Any Chance of a Teensy ++ 3.1?

Status
Not open for further replies.
Smaller - the 40 pin sized. Edison (70-pin?) connector or similar style:
tx.png

That was 40 pins with a 32 pin connector - and not shown the 144 pin K26 - you suggested the 13x13mm BGA might be the one
k26.png
 
The 144 (.5mm pitch) wouldn't work? I mean, i'm assuming it wouldn't work due to the sheer size of it, but .5mm pin pitch isn't bad at all. The current 3.1 chips are .5mm pin pitch.

Edit: the .65mm/.4mm BGA pitch would be a nightmare to route - vias-in-pad would be the only way I could think of doing it.
 
The 144 (.5mm pitch) wouldn't work?

It would be tough. Just the bare chip is wider than the current T3.x, and the widest you can go while still being able to fit in a standard breadboard (about 23.5mm) only gives 2mm from end of pins to end of board[*]

Screen Shot 2015-07-12 at 7.46.12 PM.png

[*] using the 2nd pins from the edge on each side -- you can go wider by using the outside-most pins but that means you'd have to pull the board off every time you wanted to add/change a jumper. I suppose you could require double-wide breadboards ....
 
I copied those p#227 images from thread Is-there-a-market-for-a-Teensy-3-1-48-pin-ARM-stamp ... that 32 pin connector is used on this device family: https://tiny-circuits.com/tiny-shield-proto-board-all-159.html - as noted they offer a $4 minimal proto board with two connectors installed.

Putting one of those on each side of the K26 might ease routing to the primary pins & the takeoffs - and give options for twin connects (bias one to Analog?) w/larger number of exported I/O from the 144 pins:
tx2.png- but they may never be used and would add to cost - and any project using them would need the mate. A ribbon board to board is $15.
 
Last edited:
I don't want to completely rule this out, but I'm not a fan of these high density connectors.

I know Intel Edison uses one. I know others have gone down this path, some more successfully than others have tried.

Some of my reservation is a concern this doesn't fit PJRC's business model. Teensy has never been about making a comprehensive collection of "shields", or whatever you'd call such a plug-in peripheral board. If I were to start down this path, odds are very likely the dev time that I currently put into creating more advanced software (OctoWS2811, Audio, PulsePosition, ILI9341_t3, SD_t3, SerialFlash, Biopotential, new USB types, FreqMeasureMulti, etc) would almost certainly be consumed by making and supporting lots of hardware. Many other companies in this "maker market" do this, and indeed it seems to be far more profitable than PJRC's approach. My decision making is really not about what will make the most money or conquer more of the market, but rather how to maximize (or at least sustain and continue to fund) my ongoing efforts to develop awesome stuff.

I'm also reluctant to create yet another plug-in form factor. Teensy is probably among the more successful "unofficial" Arduino compatible boards, at least in terms of actual usage in real projects. But so far, to the best of my knowledge, Teensy's pinout & form-factor has yet to be adopted by any other board. Even Arduino Zero clones that could have benefited from Onehorses's many "appallingly small" peripherals haven't been swayed to align their power and signals similarly.

So to even begin to consider this, I'd really want to be interchangeable with an already-established format. (hint, hint... to anyone who wants to persuade me....)

Cost is also a concern. I've always tried very hard to keep Teensy affordable. These connectors vary considerably in cost. Some of the cheaper ones may be designed for only a very low number of "cycles".

PCB real estate is also a possible issue. There are a lot of possible ways to use the space. One space-hogging idea I have in mind involves using the right hand side for the same SD socket we currently put onto the audio board and Wiz820 adaptor, with wiring to the native SDIO pins on this K26 chip. I also really want a convenient way for people to connect to the other USB port (the 480 Mbit/sec one), which probably will end up being a location to solder a 4 or 5 pin header that mates with the common USB-to-motherboard cables made for PCs. These, plus the large 13 mm chip and other stuff required all add up to not much PCB space left over.
 
Lots of Pros & Cons to see and weigh. A T3k26 would offer awesome compute power - new fast protocols in larger number. Though: too big with unused pins is as bad as not enough pins depending on the use case.

Teensy seems to be the DIY for DIY'ers - I wasn't suggesting a PJRC emulation of that 'Tiny' ecosystem - just the ability to get needed pins from a smaller format - while largely preserving the T3.1 layout. No matter what the layout is it will be new and need adapted to.

If you decided to design it with a mind to make a 52 pin T_monster version for bench use and design, then lop off the end for a 40 pin T_svelte version. Of course the width limits routing options and only gains 12 pins. In which case the T_svelte 40 pin with a 32 pin connector would beat that handily, and a sample daughter card could be designed with 5v tolerance built in - a perfect 16 pin onehorse project.
 
Another issue is cost. I'd wager that the BGA version would likely require a 6-layer board to maintain the skinny form factor yet extract all those signals.
 
Paul,

One thing that I would love to see on the following teensy is a mix of Castellated holes and Through holes for the pins. Making easy to install teensy permanently on other boards (without using the headers).

The Particle Photon has something that is very interesting.

Castellated holes.png

Vitor Henrique
 

Attachments

  • Castellated holes.png
    Castellated holes.png
    14.1 KB · Views: 689
I second that. I think it's a great idea and the cost should be close to zero as long as the fab house is OK with creating castellated holes.
 
Are there any ZIF sockets for the castellated holes? For those who want to install the Teensy NOT PERMANENTLY and without pins...
 
Last edited:
Castellated holes were discussed here:

https://forum.pjrc.com/threads/28847-Request-Castellated-Edges-on-Teensy-3-1-PCB

There are 2 issues.

#1 - Teensy is made with V-score for the horizontal edges. See the photo in reply #15 on that thread.

#2 - The intended path for SMT build involves buying the Mini54 chip. Castellated make a lot more sense when you're selling a RF module, where people need to avoid their own PCB design, and where the board is based on a special part which can't be purchased by ordinary people / companies.
 
Paul, I understand that you would need to find a new process for manufacture teensy, but I don't think I'm alone wanting to embed teensy directly on a final product without the header.

I'm a chemical engineer and a electronic hobbyist, I don't have enough time and effort to put on developing my own pcb for a custom teensy, I think there are more people like me here.

But more important, Teensy is amazing, and this is just an idea, keep in mind for future versions as it would only open more doors for this amazing board. I think it would be worth a talk with your pcb manufacturer :)

Vitor Henrique
 
Paul, I understand that you would need to find a new process for manufacture teensy, but I don't think I'm alone wanting to embed teensy directly on a final product without the header.

I'm a chemical engineer and a electronic hobbyist, I don't have enough time and effort to put on developing my own pcb for a custom teensy, I think there are more people like me here.

But more important, Teensy is amazing, and this is just an idea, keep in mind for future versions as it would only open more doors for this amazing board. I think it would be worth a talk with your pcb manufacturer :)

Vitor Henrique

+1

There are some microprocessor modules out there with Castellated edge connections.
Example
http://www.espruino.com/
http://www.espruino.com/Pico
 
Last edited:
If you look carefully at the espruino photos, you can see the 3 "mouse bite" locations, 1 in the center of the USB area, 2 on the far side.

Teensy isn't made this way. You can see the raw panels in the photo I posted.

Normally you never need to worry about how PCBs are actually fabricated and assembled on panels. You get the board after it's been broken out of the panel, and all secondard processing and testing and programming has been done. Likewise when you buy food at a grocery store, you have no need to concern yourself with how it was grown, harvested and processed.
 
@PaulStoffregen I'm guessing it's a big effort to switch the Teensy 3.1 to using routed instead of V-score edges, is the effort on the manufacturer's side, or your side because of panelized testing or programming?

For the new Teensy 3.1++, it should be possible to switch, right? There's no guarantee you'd start to get quantity orders customers wanting to do a their own custom PCB, but if it's not a ton of engineering effort, it's a good way to test the waters. You might also want to survey existing MINI54 purchasers, or people who have built a product prototyped with the Teensy to get their take on the idea.

From my perspective, I thought the price of the MINI54 chip in quantity 100 was too high. By the time you add up all the costs of the components on the Teensy 3.1 at qty 100 sourcing from PJRC, Digikey, and Mouser, you're only a couple bucks away from the qty 100 price of the Teensy 3.1. (based on old quotes from PJRC, your pricing may be different and I'm not going to post it here). Some of these parts will be scrapped or fail testing during manufacturing, where the yield on the Teensy 3.1 should be higher (could be wrong), so the average BOM cost will be higher. Given the choice between purchasing a Teensy 3.1 with castellated edges and the MINI54, I'd go with the Teensy 3.1.

It's significant engineering effort (as you know) to try to fit all the components into the small space of the Teensy 3.1. I would rather use a 2-layer custom PCB, and not have to go through that effort.

I would seriously consider putting castellated edges and suitable pads for extra signals on the underside of the Teensy 3.1++, and to pick a couple potential mid-volume customers to get feedback on the idea/design. If this feature turns out to be a winner, you could think about a Teensy 3.1 rev with castellated edges.
 
Last edited:
Over the last 4 years, we've tried several different panel designs. The one you see in that photo is relatively new, and I believe it's the first one we've ever used that is actually working very well. Getting to this point has taken a lot of work, many incremental improvements. This may sound easy, but let me tell you, it's quite difficult to accomplish as a small company (or even a big one) while dealing with the constant challenge of scheduling purchasing and production, and keeping up with same-day shipping, and customer service, and sometime even developing new stuff. Quality improvements are particularly challenging, because of the long lag times from initial design, through the purchasing and production cycle, and then more lag because customers waiting for products or unanswered email and forum messages are always more urgent.

With too much tab routing and too many slots, the panel loses rigidity and vibrates during high speed pick & place assembly. Obviously others are making their boards with hardly any mechanical connection from the board to its panel. Maybe they have finely turned assembly? Maybe they're experiencing lower first pass yield and enduring more rework (which is cheaper in China). From everything I've experienced over the last 10 years, I'm pretty sure the later is more likely. Hardly anyone airs such "dirty laundry" in public!

High first pass yield and low tolerance for rework are huge goals for PJRC. Much of that is driven by PJRC's purpose, which is to facilitate developing technology. Dealing with moderate fallout and quality issues is probably a net-win for many companies, if it substantially lowers cost. But I can tell you from years of painful experience, even moderate fallout becomes huge drain on development here at PJRC. It probably is for other too... but how many of the other companies in this market do you see developing so much software relative to their total revenue?

So I'm really reluctant to try a weaker panel design. Teensy3++ will be our first board with a BGA, which makes me doubly fearful of risking a less rigid panel... right after we've *finally* achieved a really good panel design that solves the long-standing problems.

I'm also skeptical castellated will improve sales. If you're feeling the Mini54 price is too much, how can you possibly think you'll enjoy the pricing for fully assembled Teensy boards?
 
Last edited:
As someone who has dabbled with developing my own Teensy 3.1 PCB for a project I'm working on, but here are my thoughts for a ++:

- If the 3.1 would have had all the pins broken out in a manner where they were easily usable, I wouldn't have even considered building my own (conversely, building my own has been considerably more difficult than I anticipated).
- For a ++ product, a 50% larger form factor is ok, as long as all the pins are broken out in a manner that makes them easy to use
- Please use a processor with built in floating point hardware - M4F or better
- A second SPI bus should be high on the priority list.
- Consider a decent size (128k or better) NVRAM built in. The Adesto CBRAM chips are pretty inexpensive and work well.
- Include the RTC crystal
- Multiple serial ports with FIFO buffer would be useful
- I could see a second DAC being useful

Hope that helps!
 
...while dealing with the constant challenge of scheduling purchasing and production, and keeping up with same-day shipping, and customer service, and sometime even developing new stuff. Quality improvements are particularly challenging, because of the long lag times from initial design, through the purchasing and production cycle, and then more lag because customers waiting for products or unanswered email and forum messages are always more urgent.
To grow and expand, one needs to give some of the hats that one wears to others!

http://www.smartva.co.uk/stuck-in-one-man-band-rut/
 
Last edited:
Status
Not open for further replies.
Back
Top