https://www.st.com/en/microcontrollers- … eries.html
https://www.st.com/resource/en/data_bri … 7a-dk1.pdf
This is what I wanted for the longest time: a RPi integrated with an ST32F4 or better. The boards will be available in April and will come with ST’s 10-yr support commitment and a full native SDK package. They will offer this thing at $69 (I wish it was around $50…. oh well)…. 
More info:
Yawn …
[BennehBoy – Thu Feb 21, 2019 3:28 pm] –
Ooh looks like a nice platform to run things like openelec, etc.
FYI.
Openelec does not seem to suooort newer RPi boards. E.g. it won’t even boot on the RPi 3B+
LibreElec May eventually support this board, since it supports the 3B+, but there are so many other competing boards e.g. OrangePI, that I doubt people will rush to support Kodi on this platform
[mrburnette – Thu Feb 21, 2019 4:32 pm] –
YALD (Yet Another Linux Distribution)Yawn …
It might not be for everyone, but I look forward to use it in my lab. My students can VPN into it, open a professional toolchain (Atollic TrueStudio with CubeMX), compile the code, debug it with the onboard ST-Link and deploy it immediately and remotely. With its Cortex A7 cores the board can analyze a massive amount of data taken by the on board uC and only return processed data. $69 for digital+analog capabilities in a professionally supported package sounds pretty good to me. It promises to be a very versatile, cheap yet professionally supported data acquisition system.
Well, that explains it. I did a year in the EE lab of a state university and the department’s professors are always wanting to play around with new toys; it is the poor student teachers that have to deal with the insanity of trying to make the lectures and labs have continuity.
The $35 Raspiberry Pi 3x can do the same at half-the-cost. Having an STM32Fx attached is “nice” but the concepts can be implemented with a $2 Blue Pill bringing the total cost to under $40. Such a low price point provides for an opportunity for students to own a complete lab’ish environment off-site. Additionally, boards will meet the magic smoke from time to time, easier to replace smaller, less-expensive commodity boards.
Ray
I really like the idea of arm cortex-a7 combined with an ARM cortex-m4. Personally, I think it is much more powerful than just combining a Raspberry PI and an off the shelf cortex-m board. The biggest advantage this board has is the MPUs and MCU are on the same die sharing memory and peripheral all connected with high speed buses. With a RPI and a standalone board, you are going to run into a bottleneck trying to communicate between the MPUs and the MCU.
To me, it seems like a better version of the beaglebone. Where the beaglebone is a single core cortex-A8 (older tech) with two proprietary MCUs, this ST board has the advantage of a modern lower power dual cortex-A7 MPU paired with a standard cortex-m4. I have a beaglebone, I have played with its PRU (Programmable Real-time Unit) In spite of that neat feature, I haven’t invested much time really learning it as it seems like a one off combination. Given TI’s habit of obsoleting chips not long after they are introduced I was afraid of spending any time learning it only to have it tossed in the bin. I will grant you that the beaglebone seems to have persisted in spite of my trepidation towards TI. If I use this ST board, I’m not wasting my time learning the quirks of some wierdo PRU thing without any peripherals, I would be using your everyday ST cortex-m4 with peripherals I’ve used before and programming it with a familar gcc toolchain.
The other unique MCU feature I like is the huge amount of memory instead of it being flash bound. No hidden performance killing flash wait states or inconsistent clock timing because of a hidden instruction cache. * At least this is what I get from reading the datasheet
This might be the perfect 3D print controller board using klipper.
[mrburnette – Sat Feb 23, 2019 12:37 am] –
It might not be for everyone, but I look forward to use it in my lab. My students…Well, that explains it. I did a year in the EE lab of a state university and the department’s professors are always wanting to play around with new toys; it is the poor student teachers that have to deal with the insanity of trying to make the lectures and labs have continuity.
The $35 Raspiberry Pi 3x can do the same at half-the-cost. Having an STM32Fx attached is “nice” but the concepts can be implemented with a $2 Blue Pill bringing the total cost to under $40. Such a low price point provides for an opportunity for students to own a complete lab’ish environment off-site. Additionally, boards will meet the magic smoke from time to time, easier to replace smaller, less-expensive commodity boards.
Ray
You can be friendlier than a Grinch and a bit less judgmental. I want these boards for my research lab, not for teaching undergrads. “Students” here was used to represent doctoral students, not undergrads. We build a wide range of scientific equipment for our research needs and in many cases, where reliability and durability are needed, we don’t use BluePills as we know they have inferior quality control compared to genuine ST boards. When we can use Blue Pills, we do. But there are situations when BluePills just don’t cut it.
That being said, $69 for a professional board compared to your $40 jerry-rigged yet workable solution is something that at least should be considered even for teaching undergrads. The price difference is not that big, considering we would be putting in front of the student a professional toolchain. And guess what? As the new board already has ST-LINK on board, it can probably program a BluePill as well if magic smoke is a concern.
[EventHorizon – Sat Feb 23, 2019 4:21 pm] –
…
You can be friendlier than a Grinch and a bit less judgmental. I want these boards for my research lab, not for teaching undergrads. “Students” here was used to represent doctoral students, not undergrads.
…
I tried being friendlier once over the past 4 years, but like a bad meal, it did not sit well.
Putting the target audience into the doctoral spotlight, I completely agree with you – it could be a great teaching aid.
Ray
[Rick Kimball – Sat Feb 23, 2019 2:47 pm] –
I like it! Where do I get one?I really like the idea of arm cortex-a7 combined with an ARM cortex-m4. Personally, I think it is much more powerful than just combining a Raspberry PI and an off the shelf cortex-m board. The biggest advantage this board has is the MPUs and MCU are on the same die sharing memory and peripheral all connected with high speed buses. With a RPI and a standalone board, you are going to run into a bottleneck trying to communicate between the MPUs and the MCU.
…
Hopefully RPi might pickup this Chip as it would fill in what the RPi is missing.
[flyboy74 – Sun Feb 24, 2019 8:21 pm] –[Rick Kimball – Sat Feb 23, 2019 2:47 pm] –
I like it! Where do I get one?
…Hopefully RPi might pickup this Chip as it would fill in what the RPi is missing.
I do not see that happening even if STM gave the uC’s to Raspberry Pi Foundation. Pi in the sky talk…
Every Pi has a heart from Broadcom.
From Tech Republic:
The Raspberry Pi 3’s mobile-phone based hardware is nippy compared to its predecessors—some 10x faster than the first-gen Raspberry Pi. However, the 1.2GHz, quad-core, ARM Cortex A53-based processor in the Pi 3 is today rather long in the tooth compared to those found in recent phones.
Boards released after the Pi 3 was launched in 2016 have packed in faster and newer processors, ranging up to Octa core. While the need to keep costs below $35 would limit the spec of the CPU, a modest bump to the clockspeed to, say, 1.5GHz or so seems feasible, and would deliver a nice performance boost when paired with a system-on-a-chip with a newer, more efficient architecture.
While the above does not preclude a change of chip suppliers, I find it nonsensical to believe STM could “jump into the Pi space” like Captain James Tiberius Kirk’s Enterprise making the transition from impulse to warp!
Ray
But what prevents them from moving away from Broadcom is not only history, but what makes the Rpi such a cash machine: the exclusivity they have on the CPU and the lock with the GPU blob. Otherwise, they would have been cloned by Chinese manufacturers a long time ago.
So it is more or less the “Clone War” and in this case, and the references are more to Star Wars 
[Squonk42 – Mon Feb 25, 2019 5:42 am] –
But what prevents them from moving away from Broadcom is not only history, but what makes the Rpi such a cash machine: the exclusivity they have on the CPU and the lock with the GPU blob. Otherwise, they would have been cloned by Chinese manufacturers a long time ago.
The GPU blob should be inside the RPi SD image that you download so a clone can be made since you will be using the same firmware. I believe that there is another reason. Maybe the automated production is already low cost, they was able to start the board production also in UK (first productions was only in China) at the same price.
[Squonk42 – Mon Feb 25, 2019 8:57 am] –
Yes, but the blob is a “lock” in the sense it is closed source, so it cannot be adapted to another chip, like Rockchip or Allwinner to make it Rpi-compatible.
https://www.raspberrypi.org/forums/view … 7#p1329813
Re: Binary Blob?
Quote
Mon Jun 18, 2018 2:12 pm
The RPF have full access to the SoC firmware source, so we do all the support in that area. The SoC manufacturer does not work on our firmware, mainly because a lot of the people who know how it works now work for us after Broadcom shut down the VC4 team.
There is also firmware for the wireless chip. We do not have access to that, all work for that is done by Cypress.
We also do a lot of the Pi specific Linux kernel code, although occasionally third parties do contribute – it’s all OSS after all.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
It is not my definition of OSS.
They keep it close so they completely control the RPi ecosystem, but also because the closed-source blob contains the GPU firmware, which in turn contains the HDMI protection scheme that nobody is allowed to see, in case it can be broken (which is pure protection by obfuscation, which is a clear demonstration that it can indeed be broken).
The wireless firmware is just an excuse, it could be replaced by another chip if required, it is not the heart of the system like the Broadcom SoC.
And most of the “Pi-specific Linux kernel code” is just APIs to the closed GPU blob 
[Squonk42 – Tue Feb 26, 2019 8:49 pm] –
Yes, the Raspberry Pi Foundation has access to all sources, but not the common mortal: don’t be afraid, we are in chargeIt is not my definition of OSS.
…
Raspberry Pi Foundation is a business, a non-profit, but still a business. The HDMI code inherited from Broadcom does have licensing preventing the public viewing… IBM essentially owns HDCP.
From Wikipedia:
In order to make a device that plays HDCP-enabled content, the manufacturer must obtain a license for the patent from Intel subsidiary Digital Content Protection LLC, pay an annual fee, and submit to various conditions.[5][6][7] For example, the device cannot be designed to copy; it must “frustrate attempts to defeat the content protection requirements”;[7] it must not transmit high definition protected video to non-HDCP receivers; and DVD-Audio works can be played only at CD-audio quality[7] by non-HDCP digital audio outputs (analog audio outputs have no quality limits).
Regardless of how one feels about OSS, the law requires license holders to keep the code private. Do not blame Broadcom or Raspberri Pi Foundation, blame Intel and the big-dollar content providers that enforce the threats of lawsuits.
As for me, I simply do not give a rat’s patuti about the video blob; IMO, it is a non-issue, the performance / $$$ is far more important: http://roylongbottom.org.uk/Raspberry%2 … #anchor25a
(Atlanta area prices)
RPi Zero-W $5.00
RPi 3B $29.99
Don’t pretend “it’s all OSS after all” when there is a closed source blob that governs the CPU boot.
Don’t pretend you are a Charity when you make money selling 10 millions boards with profit: pay your due taxes.
Don’t call for freedom when the big blob locks you from moving to another CPU, using a different LCD display or using another non-official camera module.
Lucky you that have a Micro Center nearby, but for most of us, the real prices for these boards is much more expensive. My best options from Bordeaux, France are:
RPi Zero W @ 10.44€ + 5.95€ S&H = 16.39€ ($18.65):
https://www.kubii.fr/pi-zero-w/1851-ras … 06997.html
Rpi 3B @ 34.86€ + 5.95€ S&H= $40.81 ($46.42)
https://www.kubii.fr/les-cartes-raspber … 10850.html
For almost the same price as an RPi Zero W, I can get a NanoPi Neo with a quad core Cortex A7 @1.2GHz vs. an old single core ARM11 @1GHz:
https://www.gotronic.fr/art-module-nano … -28756.htm
Of course, no “RPi Inside”
― Max Born
[mrburnette – Wed Feb 27, 2019 2:57 pm] –
“The belief that there is only one truth and that oneself is in possession of it seems to me the root of all the evil that is in the world”― Max Born
There is only one Truth, but I don’t pretend to possess it.
[Squonk42 – Wed Feb 27, 2019 3:01 pm] –
…
There is only one Truth, but I don’t pretend to possess it.
Then, my friend, we have complete agreement. Perception is a human condition.
“A votary of truth is often obliged to grope in the dark.”
— Mahatma Gandhi
I have nothing against earning money for your work, and I tend to flee as far as possible from any FSF ayatollah.
But I really dislike some businesses built on top of OSS / Open Hardware belief, which smell like pure opportunism:
https://arduinohistory.github.io/
Probably the source of all these behaviors is a mix of greed, peer recognition and the need to fulfill the people’s need to believe into something that transcend the poor reality of life.
https://www.st.com/content/st_com/en/pr … c-dk2.html
but ST bundles a pretty ‘big’ 4″ LCD, throws in gigabit ethernet and usb type C, WiFi and Bluetooth le
(but but Order from ST Order from Distributors
STM32MP157C-DK2 No availability reported, please contact our Sales office
oh mine, ST is getting too hot these days any hardware that become real immediately vaporise)
i’d think chances of this being more ‘open source’ is possible. we’d need to wait a little
i think as costs is a big consideration, st miser a little on on the number of cores but it is nevertheless a good start
http://www.orangepi.org/orangepipc/
(oh oops that price a whopping 50% higher at $15, that $10 board is here http://www.orangepi.org/orangepione/)
from aliexpress simply click the buy link on the manufacturer’s site
orange pi is unashamed to overclock and push the boundaries of 1.6ghz. it gets so hot i need to try a heat sink to contain all that heat
and anyway around 1.3 ghz is about possible
i’d think for st, being a little stingy with the cores for now is ok but that they should try to get more hz
to level up with the competition
btw quad cortex A7 x 1.3ghz can be *pretty fast*, i did get a couple of gflops when crunching matrices on quad core cortex a7.
not enough to train a deep CNN network but for a pre-trained network, i’d think it is possible to run image recognition from there
Ie can it run distros made for say a pi3?
I’m guessing not.
https://www.armbian.com/
https://forum.armbian.com/
which is a debian (with some add ons) + a kernel customised to these specific boards
accordingly much of that started here
http://linux-sunxi.org/Main_Page
but that stm32f4 is much better with IO (usb, usart, spi, i2s, i2c, adc, dac (bigger boards), sdio, fsmc and pretty much gpio, cortex-m plays a different game.
but above about $15 price / performance become more ‘gray’ in that orange pi one has 512m dram, ethernet, quad core A7 1.3ghz, the more expensive cortex-m* start to lose the ‘price-performance’ advantage as the cortex A7 quad core indeed runs much faster due to the dram, much higher clocks, cores and has virtual memory (which cortex-m lacks)
virtual memory is the main limiting factor for the cortex-m, but i’d guess it is a very big reason cortex-m arduino are simplier to work on, every thing on stm32 is ‘fixed’ so we’d only need to goto that memory location and get that ‘register’
as an example if we compare this aliexpress F746 that packs a punch in both performance and io $115
https://www.aliexpress.com/item/STM32F7 … 85301.html
vs stm’s own stm32 mp1 $100
https://www.st.com/content/st_com/en/pr … c-dk2.html
even ST’s own stm32mp1 discovery wins out the competition with that 2 A7 core + dram + another cortex m
If only Pi was fully open source (h/w too).
the cortex A series has a lot of details hidden behind non-disclosure, a lot of the io features, the graphics acceleration and things like MIPI interrfaces and the high speed camera interfaces are all either binary blobs and not disclosed. the volunteers piece together scraps of info and reverse engineering to even get some kind of open source io, even things like the hdmi interface is reverse engineered and hdmi audio has been rather recently added in the open source kernels
I pulled out my circa 2013 BeagleBone Black and loaded it up with the latest debian. I had shelved the board back then as it seemed everything I wanted to do dripped of wet paint. Then for a couple more years I didn’t do anything with it while I waited for TI to NRND the chip, as the seem wont to do. Well that didn’t happened and they have even hinted of a new BBB AI (dual core OMAP A8 + 4 PRU + DSP + 2 Cortex-M) release sometime. They also released a Altoids tin sized board stripped of ethernet and hdmi and a few pins. That is very competitive at $25.
The BeagleBones use an OMAP chip that has a single Cortex-A8 combined with two 200MHz Programmable Real Time Units (PRU). This allows linux to go on doing linux things and the PRUs to go on doing realtime stuff without any interference from the Cortex-A8. The BBB has 512K of ram and each PRU has 8 k of dedicated instruction memory and access to a shared pad between them along with direct access to some of the GPIO pins. That direct access allows you to toggle the pins in a single cycle (5ns) in addition, jmp instructions are also single cycle so you can easily get a 50MHz toggle without a problem. Contrary to what I said before, the PRU does have access to peripherals.
With great trepidation I dived into looking at doing some PRU (Programmable Real Time Unit) programming. I watched a bunch of videos and even caught a presenter answering a question about TI’s approach compared to chips with a combo Cortex-A series + Cortex-M chips and how it compares to the PRU. It reminded me that the PRU is a mostly single cycle proprietary RISC architecture that has no pipeline. That makes the PRU completely deterministic vs the cortex-m4 which is pipelined and not as predictable.
Armed with this knowledge I found out that the TI stuff still has wet paint when it comes to programming the PRU. They have added a C compiler, and they have stopped supporting the original asm compiler they provided. The new one has different syntax of course. The method for loading the PRU instructions and getting it going are different now as it is using a linux style method of remoteproc. The docs and all the previous examples serve to distract and confuse. In spite of that I was able to get a simple ws2812b pulse stream spewing from it in about a day of reading docs and monkey testing.
I’m still excited about the STM32MP1 hopefully the software to interface between the linux and the realtime subsystem will be stable and well documented. I guess my only concern is that this is ST’s first go at this Cortex-A series in a long time so there are bound to be some gotchas. However the STM32F1 was and is very usable even though it was their first chip in the Cortex-M series.
In the meantime, before they release the STM32MP1, I’ll continue to forge ahead with some TI PRU stuff so I have something to compare it to when the STM32MP1 comes out.
[ag123 – Tue Mar 12, 2019 10:50 am] – virtual memory is the main limiting factor for the cortex-m, but i’d guess it is a very big reason cortex-m arduino are simplier to work on, every thing on stm32 is ‘fixed’ so we’d only need to goto that memory location and get that ‘register’
… And a significant amount of physical memory, I/D caches, pipelines and faster CPU clocks too 
for the fun of it i found pru-duino
https://github.com/lucas-ti/PRdUino
i’m not too sure if it make sense to run ILI9341 lcd library there
