50mV drop.
Selecting from multiple input power sources, and protecting USB power, seems tricky to do with a few components, and hence small board area.Posted 2 years ago # -
We are great at breaking things! I have had several problems where the silkscreen for a library part will break the drc but have yet to have any problems with running the drc with the silk screen turned off. Its one of the few errors that the fab's seem to ignore/correct. We will try to put the run a few through osh park on our to do list (no promises as the list atm is rather backlogged)
Posted 2 years ago #
Reply »
You must log in to post.