Flex Modules

Hi, we have a few Flex Modules almost ready to sell again. With the new Hackaday Prize focused on solving problems for the world and humanity, we will be developing and entering demo projects utilizing Flex Modules. If anyone is looking to develop a wearable or utilize sensors in a small space, let us know and we would be happy to work with you.

Now that the footprints for most of the Flex Modules are set, we will begin producing libraries for various EDA tools. We will support DipTrace, Eagle, KiCAD, and maybe others. We plan to support Fritzing and Cynthia(Cindy) and I will be releasing breadboard demo designs for it assuming we can make decent SVG images for the Flex Modules.

I'll put some effort into cleaning up schematics and producing PDFs to make it easier to view from the web. We will also produce a datasheet for each board. The BOMs are complete for the new Flex Modules and we will be testing them in the next few weeks as they come in.

In between Flex Module board testing/redesign work, we will be working on testing jigs and demos as we can. I am getting quotes from the distributors and I'll also be talking to manufacturers to try to negotiate prices further. By the end of March we should be selling Flex Modules again.

The past couple of weeks I have been working through my long todo list. While doing so, I have added some more ideas. I believe it will take a month or so to test this many boards. I finished up some of the high voltage designs, which should be safe with a little care. The i2c, SPI, i2s buses for Flex Modules will be consistent.

i2c master (STM32) is laid out clockwise SCL, SDA
i2c slaves clockwise SDA, SCL

SPI is meant to fit point to point. ST has a specific layout which we will follow.
SPI: NSS, SCL, MISO, MOSI (which ST maps i2s with -)
i2s: WS, CK, ext_SD(SDI), SD(SDO)

I'll be working on a few more Flex Modules as time permits, especially better ECG designs. However, the priority will be to get documentation, working demo projects, and firmware together to bring interest to the Flex Module concept.

I am reviewing the i2c based boards I finished to make sure the SDA/SCL pairs are more consistently laid out. I was lazy about it, but it is now a concern that I am implementing an i2c switch and I want to make sure to minimize the need for vias when Flex Modules are attached to circuit boards. So some of the boards may have their pins reorganized so they can benefit. I will likely settle on slaves with top of board counter-clockwise VDD, GND, SCL, SDA pin layout. The i2c Switch slave will be the same, but masters out will be the reverse (so they connect with the slaves).

I have designed a new 12-axis sensor. This replaces the 10-axis that had the more expensive MPU-9250 and the footprint never seemed to work well. Hopefully this design with the ST LSM9DS1TR is more reliable and adds additional sensors. The BLE 3dB board has been updated and I plan to tune the matching circuit once I can rent a network analyzer. The Qi/USB charger is ready for initial testing as well. I am finishing up what might be the last designs of the USB BiPower (now with OTG) and the STM32 boards. The STM32, BLE 3dB, USB BiPower, and the Qi/USB charger Flex Modules are designed to layout next to each other in a variety of combinations for a wide range of prototypes.

I have had some questions about removing products from Tindie. I am not sure it is permanent. Tindie's new policies may ultimately improve sales, but Tindie pulling listings without notice is not good behavior (I guess I made it worse, but most of listings were of similar 'quality'). Originally I was just trying to update quantities on the PCA9685, but they wouldn't let me update any info unless it matched their new policies so I felt forced to do something.

Anyway, I am mostly concerned with sales tax issues. It worried me and is still a major compliance issue that Tindie ignores over this. I do have an e-commerce platform with our website, so as I make new listings I will likely put them up on both and request WA residents use our site.

I have been thinking about moving to a better selling platform. Tindie is very limited in that it doesn't support state sales tax, has a very simplistic shipping and inventory model, and only one type of discount. Tonight they pulled down some of our LRA devices we were reselling as they didn't meet their new guidelines for large descriptions and short titles. So, for right now I am pulling all our items off Tindie and will try to come up with a better platform.

It is hard making tiny circuit boards.

OSHPark and most PCB fabs use a silkscreen process to apply text and graphics to boards. OSHPark tweeted it has a 5mil width limitation, in that it might not print line widths smaller than that. The standard seems to be 6mil. I use Diptrace's vector font which is a much tighter and in my eyes better font than Eagle or even Altium's (which seem the same). Through a bit of ignorance and luck I have found I can get simple text in this vector font down to 3mil width approximately through OSHPark. It only works for one or two characters (like numbers), can't be next to or partially on a trace to avoid ink runs. Anyway, I plan to revert to larger fonts as it is too problematic and we plan to start doing panels in the near future not through OSHPark. So I will be bumping up the thickness. In general I have had consistent text above 4.5mil. However, that may only work for OSHPark. DipTrace lets you specify line width/thickness as well as character width as a percentage of height. On a 3 point font, 40% may still be legible, and 50-60% don't look squished. So over the next few runs I will be improving the silkscreen text. To compensate, we will publish some annotated pictures in case the silkscreen is ruined on a batch.

Of the few boards we have up on Tindie, I have been rethinking the current pricing. I was at the time, basing prices on a mix of perceived costs and hopeful scaling. To not bury the lead, I will be increasing the pricing (hopefully temporarily), but we will have a discount Tindie sale going on this weekend. We will also have about a dozen or so of each board available in the coming weeks in what is hopefully the final revision till we can panelize the boards.

To explain things better, I'll go over where the current prices come from and our costs now and in the future.

The initial prices I came up with for the boards were based on 4 times cost of goods. In this case, the cost of the boards from OSHPark and the components based on small quantity purchasing from Mouser or Digi-key. These prices were high, so I tried to set prices compared to other breakout boards with similar components. I also started negotiating with Arrow and have received reasonable pricing and planned to eat some costs while building to higher production. It was more important to see what boards people were interested in. So far, it is not many.

At this point, I think we should concentrate on producing demo projects utilizing the boards and flexible circuits. This may spark interest and eventually justify higher quantities and better pricing. So in the next couple of days, I'll be recalculating the board prices based on costs and effort. I'll also be separating out the options to individual products to make it easier to understand what we sell. If you want them cheap, get them this weekend with code 101971C (or right now before I figure out what the real prices should be).

Our current costs: With the added size of the new castellations, OSHPark is a bit more pricey going from $0.50-$1.00 to $1.00-$3.00 . I have been swapping out passives to be sourced easier by Arrow or Avnet, so when we can order from them the price may come down. Right now components cost around $4-10 and will drop to under $5 for most boards once we are buying consistently from Arrow.

Hand assembly is around 15-30 minutes, inspection and testing is also around 15-30 minutes. I will be developing bed of nails testers for some Flex Modules once volume increases to justify it. Any reworking necessary is past the pricing of the module, assuming it is repairable. Programming isn't factored in, though we may load Espruino into the STM32 Flex Modules if there is interest or we put it in demos.

We can make our own panels with multiple designs. A couple of CMs have deals for new customers and depending on how little engineering is needed it can be comparable in cost. The pricing looks to be around $1.00-$2.00 per board, but with better silkscreen. (I have been having serious problems with silkscreens on boards this small). This can be a cost effective way to proof single board panel runs with better tab routing or v-cut compared to OSHPark and we can pick the board color. So we can do runs like this and even if a board or two turns out bad, we can still use the others.

The initial single board panel runs will cost similar to the multiple design panel with NRE priced in. However, boards will then be sub $1.00 after assuming no revisions. This doesn't save much money, but a single board panel will allow us to use an assembly house. The first assembly house run will require NRE and a steel stencil which effectively double the costs per 100, but after that run boards will only be $4.50 to assemble.

I have a feeling it will take a kickstarter or similar to get us into production levels. We were thinking of producing project kits (ie. the Dance Kit) with Flex Modules and a ready to go flexible circuit board. That may be what draws interest and may also make a decent kickstarter. So we will be working on the Dance Kit and similar concepts.

Well I have been working with Arrow, TI, and Linear Tech on charger and electro-stimulation designs. That was pretty productive and I should soon have more power options for flex modules and updates to the flex modules.

I am pushing the I2C resistors to bus specific flex modules instead of the microcontrollers. This allows for those microcontroller pins to be utilized in other capacities and still allows simple plug in use of the flex modules.

I have been experimenting with improved castellations. I had utilized perf boards as jigs to quickly solder pins when breadboarding and planned to demo that for others. However, that may take more skill and patience than most people have :). So I am going to go with a whole 2.54 hole and then a castellation just 6mil after. This adds about 50mil to the outer edge.

Manufacturing quotes for large runs have been difficult. We have a local assembler Schippers & Crew that helped me refine panel designs and I received some tips from Seeedstudio. The DRV2605 LRA Driver Flex Module requires either a Via-In-Pad or a 3mil trace to route the middle DRV2605 pad. Between that and the size, I have had to request custom quotes and panelize the boards myself for proper quoting. Prices in the US have ranged from $5 to $20 per board in a panel of 100. Keep in mind this board is less than 1cm^2. TI is coming out with a slightly larger VSSOP footprint soon of the DRV2605L, so I might produce a lower cost version based on that. I plan to make an initial DRV2605 version with Seeedstudio in the next couple of weeks as they still have the lowest price per board ($3). If supplies of the DRV2605L appear in time (need to confirm with Arrow), I may just switch directly as it should fit OK.

To panelize I had made a tab routed version, but it was very space inefficient as it has a 100mil inner rail to add rigidity to the panel. Seeedstudio recommended rotating the boards 45' and attempt a V-cut or scoring along the non-castellated corners. This has produced a much more efficient panel, however DipTrace does not support 45' text in layouts. It does import gerber files, so I am working in another EDA tool to make the text and will import it in. The V-cut should provide a very clean look to the boards, hopefully I can similarly panelize all other flex modules.

We now have some initial international shipping options. They are the USPS flat rate priority, so they are pretty pricey if predictable. We will try to provide more reasonable options eventually.