Teensy 3.6 "Pro" Feedback

[brtaylor]

I'm considering creating a custom Teensy 3.6. The notional design concept would be about a 25 x 25 mm board with high density connectors exposing all of the I/O (think Intel Edison, which was 25 x 30 mm and used two high density connectors). In addition to the MK66, I would likely add 2-4 SPI flash storage IC's (for data logging, the goal would be 128 - 250 MBytes of total storage), an MPU-9250 IMU, a BME-280 environmental sensor, and an ADC (at least 8 channels of >20 bit resolution data). All components would be -40 to +80 or greater temperature range. The boot loader would be available on a daughter board.

Price, unfortunately, would be around $70. The price is much higher than the Teensy 3.6, but the advantage would be a very capable microcontroller and peripherals integrated into a small package that could be integrated directly on an end product. I'm wondering if there is any market for this? It changes my design whether I would make a more general purpose Teensy 3.6 "Pro", as described, for a wider audience or a purpose built component. Any components that I'm missing that should be added? Or components present that would not be useful?

Thanks!
Brian

[defragster]

Doing the BGA 13 mm chip? Versus the LQFP that is 22 mm in the Ref board

Sounds fun.

[dkryder]

it seems to me that someone looking to use a component in a product or a project would be looking for the lowest cost to get the concept done and also interested in the component delivering what is needed and not whatever else it can do. however, there would most likely also be a niche market for your board. so to me it would be if the goal is to create something you want and then offer it to others not being concerned about if popular or if you want to create something that would become popular. my guess is since you ask for comments about if others find your idea useful that you lean to create something that would become popular and if that is the case i think your price point is outside of todays maker market. but, if you want it, build it. to me that's what is important.

[brtaylor]

Doing the BGA 13 mm chip? Versus the LQFP that is 22 mm in the Ref board

Sounds fun.

Yes. We'll prototype components with leadless parts in-house, but our yields are too low to do that for commercial products. There's a local assembly house that I use for products with leadless parts (Screaming Circuits), I've had great experiences with them and they xray inspect every board to ensure that the assembly was good. For long lead time manufacturing runs their prices are not bad even for relatively small runs. So once any leadless part is added (i.e. the MPU-9250 and BME-280 are both leadless), I know that I'll be getting it assembled by Screaming Circuits and they can handle BGA just fine.

[tonton81]

throw in a dual CAN transceiver while your at it

[brtaylor]

throw in a dual CAN transceiver while your at it

I knew I forgot something!

[tonton81]

its one of the features i wrote a driver for that i spent alot of time on

[brtaylor]

its one of the features i wrote a driver for that i spent alot of time on

In all reality, we'll be using CAN on this design. We currently use I2C for off-board peripherals, which is still very common in the drone world, and it works great if your wiring is good and a peripheral never drops off, but it's hard to keep the I2C bus up if peripherals occasionally drop off the bus. CAN seems like a really good option for our future boards.

Brian

[PaulStoffregen]

Over $50 is a tough sell when the market is filled with ~$39 Raspberry Pi3 + ~$10 SD card.

Then again, if you're integrating high-end peripherals, maybe it can work if you manage to market well to the people who specifically need those features?

[brtaylor]

Over $50 is a tough sell when the market is filled with ~$39 Raspberry Pi3 + ~$10 SD card.

Then again, if you're integrating high-end peripherals, maybe it can work if you manage to market well to the people who specifically need those features?

Yeah, it's really quite amazing how you've managed to keep the Teensy 3.6 costs below $30.

[tonton81]

supply and demand

[mjs513]

Just noticed this thread and I really like the features. My only concern would be the high density connectors. Intel Edison also had a couple of adapter boards that broke the pins to the equivalent of the Arduino uno so you could plug other periphs into it. Think you would also have to offer some sort of EDISON type of block or breakout board. You could almost make a Mega style breakout board that the new Teensy 3.6 "Pro" could plug into so you can access normal type of headers and possibly use existing shields if compatible. Just my thoughts.

EDIT: Definitely the CAN transceivers . I agree with defragster, sounds like fun. I would definitely buy one.

[brtaylor]

Just noticed this thread and I really like the features. My only concern would be the high density connectors. Intel Edison also had a couple of adapter boards that broke the pins to the equivalent of the Arduino uno so you could plug other periphs into it. Think you would also have to offer some sort of EDISON type of block or breakout board. You could almost make a Mega style breakout board that the new Teensy 3.6 "Pro" could plug into so you can access normal type of headers and possibly use existing shields if compatible. Just my thoughts.

EDIT: Definitely the CAN transceivers . I agree with defragster, sounds like fun. I would definitely buy one.

I really like the idea of daughterboards to breakout to breadboard size pinouts. Plus they'd be relatively low density and cheap to make.

[brtaylor]

I appreciate the feedback! Here's the crux of the problem - a rough BOM assuming that there is enough demand to build 100 boards at a time:
* MK66FX1M0VMD18: $14
* MPU-9250: $6
* BME-280: $6
* Flash: $14
* ADC: $10 - $15

That doesn't include the CAN transceiver, crystal, passives, PCB manufacturing, or assembly. The CAN transeiver, crystal and passives are cheap enough to ignore specifics, but likely around $4 total. Shipping, PCB manufacturing, and assembly would probably add $12 per board. So that brings us to the magic $70 estimate.

It’s a good deal if you need all of those components. Otherwise, hard to justify. One option would be stackable headers and splitting up modules - a “motion module” with BME-280 and MPU-9250, a “storage module” with the flash memory, an “ADC module”, and a “core” module with the Teensy. Other modules, like GPS, as demand arises.. Each would be about $20 - $30. So it would cost more total and the packaging would not be as good, but might be easier to justify buying one or two at a time. Then the question would be whether there is enough demand to support a diverse lineup of prefab'd modules.

It’s an option, I think the packaging would still be good enough for my purposes.

[mjs513]

I have been mulling your post over before answering and to be honest using multiple addons is not really that attractive with the added costs. I Think the attractiveness of the your original design is everything is on one board. My only concern was the high density connector and getting access to a set of header pins that I could use for other purposes. To reduce costs maybe delete the ADC? That would bring it done to $55 to $60.

Hope someone else jumps in here.

[KurtE]

I have also been following this thread. And wonder about how I would use such a board...

If I am thinking Edison like, I don't think things like IMU, sensors and the like on the main board fits that model. The majority of the time I used the Edison, I used their main Arduino breakout board, which sort of defeats the purpose of the small size. I did also use the mini breakout, but then often times you needed multiple daughter boards, which again maybe you start off small and maybe in two dimensions it is still small, but then the third dimension grows, and then you often needed the support parts to make sure the contacts stayed well connected...

Now if you are looking for a board for a specific purpose, like maybe something like a flight controller than it sounds like a nice board...

But again how would I use it for my own tinkering? And what would the costs be?

Example: if I want a board to control a robot...

Then IMU comes in handy, but I want to be able to program it...

So I need a daughter board. And since I don't like soldering things with real tight connectors, probably rules out me making it, both for the connectors as well as the chip to program the board and other supporting hardware...

Then I need to do a board that breaks out all of the other things I would typically use? Serial ports? SPI? Wire? USB Host? SD Card?

So again wonder what the total cost would be if what I wanted was something like a T3.6+, where all of the pins are brought out to standard set of pins, with all of the current functionality of T3.6?

But then again if I were wanting to mass produce some boards using the T3.6 like functionality, it would be great if you could simply plug the main board in and not have to solder in IO pins and deal with the pins on the bottom...

So again I can see where a board like this could be very useful. But still wonder if in many of those cases if you should be more like Edison and more bare bones...

Sorry if I did not make any sense here.

[tonton81]

i hope your still around when T4 comes out , beef it up

[HEE KEUN]

i hope.. add onboard BT5.0(or BT 4.x) or just wireless communication only chip (wifi , Sub-1 GHz band , zigbee... more type)

[brtaylor]

Thanks for the feedback. I have two thoughts on this:
1. Develop a set of hardware breakout boards fitting the current Teensy size(s) and designed to use stacking headers. Yes, OneHorse has several breakout boards for Teensy, but they are not meant to be stacked. This wouldn't really meet my needs for a smaller, embedded Teensy, but could be a useful product for the Teensy community and relatively low cost and easy to develop.
2. Develop an "Edison-like" version of Teensy. 30 mm x 30 mm footprint with stackable high density connectors to breakout all of the pins. A basic core board consisting of the MK66FX and then add on boards for different sensors and components. Daughterboards to breakout pins to 0.1" spacing. If I'm not trying to pack a ton of stuff on the core board, I could go with the LQFP version of the MK66FX and assemble in house rather than sending out assembly for the BGA version. Also could probably get by with a 4 layer board instead of 6 layer, keeping costs down.

[brtaylor]

All, I'm designing a Teensy 3.6 "Pro" to enable access to all of the Teensy 3.6 pins and provide better packaging for placement directly in a final product. This is a project that I'm doing, but I'd like to make it available for sale through my startup as well. The goal is to have about a 30 x 30 mm board with the MK66FX1M0. I'm hoping that I can fit a micro USB connector on board and both the 16 MHz and 32.768 kHz crystals. My plan is to have castellations on the board edges and through holes available inboard for 1mm spaced pins. This would give the option to reflow the board to a daughter board or to hand solder pins for those that would rather hand solder. The programmer would be on the daughterboard along with any other sensors, connectors, or peripherals.

Attached is a preliminary schematic. A few things to note:
1. The circuit can be powered by either the onboard micro USB or an external input (VREG_IN1) - the MK66FX1M0 has a built in regulator up to 5.5V.
2. VBAT can be supplied externally, otherwise power is from VDD.
3. Right now, I'm at 29 connections per board side, which should fit...

Any feedback in terms of the design approach or schematic would be greatly appreciated!

Thanks,
Brian

[kjn]

All, I'm designing a Teensy 3.6 "Pro" to enable access to all of the Teensy 3.6 pins and provide better packaging for placement directly in a final product. This is a project that I'm doing, but I'd like to make it available for sale through my startup as well. The goal is to have about a 30 x 30 mm board with the MK66FX1M0. I'm hoping that I can fit a micro USB connector on board and both the 16 MHz and 32.768 kHz crystals. My plan is to have castellations on the board edges and through holes available inboard for 1mm spaced pins. This would give the option to reflow the board to a daughter board or to hand solder pins for those that would rather hand solder. The programmer would be on the daughterboard along with any other sensors, connectors, or peripherals.

Attached is a preliminary schematic. A few things to note:
1. The circuit can be powered by either the onboard micro USB or an external input (VREG_IN1) - the MK66FX1M0 has a built in regulator up to 5.5V.
2. VBAT can be supplied externally, otherwise power is from VDD.
3. Right now, I'm at 29 connections per board side, which should fit...

Any feedback in terms of the design approach or schematic would be greatly appreciated!

Thanks,
Brian

I'm keen as - would it include an MK02 or MK04 onboard for programming? What is the timeline for prototyping/production?

[defragster]

text posted : 'The programmer would be on the daughterboard' is saying the programming chip will be external.

[brtaylor]

text posted : 'The programmer would be on the daughterboard' is saying the programming chip will be external.

Correct, I'm not planning on including the programmer on board, I doubt it would fit without moving to BGA. The MKL04 is pretty wide pitched and should be possible to hand solder or reflow onto your carrier board.

Best guess on timeframe is early to mid December before being generally available.

[brtaylor]

Updated schematic attached. I changed the inboard through holes to 1.27 mm spacing, 2 rows with 15 positions per row. This is a more common and cheaper header to purchase; I added a ground to the extra pin on each. I also added the MKL04 programmer - I'm going to try to fit it on the bottom of the board as an option. Would of course negate use of the edge castellations, but would still work fine with the through hole headers, plus it might make integration into the final product easier.

Brian

[Frank B]

Maybe I'd buy such a board with 2.54mm pin spacing. To tinker with. With bootloader. But not with 1.27mm.
But I think I don't belong in your target group.