Ambiguousity

Building what I can because I can.

[fyberlabs]

CLICK HERE to support Fyber Labs Light Square Flexible RGB LED Panel

64 APA102C RGB LEDs on a cuttable flexible PCB. | Crowdfunding is a democratic way to support the fundraising needs of your community. Make a contribution today!

Our new Light Square RGB LED panel is live on Indiegogo!

[ambiguousnet]

Our new APA102C based flexible Light Square panel.

[ambiguousnet]

We put some of our designs up on Tindie

Fyber Labs has Flex Module prototypes for sale at our new Tindie store. We will be releasing more Flex Modules and demo projects utilizing them.

[ambiguousnet]

More updates on Hackaday

We have been putting up a significant amount of updates on our development progress and process on Hackaday. Please feel free to join up and follow us there to learn how our design is evolving. We are looking for feed back and contributors.

Eventually I'll be adding demo projects for Flex Modules as well as turning some of the posts here into more formal how-tos on Hackaday.

[ambiguousnet]

Posting project details on Hackaday

I'll be posting personal and Fyber Labs projects to Hackaday. I'll keep posting notes and process information here and progress on the Fyber Labs blog.

[ambiguousnet]

Solder paste refrigerating

There are many who recommend storing solder paste below room temperature (ie. refrigerating). Recent discussions about it and experimentation tells me that you get better results if not longer lasting solder. Solder paste that is refrigerated tends to dry/bake in the stenciled shape, while warm paste will run during preheat (or have severe "slump"). For most designs, this doesn't matter. During reflow the solder surface tension and solder resist will pull the solder back in place. However, it does increase chances of excess solder bridges and solder wicking depending on pad size, heat absorption, component proximity. So, I think it is good to keep the paste cool.

I tend to buy solder in syringes which limits air exposure. So far my year old solder paste is still going strong. Suppliers don't appear to store or ship the paste with any environmental consideration, so I think refrigerating only matters before you work with it.

[ambiguousnet]

For making the small modules, solder was stenciled on, the board populated, and then a heat gun used to reflow the solder. OSH Stencils tend to be large and their jigs are too thick. So I am using ‘bad’ unpopulated boards as jigs to stabilize the stencil.

[ambiguousnet]

Tag Connect

You may notice on my board designs that the SWD connector is a tiny bed of nails (well Tag Connect call it Plug-Of-Nails). These are designed to work with Tag Connect's TC2030-IDC. These can be used by most JTAG programmers and minimize the space used for programming. I am just starting to use them, but it looks like a quick and simple solution that doesn't cost too much.

[ambiguousnet]

Organizing components

I am hand assembling my module boards. I sent Gerbers to make stencils from OSH Stencils. They are quick and ship within a day or so. OSH Park is incredibly cheap for my tiny modules, but will take up to two weeks. Castellations from OSH Park come unfinished, so expect to spend some significant time clipping and sanding them.

I wanted to note how I organized assembly. It isn't too interesting, but I think it works well. I have sample books for small components and various sized supposedly ESD safe SMD containers that connect together. My method for organizing is to place the components near similar component types and then by size. So a page of 0201 would go before 0402 sized capacitors. Resistors or inductors would have their own sections.

Larger component cut tapes in quantity or size as well as loose components go in the SMD containers. Here are some of the bigger SMD containers available that still lock together.

Not too difficult, but I also made SMD containers to hold each part in a board's BOM along with their boards (the boards are really tiny). For optional or test parts, they are similar numbered and labeled with only the component name changed. When I build something I cut single components from the sample storage and place them in the BOM container set. I obvious log which optional components go in which build and mark the board. I made the labels with full sheet labels and the BOM orders by running them through Microsoft Publisher as a catalog. I don't think Publisher is well maintained as it is a bit buggy and not too featureful compared to Word. I made the label sizes in Word and them imported them in to Publisher.

A finished BLE 2dB board on a honeycomb soldering pad:

[ambiguousnet]

Hardware Design Notes and Oopses

More confessions on how not to do things.

I am designing a complicated prototype that has two microcontrollers, a variety of sensors, Bluetooth, and USB power for Fyber Labs. These were designed as modules that could be individually breadboarded and well also be mounted on flexible PCBs.

I rushed out designs before double checking the bill of materials (BOM), so I missed a unique component pattern. I made several versions of the same board to test different design decisions (assuming everything else would work). I would not recommend either decision.

I used Mouser's BOM loader to parse and correct them and manually re-imported them to the schematics.

I then merged and sorted the BOMs in Excel so that I could make sure I could order enough parts for all of the modules. I planned about 10 sets of components to give me plenty to screw up on and to hit some quantity discounts. I loaded that BOM into Octopart and selected the cheapest sources manually. I assume that Octopart charges for (or should charge) instant buying. One issue is that many of the Digi-Key components selected are on reel instead of cut tape. So you get a $7 charge and a useless reel for small component orders if you are careless like myself. Otherwise, I think that was a good method to easily consolidate the cheapest order possible.

0402 and smaller footprints need space. The components will bridge and merge with near components. I would recommend around a .20mm separation at least. It is hard to keep perspective when working this small, so now I know to spread footprints out more. You also really need a microscope. I purchased the AmScope SM-4TPZ-FRL-3M which is great for soldering, examination, and testing. You can see two components touching each other on the bottom left. Luckily it is on a shared trace, but this board revision does not have a good yield because of spacing.

[ambiguousnet]

DipTrace defaults

Just some notes on updating PCB layouts in DipTrace. I made some errors and updates to patterns. In the case of changing paste or mask layers, those changes will not propagate to an existing pattern. Similarly if you stupidly reverse a pin layout like I did, updating the PCB Layout will result in only a notice that the pattern is 'flipped'. I can see how you might not want to force updates that may screw up a set pattern, but it would be nice if unmodified default options could be updated. That and some weirdness with library paths after updating saved files from the Beta are all the issues I have had. DipTrace 2.4 is pretty amazing. I plan to discuss some of the cool 3D preview capabilities later. You can see an example on the Fyber Labs blog with my module boards applied as components on another board. I think it is an incredible value for an EDA and blows all low cost solutions away.

[fyberlabs]

Sorry we haven’t made any updates for awhile. We have been working on some new concepts.

It is common now to use castellations for module board designs. WiFi, Bluetooth, and other communications modules can be found with them. Since we are trying to release as much as possible as Open Source...

[ambiguousnet]

We have been working with DipTrace for the past year and the new 2.4 version is just amazing. Fyber Labs will be releasing more board designs soon -

[ambiguousnet]

Hardware Workshop Review

The Hardware Workshop was quite informative. I would recommend it to anyone that is making a physical product (hard good, soft good, electronics, etc.) for the first time. There was a good amount of advice for regular products as well as electronics.

I learned that there are very emotional and old school business policies wrapped around retail and that it is an incredible channel worth the margin sharing, but not something for electronics for a variety of reasons. Some of our hardware product concepts may sell through specialized retail, so I think we should research more about those channels.

Another depressing reality is with crowd-funding. No real funding partners would take a company seriously for later funding rounds if they ever did equity based crowd-funding. Between that and the rules around it, I would say it would be quite a risk to a business with little temporary reward.

And when it comes to pre-order/donation systems like kickstarter, it is pretty tough. The marketing campaign has to be set and ready to go well before with either an exclusive journalist agreement or well handled embargos with lots of hand holding and meetings. Kickstarters cause tax issues and using a major funding site is costly and does not necessarily bring any true value vs. your own site. For most hardware projects, the success goals are way too small. Even on extremely successful projects there is substantial risk from experience and unexpected costs/delays and a low chance to break even.

For those that don't know, small batches need small factories which means possibly unreliable partners. Most stories indicate that it isn't likely your factory will try to cheat you. It is just more likely that there will be a miscommunication or lack of specification through your and their lack of experience that could cause huge problems. Also, all electronics need FCC certification in the US and possibly additional licensing and registrations. You should plan for a minimum cost of around 20k just for the FCC. Other countries may require additional certification and registrations.

A large amount of the discussion was on Branding. Branding is very important and a way to make an emotion connection with customers. Treat your brand and company as a person and invest consistent values and ethics into its marketing to show who this person is. There were many successful examples of this demonstrated, but it might have been good to discuss marketing failures more.

If anyone would like a copy of my notes or advice, please contact me. I would highly recommend attending the future workshops.

[ambiguousnet]

Macs (and Parallels) are crashy as heck

OK, so with my old Windows machine it had a badly designed video card in it for over a year. It was yet another victim of European no lead solder laws and BGA size requirements which is getting to be the norm. (Hey EU, which pollutes the environment more - lead mixed solder or hardware that only lasts 6 months to a year?). Anyway, it would lock up at the driver level if I ran a game too long or had flash running for over an hour... Otherwise it never really crashed other than when the Seagate hybrid drive would screw up (PROTIP: never ever buy hybrid drives).

With the new shiny top of the line Macbook Pro I have, I see a crash at least once a week. Either Cocoa crashes and the windows disappear or go weird (like explorer dying in Win98 or Xwindows dying), or the whole thing locks up completely like good old non multitasking non memory stable/safe MacOS pre 9. What I would give for a f'ing bluescreen so I might know where things are going wrong. Heck if you are going to moronically fill system.log with every damn event, at least make an event viewer like Microsoft did. Of course you could be less idiotic and parse them out to more relevant log files. I assume much of this is Parallels fault along with Apple's complete lack of supporting virtual machines well. At least the file systems have journals, though I am still a bit worried about that implementation too since I doubt they test it well given the average Mac user.

[ambiguousnet]

PCB Library Expert Lite

I tried the shareware version from PCB Libraries. They are trying to build a community around their expertise on footprints. They look knowledgeable, but maybe a little too closely related to IPC. Anyway, they claim to support Diptrace, however their product only outputs in an importable file format. So it isn't directly compatible. The results are more than a bit iffy with very complicated splines and the pads appear to not even follow their own interpretation of correctness (every pad should be a rounded rectangle). In fact their D styled pads don't seem to have consistent arcs that line up with the various board layer types in Diptrace. Anyway, the lite version doesn't appear worth utilizing much less buying the full version. It seems slightly better coded than the Ultra Librarian .Net abortion from Accelerated Design, Inc.. In that it doesn't crash randomly or have a database full of duplicates. Ultra Librarian does export to many formats (again only import to Diptrace) and handles the bxl format that many manufacturers will release. So the Ultra Librarian Free Reader is worth having for its conversion capability though I don't think I would trust its full version with the supposedly huge library. Then again, they may keep it more up to date than the free web database. I guess I am doomed like most engineers to make my own drawings.

[ambiguousnet]

Marc Barros Hardware Workshop

I'll be attending the Seattle Hardware Workshop at Synapse on March 14th and 15th. It sounds pretty good and I have seen some of the speakers involved with incubators before.

[ambiguousnet]

0402 footprints

Most reference designs these days utilize 0402 passive components. It is fairly common and supposedly isn't too hard to reflow. I do have some misgivings about using it in our design without paste stencils and solder resists (hey, I like to prototype cheap and fast). Luckily, you can get cheapish Polyimide stencils by order from OSH Stencils. Mike Purfield recommended it to me and I will try them out. They use a 3mil or 0.003in thick sheet. I plan to order some brass at the same 0.003in thickness to try to etch my own paste stencil. According to the IBS SMT notes this should be thin enough for even 0201. In fact it may be a little too thin as 7 mils would be small enough for 0402, though 3 mils should etch quick I hope. I would also check IPC recommendations, but those cost a ton of money for a PDF. From what I have seen their documentation is dense and unclear, and it appears few manufacturers pay attention to their authority on most matters. I can also try to make my own solder resist, but I need to experiment with some coatings first.