PDA

View Full Version : Teensy 3.1 - 2mm edition



scswift
01-06-2014, 01:12 PM
I know you're probably working on a Teensy 3.1++ which breaks out all the pins on a longer board, but what I'd really like is one which breaks out most of the pins on a board that is the same size using 2mm pin headers.

A 2mm teensy would ideally be just a smidge longer and a smidge wider, with 18 pins along each side, and 7 pins between them on the end. This would break out all 10 digital pins which are on the pads underneath.

2mm pin headers, in addition to breaking out more pins along the edge, would also greatly reduce the stack height if you seat the board on top of another. I've been looking at ways to break those pins out on the underside of the board with a board of my own design that sits on top of the audio board and with the 2.54mm pin headers the stack ends up being over 15mm tall. And that's something I can no longer call Teensy. Or fit in my enclosures.

Constantin
01-06-2014, 03:26 PM
I believe Paul designed the boards to use standard (for the US) 2.54mm-spaced headers in order to facilitate bread-boarding. Hence also the form factor, i.e. looking like a long gumstick.

You may want to consider rolling your own Teensy solution if you are very space-constrained - Paul has been kind enough to make three different-sized Teensy 3 bootloader chips available, making it relatively simple to roll your own solution. Between the bootloader chip, the schematic, a copy of Eagle, and a fab like Laen's, and a stencil from Pololu, you should be good to go.

I've gone that route because of space constraints myself. A re-purposed toaster oven (with a big heat-sinked SCR and a Arduino + reflow shield) solders it all together. If space is a consideration, I'd consider going to a square form factor with a board whose pins are on the edge. Hopefully, you'll be able to do it in two layers and single-sided SMD mounts, because 4-layer boards are not only more money, but they usually take longer to assemble if you use a service like Laen's.

I haven't attempted dual sided SMD work, but it could be interesting to mount the Bootloader on one side and Teensy chip on the other. The largest bootloader chip may not even require a reflow operation / stencil if you feel confident enough to drag solder a connection yourself.

PaulStoffregen
01-06-2014, 05:17 PM
Hopefully, you'll be able to do it in two layers and single-sided SMD mounts, because 4-layer boards are not only more money, but they usually take longer to assemble if you use a service like Laen's.


I'll be very impressed if anyone manages to do it in just 2 layers. I couldn't. Teensy 3.0 & 3.1 are 4 layer PCBs.

stevech
01-06-2014, 06:13 PM
would square form factor need an adapter PCB so it can be used with the layout of solderless breadboards?

I began using WizNet ethernet modules when they had only 2mm pin spacing (prior to the 812MJ). Being a US person, this was a major PITA. Maybe because I didn't have 2mm spacing breadboards, perf boards, etc.

I suppose one could do 2 layers with enough sq. inches of board size! :p

PaulStoffregen
01-06-2014, 06:42 PM
Does anyone make solderless breadboards with 2mm spacing?

Constantin
01-06-2014, 06:49 PM
I'll be very impressed if anyone manages to do it in just 2 layers.

:D Spot on. For sure hard to do, especially if one tries to break out all analog pins the way you did, in a nice tidy row. Oh, and they're supposed to hold tolerances well enough to allow ~ 13 bit performance too!

If analog performance is a non-issue (either because of loose tolerances or no need at all), then it might be easier to get it done with a dual-sided PCB that has the relevant SMDs mounted on both sides. I've not even attempted it, dual-sided SMD reflow soldering seems like an invitation for trouble. It would be pretty funny to throw the chips and the relevant components on a dual-sided board in a layout program like Eagle, place them semi-coherently, and then see what the Eagle autorouter does (I like its performance so much that I never use it).

Freeroute.net used to offer a pretty good autorouter to help establish the efficiency of a particular layout. Pity that it didn't distinguish between analog and digital lines, power supplies, etc. making some of the layouts very efficient re: via reduction and fairly scary re: the paths that signals had to take in order to get from point A to B. Now the project seems to languish as the principal has found other amusing things to do and doesn't keep the code updated to work with Eagle, Java 7, etc.

PaulStoffregen
01-06-2014, 07:07 PM
I spent about 1 week to manually route it in 4 layers. Teensy3 was by far the most difficult PCB routing I've ever done. But much of the difficulty was due to having the 14 pads on the bottom side, which create a lot of obstacles for vias. The Arduino style pinout also made things much harder.

Edit: on Teensy 3.X, one of the 4 layers is dedicated to ground. The other 3 are routing. If you routed on all 4 layers and include only minimal traces for ground, more might be possible?

If you eliminate the 14 bottom-side pads and change the board's pinout so match the pinout on the MK20 chip, 2 layers might be feasible.

As far as 2mm pin spacing (or any alternate pin pitch) is concerned, I do not believe it's commercially viable for PJRC unless solderless breadboards and other prototyping supplies are readily available.

Markus_L811
01-07-2014, 11:28 AM
Does anyone make solderless breadboards with 2mm spacing?

I didn't see that even in germany, there are some underdog boards with 2mm spacing or even smaller. But breadboards no, not seen and btw. all is in 0.01" (2.54mm) spacing, so there is no need indeed.

christoph
01-07-2014, 12:13 PM
I see a huge market for 2mm to 2.54 mm adapters rising at the horizon! Maybe even with level shifters... ;)

scswift
01-09-2014, 01:17 PM
Why should it matter if a 2mm breadboard exists or not? We have the 2.54mm version to use for prototyping. The 2mm version would be for after you've finished your prototype and you want to embed the board into your final product, but a 2.54mm version which breaks out all the pins would take up too much space. Besides, making an adapter for breadboards would be a piece of cake, and cost next to nothing. You could even stick some of those super low profile 2mm sockets on it so you don't have to solder the board to it permanently and can just socket it into your final board when you're done.

As for commercial viability, obviously I can't vouch for that, but the XBee only comes in a 2mm pitch and that doesn't seem to have hampered its popularity. And I don't know how many of these you'd need to sell a year to consider it worthwhile, but I sell 150 kits a year that I would happily pop these things into.

It seems like you're focusing on the hobbyist market, where someone might just want to learn electronics, or build one of a particular thing. But is that your real market for these things? The Teensy seems like it's designed for someone who started with the Arduino but has grown beyond that and needs a lot more power, and in a very small space, at a reasonable price point. That sounds semi-professional to me, and I don't think someone using these in a commercial setting is gonna balk at adding a 2mm pin header to their host PCB.

I could be wrong, but that's what I'm planning to use these things for. I've already got a buyer lined up who wants 25 boards to power an art installation. That's why I grabbed a bunch of Teensys and audio modules when you first put them up for sale. (And that's why I was really hoping the 3.5mm jack on the audio board could be connected directly to an amplifier using the same power source.)

Donziboy2
01-09-2014, 02:22 PM
Why should it matter if a 2mm breadboard exists or not? We have the 2.54mm version to use for prototyping. The 2mm version would be for after you've finished your prototype and you want to embed the board into your final product, but a 2.54mm version which breaks out all the pins would take up too much space. Besides, making an adapter for breadboards would be a piece of cake, and cost next to nothing. You could even stick some of those super low profile 2mm sockets on it so you don't have to solder the board to it permanently and can just socket it into your final board when you're done.

As for commercial viability, obviously I can't vouch for that, but the XBee only comes in a 2mm pitch and that doesn't seem to have hampered its popularity. And I don't know how many of these you'd need to sell a year to consider it worthwhile, but I sell 150 kits a year that I would happily pop these things into.

It seems like you're focusing on the hobbyist market, where someone might just want to learn electronics, or build one of a particular thing. But is that your real market for these things? The Teensy seems like it's designed for someone who started with the Arduino but has grown beyond that and needs a lot more power, and in a very small space, at a reasonable price point. That sounds semi-professional to me, and I don't think someone using these in a commercial setting is gonna balk at adding a 2mm pin header to their host PCB.

I could be wrong, but that's what I'm planning to use these things for. I've already got a buyer lined up who wants 25 boards to power an art installation. That's why I grabbed a bunch of Teensys and audio modules when you first put them up for sale. (And that's why I was really hoping the 3.5mm jack on the audio board could be connected directly to an amplifier using the same power source.)

Why not just build the Teensy right into the board instead of having to soldier it in afterward?

Constantin
01-09-2014, 02:54 PM
Why not just build the Teensy right into the board instead of having to soldier it in afterward?

Exactly.

Paul has been kind enough to post a schematic of the Teensy 3.x boards he builds (http://www.pjrc.com/teensy/schematic.html) and the bootloader chip that is needed to make it all work can be bought in three different sizes (http://www.pjrc.com/store/ic_mini54_tqfp.html) as well. If someone feels a strong urge to build a board with a 1mm, 2mm, or 5mm pin pitch, they can if Paul doesn't have the desire/bandwidth/etc. to develop an alternative form factor. Nothing prevents them from reselling those solutions either - as long as the expectation is that they will provide the technical support if something goes wrong, not Paul.

However, there are risks of rolling your own solution. The benefit of something that can be breadboarded easily is that you can try out the circuit before you commit yourself to a $$$ solution by making a PCB and stencil. You may also discover hidden 'quirks' like the need to have 3.3V connected to the voltage regulator output in order to have USB connections or that D33/PTA4 cannot be pinned permanently low or the K20 chip will enter EZ-mode on startup instead of acting as expected, etc. Being able to plug the Teensy into a standard breadboard (here or in the EU) is a big plus for those developing new circuits.

As much documentation as Paul has provided, even more will likely emerge on this forum in terms of the do's and don'ts as folk explore these amazing chips. Please don't take this as a criticism, BTW, it's a simple reflection of the complexity that these chips have reached. The Freescale K20 hardware manual is over 1200 pages and much of it is geared towards EE's who have been doing embedded design for years. Some of the Atmel 328P literature could be considered hand-holding for beginners by comparison. I imagine that if Freescale actually tried to explain sample implementations of various hardware functions / software registers with the same level of detail that the K20 manual could be 5x bigger.

Last but not least, I wonder how much more successful the Xbee folk could/would have been with the hobbyist community if they had chosen a 2.54mm pitch from the beginning instead of standardizing around 2mm (which may make more sense for commercial customers - and that may be their target market). There is a reason that there are a bazillion Xbee-shields out there, just as there are a bazillion Arduino shields - hobbyists sighed and made adapters / shields to have access to the neat technology.

No argument, BTW, regarding the success of either product (Xbee or Arduino) despite the pin issues - the development environment, the forums / user community / etc. more than made up for the hardware quirks.

Markus_L811
01-09-2014, 04:40 PM
Exactly.

Paul has been kind enough to post a schematic of the Teensy 3.x boards he builds (http://www.pjrc.com/teensy/schematic.html) and the bootloader chip that is needed to make it all work can be bought in three different sizes (http://www.pjrc.com/store/ic_mini54_tqfp.html) as well. If someone feels a strong urge to build a board with a 1mm, 2mm, or 5mm pin pitch, they can and should instead of asking Paul to develop a form factor that is unlikely to serve his primary market. Nothing prevents them from reselling those solutions either - as long as the expectation is that they will provide the technical support if something goes wrong, not Paul.

However, there are risks of rolling your own solution. The benefit of something that can be breadboarded easily is that you can try out the circuit before you commit yourself to a $$$ solution by making a PCB and stencil. You may also discover hidden 'quirks' like the need to have 3.3V connected to the voltage regulator output in order to have USB connections or that D33/PTA4 cannot be pinned permanently low or the K20 chip will enter EZ-mode on startup instead of acting as expected, etc.

As much documentation as Paul has provided, even more will likely emerge on this forum in terms of the do's and don'ts as folk explore these amazing chips. Please don't take this as a criticism, BTW, it's a simple reflection of the complexity that these chips have reached. The Freescale K20 hardware manual is over 1200 pages and much of it is geared towards EE's who have been doing embedded design for years. Some of the Atmel 328P literature could be considered hand-holding for beginners by comparison. I imagine that if Freescale actually tried to explain sample implementations with the same level of detail that the K20 manual would be 5x bigger.

Yes btw. I would make an board with an K20 and an 54TAN seperate as an portconverter for porgramming from USB

scswift
01-09-2014, 05:23 PM
Integrating the Teensy into my own design would actually be more expensive than buying the boards for $17 a pop. I can't order a thousand of them at a time like Paul does.

And separating the Mini54tan from the boards to reduce the cost isn't an option as I want to retain the ability for my end users to upgrade their firmware.

Constantin
01-09-2014, 07:51 PM
Integrating the Teensy into my own design would actually be more expensive than buying the boards for $17 a pop.

You're right, the cost largely depends on the production quantity. I did a quick look-see and came up with the following budget for 10 boards:

The K20DX128 is less than $5 in single quantities, $4.5 ea. in quantities of 10, and $2.66 ea in quantities of 100 (http://www.digikey.com/product-detail/en/MK20DX128VLF5/MK20DX128VLF50-ND/3507511), for example.
The bootloader chip is $8. Paul may even offer a quantity discount for all I know.
Add $1 ea for for a 2x2" dual-layer PCB (may need to be a 4-layer or bigger, if so, add $2),
Add another $3 ea. for stencil, and
Add another $4 ea. for the other components (USB port, crystal, capacitors)

Some component prices (MCU in particular) come down quickly with quantity and fixed costs (like the stencil) drop proportional to shipments.

So even quantities of 10, you can likely assemble a Teensy 3.x for ~$21 ea. as long as can you do the placement work yourself on a 2x2" dual layer board. If a 4-layer board is needed, it is indeed more money at 10 units but likely less than Paul's price at 100 units. If a specific form factor, pin spacing, etc. is important to you, just do it. The time it takes you to solder 30+ teensy pins vs. placing SMD components + reflow solder is about the same. A buck or two in price difference shouldn't be the deciding factor.

scswift
01-12-2014, 01:34 PM
You're right, I hadn't considered how much soldering I'd need to do. I have no interest in hand assembling boards though.

That being the case, I'd like to make a new proposal: A 2mm Teensy with headers already in place, like the Xbee.

A 2mm Teensy with headers could be soldered directly to the board if one chose, or one could design their board to have sockets so once they have them assembled they could just pop the Teensy into place. A breadboard adapter with sockets would also be easy enough to produce, so you go straight from breadboard to the final product.

scswift
01-12-2014, 01:49 PM
Also, I'd like to mention that part of the reason I don't want to integrate the Teensy into my own design, besides the cost, is the complexity. It would take me a month to design a PCB for the Teensy. And Paul hasn't provided a BOM or even specified values for certain components like the ferrite beads. I don't even know what the function of the ferrite beads is, let alone how to determine what part I should use. I know they filter noise, but I don't know what the source of the noise they're supposed to be filtering is. Paul said in another thread that he thinks they're 600ohm at 100mhz. Maybe they're not intended to filter noise getting to the microcontrolelr, but to filter noise coming from it instead? I don't know.

I've only been designing circuits since 2008, all self-taught, and I still have a lot to learn. So by integrating the Teensy wholesale into my own design, I greatly reduce the chances of my screwing something up.


Well actually looking back at the post I see he does mention that if you don't need high resolution from the A/D, you could probably just eliminate the inductors completely. I'm not sure what constitutes "high resolution" though. If one were using it as a line in from an electret mic preamp, ie, at audio frequencies, would it matter then? No idea.

scswift
01-12-2014, 02:00 PM
Anyway I understand if Paul doesn't think it makes financial sense to offer such a board. I just wanted to put the idea out there. I think there would be a market for it.