Is there a market for a Teensy 3.1 48 pin ARM stamp?

Status
Not open for further replies.
I had mentioned prior the idea of having one board that was long but had a place one could cut to make smaller (if you put a little extra space and have the holes in a grid it should be really easy to cut). I have seen few other microcontrollers with this feature.

Perhaps a single board of 2.1" with 42 pins up from 28 - and then the 32 pin connector like on the gif post taking up .6" allowing pulling out 12 redundant (power/gnd/SPI/I2c/Serial1) pins and 20 novel new pins. But if the old end pins stay there - routing 40 lines though 6 pin gaps would be tough (similar for any pins run off the button end). Or the USB plug end could move and then new stuff could happen up there - but that could break some designs keyed off USB rather than board pins. That board might be stocked with/without the connector - but would always be smaller.
 
As we talk about boards with 100 or so pins, I wonder whether you get to a state where you just have too many pins for the hardware to manage (particularly for inputs). I would think it might be better long term to think of using multiple Teensies, each of which manages 10-30 sensors or outputs, and then they communicate to an aggregator Teensy (or a single board computer) that combines the output from each of the slaves.
 
Often, the application needs more I/O devices or needs to simplify such as using 2 SPI ports instead of trying to time-share one SPI port which can cause intractable issues like blocking an SPI device's service request interrupts too long.

Take the same ARM chip from, e.g., a 64 pin package to a 100 pin package - and the number of available on-chip peripherals increases, appealingly. Such as 2 or 3 SPI ports can be simultaneously used. For me, I'd vote for a large pin-count MCU package. Then decide which (of say 100) get used and on what PCB interconnect. Some primary PCB interconnects can be arranged as in T3.1 and secondary might be more on the interior, or not through-hole but rather an SMD connector for a cable or mating card.

Same, for 144 pin packages.

If a new Teensy PCB format is being contemplated, for more I/O peripherals .... perhaps we should abandon the time-worn 0.1 in. pitch PCB edge through-hole notion. think more about higher density PCB interconnect solutions, SMD connectors, etc.
 
Last edited:
As far as an SMD reference to posts #76 & #78 this one gives a view of size and $4 gets a Stackable feed through proto board with 1M and 1F SMD - dropping $1 gets a single bottom connector. Functional boards start at $14. Not user solder able? With ~20 new pins in 32 pin connector that could include SPI2 as well as old SPI1.

tx.png
Here's a minimal 2" 40 pin version - with updated bootloader but old MCU needing upgraded ;-) - I tried being visually proportional on the pins in mspaint, if not the connector may have to go longways unless a side pin were dropped on a 7 pin wide board format. Don't know the footprint of the new MCU.
 
Last edited:
With close to 2000 views on this topic, there is a great chance that, just maybe, there is a keen market interest for a Teensy 3.x 48 pin ARM stamp or clone.
:)

BTW ... For now, Talldog's Tindie 48 pin expanded GPIO solution works just fine.
:cool:
 
Last edited:
standardized expansion connector with LOTS of signals... to me, that's the good part - more so than going up to 48 pins on the base board.
 
I think stevech and I agree - at least in part - I'm not excited by a clownishly large teensy - although an excessively fast and overpowered MCU would be desired, actually it is a LAW (Moore's)- having a standard expansion to draw out more I/O as needed - or to mate and connect in a modular (non-destructive) way - as indicated in my post #79 drawing.

Offered my input so PJRC can see options - adding the connector would generally add $1 to each unit - stocking an identical version without that would mean two stocked parts - but from a single new design. At this late date re-working a T_3.1 (esp. given working/evolved options) with a new footprint would be a distraction of effort better put into whatever the new T_3.1 will be.
 
1st priority: standard expansion connector - SMD. Lots of pins. Easy to buy.
2nd priority: MCU with higher MCU package pin count so that there are more on-board peripherals: SPIs, I2C, SDIO port, etc., mostly all simultaneously available. 512K-1MB flash is fine. 128K or more RAM.
retain current board pin count and size and even same signals. Higher pin count MCU's new peripherals can use only the expansion connector.

But despite title of this thread... for compatibility, use expansion connector, don't change to 48 pin board.
 
Last edited:
I don't care if you bury the GPIO in 4 layers or 6 layers, you will probably still need a larger Teensy footprint in any case when you re-surface the GPIO on the top layers.
:cool:
 
That's where the dense SMD connector would help expose the pins and add modularity. I tried to find the k26 - it is 20mmx20mm in pin style and 144 pins [100 GPIO] - the BG variant can be 13mmx13mm or under - the T_3.1 is about 13mm? So the board goes wider for a this thing in non-BG- tough to maintain the foot print adding 7mm.
k26.png
 
Paul said, the chip (package) is larger, it will not fit on a board of the same size. Increasing the layers will not help to retain the size. It does not fit.

A expensive high-density connector means that in the most cases you'll need a interface-board-> same situation as before. What is the benefit ?
 
Smaller primary board user of ~40-44 pins may never need pins - just RAM/FLASH and MHz - still TEENSY'ish. Without BG MCU - it will be 3 pins wider. Even going to 48 pins orphans a lot of signals a SMD connector could expose in less space.

Anyone needing more pins: a pluggable interface board just got 32(?) pins of whatever that holds - some mix of old&new.
> Dual SPI, Serial1 to an ESP8266, SDIO FLASH

Modularity as noted - the linked item - you can build one off boards that make your project done - or can stack boards.

For instance: new MCU may not be 5v tolerant so a higher mA daughter board of 5v tolerant/capable IO for LED's could be critical to it being as useful as the T_3.1 was. But it would run at 180MHz with FPU, 2MB flash, 256KB RAM.
 
Yes the new chip is awesome. It allows to store some minutes of aac or mp3 on board... maybe 2 or 3 mp3/aac simultanous playing.. flac with std. blocksize..
Perhaps aac-encoding in realtime.. (or faster!)
I cant wait for it :)
Mabe it can be overclocked to more than 200MHz

Does anybody have detailed info about dolby digital or dts ? How many mips are needed to encode/decode it ?
Is there any sourcecode available?
 
Last edited:
So many possible uses like that would be happy with any expansion to the T_3.1 footprint of pins - given the better MCU. Would be a shame to not consider the best way to preserve the Teensy Part of it - and expand on what the T_3.1 started but traded in 'general connectivity'

$1 a connector isn't a good thought - but the minimal linked item was $3 for a single connector proto board. Compares well to a $5-$10 T_3.1 breakout.

If the T_3pp put the fatter MCU (a 20mm/8 pin extension) on the new end carved out the inner pin footprint of the T_3.1 then a socketed solution to T_3.1 simulator would be there in a 44pin unit 10 pins wide - though that could cover the SMD usage. Of course BG MCU fits better.
 
BGA is too expensive i think. Needs more precision, more signal-layers.
I read that, if it is not mounted properly it tends to fail after month/years, so it's required to check the soldering with x-rays for each new batch.
Do you remember the problems with the x-box ? I owned a laptop ... same problem..
 
Yeah that is a huge pile of pins in a small space with most buried inside, even with good place and connect. So 10 pins wide
 
Yeah, a BGA would be needed and it would have a pitch that's hard to route.
We did a small BGA chip recently -- in a low cost oven. Worked fine. But the pin count was comparatively low.

The breadboard-inspired 0.1 in. pin pitch boards... I think are ready to be retired. But that's more than may make sense for PJRC at this point.
 
While there are a number of advanced users who need ever more pins, spi/i2c/uart/i2s/can busses, I suspect a lot of users are like me, and like the Teensy because it is easily used on a breadboard or perfboard. What little soldering that we do is all through hole soldering (and some of us have no soldering skills). If the next Teensy needs SMT soldering, it will leave a lot of us out.

Similarly, in terms of size, the current DIP size is just about right for me. It fits on those mini-breadboards for just the chip, and on a normal 1/4 breadboard with 30 rows, you have enough for some components on your build.

I don't have any solutions that would work with those that need bigger numbers of pins and more busses that would also allow hobby microcontroller programmers to still get their simple projects done easily and fairly cheaply.
 
The T3.1's single SPI port, and no SDIO ... considerations for this goal we're discussing. While many would like to avoid the big 48 pins @ 0.1 spacing, versus a high density SMD connector for added I/O.
Perhaps this is all moot, as an MCU package with 100 pins or so is needed. Keeping that "Teensy" is the challenge, else we have yet another dev board like WaveShare.com sells.
 
Tough choices - I can maybe see why PJRC doing debug was the easier/better choice. Won't break anything (i.e. a new footprint) and might add life to T_3.1 as it will allow going down the razors edge maximizing I/O & CPU utilization to stay HYPER effective. [given how many use AVR Arduinos] Even if that only helps the gurus here - that helps everyone who can benefit from their better work. Today I saw a blog about Teensy Debug - hacking off the MiniT bootloader chip and using standard tools - it was suggested that wasn't needed if you froze/reset out the MiniT? - but still needed soldered wire to pins to pull out the extra signals. Maybe PJRC is refining a clean way to do that without bricking the MiniT and the value that adds.

I vote for the better/bigger MCU on a modular frame as Teensy as is reasonable - or at least a new thread for the T3.1pp

Just for scale here is a $7 k26 MPU versus the current <$4 k20 - if I found the right links:
txpp.png
mspaint duplication is ugly but should be near scale where 10 pins high leaves 8 in the middle for the 20mm MCU with about 36 pins/side and a 32 pin SMD - destroyed the T_3.1 footprint
 
Last edited:
the BG variant can be 13mmx13mm or under - the T_3.1 is about 13mm?

Yes, I've had my eye on this 13mm size.

It's a very tight fit. It will not allow soldering header or sockets on the top side of the PCB, at least not with their plastic part flush to the PCB surface. But other than that caveat, and the incredible difficulty of PCB routing, it might fit. Maybe?

The smaller sizes are 0.65 mm ball pitch, which requires insanely expensive (at Teensy's modest volume) PCB processes, mainly via-in-pad, but also smaller than 5 mil traces and other high end specs.


Yes the new chip is awesome.

Yes, indeed it's awesome. The flash is also in 4 banks, which might really help applications wanting to write to flash.


The breadboard-inspired 0.1 in. pin pitch boards... I think are ready to be retired.

No, 0.1 inch pins are here to stay.

While there are a number of advanced users who need ever more pins, spi/i2c/uart/i2s/can busses, I suspect a lot of users are like me, and like the Teensy because it is easily used on a breadboard or perfboard.

Yes, the vast majority prefer breadboard compatibility with 0.1 inch pins.

If you look at the number one complaint about regular Arduino (well, ignoring the retail prices) it's the 0.16 inch spacing between headers!


Keeping that "Teensy" is the challenge, else we have yet another dev board like WaveShare.com sells.

Yes, it is a tremendous challenge. I'm considering a lot of different options....


Tough choices - I can maybe see why PJRC doing debug was the easier/better choice.

I do intend to do both. Currently debug is considered a higher priority. Believe me, that's not necessarily because it's easier.


When a ++ model happens, it'll almost certainly use one of the BGA packages. I'm not going to make a huge board like this with the giant TQPF chip.
 
Just my single customer data point, I like the current Teensy size. Fundamentally I run out of pins about where I run out of programing skill and use another Teensy. For those things where I genuinely need more horsepower and am not worried about space I'd reach for a Pi or some other device that you natively interact with via a keyboard and screen. For me the critical point for a Teensy class device is small size to fit into fun things, low power options and low enough cost I can afford to learn by screwing up. The pin flexibility and current capacity to breadboard without wires by dropping modules directly aligned with the Teensy also helps (they then get soldered directly to it onehorse style).

While I love the idea of stackable modules as done by Paul that moves you into the territory of Edison and others in the 'SMT module' class and that's a crowded place to be even if none of them have got it right yet.

Do wish the default headers spaceing was just a bit tighter than 0.1, but that's something that'd take a time machine to fix.
 
Debug .. why not have a switch or jumper or software controlled MUX to change the two SWD pins during debugging. That way we can use the great off the shelf SWD debugging tools and software, and PJRC doesn't have to develop software?
With larger pin count MCUs, one can justify setting aside the two SWD pins (not having to sacrifice I/O peripherals to do so, as is the case with, say, 64 pin MCUs).
 
Last edited:
Status
Not open for further replies.
Back
Top