Teensy 4.0 released

[hamster_nz]

If you  haven't seen it, the Teensy 4.0 being released:

https://www.pjrc.com/teensy-4-0/

Gosh, it looks powerful:

Dual-issue superscaler ARM Cortex-M7 at 600 MHz, with branch prediciton and all the other good stuff.
1024K RAM / 2048K Flash
H/W float and doubles.

Runs at about 100mA, and costs $20.

[SiliconWizard]

Not bad. The MCU also supports double precision FP! (Edit: Missed that you mentioned it )

Those boards have very few IOs broken out though, so for an MCU that powerful, I think that would limit a lot what you would do with it. Still pretty nice stuff.

[rstofer]

Thanks for the tip!  I ordered a couple...

[ebclr]

What are the tools to develop on tennsy ?

[SiliconWizard]

Teensy boards can be used with the Arduino environment through an add-on: https://www.pjrc.com/teensy/td_download.html

You can also of course directly use the relevant NXP SDK instead.
See: https://www.nxp.com/products/processors-and-microcontrollers/arm-microcontrollers/i.mx-rt-series/i.mx-rt1060-crossover-processor-with-arm-cortex-m7-core:i.MX-RT1060?tab=Design_Tools_Tab

[firehopper]

Teensy boards can be used with the Arduino environment through an add-on: https://www.pjrc.com/teensy/td_download.html
You can also of course directly use the relevant NXP SDK instead.
See: https://www.nxp.com/products/processors-and-microcontrollers/arm-microcontrollers/i.mx-rt-series/i.mx-rt1060-crossover-processor-with-arm-cortex-m7-core:i.MX-RT1060?tab=Design_Tools_Tab

actually you have to use teensyduino 1.47 beta 5 to use the 4.0 for now till they make a non beta version
https://forum.pjrc.com/threads/54711-Teensy-4-0-First-Beta-Test?p=193715&viewfull=1#post193715

[KE5FX]

Pretty intriguing piece of hardware for $20.  What does it take to support 100 Mb Ethernet with these?  Any good RJ45 PHYs available with ready-made software support?

[Nominal Animal]

What does it take to support 100 Mb Ethernet with these?

The WIZ85io ought to work. (You'll need the $6 adaptor, and a WIZ850io module.)
You could also use the other 480 MBit/s USB port in host mode, and a suitable USB-Ethernet dongle, I guess.

I don't have a 4.0 yet, and I wasn't a beta tester either.  There are also some sharp corners left (having to use a beta version of the Arduino environment, linker script currently copies code so it's run from SRAM and not from Flash directly, and so on), but having been an original Teensy 3.0 Kickstarter backer, my experience says most of those will be ironed out pretty quickly.

[legacy]

if I ever made a debug cable with that Tensee v4.0 stuff for my ARM7TDMI@16.78MHz board, the debug cable based on ARM-Cortex-M7@600MHz would be more light years head than the machine under debug.

That's illogical!   

No doubt about! If the debugging cable is light years ahead than the machine where the same debugging cable is used, then it's logical that you trash it and build a new machine with the same chip used for the debugging cable (or even a better chip), why do you persist in supporting that old stuff?

Good question, no AI will ever able to answer. That's human's peculiarity! We are highly emotional individuals, usually arrogant and prideful to the point of blindness, so assured of fun (or victory, or whatever) that we make decisions based on pride and other emotions.

So ... let's make it, let's build it

[ehughes]

The RT family is quite powerful.     Execution from TCM is quite nice.    Execution from QSPI is very "Meh"

The 1050 Families and below had an GPIO unit that was dorked up.     The internal AHB connection to the GPIO unit was quite odd and was very slow. 

The price of the teensy looks nice.  I hate the fact that there isn't SWD access and it is crippled by by the Arduino system.

[Sal Ammoniac]

if I ever made a debug cable with that Tensee v4.0 stuff for my ARM7TDMI@16.78MHz board, the debug cable based on ARM-Cortex-M7@600MHz would be more light years head than the machine under debug.

That's illogical!   

The Wi-Fi module I use on my Cortex-M3 board (@72 MHz) has a 700 MHz ARM A9 on it!

[legacy]

Anyway, the Tensee-v4 looks "perfect" for building a graphic calculator! I have already reverse-engineered a couple of pocket calculators made by CASIO and Sharp, so ... maybe I will reuse their keyboard and LCD, replacing the motherboard with a Tensee-v4, which has a lot of horsepower for doing mathematical calculous and even more horsepower than the latest CLASSPAD 400.

Currently, I have made a kind of dumb remote terminal with a mathematical keyboard, which sends requests to an RPI running Mathematica and getting back the screenshot with the answer, sent via serial port at 115200bps.

Which is not a "fast" link, but from the user point of view, it looks like a super-fast calculator with a lot of horsepower under the hood, ... just ... wtf is that long cable that connects the pocket calculator to a ... RPI?!? Is there any trick?

Yes, there is a trick, people usually take 25 sec to get that there is a trick, and the true "trick" only works if you set the application screen on the RPI to a quarter of the VGA screen since the LCD is 320x240, with 2 bit of color.

[SiliconWizard]

I hate the fact that there isn't SWD access and it is crippled by by the Arduino system.

As I hinted above, you don't have to use the Arduino system at all. You can perfectly develop using NXP tools directly.

As for SWD: I'm not sure about the Teensy 4.0, but all other versions include an USB bootloader, so you can Flash them directly through USB with a dedicated app that takes a HEX or bin file. Sure you can't debug... but you can Flash it with just anything you like.

Anyway, yeah as I said there are a bit too few IOs broken out to be really useful. I actually made a custom keyboard (HID) out of a Teensy 3.2, and it was rather quick to do, the number of IOs was alright for this application, but barely. I didn't use the Arduino environment whatsoever.

If you're going to want to drive an LCD screen with it and still have enough IOs left for anything else, you're probably going to have to stick to SPI. Not necessarily ideal for an high-res LCD display, probably adequate for a lower res one.

So yeah don't expect to be able to do anything really fancy around this board (unless of course your project needs very few IOs), but as an evaluation platform for this MCU, it's fantastic and very cheap.

[brucehoult]

The price of the teensy looks nice.  I hate the fact that there isn't SWD access and it is crippled by by the Arduino system.

Huh?

Even AVR Arduinos aren't "crippled by the Arduino system" as you're free to use the IDE but not the libraries, or use neither and use avr-gcc and avrdude directly from the command line if you want.

Same here.

Arduino compatibility is a helpful feature for those who want it, not a jail.

[thm_w]

Huh?

Even AVR Arduinos aren't "crippled by the Arduino system" as you're free to use the IDE but not the libraries, or use neither and use avr-gcc and avrdude directly from the command line if you want.

Same here.

Arduino compatibility is a helpful feature for those who want it, not a jail.

No its absolutely not crippled by the arduino system, but if there is no debug its crippled by the idea that "We have a USB/serial bootloader, so its good to go" that applies to some Arduino products.

It looks like the relevant pins are: GPIO_AD_B0_06 to GPIO_AD_B0_10
I don't see a schematic posted anywhere, so some digging into the code is likely necessary to see if these pins are broken out and just not labeled.

[PCB.Wiz]

...
I don't see a schematic posted anywhere, so some digging into the code is likely necessary to see if these pins are broken out and just not labeled.

A  schematic would be great, but the web seems to only have schematics up to 3.6 ??
Maybe someone can post a link to a Teensy 4.0 Schematic ?

[BroMarduk]

Per Paul Stoffregen in the Teensy 4.0 Release thread (posted yesterday):

"The schematic is likely sometime next week. No schematic has been drawn yet, so it's going to take a little while."

[maginnovision]

Yea, he always takes some time to draw a schematic for the website. He doesn't use the board schematic. I think the 4.0 looks really good but I'm sort of disappointed to see the 3.6 form factor with uSD go. I thought it was really nicely done and very useful for a number of projects.

[Fire Doger]

if I ever made a debug cable with that Tensee v4.0 stuff for my ARM7TDMI@16.78MHz board, the debug cable based on ARM-Cortex-M7@600MHz would be more light years head than the machine under debug.

That's illogical!   

No doubt about! If the debugging cable is light years ahead than the machine where the same debugging cable is used, then it's logical that you trash it and build a new machine with the same chip used for the debugging cable (or even a better chip), why do you persist in supporting that old stuff?

Good question, no AI will ever able to answer. That's human's peculiarity! We are highly emotional individuals, usually arrogant and prideful to the point of blindness, so assured of fun (or victory, or whatever) that we make decisions based on pride and other emotions.

So ... let's make it, let's build it

Price
Package complexity
Power requirements
Minimum external components requirement
Availability
Life span
Already used family in other projecs

And algorithms much simpler than AI can choose an MCU based on required IO-features.

[ali_asadzadeh]

Do we have Gerber files for Teensy 4.0!

[luiHS]

 
Very interesting, I have already ordered two units, to try it with the new Teensyduino. We lack the scheme to know which ports are being used, mainly for some applications that require DMA with several contiguous port addresses.

Although it had been known for a long time, I was totally disappointed that it was not done with the format of size and quantity of ports available in Teensy 3.5 and 3.6. Such a small format, Teensy 3.2 style, for a processor as powerful as the RT series, is a series limitation. I find it incompressible, to make such a powerful product, but to limit it so much in terms of available ports.

[SiliconWizard]

Do we have Gerber files for Teensy 4.0!

Well, if the author is not going to release the original schematic but some hand-reprocessed version of it for publishing needs, I kind of doubt they will release the Gerbers or any EDA file directly. Just a guess though.

[SiliconWizard]

Although it had been known for a long time, I was totally disappointed that it was not done with the format of size and quantity of ports available in Teensy 3.5 and 3.6. Such a small format, Teensy 3.2 style, for a processor as powerful as the RT series, is a series limitation. I find it incompressible, to make such a powerful product, but to limit it so much in terms of available ports.

Yes, that was my point too.

OTOH, I can completely understand the objectives of the author. They aren't looking to compete with full-blown dev boards. There are thousands of them on the market already. They are releasing small and easy to work with boards at a very low cost.

As they did for older versions, they may release an ulterior version later on with a slightly larger form factor and more IOs broken out.

If you're looking for almost as cheap and a lot more flexible dev boards, you can go for the STM32 Nucleo boards. Pretty much all between $20 and $30. Pick a Nucleo board with an STM32H7 on it, relatively close in performance with this NXP chip (although not quite). Currently $27 I think.

[luiHS]

If you're looking for almost as cheap and a lot more flexible dev boards, you can go for the STM32 Nucleo boards. Pretty much all between $20 and $30. Pick a Nucleo board with an STM32H7 on it, relatively close in performance with this NXP chip (although not quite). Currently $27 I think.

STM32H7 microcontrollers are much more expensive than RT. I'm starting to use the RT1020 in LQFP144, they cost about $ 5, while the STM32H7, which are also less powerful, cost about $ 19 or more. I will not use the Teensy or any other premade board, to make my products. I only use those boards for testing.

The good thing about the Teensy environment, is the large number of libraries available, and a lot of sample source code, from which many developments can be made.

It's been a while since I ruled out using the H7, my bet is on the NXP RT series.

[SiliconWizard]

Didn't notice those NXP parts were this "cheap". That's impressive.

[mark03]

The STM32H7 also comes with a really long errata.  I'm not sure how the RT series compares in that regard, but we recently had to rule out using the H7 for a project after reading through the ST forum discussions where advanced users are trying to reverse-engineer their way to a working Ethernet peripheral

[JOEBOBSICLE]

Th

The STM32H7 also comes with a really long errata.  I'm not sure how the RT series compares in that regard, but we recently had to rule out using the H7 for a project after reading through the ST forum discussions where advanced users are trying to reverse-engineer their way to a working Ethernet peripheral

The H7 and most other new Cortex M7 based microcontrollers have DCacheand multiple memory buses, I.E AHB1 and AHB2 etc...  Which makes it a hella complicated chip, although very fast if you know what you're doing.

It's not that reverse-engineering is required, it's that you just have to know what you are doing regarding getting the MPU setup correctly or invalidating/cleaning DMA buffers before transmitting/receiving.
The cleaning/invalidating buffers doesn't work for transmitting unless all of your packets are coming from 32 BYTE ALIGNED BOUNDARIES



Which is unlikely in most TCP/IP stacks as you can't enforce the alignment of buffers they create... So you have to go with the reserving a custom pbuf pool approach and using the MPU.
See: http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.dui0646b/BIHIIJDC.html

Are microcontrollers even micro considering their complexity these days? When they run at 500MHz and have caches? I think they should be called mini controllers...

[donotdespisethesnake]

Do we have Gerber files for Teensy 4.0!

None of the Teensy boards are Open Source. That, and the proprietary bootloader chip, prevent any cheap clones so the creator can stay in business.

[PCB.Wiz]

That, and the proprietary bootloader chip, prevent any cheap clones so the creator can stay in business.

Which is the boot loader ?

Addit: Found it, they moved to a Freescale 3x3 package part, as 'smart memory', but that seems a more costly way to get loader memory, as other MCUs are cheaper (eg SiLabs EFM8 3x3's)
Still, if they have a good line to Freescale, there are benefits in single vendor.... (some boards show Nuvoton parts, larger package)

[maginnovision]

Do we have Gerber files for Teensy 4.0!

None of the Teensy boards are Open Source. That, and the proprietary bootloader chip, prevent any cheap clones so the creator can stay in business.

The bootloader is exclusively the reason they aren't cloned en-masse. He's typically released boards based on the beta versions through his osh park account. The only complicated part of the boards is the size and that's not much of a problem for clones. The 3.2 board you have been able to add to an oshpark order for years. He also sells the bootloader for people who want to make their own boards that are 100% compatible.

[langwadt]

Do we have Gerber files for Teensy 4.0!

None of the Teensy boards are Open Source. That, and the proprietary bootloader chip, prevent any cheap clones so the creator can stay in business.

it might slow down dumb verbatim clones but i doubt it is much of a road block to anyone else

[splin]

STM32H7 microcontrollers are much more expensive than RT. I'm starting to use the RT1020 in LQFP144, they cost about $ 5, while the STM32H7, which are also less powerful, cost about $ 19 or more.
It's been a while since I ruled out using the H7, my bet is on the NXP RT series.

I think you are referring to the older hi-end parts. The LQFP100 STM32H750 is cheaper than the LQFP144 RT1020 - $3.99 @ 90 off (Avnet) v $4.21 @ 60 off (Future Electronics)

On paper the H7 seriously outclasses the RT1020 in many areas:

128k FLASH v 0
1MB RAM v 256k
3 x 3.6MSPS-16 bit/4.5MSPS-12 bit ADCs with 83dB SNR, 1 LSB INL*12 bit mode) v 2 x 1.4MSPS 12-bit, 66dB, 2.8 LSB INL
2 x USB v 1 *BUT* the RT1020 has an integrated HS PHY

plus:

2 x 12 bit DACs
Camera I/F *BUT* the RT1020 Flexio can provide this.
LCD I/F
High resolution timer (2.5ns)
JPEG Codec
Chrom-ART Accelerator
HDMI-CEC
Digital filters for sigma delta modulators
2 x opamps

Timers, PWM, external memory I/Fs, SPI, SAI, I2C, UARTS, SDIO/MMC, Crypto, comparators etc. seem similar

The much larger RAM is one of the most important advantages - if you need it of course - but applications needing a high performance MCU and complex/fast peripherals typically need a lot of RAM - especially if the code has to be loaded into RAM for performance reasons.
 
The RT1020 has at least one killer feature though - FlexIO which looks similar to the LPC4370. I wish all modern Cortex had something like this - especially if they supported LVDS for interfacing high speed external serial ADCs. Anybody know how fast the FlexibIO can be clocked?

The RT1020 also has an 8x8 keypad peripheral.

As to performance, I've no idea why you state that the H7s are less powerful - both are 400MHz M7s so should be identical executing from RAM (but peripheral buses may impact some peripheral/DMA performance). The H750 will vastly outperform the RT1020 for any code which doesn't fit in the latter's 256k RAM, but does fit in the H7's 1MB, unless the code takes great advantage of the 16KB instruction cache which can be very hard to design and maintain.

The RT1020 is supposed to be available at 500MHz but not yet (from distributers at least). I don't know if the 500MHz parts will replace the 400MHz ones or if there will be a price premium. The H7's are becoming available at 480MHz for revision 'V' parts - it seems to me that they are intended to replace the 400MHz parts as they don't appear to have different part numbers which is pretty confusing since distributors don't specify the silicon revisions. You have have to rely on the distributor's descriptions to determine if it is a 400 or 480MHz part, but they frequently make mistakes in descriptions. If you click through the 'buy direct from ST' links on their web site they don't even tell you which speed version you'te buying - how crap is that? 

I see that Mouser are offering a 400MHz and 480MHz version of the STM32H743ZIT6 (pretty much identical to the H750 but 2MB FLASH) for the same price, so another indication that the 400MHz part is being superceded. The 480MHz has the part number STM32H743ZIT6U but the only place you can find that referenced on the ST website is in the NUCLEO-144 User Guide. This is ridiculous ST - sort it out!!!

The RT1020 is available in an LQFP144 package whereas the H750 is only in 100LQFP or 176 and 256 pin BGA packages so that will favour the RT1020 in some cases, but ST could add a LQFP144 in the future. The H743Zi is an 144LQFP but is more expensive given the 2MB FLASH - $10.27 @ 100 (Mouser)

I said "on paper"; as mark03 pointed out the feature list is irrelevant if they don't actually work! Quite a few of the errata items have been fixed in the 'V' (480MHz) revision but still plenty remain - including all the Ethernet ones. The RT1020 has quite few errata items but not as many - perhaps because fewer developers are using it so they haven't been found yet?

Another issue with ST parts has been pin blocking severely restricting the combinations of peripherals that can be used together. I don't know if the H7 series is much better in this regard or how the RT1020 compares.

[ebastler]

Anyway, the Tensee-v4 looks "perfect" for building a graphic calculator! I have already reverse-engineered a couple of pocket calculators made by CASIO and Sharp, so ... maybe I will reuse their keyboard and LCD, replacing the motherboard with a Tensee-v4, which has a lot of horsepower for doing mathematical calculous and even more horsepower than the latest CLASSPAD 400.

But it would no longer be battery powered, right?
(Or at least that battery would no longer fit inside the case. )

Runs at about 100mA.

That Mathematica terminal is neat, by the way! 

[legacy]

The bootloader is exclusively the reason they aren't cloned en-masse.

Yet another evidence that Open Source does not work as expected.

edit: just to avoid confusion
"does not work as expected" is not equal to "doesn't work", it doesn't mean "failure of Open source", it means that it does work but with some "reserve" -> when used for business, it needs tricks and legal staff to protect your rights.

[Nominal Animal]

The bootloader is exclusively the reason they aren't cloned en-masse.

Yet another evidence that Open Source does not work as expected.

Ahem.  Do you realize that all the support libraries and so on are open source?

As I recall, Paul Stoffregen has said that the only reason the bootloader and hardware board files are proprietary, is to fund the development of the open-source support software and further microcontroller models.  (I apologize for not finding the source, but if you are interested enough, you can go on to forum.pjrc.com and ask him directly to verify.)

Assuming I recall right, this is just one business model that incorporates both proprietary and open source parts.

A better question is, why hasn't any microcontroller manufacturer employed somebody like him, to generate not just evaluation boards, but to create a full ecosystem around the boards?  Is their profit margins truly so slim that having less than half a dozen employees to set up and support a whole ecosystem is too expensive, compared to the goodwill/long term profits engendered?  It seems to me that most manufacturers are still stuck to pretty limited business models, where they try to extract value out of their customers at every single turn, from tools (compilers and environments) to documentation (by requiring them to register, so their information can be sold to advertisers, or at least used internally in advertising), and that PJRC and various Arduino board manufacturers are actually exploring the business models open source allows.  (Atmel being the odd one out, in my humble opinion; I've never liked Microchip's approach, and worry about that a bit now.)

My point is, are you sure you are not just whining because you'd like to be able to take advantage of others work without giving back in turn?  If so, that's not how share-alike open source works; and the permissive open source licenses are less than half the open-source world.  So, perhaps it is not open source that is the problem, but a limited imagination regarding business models?

As I understand it, PJRC is doing quite fine, keeping that small part proprietary, but open-sourcing everything else; actively contributing to the Arduino environment, and even making it possible for others to manufacture compatible clones of their boards.  Those contributions are definitely the main reason why I use and like Teensies, and that business model / pattern is one that I recommend others consider as well, if doing something similar.

[legacy]

edit: just to avoid confusion
"does not work as expected" is not equal to "doesn't work", it doesn't mean "failure of Open source", it means that it does work but with some "reserve" -> when used for business, it needs tricks and legal staff to protect your rights.

Ahem.  Do you realize that all the support libraries and so on are open source?

yes, but they cannot release the bootloader due to the attitude of *cloning* it for mass production, which is a serious problem, especially for "Copyleft" places (Russia, China, ...), so they keep libraries opensource (which is good), and the protect their business by making this strategy (which is also good, for them).

I do believe that it's a good trade-off.

[legacy]

I consider you like a friend, but I did find this point a bit "nah", so ... apologize in advance for the crude tone of this answer.

My point is, are you sure you are not just whining because you'd like to be able to take advantage of others work without giving back in turn?

All the projects hosted at DTB have been written by my crew and me from scratch! We have our own libraries, our own code, our own HDL code, and none in my team has ever used anyone's library for any projects or business! And personally, I am not willing to release anything to the public because the last time I did it my work got cloned and used for business without respecting any of my "rights".

I do have neither the time nor the money to pay a lawyer against someone who probably lives in the "copyleft" area, therefore not punishable, and frankly I do think that the OpenSource is bullshit if you are not willing to take legal actions, this even because I really do not like to quote people in a court case!

I am not saying that I do not trust OpenSource for large projects; in fact, I am contributing to the development of a few Linux/RISC-ish kernels by testing stuff and by reporting bugs to Bugzilla. GNU and kernel dot org have a squad of lawyers, and a lot of money, so defending the rights is not my problem.

In short, my position about "OpenSource" has constraints, and it's under trade-offs, but for sure I have never taken any advantage of others work.

[SiliconWizard]

The bootloader is exclusively the reason they aren't cloned en-masse.

Yet another evidence that Open Source does not work as expected.

Well, maybe not Open Source as a whole, but OSHW, quite possibly indeed. For obvious reasons. Hardware and software are definitely two different beasts.

If you're designing and selling your boards for a living - how the hell can it work if you release the whole thing as OSHW? If the boards become half successful, they will quickly get cloned in countries where costs are dramatically lower, will get sold at half the price or less, and you'll be left with no business. How could it ever work? I'm all ears. Usually the only way you can not be affected by this is when what you sell is actually NOT successful, so doesn't spark any interest in getting cloned. Is not being successful a business plan?

(I guess the proponents will answer that you shouldn't rely on sales of OSHW to make a living? But what exactly can you sell if both the hardware and software can be had for practically nothing?)

I know we have an example here with Dave having open sourced some of his projects such as the µCurrent. But it has gotten cloned a lot now, and I wonder how much he really manages to make out of this? I'm sure if his whole business revolved around it, that wouldn't get him very far. And still, this is a bit different kind of product IMO: it's in the "precision" category, so serious users may feel more confident buying the real thing instead of a cloned version that could contain fake ICs. But for a simple dev board? Most users don't give a damn, they'll buy whatever is cheaper as long as it appears to work.

[Nominal Animal]

Dude, I have never taken any advantage of others work, and I have not even used Arduino because I am aware that Massimo Banzi is still promoting boards "made in Italy" where they were manufactured in China, and I do not like to buy "clones".

OK, so your complaint is that open source cannot work because Russia/China/other markets do not respect copyright law.  On the face of it, that looks like a fair argument.

Fake products are a huge issue, regardless of the license or market, yet nobody claims that brand-building cannot work, because the market will be flooded by cheap clones.  Yet, that same argument should somehow mean open source cannot work.  This is a severe dichotomy I do not accept without a strong argument.

If we look at Teensies, the fact that the software side is open source, and even clones allowed, except for the bootloader that resides on a separate chip (although you can buy those off PJRC as well, and make your own compatible boards), across four generations of microcontrollers, shows that it is a financially viable pattern.  To reiterate, the open source codebase and non-exclusive hardware approach means it is sufficiently interesting even for those who design and build their own variants, while the proprietary bootloader ensures continued tech development funding.

If you had written something like "It is difficult to run a business using OSHW and FOSS alone, without any proprietary bits", I would have just nodded or posted a "Fully agree".  You didn't, you claimed that it is proof that open source is not suited for business.  And that is demonstrably not true.

The reason I "attack" that notion so aggressively, is because I think it is misleading, and limits the ideas people have for potential business approaches; you included.  Such misleading beliefs limit people, and I don't like that.
To me, the notion that "open source is not suited for any business models" is as silly and harmful as demanding people stop riding bicycles because it is dangerous.
You can definitely show examples of bicycle accidents, but to make a sound argument, one must look at the whole picture.
Yes, a pure open source approach with no proprietary bits is very hard if not impossible right now, because of bad actors who do not care about copyright law or international agreements.  That is not sufficient argument to say open source is not suitable for business.

(As to MCU manufacturers, I consider the actual chip design to be the proprietary part, and the I/Os and operating conditions and programming information to be perfectly suited for an open source approach.  I've mentioned this elsewhere, but let's just say that I have a lot of successful examples of how this kind of business approach does work right now.  The approach I do not like is the one where the manufacturer charges you for the chips, then again for the development tools, then again for development tool plugins that allow for run-time debugging, require your information and the right to use or sell that information for marketing reasons to provide you any documentation, often sign an NDA to see the full programming/usage information, and so on.  I see that value extraction as a bottleneck, hurting both the developers, and the manufacturer themselves.)

Now, as to personalities, my experience is that often a sharp poke is needed to get people to actually reconsider their position.  A friendly reminder is too easily ignored.  Because I believe your assumption is limiting your future possibilities, I took the risk of poking sharply in the hopes that it would make you think about this from another point of view.

I hope you don't think of this as a personal attack, but something like a hard nudge so that a friend does not step into a puddle on the sidewalk.  That is exactly the sense it was written in, anyway.

[langwadt]

If we look at Teensies, the fact that the software side is open source, and even clones allowed, except for the bootloader that resides on a separate chip (although you can buy those off PJRC as well, and make your own compatible boards), across four generations of microcontrollers, shows that it is a financially viable pattern.  To reiterate, the open source codebase and non-exclusive hardware approach means it is sufficiently interesting even for those who design and build their own variants, while the proprietary bootloader ensures continued tech development funding.

I don't see how a proprietary boot loader is much of a road block to someone who knows what they are doing, all the hard work has been done, just replicate the functionality

[maginnovision]

If we look at Teensies, the fact that the software side is open source, and even clones allowed, except for the bootloader that resides on a separate chip (although you can buy those off PJRC as well, and make your own compatible boards), across four generations of microcontrollers, shows that it is a financially viable pattern.  To reiterate, the open source codebase and non-exclusive hardware approach means it is sufficiently interesting even for those who design and build their own variants, while the proprietary bootloader ensures continued tech development funding.

I don't see how a proprietary boot loader is much of a road block to someone who knows what they are doing, all the hard work has been done, just replicate the functionality

That requires work and effort. Those of us who can do that respect Paul/teensy enough not to do it. Cloners don't want to go through the effort they just want easy money. You can, or could, find MANY teensy 2.0 clones because it was just hardware thrown together and used all the software paul and contributor's wrote.

[legacy]

If you're designing and selling your boards for a living - how the hell can it work if you release the whole thing as OSHW?

My opinion is that it cannot work unless you have a legal staff or a very large commercial distribution which can compensate the leakages.

Years ago I released a couple of projects and they got cloned without respecting my rights. Who did defend my rights? I was a small developer, so ... nobody did.

It also happened again to my brother when he designed a couple of kits for 3D-printers, and yet again he lost time and money because his kits got cloned and sold on the internet. Who did defend his rights? He was a student, so -yet again- nobody did.

[SiliconWizard]

If you're designing and selling your boards for a living - how the hell can it work if you release the whole thing as OSHW?

My opinion is that it cannot work unless you have a legal staff or a very large commercial distribution which can compensate the leakages.

I don't quite understand what the legal staff would really do. Of course it depends a lot on the particular situations. But a typical OSHW project "cloned" integrally doesn't in itself infringe on anything. It's the right of anyone to build it, and they can also legally sell artefacts of those ("clones") perfectly legally. That's the problem with the model. AFAIK, the only thing they could not legally do is modify your project without giving back their modifications. For a purely hardware project, this is not really a showstopper for anyone willing to sell clones. (For a software project, this may be.)

You have several ways of "working around" that.
- Protect a few trademarks. Use them liberally in your products. Hope that people tempted to sell clones will not bother removing the trademarks, or conversely will use them to capture your image and credibility. Then you have grounds to sue. Thing is, trademarks are a bit expensive to maintain, so many very small businesses can't protect a whole lot of them.
- Only make your products partially open source. Open source part of them, and keep some critical parts closed. Whatever you released as open source can be useful to some people, but they could not "clone" your product as a whole, at least not without a lot of effort. Still, many people will criticize this approach and say you're not playing the open source game right.
- Only design products that would not be cost-effective to clone. Use specific and expensive parts that have no lower-cost equivalent. Make them hard to produce, assemble and test.
- Don't base your business on the sales of hardware products. Just sell them as a means of getting business from other avenues. (This option is probably one of the most used.)

But again, OSHW is very different from OSSW. One thing that makes the OSSW model work well is the contributions. Software gives a lot of opportunities for contributing, which benefits everyone including the initial authors. OTOH, there are a lot less opportunities for contributing on pure hardware projects. That can't even be compared. So the benefits are typically a lot less obvious. Maybe it could for very specific, and very modular projects, in which new "modules" could be contributed. That works only for a very small subset of hardware projects IMO.

OSHW has educational value though. Sure if you're doing that as a hobby that may be good enough incentive. If you're running a business, well. You'll be glad some people learned something thanks to you, but it will not fill up your bank account.

[nimish]

Lame, no usb-c. For 2.0 sinks it's not even that much more complex and needs a little bit of board design to enable the flippable connector.


Sent from my iPhone using Tapatalk

[Nominal Animal]

- Only make your products partially open source. Open source part of them, and keep some critical parts closed. Whatever you released as open source can be useful to some people, but they could not "clone" your product as a whole, at least not without a lot of effort.

This is exactly what I recommend, and I'm more "FOSSy" than most.  The key is to make the proprietary part something that people do not need to interface with directly, or that has a published interface.  (Teensies use HalfKay via Teensy Loader; on the 32-bit Teensies, the bootloader implements that and MCU initialization.)  The key here is to realize that anyone with the resources/knowledge how to reimplement that bit, is able to create their own brand and variants with the same amount of effort. 

Still, many people will criticize this approach and say you're not playing the open source game right.

The "trick" is to make the proprietary bit either implement a published interface, or have it be something that users do not interface with directly at all.
This is quite accepted among e.g. Linux kernel developers, who only want devices programming information, and do not demand their internal firmware to be open sourced.
(Binary blobs executed by the main processor are disliked, because they make debugging kernel code impossible: those binary blobs can do anything, because they run in kernel mode.  Firmware running on the separate board, however, is just firmware; all the kernel developers ask is for their kernel interface to be public, and the firmware binaries to be freely redistributable.)

When doing this, "many people" is replaced by "only a few vocal ones (that you can safely ignore)".

If you disagree or do not believe me, look up Greg Kroah-Hartmann (gregkh) and the various Linux driver programs he has helped run.  There are even ways how the programming interface can be extracted from internal documentation under NDA, so that Linux drivers can be created.

To repeat, I consider most chips and processors to fall into this business category as well: their internals being the proprietary part.

The same approach can also work for software projects as well.  Since most bugs/issues people have are related to the graphical user interface, it makes sense to make it open source (say, written in Python, or include C/C++ sources and build facilities); but keep the core computation in a self-contained dynamically linked library.  (This also makes porting to different OSes easier.)

[donotdespisethesnake]

There are quite a few misconceptions about Free or Open Source. This leads people to think there are flaws in Open Source, but really they are just an incorrect perception.

To be clear, Paul Stoffregen has never claimed his Teensy hardware is "Open Source Hardware". So I don't see any reason why this would be considered a "failure of Open source".

Free or Open Source was never designed to support a business model, it was designed to support an ecosystem of user supported software, and it has done that fantastically well.

OSS is based on copyright law, which does not apply to hardware, so all the "Open Source Hardware" licenses don't have any real legal backing, they can be considered as "social contracts".

Regardless of whether an author creates proprietary or free designs, they can both be exploited by bad actors who don't respect the legal requirements. In both cases, you will need lawyers to chase down and enforce license terms.

I published an Open Source Hardware design and it was quickly put into production by a Chinese company. I've never had any contact with them, they must have found it on the web. I have no problem with them making money out of it, I never had any expectation of that. For me, the system worked, I created some IP which I Open Sourced, then commercial companies step in and make it available to everyone at a cheap price, without them "owning" the IP. In my book, everyone wins.

[SiliconWizard]

Free or Open Source was never designed to support a business model, it was designed to support an ecosystem of user supported software, and it has done that fantastically well.

Oh, I agree with that.

legacy's statement was a bit too blunt maybe, but I think he was mostly talking about the business side of things!

Basically, back to the start of this discussion: the question was about the Teensy boards being OSHW, and some of us just explained why this wouldn't work for the author since they are doing business around selling hardware. Again, no benefit from open sourcing the hardware whatsoever.

So again the point was not to discuss open source in general (IMO), but open source in a business context. There are quite a few ways to make it work for software, but for hardware, not so much.

I published an Open Source Hardware design and it was quickly put into production by a Chinese company. I've never had any contact with them, they must have found it on the web. I have no problem with them making money out of it, I never had any expectation of that. For me, the system worked, I created some IP which I Open Sourced, then commercial companies step in and make it available to everyone at a cheap price, without them "owning" the IP. In my book, everyone wins.

You're exactly illustrating one of my points. Everyone wins here because you didn't care about selling your design as products. If you wanted to make a business out of it, it wouldn't have worked. End of the story. That was my point, legacy's point and probably the point of the author of the Teensy boards.

[ebastler]

I published an Open Source Hardware design and it was quickly put into production by a Chinese company. I've never had any contact with them, they must have found it on the web. I have no problem with them making money out of it, I never had any expectation of that. For me, the system worked, I created some IP which I Open Sourced, then commercial companies step in and make it available to everyone at a cheap price, without them "owning" the IP. In my book, everyone wins.

But did the company give you credit for the design? In my experience with various open-sourced designs picked up by Chinese manufacturers, they don't acknowledge the designer in any way.  If you (as the designer) get neither revenue nor recognition, I don't see how "everyone wins".

[SiliconWizard]

I published an Open Source Hardware design and it was quickly put into production by a Chinese company. I've never had any contact with them, they must have found it on the web. I have no problem with them making money out of it, I never had any expectation of that. For me, the system worked, I created some IP which I Open Sourced, then commercial companies step in and make it available to everyone at a cheap price, without them "owning" the IP. In my book, everyone wins.

But did the company give you credit for the design? In my experience with various open-sourced designs picked up by Chinese manufacturers, they don't acknowledge the designer in any way.  If you (as the designer) get neither revenue nor recognition, I don't see how "everyone wins".

Exactly. Well, there's still one way he could have won. Let's suppose he doesn't intend on making revenue from it. Let's suppose he doesn't really care about recognition either. But maybe he just designed something and maybe built a hand-made prototype, and would benefit from being able to buy for cheap one or more industrially manufactured products out of his design, something he may not have the funds to do. But that's a pretty limited "win".

[langwadt]

I published an Open Source Hardware design and it was quickly put into production by a Chinese company. I've never had any contact with them, they must have found it on the web. I have no problem with them making money out of it, I never had any expectation of that. For me, the system worked, I created some IP which I Open Sourced, then commercial companies step in and make it available to everyone at a cheap price, without them "owning" the IP. In my book, everyone wins.

But did the company give you credit for the design? In my experience with various open-sourced designs picked up by Chinese manufacturers, they don't acknowledge the designer in any way.  If you (as the designer) get neither revenue nor recognition, I don't see how "everyone wins".

recognition could also back fire in people expecting support from the designer

[SiliconWizard]

recognition could also back fire in people expecting support from the designer

Good point.

But just a thought about the recognition thing.

If said company makes products from an open source project and sells them, it has to clearly show that it's open source, and give references of the project (which in turn would give automatic recognition...)

So if the company actually hides the fact it's open source, it's infringing laws. You can sue. (But good luck with that if you're a lone developer against a big chinese company )

Anyway, the point was interesting. Even if the company does things right, it could pose issues you aren't prepared for.
A good example would be for development tools. A company selling them could actually use the fact that they are open source as a big selling point (engineers and hobbyists like open source stuff), and be perfectly open about it. Then yes you'd get some recognition, but you could have to suddenly deal with thousands of people using your project and opening myriads of tickets on github.
Certainly something to consider as well...

[Nominal Animal]

(FWIW, I do agree with all posts between this one and my own above.)

It is also interesting to consider the various open source software licensing models as reflecting completely different approaches to "bad actors".  Might make it easier to understand their fundamental differences as well.

[Nominal Animal]

I think it's more of a low hanging fruit theory.

Well, yeah.  Those who learn the stuff just to create cheap clones only learn enough to do that.  You can tell from the stupid mistakes on the fakes that even I, a complete beginner in designing my own boards, would not make.

If Teensy is really that in demand, someone can simply use Paul's Arduino port and modify the download script so that instead of using Paul's downloader and bootloader, it uses a new downloader and a downloader chip designed for that.

But for exactly the same effort, you can make your own brand.  Teensy 4.0 is tiny, and that's the niche Paul targets, so one could use a bigger version of the same chip, and break out more pins, and use ones own bootloader/uploader, but otherwise use the same base libraries as Arduino/Teensyduino uses -- and make lots more money than by cloning.

I'm sure programming a USB downloader chip that talks to factory bootloaders in those chips is not a hard task. It just takes more time to do than making Arduino clones.

That, and when you know enough to do that, you can do much more profitable things than you can make money off cloning existing products.
To make Arduino clones, you basically need to know how to order PCBs, and how to get the parts and solder them onto the PCBs.

[ehughes]

etter question is, why hasn't any microcontroller manufacturer employed somebody like him, to generate not just evaluation boards, but to create a full ecosystem around the boards?  Is their profit margins truly so slim that having less than half a dozen employees to set up and support a whole ecosystem is too expensive, compared to the goodwill/long term profits engendered?  It seems to me that most manufacturers are still stuck to pretty limited business models, where they try to extract value out of their customers at every single turn, from tools (compilers and environments) to documentation (by requiring them to register, so their information can be sold to advertisers, or at least used internally in advertising), and that PJRC and various Arduino board manufacturers are actually exploring the business models open source allows.  (Atmel being the odd one out, in my humble opinion; I've never liked Microchip's approach, and worry about that a bit now.)

Unfortunately, hobbyists don't generate enough revenue to justly this.       Having seen the "inside" of one of the largest MCU vendors,   this will never happen. Why?   Because the the paying customer who will be issuing large purchase orders want real evaluation boards with IO broken out.    Most of these customers get direct support and start working on their own boards very early.      The sales due to hobby stuff barely a blip on a spreadsheet.

Is their profit margins truly so slim that having less than half a dozen employees to set up and support a whole ecosystem is too expensive, compared to the goodwill/long term profits engendered?

Because those 6 people are dedicated to solving problems for large customers who will spend millions on parts for a single PO.         What goodwill are you thinking of developing?   Most customers are willing to pay for solutions and outcomes.   Most of the negative feedback on the dev tools are not price or cost of compilers.     It is all about showing the MCU in a working configuration with as many IOs provided as possible to test customer configurations.         


There is zero chance any semiconductor company is going to dedicate a half dozen engineers to support a $20 dev board "ecosystem" for hobbyists.   The economics simply don't make any sense.   

[langwadt]

A better question is, why hasn't any microcontroller manufacturer employed somebody like him, to generate not just evaluation boards, but to create a full ecosystem around the boards?  Is their profit margins truly so slim that having less than half a dozen employees to set up and support a whole ecosystem is too expensive, compared to the goodwill/long term profits engendered?  It seems to me that most manufacturers are still stuck to pretty limited business models, where they try to extract value out of their customers at every single turn, from tools (compilers and environments) to documentation (by requiring them to register, so their information can be sold to advertisers, or at least used internally in advertising), and that PJRC and various Arduino board manufacturers are actually exploring the business models open source allows.  (Atmel being the odd one out, in my humble opinion; I've never liked Microchip's approach, and worry about that a bit now.)

how much more than something like STMs Nucleo do you want?

[Nominal Animal]

It is just a common pattern: get their minds while they're young, and they're yours forever.
Just look at the mindshare Arduino has garnered, even though it is technically/technologically pretty crappy.

Remember: I am not an EE, and don't know what it looks like on the professional side.  I only see how the companies treat hobbyists and open source developers.  Do not be surprised, if my understanding of the professional side of electronics design is completely wrong; just correct me, please.

[SiliconWizard]

It is just a common pattern: get their minds while they're young, and they're yours forever.
Just look at the mindshare Arduino has garnered, even though it is technically/technologically pretty crappy.

Very true, that's why many vendors have partnerships with universities, and even give away stuff for free.

[rstofer]

A better question is, why hasn't any microcontroller manufacturer employed somebody like him, to generate not just evaluation boards, but to create a full ecosystem around the boards?

Well, we had (and still have) mbed.org and they originally developed the ecosystem for the NXP LPC1768.  Now the project is owned by ARM and many boards are included.  STM32F Nucleo boards are well represented, for example.

I have always liked the original LPC1768 version.  It's reasonably fast, does ethernet by just adding a MagJack and the TCP/IP library is based on lwIP.  The only part I really care about is the Berkeley Sockets interface and it works really well.

Online tools is a concept that I support but many want to use a local toolchain.  No big deal, try Rowley CrossWorks or even gcc (because CrossWorks is based on gcc I believe).  Once the binary is created, just drag and drop to the board.  The board itself looks like a USB Mass Storage device.  Drag and drop works well.

Lots of boards...

https://os.mbed.com/platforms

[thm_w]

It also happened again to my brother when he designed a couple of kits for 3D-printers, and yet again he lost time and money because his kits got cloned and sold on the internet. Who did defend his rights? He was a student, so -yet again- nobody did.

Can you link to said kit being sold?

I know we have an example here with Dave having open sourced some of his projects such as the µCurrent. But it has gotten cloned a lot now, and I wonder how much he really manages to make out of this? I'm sure if his whole business revolved around it, that wouldn't get him very far. And still, this is a bit different kind of product IMO: it's in the "precision" category, so serious users may feel more confident buying the real thing instead of a cloned version that could contain fake ICs. But for a simple dev board? Most users don't give a damn, they'll buy whatever is cheaper as long as it appears to work.

ucurrent is sold out everywhere, not sure if that tells you anything or not. But yeah his whole business doesn't resolve around it, its a huge stretch to expect this.
He sold about 1,200 in the kickstarter: https://www.kickstarter.com/projects/eevblog/current-gold-precision-multimeter-current-adapter

A better question is, why hasn't any microcontroller manufacturer employed somebody like him, to generate not just evaluation boards, but to create a full ecosystem around the boards?  Is their profit margins truly so slim that having less than half a dozen employees to set up and support a whole ecosystem is too expensive, compared to the goodwill/long term profits engendered?  It seems to me that most manufacturers are still stuck to pretty limited business models, where they try to extract value out of their customers at every single turn, from tools (compilers and environments) to documentation (by requiring them to register, so their information can be sold to advertisers, or at least used internally in advertising), and that PJRC and various Arduino board manufacturers are actually exploring the business models open source allows.  (Atmel being the odd one out, in my humble opinion; I've never liked Microchip's approach, and worry about that a bit now.)

I agree it would be a good investment.

Look at Atmel, what funds have they put into Arduino? Basically nothing, even though it turned out to be huge.
These companies are often clueless, or at least focus elsewhere, "dev board" is not a money maker so why invest money into it. "Samples" department does not directly make money so why offer them. etc. Maybe all their employees are busy solving problems for their top 10 customers. The kind of people that order 1mil+ quantity. The compiler thing also has only started to turn the past few years, TI went free, others I'm sure as well.

[L1L1]

If you're designing and selling your boards for a living - how the hell can it work if you release the whole thing as OSHW? If the boards become half successful, they will quickly get cloned in countries where costs are dramatically lower, will get sold at half the price or less, and you'll be left with no business. How could it ever work? I'm all ears.

You may not have heard of them but there are companies like Adafruit or Sparkfun that do exactly that: they sell OSHW for a living. They publish design files, schematics, and source code, all under open source licenses. According to Wikipedia, in 2015 Adafruit had 45 Million USD in revenue and over 100 employees.

[brucehoult]

You may not have heard of them but there are companies like Adafruit or Sparkfun that do exactly that: they sell OSHW for a living. They publish design files, schematics, and source code, all under open source licenses. According to Wikipedia, in 2015 Adafruit had 45 Million USD in revenue and over 100 employees.

As well as those, Seeed Studio has been reported with the following revenues:

2008: $4.1m
2014: $10m
2016: $30m

[Gribo]

TI sells their dev boards at cost, and many times, you can get them for free in a conference or a seminar.
They also provide support for these boards.

[SiliconWizard]

If you're designing and selling your boards for a living - how the hell can it work if you release the whole thing as OSHW? If the boards become half successful, they will quickly get cloned in countries where costs are dramatically lower, will get sold at half the price or less, and you'll be left with no business. How could it ever work? I'm all ears.

You may not have heard of them but there are companies like Adafruit or Sparkfun that do exactly that: they sell OSHW for a living. They publish design files, schematics, and source code, all under open source licenses. According to Wikipedia, in 2015 Adafruit had 45 Million USD in revenue and over 100 employees.

I have heard of them. I know about Adafruit's history a bit more than Sparkfun's, but still not much.

Note that I actually asked how, not who.

I don't know exactly what their business model is, what their sources of revenue are (beyond selling boards). But I know they definitely do a lot more than just design and sell stuff. They have a lot of online and offline presence. This all takes a lot of people and energy. Certainly not something you can pull off if you're a single-person company. And I don't quite know how Adafruit managed to get there from scratch, but I suspect there is a factor here: being there at the right time. Compared to when they started, I think it would be much more difficult to start this kind of business now, and the probability of having one's products quickly cloned was much lower than it is today. Their current situation doesn't tell much about how they got there. Of course now they have all this presence AND a ridiculously gigantic catalogue of products, they have just so many that even if some get cloned and sold at lower prices, that won't even tickle them.

[L1L1]

If you're designing and selling your boards for a living - how the hell can it work if you release the whole thing as OSHW? If the boards become half successful, they will quickly get cloned in countries where costs are dramatically lower, will get sold at half the price or less, and you'll be left with no business. How could it ever work? I'm all ears.

You may not have heard of them but there are companies like Adafruit or Sparkfun that do exactly that: they sell OSHW for a living. They publish design files, schematics, and source code, all under open source licenses. According to Wikipedia, in 2015 Adafruit had 45 Million USD in revenue and over 100 employees.

I have heard of them. I know about Adafruit's history a bit more than Sparkfun's, but still not much.

Note that I actually asked how, not who.

I don't know exactly what their business model is, what their sources of revenue are (beyond selling boards). But I know they definitely do a lot more than just design and sell stuff. They have a lot of online and offline presence. This all takes a lot of people and energy. Certainly not something you can pull off if you're a single-person company. And I don't quite know how Adafruit managed to get there from scratch, but I suspect there is a factor here: being there at the right time. Compared to when they started, I think it would be much more difficult to start this kind of business now, and the probability of having one's products quickly cloned was much lower than it is today. Their current situation doesn't tell much about how they got there. Of course now they have all this presence AND a ridiculously gigantic catalogue of products, they have just so many that even if some get cloned and sold at lower prices, that won't even tickle them.

I guess this discussion should probably move to the "open source hardware" section :-)

You are probably right: these companies probably started at the right time, in the right place. On the other hand I think their success is also built on the quality of their catalogue and their support.

Nathan Seidle, the founder of Sparkfun, has a great talk on the business open-source hardware, including dealing with clones:

[SiliconWizard]

I guess this discussion should probably move to the "open source hardware" section :-)

True. It was interesting to see this from our perception of the POV of the Teensy boards' author though, rather than as a general discussion without real examples.

You are probably right: these companies probably started at the right time, in the right place. On the other hand I think their success is also built on the quality of their catalogue and their support.

Yep. As I said, there needs to be much more than just selling hardware to pull that off. Not something that can be easily done for a small business. And for a bigger one, it pretty much needs to be dedicated to that IMO, and not just release OSHW as if nothing happened and get on with business "as usual"...

Another point, as I also talked about, is the contribution factor. Again, whereas for software it's working well and the benefits can be rather obvious, it's a lot less from this POV for hardware. A lot fewer opportunies for contributing to OSHW compared to OSS (not saying there is none, but clearly a lot fewer). So this is hardly an incentive to do it, whereas it is for OSS.

Nathan Seidle, the founder of Sparkfun, has a great talk on the business open-source hardware, including dealing with clones:

I'll watch this, thanks.

[Nominal Animal]

It is also important to realize both SparkFun and Adafruit target the hobbyist market, not the professional market (except for proof-of-concept and idea exploration aspects for professional developers, I guess).

While a big company like TI can reach a lot of professionals at a conference or trade show, and raise developer interest there, they almost completely miss the hobbyist market.  As these smaller companies have shown, there is a valid business niche in targeting hobbyists.  Exactly how that is done, I do not know, but I do recognize a few key aspects looking at them.

My earlier point was just an observation that it seems like me that using that hobbyist niche as a way to establish a presence in the mind of those that will eventually become EEs and gadget designers ougth to be worth the financial risk, because as Microsoft has shown in the software world, just getting the mindshare early is sufficient to grab most of the market.  Even when your product is not superior technically, and you acquire new product families through business acquisition, not in-house development.  As SparkFun, Adafruit, Olimex, Watterott, and others have shown, this market niche in itself is profitable; it is just completely different to how larger companies like Texas Instruments and Analog Devices are used to operating.

I bet (but could be wrong) that the FOSS/OSHW approach is a big part of successfully operating in that hobbyist niche; that it is part and parcel of building a successful "brand" in the hobbyist market.  Because of how different the operations are in the professional business world, I believe successful application of FOSS/OSHW there is necessarily different, probably "less pure" because of practical reasons.

However, what I keep harping about here, is that that does not mean that the big business approach is the only one that works.  There are different business niches, and I believe PJRC and others have shown that mixing proprietary and open approaches (specifically, keeping all interfaces open and documented, but some inner details proprietary) has the widest variety of business niches it can be applied in.  It is pretty important to notice that the Linux kernel itself has a very similar practical approach: the developers don't demand the internal details of the devices, or sources to the firmware binaries that auxiliary processors or microcontrollers run on various devices; it is the interfaces that are the key, and that all code run by the main processor is open source so that developers can examine, understand, and fix any issues.  Also, not many users realize how stable the userspace-facing binary ABI of the Linux kernel is.  (Except for idiotic subsystems like Alsa. But that's a different story.)

It is important to look at the bigger picture, future opportunities, existing successful businesses, and so on; and that making sweeping claims like "a business cannot survive on OSHW/FOSS" are nonsensical, similar to claiming that everybody must speak English, because <reasons>.

As to those who believe the largest companies are the most important, I'd like to point out Mittelstand in Germany, the strongest EU economy: SMEs account for 35% of total business turnover, but about half of the total added value.

[Nominal Animal]

If you happen to be in Finland or nearby, Mandu has Teensy 4.0s from PJRC for 22€ apiece.  Just got two myself to play with .

[hansd]

Another option https://www.seeedstudio.com/Arch-Mix-p-2901.html

[ralphrmartin]

Buyer beware. I just bought one, supposing that all of the libraries for the previous models would work. They have not all yet been ported, e.g. the only USB functionality provided for now is Serial. I didn't see any statement about this on their website (nor, therefore, when they hope to have them ported by). I'm a bit miffed and feel this is rather misleading.

[Nominal Animal]

I just bought one, supposing that all of the libraries for the previous models would work.

I should have been more clear in my first post, as I was fully aware of this; and not just say that there are sharp corners left.  Me fail English once again.
It was a very similar situation when Teensy 3.0 was Kickstarted.  (Even the NXP iMXRT1062 chip in the Teensy 4.0 is very new, released only this year I believe.)

I agree, PJRC could and should have informed buyers much better.  I think they should also redo the store web pages; they're puzzling to navigate.

You can follow the progress of the support at Paul's GitHub account, especially the cores repository (for commits to master).

For what it is worth, the extreme speed of this thing has produced quite new problems.  For example, the serial monitor in Arduino is not fast enough, especially on Macs, to keep up: the serial monitor just freezes after ten minutes or so on Macs.  In particular, I believe Teensy 4.0 is the very first microcontroller with proper Arduino environment support that can/will do actual high-speed USB data rates (480 Mbit/s including USB transmission overheads, so about 48 MiB/s).  (Currently, the speed is not yet optimal.)

I'm a bit miffed and feel this is rather misleading.

Completely understandable.

The USB HID types all use full speed USB (12 Mbit/s) -- and HID devices are limited to 64,000 bytes per second each anyway; one max. 64-byte packet every millisecond --, so I am a bit puzzled why you need those slower modes with such a fast microcontroller.  (It also explains why only high-speed USB is implemented thus far: supporting different speeds on different endpoints is probably quite hairy.  I just hope that it does not mean that USB serial + HID can only be supported at full speed; I haven't checked out the iMXRT1062 datasheet to find out, though.  That datasheet itself was not available freely on the web until a month and a half ago.)

[SiliconWizard]

Yep, that's not very nice of PJRC.
That said, the software part is basically open source. So I don't necessarily blame PJRC to have released the board early. Among early users, some may very well help with completing the support libraries... but yes PJRC should have stated so clearly.

This is a very common practice though... the worst being with all those SBCs. Most often they are released with only partial software support, partly broken Linux distros, etc. And you have to wait until that gets better, and they are definitely much more complex so writing libraries/drivers yourself is a LOT more complex, sometimes even impossible due to the lack of documentation. So... you don't have it too bad with PJRC...

[westfw]

the serial monitor in Arduino is not fast enough, especially on Macs, to keep up: the serial monitor just freezes after ten minutes or so on Macs.  In particular, I believe Teensy 4.0 is the very first microcontroller with proper Arduino environment support that can/will do actual high-speed USB data rates (480 Mbit/s including USB transmission overheads, so about 48 MiB/s).

Not that a "terminal" display was ever meant to go >>6Mbps (the sort of data rate achievable with older-generate native-USB Arduino-like boards) (and the Arduino Serial Monitor is hardly "optimized"), but shouldn't SOMETHING in the path be self-throttling?  This is USB, which has thousands of pages of standardization, not just some wire where you dump bits...  (sigh)

[Nominal Animal]

Yeah.  The first 4.0 beta boards (around Christmas) used the iMXRT1052 chip, and the first beta boards with the actual iMXRT1062 chip were manufactured in mid-April, just five months ago.  And, like I said, the manual wasn't freely downloadable until a month and a half ago or so; before that, it was under an NDA.  Making it pretty difficult for any users to help with the software side.

For example, Defragster and KurtE are just contributing users (with quite a few contributions to the Teensyduino core I liked to above), they don't work for PJRC.  (There were a couple of dozen frequent users on the forums that got free beta boards, but that's about it.)  So yeah, users do help a lot, as the only proprietary bit is the bootloader, and the rest of the software is free open source.  Stoffregen contributes a lot to the Arduino environment, too, and not just within the Teensyduino add-on.  It is pretty telling that even Teensy 2.0 is still supported by Teensyduino; ten years after it was introduced (in 2009).

For what it is worth, I do expect the software support to get to parity with the earlier Teensies; but "when", that depends on what kind of unexpected issues crop up, especially with the iMXRT1062 datasheet being version 0.1.

Not that a "terminal" display was ever meant to go >>6Mbps (the sort of data rate achievable with older-generate native-USB Arduino-like boards) (and the Arduino Serial Monitor is hardly "optimized"), but shouldn't SOMETHING in the path be self-throttling?

As far as I know, it is the application that gets stuck, or possibly the Mac USB-serial kernel driver.  Restarting the Arduino Serial Monitor continues to show the output from the Teensy 4.0.  Note that the USB serial interface does provide the info on whether data was received or dropped; but the microcontroller program can simply ignore that (and usually do).

I do have a Teensy 4.0 right now, so I could write and run some microbenchmarks on send, receive, and roundtrip USB serial bandwidth on Linux, if there is interest.  (Note that thus far, the USB stack is not optimized yet, as only the high-speed USB serial interface type is implemented.)

[maginnovision]

I've never really used too many of the libraries outside of the I²C library but the contributors mostly have their one or two libraries they "own" and then anyone and everyone works on the rest(Sometimes just Paul). There are so many libraries that I'd suggest you follow along in the forums, or GitHub. Forums have ways been my preferred way since you can follow progress even between commits. I have not bought a 4.0 yet as I dont actually have a use for one so haven't followed along to know where any of the development is at.

[cgroen]

Yeah.  The first 4.0 beta boards (around Christmas) used the iMXRT1052 chip, and the first beta boards with the actual iMXRT1062 chip were manufactured in mid-April, just five months ago.  And, like I said, the manual wasn't freely downloadable until a month and a half ago or so; before that, it was under an NDA.  Making it pretty difficult for any users to help with the software side.
.

I don't think so. I made the first 1060/1064 designs almost 7 months ago, and there were UM and datasheets freely available on NXP site. Also the SDK has been around for longer than that I think ?

[Nominal Animal]

I made the first 1060/1064 designs almost 7 months ago, and there were UM and datasheets freely available on NXP site.

I glossed over the details, it seems.  See this message on the PJRC forum.

Essentially, the security manual was/is under NDA, but the datasheet and user manual were accessible from NXP if one registers there.  PJRC put the PDFs as freely downloadable from the PJRC.com site, but Paul notes that NXP may demand they take it down.

(No way in hell would I register at NXP for just the datasheet and/or manual for this chip, BTW.  The dev board isn't that impressive.  Remember, I am not a board developer, but a dev board user.)

Also the SDK has been around for longer than that I think ?

The SDK is useless to anyone not using NXP's hardware abstraction, especially if it has a no reverse-engineering clause (that is binding to PJRC, because of their agreements with NXP).  Remember, Teensyduino provides its own HAL for Arduino environment; the NXP SDK is not used.

Or are you seriously suggesting that us users should reverse-engineer their SDK instead of the chip manufacturer providing datasheets publicly, without registration/agreement?

[cgroen]

I made the first 1060/1064 designs almost 7 months ago, and there were UM and datasheets freely available on NXP site.

I glossed over the details, it seems.  See this message on the PJRC forum.

Essentially, the security manual was/is under NDA, but the datasheet and user manual were accessible from NXP if one registers there.  PJRC put the PDFs as freely downloadable from the PJRC.com site, but Paul notes that NXP may demand they take it down.

(No way in hell would I register at NXP for just the datasheet and/or manual for this chip, BTW.  The dev board isn't that impressive.  Remember, I am not a board developer, but a dev board user.)

Also the SDK has been around for longer than that I think ?

The SDK is useless to anyone not using NXP's hardware abstraction, especially if it has a no reverse-engineering clause (that is binding to PJRC, because of their agreements with NXP).  Remember, Teensyduino provides its own HAL for Arduino environment; the NXP SDK is not used.

Or are you seriously suggesting that us users should reverse-engineer their SDK instead of the chip manufacturer providing datasheets publicly, without registration/agreement?

Well, to be honest I don't remember I felt severely abused by registering on their site to get access to the datasheets/UMs, but then again, there are many other chips "out there" so enough to choose from.
For me personally (and the projects I work on) it did not have any catastrophic impact to register, something I was already as I use the forums on NXP already. (BTW, you also have to register to get their SDK, so no need to reverse engineer that one instead of registering for datasheets/UM )

[magic]

I sign up for that kind of things with fake name and a disposable email like guerillamail or mailinator.
Some companies block those so I find some more obscure disposable email service that isn't blocked, download what I want and leave them a support ticket to please unblock mailinator

[Nominal Animal]

Well, to be honest I don't remember I felt severely abused by registering on their site to get access to the datasheets/UMs, but then again, there are many other chips "out there" so enough to choose from.

I am not sure if I would either, if I was designing a circuit.

Having to register to NXP just to be able to program a development board using their processor chip, that I bought from some other company; nuh-uh.  That ain't going to fly.

I can understand registration for interactive tools like TI's webench, or EasyEda; I mean, I can provide my name and email address, but I won't provide my postal address as I really, REALLY don't want paper-based spam.  Dealing with email spam is annoying enough already (I got lots of active filters).  Requiring registration just for PDF datasheets and manuals seems suspicious to me, enough so that if I had had to register at NXP for the datasheet and manual, I suspect I might not have bought Teensy 4.0.

All that said, I believe it is just some marketing folks insisting they need to know everyone using their chip, so they can extract maximum value from their customer base.  Hopefully someone higher up understands that that attitude will hinder any entrance to the hobbyist/Arduino market ... but perhaps they don't care.  Perhaps the user information is more valuable to NXP.

[mac.6]

Registration is for some part of SDKs that are not under BSD-style license, so it requires you to "click and accept" those license if needed.

Does paper spam still exist? I barely got snail mail today and I run spamassassin...

[Bud]

Hell yes in Canada. 

[SiliconWizard]

Not much snail mail spam anymore these days, too expensive. You still get some in your mailboxes not mainly through mail, but directly from people distributing paper ads as very low-paid jobs...

It's annoying when there's too much, but I've grown not to bother (except that it may "hide" real mail when there's too much!)
I use all this paper as protection for paint jobs and workshop stuff in general. There's always a need for that, and it's free!

[langwadt]

I published an Open Source Hardware design and it was quickly put into production by a Chinese company. I've never had any contact with them, they must have found it on the web. I have no problem with them making money out of it, I never had any expectation of that. For me, the system worked, I created some IP which I Open Sourced, then commercial companies step in and make it available to everyone at a cheap price, without them "owning" the IP. In my book, everyone wins.

But did the company give you credit for the design? In my experience with various open-sourced designs picked up by Chinese manufacturers, they don't acknowledge the designer in any way.  If you (as the designer) get neither revenue nor recognition, I don't see how "everyone wins".

recognition could also back fire in people expecting support from the designer

related, http://wiringpi.com/wiringpi-deprecated/

[vk4ffab]

Registration is for some part of SDKs that are not under BSD-style license, so it requires you to "click and accept" those license if needed.

Does paper spam still exist? I barely got snail mail today and I run spamassassin...

For all those things I live at 2A George St, Brisbane City QLD 4000. That is the State Government parliament building. My real address, yeah bugger that.