Discussing new STM32F boards « LeafLabs Garden

LeafLabs Garden » Community » The Lounge

Discussing new STM32F boards

(26 posts) (9 voices)

Started 3 years ago by gbulmer
Latest reply from gbulmer

gbulmer
Moderator

(This is moved from http://forums.leaflabs.com/topic.php?id=1303&page=9#post-8263)

olimexino -

do you think the addon ethernet/usb host as bad idea?

I don't think it is automatically a bad idea, but until I see how an ethernet or USB Host add-on board is fitted, I wouldn't like to give a definitive answer.

Summary:
1 If adding Ethernet or USB Host daughterboards prevents Arduino 'daughterboards' (shields) being fitted, then there is little point (for me) in being Arduino shield compatible.
2. If the board becomes much bigger so that Ethernet or USB Host daughterboards don't prevent Arduino shields being fitted, then maybe put Ethernet or USB Host on the main board.
3. If an Ethernet or USB Host on a daughterboard is less robust, or less reliable than on the mainboard, maybe don't bother.

Detail
An ethernet daughterboard might get in the way of other daughterboards. In that case it was inferior to on-board ethernet.
It might be sensitive to noise, or generate enough noise that it isn't practical to use other daughterboards. In that case their is no benefit to having it on a daughterboard except a lower price to people who don't want it!

In general, I don't like high-speed interfaces (I am hoping Ethernet will be 100Mbits, or USB, 480Mbits) running through header pins without ground. If the board is Arduino header pins compatible, the Ethernet/USB Host board connecter isn't Arduino compatible, so the board should be outside on an Arduino daughterboard footprint. If it interferes with the Arduino footprint, preventing normal Arduino 'shields' being used, then there was little point making the board Arduino footprint compatible (I have no problem with that, if I can make daughterboards on a 0.1" pitch, but some folks may want an Arduino footprint).

So, if Ethernet/USB Host is outside the Arduino daughterboard footprint, there is the problem of making that connection robust. The simple solution is to make the MCU board so big that the daughterboard is within its footprint. In that case, it might be as well have routed the MCU board for Ethernet/USB Host, and have a reduced cost version without the parts fitted.

----

siy - sorry, I wasn't clear. I interpreted olimexno's cost as the retail price increase of $6 on the board, and your costs were based on part cost were about $5.50. I think quantity is hard to guess. I'd start by assuming 1000-off, where some of the parts are about 50% cheaper (e.g. at Farnell), but vary by quite a large margin.

I absolutely agree, it is well worth considering a high-speed interface. Several people have posted on this forum about projects which need a lot more throughput than Maple's USB. STM32F4 has a theoretical ADC bandwidth of 7.2M 12bit samples/second, which is way beyond full-speed USB.

Something 10x faster than an Arduino is a different machine for a potentially different market, and IMHO folks can't just extrapolate from Arduino. IMHO Ethernet might be a distinctly different market from USB.

(Full disclosure: I am not a member of LeafLabs or Olimex staff.)

(I am trying to move this discussion off the old thread. I think that thread is an inappropriate title)

Posted 3 years ago #
bubulindo
Member

I think that the Arduino footprint should be ditched. I mean, this is a 100+ pin plus processor so choosing which signals to fit in those 20 odd pins is hell and in the end won't serve everyone. Plus, there are better designs for this type of processor (Native, for example).

Even when talking about compatibility with shields, I think Arduino has stagnated (when did you see a new shield coming out lately?), and I would guess that with the new ARM based board coming out, the Arduino team won't keep the same format. So it would be nice to have the same format as the new Arduino (just to pick up on their momentum) if it is different and easier to work with a lot more pins outside.

As far as interfaces go, USB is a must especially because that is the most likely way of downloading programs, even if there was an Ethernet bootloader (or SD card), the easiest way to download programs to the board would still be USB. Also, because I think that USB projects will be far more than Ethernet projects.
Although, that being said, I would love to have fast Ethernet on this processor because its speed and data processing capability make it so much better to put data on the web.
If you used the F417, Ethernet interface would have to be installed for sure since that one has all the cryptographic, hash goodies. :)

Resuming, I think that an STM32F4 based board should have the following features:

- single row breakout to be able to use in breadboard, or if double row is used a breakout shield to be sold at the same time. Although, if a lot of peripherals are added, the left over pins would make this easier to decide.

- Vbat for RTC applications. Not very important to have the battery holder in the board, but it would be very important to have a battery connector for this. Maybe with a jumper or solder bridge in case the board was already supplied by a battery.

- 32kHz LSE clock.

- USB communication/Ethernet. Why not design the board with both and only populate one or the other (or both)?

- Sd card connected to SDIO pins. This would make perfect sense with the hardware interface and DMA possibility for the SD card access.

- Some sort of LCD connector, to make use of the manuals LCD interface (haven't looked into it yet).

Did anyone from LeafLabs looked into these (and the linked thread) suggestions yet?

Posted 3 years ago #
siy
Member

@bubulindo, unfortunately there is no standard pinout for the LCD (even very common 16x2 characters LCD's I did saw in at least 3 different pinouts). Also, for small graphics displays often you can meet "4-wire SPI" (SPI interface with additional pin for data/command selection), larger ones usually have parallel interface (mostly with 8080-bus type or, at least, this mode is selectable, so this is a kind of standard). TFT displays also may have different data bus width (16 or 18 bits). Not looked into segmented LCD displays, but guess that there also no standard. But I think that separate LCD connector is not necessary as long as we have FSMC and SPI broken out.

Posted 3 years ago #
siy
Member

@bubulindo, by the way, for my own serial display boards I'm using same 8-pin connector with identical pinout (kind of homebrew standard) and it appeared quite convenient. Also I have finished work on cleaning up/refactoring of two libraries for serial displays (one is for JDT1800 TFT display based on ST7735 controller and one is for UG-2864HSWEG01 OLED display in serial mode). If there would be a place for such designs/software I may share these with community.
@gbulmer, perhaps it worth to create third repo, for shields/boards and relevant libraries? In some cases there might be library only if shield/board is purchased somewhere or just works with some IC which don't need shield or board.

Posted 3 years ago #
gbulmer
Moderator

siy - a third repo would be for shields/chip code? Probably needs to be a repo/board?

bubulindo - I was not suggesting the board would NOT have USB. The discussion was about USB HOST and Ethernet.

Posted 3 years ago #

@bulmer, yes, I think that separate repo for code which works with particular chip/module (I meand, for example, that GSM/GPS/Bluetooth usually used as modules rather than chips). After some thinking I got conclusion that basic split up by categories might be more convenient, for example

- human-interface (everything related to LCD/OLED/TFT/LED displays, keyboards,  and such)
- communication+navigation (GSM/GPS/BT/ZigBee/etc.)
- networking (Ethernet, Wi-Fi, etc.)
- sensors (this one, pehaps will need deeper separation, but definitely not right now)
- control (this one is for all things related to power control/lighting/motors/etc.)
- algorithms (this one is dedicated to non-board specific libraries, such as PID control, statistics, signal processing, etc.).

The proposed separation is just a brief example to illustrate the approach, it can be discussed deeper and improved.

Each library then can be placed in the appropriate category and should follow some common internal structure:

- <chip or module or board name>
  - wirishlib (here comes all code in the format suitable for 'copy and use' in Maple IDE)
  - examples (optional additional/extended/advanced examples which not fit/inconvenient to use with Maple IDE or wirish, perhaps here might come examples which use libmaple directly)
  - lib (optional library which uses libmaple directly)
  - hardware (optional hardware designs)
    - <board/shield name> (design files, docs, etc.)

I think that enforcing structure from the very beginning is essential, without that we will get completely unusable repo. Also, predefined repo structure enables further development of things like automatic downloading of dependencies. Adding such a feature into IDE might be convenient for beginners and allow provide rich hardware library without need to package everything along with IDE.

Posted 3 years ago #

gbulmer
Moderator

siy - Has Arduino already defined the software structure?
I haven't checked recently, but there was work on user libraries, a while ago.

I agree with a predefined structure, but I'd prefer to keep it simple initially.
I think I'd much prefer self-contaned repo's with no dependencies, i.e. all of the parts are in the repo.
Trying to test and debug embedded code which has pulled in the wrong version of dependencies sounds like a nightmare.

Posted 3 years ago #
siy
Member

@gbulmer - I think yes, libmaple/libraries is organized identically to adruino/libraries. Each library in separate subdirectory and consists of C++ class and header, optionally there might be present keywords.txt file and examples subdir.
About dependencies. I did mean only dependencies between application and libraries in repo, not between libraries in repo. And IDE may just have something as simple as just library repo browser. Also, along with releasing IDE we can create branch in library repo and by default IDE will browse libraries in this branch only. Using other versions of libraries will be explicit user decision in this case. Anyway, I think it is too early to discuss possible issues with feature which is not even planned and exists only as raw idea.

About structure. Yes, I'd like to keep it simple too, but I believe that basic separation by category will help keep repo manageable and will simplify further transitions to more sophisticated structure as needed. From the flat structure (i.e. repo/library) it might be much harder to do and will create too much confusion. Also, predefined internal structure for each subdir (i.e. library) simplifies submission of new libraries because provides clear template so author of the code/design don't need to spend time on thinking how to organize things to make them usable. Beside that I think that it will be very helpful for beginners if each library include simple wiring diagram between some Maple/Maple-like board and hardware for which library is written. This will provide information on how library is supposed to work and greatly simplify further adaptation to particular user needs. I don't think this might be set as strict requirement, probably this might be just a suggestion to authors.

Posted 3 years ago #
olimexino
Member

Hi,

I think we maybe will go just with two boards one with bare minimum:

- USB device, USB Full Speed Host, USB OTG
- micro SD card
- UEXT
- two buttons, two LEDs
- power supply
- Arduino 20 pin platform (we love Gameduino, Hackavision etc shields so must keep this "ancient" shields layout for compatibility)
- new connectors platform, maybe split on ports PA PB PC PD PE PF PG i.e. 16 GPIOs+GND+3.3V+5V on 20 pin connectors upper and lower around the Arduino layout

and another board with same as above + USB HS host and Ethernet

right now we are preparing for Embedded World next week and have no free manpower, so we can work on these boards when we are back early March.

we will need 1-2 weeks to make the first prototypes and if someone of you guys want to work on the software library etc implementation please let me know so I can mail you prototype to have

if everything goes well with the prototypes we could even run mass production by end of March.

The goal/target is to release the bare minimum board for EUR 30 and Ethernet/HS USB board for EUR 40 or less.

the project will be Open Hardware and the Eagle etc files will be released so anyone could make their own board based on these two

Posted 3 years ago #
gbulmer
Moderator

olimexino - I am working on ala42's libmaple and ST Micro's STM32F4 peripheral library examples, as well as writing new examples. I am mainly focused on stuff to support sound output at the moment (DAC, basic timers 6&7, etc).

I'm using an STM32F4-Discovery, but I'd be interested to try a prototype. Please be aware that I am revising my own board design to support both STM32F103 and STM32F405, which might make a difference to you. I don't want to be 'infected' by a closed source design.

I am using STLINK/V2 for programming and gdb debugging, so I would like a header to carry SWD signals, power and ground.
It could be 4 pin 0.1" pitch, 6pin 0.1" pitch like the STM32F4-Discovery, but you might want to use 2x5 0.05" pitch, like the 'official' 10 pin JTAG (e.g. described in The Definitive Guide to the ARM Cortex-M3 (second edition) by Joseph Yiu, Appendix H, page 448, Figure H.4). The 'official' 2x5 pin might be attractive to engineers using JTAG hardware debugging.

If you want to get in touch, it should be very easy to guess my gmail dot com address.

(Full disclosure: I am not a member of LeafLabs or Olimex staff)

Posted 3 years ago #
rugbyguy
Member

Hi there,
As a new user to ARM, I have just received a STM32Discovery from Farnell (UK)approx£10 inc FREE P&P and came next day.
If I may add to your interests and to "Olimexino" these starter kits are an excellent way to go and to make best use of them, these boards have 2x50 pin connectors that fit the flat ribbon cable.
This makes them very easy to connect to other boards and to fit to a universal/motherboard type system.
I would buy if available a board to plug this unit into.
Suggested options:- Wireless option for WiFi,Zigbee, Bluetooth, LoPower wireless.
USB Host/OTG,Ethernet, SD Card, Power 5-24V.
LVDS,HDMI,RS232/485,ADC connectors.
Maybe even some space for 12-240V relays.

As I recycle kit for voluntary sector I would like to use some of the LCD screens I have aquired and build some test kit.
This is a most useful forum.Thanks to all.

Posted 3 years ago #
bubulindo
Member

Some posts ago I made a similar suggestion. Basically, design a new board with the same pinout as the F4 Discovery board so that people could use the Olimex board or a Discovery to use that. I think that idea was dropped, mainly because some of the peripherals should be in the board and not through connectors that may or may not be good.

One other thing is price. The Discovery board price follows the gillete model. Losing money on the hardware to make it on the replaceable parts. In this case, in an increase in sales volume on this chip family by ST. If you look at Olimex website, you'll see that a board with all (or part of) that hardware can go to 100$+.

I may do something similar to this with a veroboard, but I have low expectations... basically try to get the SDIO interface working (and who knows, adapt it to the sdFAT library) and get a communication channel to the outside so I can get the RTC working.

Who's keeping a repository for new F4 software stuff? I've created a driver for the random number generator so if someone could point me to where I should leave it to be included in a next libmaple delivery, I'd appreciate it.

Posted 3 years ago #
siy
Member

@bubulindo, take a look at https://github.com/gbulmer/ , but I think such a library is good candidate for the other repo which is discussed in another thread.

Posted 3 years ago #
jbaker
Member

Olimexino,

I would be interested in one of the prototypes to work on the software library. Let me know where to send mailing info. Also, I assume you have an existing library in version control somewhere?

Posted 3 years ago #
olimexino
Member

@jbaker contact me at info@olimex.com, I'm in Germany next week and will be able to respond when I'm back next Friday

Posted 3 years ago #

RSS feed for this topic

12 Next »

Reply »

You must log in to post.