open-source teensy-compatible - what features do you want?

[ptorrone]

limor and i can’t purchase teensys from sparkfun anymore, they now decide who can buy, using a secret set of rules. since sparkfun is now the exclusive maker and seller of teensy, and paul and robin apparently have no say, limor and i were informed we’re not allowed to purchase.

instead of being sad about it, we’re doing the obvious thing. we’re making an open-source, teensy-compatible board.

we also just had our second kid a few weeks ago, so the timing is not great. that’s fine. last night, after a cluster breastfeeding session (parents know exactly what that means), limor and i started this project. “freensy” is the working name. it is not the final name. maybe.

features​
open source. anyone can purchase it and or manufacture it
lower cost: $9.95

core
rp2350 32-bit cortex m33 dual core
150 mhz at 3.3v logic and power
overclocking:
300 mhz often achievable with a small voltage bump, vreg to ~1.20v or 1.25v
350 mhz and up requires max safe internal voltage around 1.30v to 1.35v plus some silicon lottery luck
memory and storage
520 kb ram + 8 kb otp memory
8 mb spi flash for files and circuitpython or micropython storage
no eeprom
appears as a usb mass-storage device on your computer

usb and power
usb host on any two pins
modern usb-c device
built-in lipo charging with optional jst connector footprint (or use the vbat pin)
600ma peak 3.3v regulator
4 layer board with inner ground and 3v planes

io and debug
stemma qt i2c port for hundreds of sensors, oleds, breakouts, etc
swd debug footprint on the bottom

indicators
onboard neopixel
red status led

software
open source, non-proprietary bootloader
supports arduino, micropython, circuitpython, pico sdk, go, rust

mechanical
color: red
optional enclosure
full enclosure design published as free open source so anyone can make or modify it

things we’re thinking about​
should we add an optional footprint for 2 mb or 8 mb psram?
also, we’ll be doing an esp32 p4 version.

nothing says “new baby energy” like starting a new board at 3am and making sure no single company gets to decide who is allowed to build.

[Fluxanode]

I'd say
1. a faster processor speed without overclocking say 600 MHz to 1 gig.
2. 16 bit A-D's - minimum 4
3. SD card for storage.
4. 6 hardware serial ports.
5. match the amount of IO pins as the teensy 4.1
6. industrial rated version.
7. SPI ports minimum of 2.
8. LCD direct connection port and support for Adafruit LCDs.
9. match Teensy 4.1 form and functionality.
10. Ethernet port like the teensy 4.x

I like the 4.1 and its features but for a better price...

[SiliconWizard]

Well, which Teensy are you trying to re-implement?

You picked a RP2350 which is no match for a Teensy 4.x. But probably for the 3.x series.

[aeg]

Did you already ask Paul about becoming a second source for Teensy?

If he says no, just reverse engineer the Gerbers, make a verbatim copy, and call it something other than Teensy. PCB artwork is not eligible for copyright protection in the United States.

RP2350 is not form fit and function compatible with RT1062. What you've dreamed up is yet another Pico 2 variant. Why would anyone buy that from you instead of from Waveshare or Seeed for half the price?

[westfw]

Oh come on!

• The last thing the Open Source Hardware world needs is a "battle" between Adafruit and Sparkfun (and Arduino, I guess.)  Y'all each have enough unique products, and enough similar-but-not-quite-the-same products to keep us all happy, and it ought to be enough to keep you happy too.
• Proposing an rp2350-based board as "teensy-compatible" is ridiculous.  @Fluxanode pointed out most of the issues.

Reconsider after you've gotten a good nights sleep.  (parents know what THAT means, too.)
(and congratulations! - my second grandchild must have been born about the same time, so my daughter and SIL are providing reminders of THOSE issues!)

[bingo600]

Wrt:
Teensy 4.1 compatble , i totally agree with the above users.
I would chose the Teensy 4.1 , for the "oomph" (and the ethernet) - If needed.
A (current) pico wouldn't be able to replace it.


Wrt: A new RP2350 board i'd say :
Make sure it's an A4 stepping RP2350 (The one w. the silicon bugs fixed)
Aren't many of those A4's on Ali/ebay with the extras you'd add.

That said :
A $10 price would prob. end up as $20 i Europe.
I'm not even sure i'd buy one.

[nimish]

The only real competition would be to make something using, say, an STM32H7 or RA8 series high-end microcontroller. But they are very expensive. Perhaps the ESP32-P4, but that has a really terrible ADC. HPM6750? Not available in English. CH32H417 would be interesting with USB 3.0 SuperSpeed.

Teensy has a closed source bootloader and locked down debugging which is annoying.

[westfw]

Teensy has a closed source bootloader

Why?  Doesn't the i.MXRT1060 chip have a built-in bootloader capable of loading over USB?  (I guess it can only load to RAM, and then you'd need a secondary process to write QSPI flash, but still...)

[bingo600]

Teensy has a closed source bootloader

Why?  Doesn't the i.MXRT1060 chip have a built-in bootloader capable of loading over USB?  (I guess it can only load to RAM, and then you'd need a secondary process to write QSPI flash, but still...)

I don't have much experience w. the Teensy's ..
But i'm quite sure the the Teensy 4.1 has a closed source "Co processor mcu". It's  handling power sequencing of the iMX , fw loading , and maybe debug.
Kind'a like the debug chip on "demo boards", i just don't think you can easily bypass it.
Should make "fw bricking" the main cpu, almost impossible.

That's prob. why they're not cloned on Ali/ebay.
Wo. the loader chip, you might have power on issues , and can't use the std. Teensy fw loading programs (I think the loaders are closed source).

[newbrain]

and maybe debug

Unfortunately, no debug.
The loader chip uses JTAG (not an issue in itself), but its source is not available and it does not support debugging...

In addition to that, the OTP configuration of the 1062 in the Teensy disables Debug Monitor exception, making writing a generic GDB stub more complicated that it needs to be.

It's still possible to create a working one for code in (ITCM) RAM, with support for breakpoints and threading (e.g. FreeRTOS).
I had written a 99% functional one (it would still gets confused if one tried to single step a non-active thread), using a number of SVC instructions to implement soft breakpoints and single stepping.

It was tailored to integrate with the specific needs of my project, but it's still quite general - about 3000 LOC (less with no RTOS support), IIRC.

As for the original proposal, I agree that would be YAP2C, yet another Pico2 clone and not a Teensy, slower at default speeds (IME all Pico2 reach 300 MHz with no problems, but the Teensy can usually be overclocked to about 900 MHz so  ), not compatible with the Teensy libraries and tooling etc. etc.

From the OP:

usb host on any two pins

This sounds odd, how?

[SiliconWizard]

Yes, overclocking the RP2350 and cleverly using the two cores (which is definitely not trivial and likely beyond the skills of the average user of this kind of boards) may get you the same kind of "processing performance" as a Cortex-M7 @ 600 MHz, but the RT1062 still has way more peripherals, including Ethernet and USB *HS*.

[upbeat]

Just a nuance on the request for Ethernet is for it to be PoE - this would be an additional capability over the Teensy 4.x, afaik.  Also, request I2S support. 

[SiliconWizard]

Ethernet with a RP2350 could only be reasonably implemented using an interface chip such as the W5500.

As to I2S, though, it can be easily implemented using the PIO and there are numerous examples online for that. Actually, for I2S, the RP2350 PIO is pretty nice as it allows implementing I2S with pretty much any number of channels, while many I2S peripherals in MCUs are restricted to two (stereo, a single data line).

But regarding the possibility of them using something else than the RP2350, we don't know.
Currently, the RT1062 seems no stock at Digikey (and AFAIK, Adafruit use Digikey) and is > $7 per 1000. That's not cheapo.

[nctnico]

Yes, overclocking the RP2350 and cleverly using the two cores (which is definitely not trivial and likely beyond the skills of the average user of this kind of boards) may get you the same kind of "processing performance" as a Cortex-M7 @ 600 MHz, but the RT1062 still has way more peripherals, including Ethernet and USB *HS*.

I agree. Especially with some less than optimal libraries and running Micropython you will want lots of processing power and memory. Sticking with the RT1062 or similar is probably the best way to stay clear from yet-another-Arduino.

[EEVblog]

If he says no, just reverse engineer the Gerbers, make a verbatim copy, and call it something other than Teensy. PCB artwork is not eligible for copyright protection in the United States.

Doesn't matter when that's not considered a cosher thing to do, especially in this industry. The layout is borderline trivial to re-do.

[zapta]

But i'm quite sure the the Teensy 4.1 has a closed source "Co processor mcu".

It's the licensing chip and they sell it also separately.

[EEVblog]

I created a new thread not knowing this one exited from Phil.
Use this new thread of mine here for the drama aspect:
https://www.eevblog.com/forum/chat/sparkfun-splits-with-adafruit/
and keep this thread for technical discussion of the clone.

[ebastler]

we’re making an open-source, teensy-compatible board.

When did "has the same form factor" and "rhymes" become an acceptable definition of "compatible" for embedded computers? 

[nimish]

Teensy has a closed source bootloader

Why?  Doesn't the i.MXRT1060 chip have a built-in bootloader capable of loading over USB?  (I guess it can only load to RAM, and then you'd need a secondary process to write QSPI flash, but still...)

Last I checked part of the whole Teensy offering was a custom bootloader and chip from PJRC. That's what makes it a Teensy and not a generic dev board. Personally I don't think it's necessary these days.

IMO, to compete with a teensy 4 a better microcontroller like STM32H7R, STM32N6 or something else would be the option. Not another rp2350 board, why wouldn't I just use a pico 2?

[tellurium]

Sticking with the RT1062 or similar is probably the best way to stay clear from yet-another-Arduino.

If I were to make Teensy, I'd go for RT1170, 1GHz M7 + 400MHz M4, 2M SRAM, etc etc.
Not the rp2350 which sounds more like a pico2 variant.

[profdc9]

I think there's something to be said about using a microcontroller that is more community supported like the Pico 2 rather than the latest offering of NXP, STM, etc.  Having the maximum number of peripherals or the top clock speed isn't always important for many projects, but wide availability now and into the future is very important for many projects.  This is especially true of open hardware projects where the device may be constructed at an unknown time in the future by someone with generally unknown and limited resources, rather than being a product where the number of parts guaranteed only needs to be as many to manufacture the next run of a device.

In particular, if a vendor can decide they will simply stop supplying the product to someone else, they can do the same to you.  I am not going to take sides in the Sparkfun vs. Adafruit row, but if your product depended on Teensy, and now you can't get it, you're in trouble.  This situation is exactly how proprietary licenses can be weaponized, and doing so should give anyone pause in their future dealings with Sparkfun.

That said, a device for prototyping and as a basis for open hardware projects needs to lower the threshold to becoming familiar and using the device.  That is something that Adafruit excels at through its libraries and tutorials.  Adafruit will have to do much work to ensure that there is the variety of options and peripherals that the Teensy provides.  Starting with the Pico ecosystem gives them a leg up in making that diversity possible, and if I think if such options existed now they would be significant competition to the Teensy.

[SiliconWizard]

Thinking about it some more, we may be looking at it from the wrong angle.

We have reacted as potential *users* would. And clearly, a RP3250-based board would be no match for a Teensy 4.x. But people who want to get a Teensy, with its processing power, peripherals and support libraries can still get it from Sparkfun. It's not like they have disappeared.

From Adafruit's perspective, they want to replace a product in their catalog with another product they can sell. And I don't doubt they have enough business experience by now to at least be able to figure out what will sell or won't. My external perspective on this is that the average Adafruit customer may not actually need the processing power of a Teensy 4 and would be perfectly content as long as Adafruit provides sufficient libraries for the new board. OTOH, if they released a Teensy clone, they'd be dependent on all existing libraries for the Teensy and they may not want to have to bother contributing to these libraries (knowing Adafruit is quite active on the software side of things, contrary, AFAIK, to Sparkfun), given how things went both with Sparkfun, and indirectly with PJRC.

Contributing on the software side for the Teensy, which Adafruit would inevitably do as that's what they do for pretty much all their products, would likely be very uncomfortable given the situation. Because, while the hardware is not open, the software is (except maybe the bootloader, that they would likely replace with their own?). At least, that's the kind of mess I would not like to get into.

So, just a few thoughts. But OTOH, the reactions here were all natural, as they specifically asked what features people want. Maybe the thread would have been more fruitful to them if they had asked "what features do you want, given that our constraints are this, this and that, for this and that reason."

Edit: to be fair to them, regarding my last paragraph, they did at least give one constraint: $9.95 max (which I assume would be the retail price), which is pretty aggressive. There's no way you can sell a product at that price using a RT1062 or similar. So I guess that closes many doors and the only thing that remains questionable is calling that "teensy-compatible". (That might be possible using some chinese MCU, but as an american company that has most of its partnerships with other american companies, they probably don't want to take that risk.)

[Sacodepatatas]

Starting with the Pico ecosystem gives them a leg up in making that diversity possible, and if I think if such options existed now they would be significant competition to the Teensy.

And why not looking for a collab with Raspberry Pi Ltd in which a new Pico 3, more capable than a RT10XX, would be released on purpose of competing with Teensy specs?

[ejeffrey]

I think there's something to be said about using a microcontroller that is more community supported like the Pico 2 rather than the latest offering of NXP, STM, etc.  Having the maximum number of peripherals or the top clock speed isn't always important for many projects,
<Snip>
 if your product depended on Teensy, and now you can't get it, you're in trouble. 

But the point is that if you depended on a teensy, a product that rhymes but can't do the same job isn't a substitute.

There are no shortage of inexpensive microcontroller boards in different form factors.  One of the defining features of the teensy line has been high performance.

[profdc9]

There are no shortage of inexpensive microcontroller boards in different form factors.  One of the defining features of the teensy line has been high performance.

There's another major factor though: shortcutting the effort of getting up and running on a complex microcontroller, including all of the libraries.  One could just use the STM32 and CUBE IDE to do development, but the learning curve is quite high here.  Teensy provides hardware libraries and tons of other useful libraries and it has been designed to get one up and running quickly with the Arduino environment or other simplified development tools.  That is one of its main selling points.  If performance is completely not an issue, then using the standard ATMEGA328P is useful because of its ubiquity and because the Arduino environment comes with completely open source libraries, and I do use it still for many projects because it is adequate.  The Pico 2, while not having the performance of the Teensy, already has an open source ecosystem around it that could still use simplification to be more like the ATMEGA328P-based Arduino which is about the simplest thing you can set up as microcontrollers go.  Adafruit is in a good position to do this.  Given Qualcomm's acquisition of Arduino, this is more than just an academic exercise; this is something that holds Qualcomm's feet to the fire because there would be an easily available alternative in case Qualcomm tries to "monetize" Arduino by, for example, creating proprietary only extensions of Arduino or trying to shove AI down our throats.

Of course, there's nothing stopping Adafruit from using the most advanced and fastest processor available now, but that could have repercussions when such parts are discontinued.  The ubiquity of the RP2040/RP2350 chips makes their continued availability much more likely as well.