I want RET6 Mini or, better yet an F4 Mini.

LeafLabs Garden » Community » The Lounge

I want RET6 Mini or, better yet an F4 Mini.

(32 posts) (6 voices)

Started 3 years ago by MediumKahuna
Latest reply from gbulmer

12 3 Next »

MediumKahuna
Member

I just found Leaf (Hello my name is Jud) so first let me say that, I'm ecstatic about it.
You guys are great! Thank you so much.
I hope to become a solid contributor soon (still reading the excellent documentation).

I can't believe the RET6 isn't a huge hit. The 64K of RAM, alone, makes my day.
However, I really want it in a Mini format.
I guess there isn't 48 pin chip any better than the STM32F103CB

Make the F4 in a compact version first. The 64 pin version (STM32F415RGT6) has plenty of pins.
Then make an Ardunio adapter.

Posted 3 years ago #
siy
Member

@MediumKahuna, take a look at https://github.com/gbulmer/openstm32hw I guess that one of boards for 64-pin chip will fit into your requirements.

Posted 3 years ago #
MediumKahuna
Member

Wow, so much cool stuff.
Its going to take me a while to get a feel for it.
I'll try not to shoot off my mouth too much in the meantime.
But, Its hard to control myself when I get excited.

I guess I should work on formalizing my requirements but this is a bottom up situation where what I end up doing depends on what I can get to do it with.
A compact form factor, something like 4 million double precision floating point macs per second (my RET6 only does 500K) and 64k of RAM but, the more the better as both of these dramatically simplify coding of the algorithms I want to use. Just about any modern chip has plenty of ROM for my needs. A high speed programming interface (I prefer Ethernet) would be nice.

Of course, the most important requirement is support of a great community and I think I've found that here (I'm crying and I just peed all over the floor).
So, I want to use a chip that you guys like and plan to support (I plan to help of course).

I do have lots of experience and many skills but, a lot of it is dated or not main stream (millions of lines of code but mostly C and assembly, lots of BSD but very little Linux and so on). While I've written a ton of C++, a fair amount of Java code and I'm a thoroughly object oriented guy now, I only, really, truly, grasped OO a few years ago.
I have an electrical engineering degree (University of Michigan) and started out as an embedded hardware designer but soon turned traitor and became a software developer, did a stint in management and now I'm a machine vision software designer. I say designer (as opposed to developer or programmer) because nowadays I work more in high level concepts and let guys (and girls) that know what they're doing, do the detail design and coding.
I'm more a jack of all trades and master of few.
While I work hard to learn I'd rather find people that know what they're doing (or are trying harder to learn) than to know everything myself. So, I know a lot of professional experts that I can sometimes persuade to help a little.

I'm here as a hobbyist and not for professional reasons so, another big requirement is to have fun.
I hope to experience the joy of giving back to the community but, these days, I have more cash than time or energy. I'm prepared to throw down, fairly, big time as soon I get a better feel for what's going on.
In the meantime I'll entertain any request for cash to make boards, buy Eagle software, jtags, and the like. Also, I have a fairly huge stash of ST discovery boards, Freescale tower stuff and other development and breakout boards (including a stack of Maple RET6s) that I'm willing to trade, lend, or even give away.

If anyone wants their ears talked off to the point of real pain, just ask me about my projects.
My ex says, "I'm not a good listener" but I'll try to listen to your ideas too and I'm working through every scrap of doco on this site.

I know a guy that's a real, professional, embedded hardware expert. He says that ST is the laggard of the industry on all fronts. I'm trying to get him to show me what competes with a STM32F415RGT6 and why. I assume that the reason to go with an F4 is to minimize the porting effort, have a range of pin compatible chips etc. And, switching vendors will be a hard sell but, I'll be reporting the facts as I acquire them.
While, I'm not an expert, It seems to me the STM32F415RGT6 would be perfect for the Mini I have in mind and will be at least nearly supported by the Leaf community.

At any rate my outlook is much much brighter for having discovered Leaf.
And thank you siy for your reply.

Posted 3 years ago #
MediumKahuna
Member

I looked at Minis on GitHubas as per your link Siy.
Awesome stuff.
There's admirable consideration for making your own boards.
I'm ready to throw down for a nice 4 layer board
I can supply:
Professional skematic review
Professional layout review
Two passes at making a batch of boards
At that point we can get bare boards for $5-$10 a piece (you might have to cut them apart yourself). If we do a nice job and its popular, maybe we can persuade Leaf or somebody to take it on.
I've got enough friends to make that all happen for a few hundred bucks which I'd be happy to donate (And no, I'm not the "professional" above. I'm talking real professional that I'll have to baby-sit his kids and take them to the zoo while he does it).

I'd like to start with one of your designs, kick it around some more and then make boards.
The reason I'd like to go with the 4 layers is to minimize noise getting into the analog stuff and stability.

Some requirements I'd like to see satisfied:
Compact.
Leaf compatible software/dev
Hand assembly (by advanced user)
Hardware floating point.
Quiet analog.
100MB (at least) Ethernet on board possible mini connector.
USB High speed to go (host/client), on board.
Extra RAM
Possibility of Maple compatible carrier board.

I'm not sure of the best way to start a project around here.
I'd be grateful for any advice, including, how to be politic.
I want to do some more research before really shooting my mouth off.

Posted 3 years ago #
siy
Member

The simplest way to start project in "OpenSTM32 way" is to just clone openstm32hw repo at github to repo under your account at github. Then add your new design to your repo and submit (again, via github) a pull request. That's it. All existing designs are covered by quite permissive Creative Commons license, so, you can use one as a starting point for your design.

Few comments on the planned feature list above: satisfying it with 64-pin chip will be problematic, since it has no FSMC. So, as a bare minimum you'll need 100-pin chip. Probably you'll need even larger chip, since some FSMC signals share pins with Ethernet signals in 100-pin package.

Posted 3 years ago #
gbulmer
Moderator

MediumKahuna - no 64pin STM32F4 supports ethernet.

You either need:
a. larger package, and hence a larger board, or
b. STM32F107 if you want to stick with 64pin device, or
c. external ethernet support

... 4 million double precision floating point macs per second ...

Do you need double precision?
If you use single precision floating point, STM32F4 has single instruction, three cycle, 'fused' MAC (VFMA)
http://infocenter.arm.com/help/topic/com.arm.doc.ddi0439c/DDI0439C_cortex_m4_r0p1_trm.pdf
Which would be significantly faster (i.e. theoretical peak at 168MHz/3 = 56 million single precision MACs/second). Obviously loop overhead, and loading data drops that a lot, but still *much* faster than double.

I think, with care, the ADC noise on double layer PCB isn't too awful, and it is *much* easier to test and prototype. When we get moving, my chum Chris and I can make a DIY double-sided PCB, and assemble and test it in a day. Also, Eagle is good enough :-)

I won't be doing much on my board this month, I am making progress using STM32F4Discovery, so I don't need a new STM32F4 board yet.

I am not using an Arduino/Maple header layout so that I can keep it compact, and use breadboard and veroboard for add-on boards. My intention is to make a carrier PCB instead.

When you say 'Hand assembly' do you really mean that? I did that at college in the late 70's using an 8-bit (6800) microprocessor. There weren't many instructions, and converting from assembler to hex by hand was pretty simple (i.e. this is what I understand by "hand assembly"). Nowadays, I'd recommend using an assembler. I can't think of a good reason for doing it by hand, and the GNU ARM assembler is free, and very good. AFAIK, it gives access to all the STM32F4 instructions.

Posted 3 years ago #
Rod
Member

MediumKahuna - My MM4 board is now at rev 3. I was holding off up loading it until its fully tested. The new set of functionality wrt the Maple Mini is as follows:

- Maple Mini 40 DIP pin out, but 0.7" instead of 0.6" width. (To use it I'm typically putting two 20 pin sip connectors down on the right (i.e. at 0.6" and 0.7") and installing either a Maple Mini or a MM4).
- Two sided Eagle layout with a lot of ground plane on one side and a lot of analog ground plane on the other.
- All SMD components on the bottom side of the board (only LED, buttons, Micro USB and 32 kHz crystal on top surface + Maple Mini like pin labels). This is from a suggestion from gbulmer to allow 'toaster oven soldering' and is very similar to the Maple Mini.
- 64 pin CPU pad with solder straps allowing F1, F4 and L1 variants. (e.g. The STMF103RG model has 96Kb of RAM and 1 Mb of on-board flash)
- On board 32 kHz crystal for RTC (PC14/15 are remapped from Maple Mini to allow this)
- On board peripheral (optional stuff): SPI flash to 16 Mb for data logging.
- On board peripheral (optional stuff): precision time base to 0.5ppm.
- The on-board peripherals do not consume externally visible pins.

My plan was to hold onto this until I'm done unless someone else wants to get in on its development...

So, no Ethernet or external RAM, but you can go to 96Kbyte with existing Leaf Labs Maple software (not yet tested).

Posted 3 years ago #
bubulindo
Member

@Rod,

Do you have a board where you can vary the power supply? I made a bit of software for the PVD for the F103, but I haven't got the hardware to vary the supply voltage below the possible selectable levels. So I have no idea if it's good or not. The idea came from the fact that a lot of these STM32 boards have a battery connector to them (didn't check the behavior of the supply regulator in undervoltage situations though) and would be quite handy to know if the supply drops.

If you have one where you can test, I can mail you the file to include and test this.
No hurry though... I'm stuck at sea now and will only make it home on Wednesday.

Posted 3 years ago #
Rod
Member

I do have a variant where I've put a schottky diode after the regulator and a battery input also supplying that point. That board has a bunch of other stuff on it as well, though. The MM4_r3 and its host board are what I'm primarily working on at the moment.

Posted 3 years ago #
MediumKahuna
Member

Thanks for your interest and solid input.
Siy
We've been using Git for about a year now at work (and I can't imagine how I ever did anything before) but, I haven't tried an open hub yet but I'll do just what you said straight away.
Also, the last time I did serious hardware design or layout, we still used paper (yikes).
I'm bought Eagle, I'm getting up to speed and I understand the theory but, I'll never again be anything like good at it.

I haven't read any more than 2000 MCU data sheets in the last few weeks so I should have known we can't get a FSMC on anybody's 64 pin chip,
When I started working on this (a month ago), I was looking for the perfect processor (so I didn't look at ST much). But I soon realized that's its much more important to find the perfect community. Especially since the perfect community exists and I found it right here. The ST stuff looks good to me, but, guys I know that, know say, "ST stands for yesterday's technology, tomorrow". It doesn't matter anyway, the primary requirement is to use what Leaf uses so I'll be learning my ST chips.

Extra RAM can fall off the bottom of the list though.

Ethernet 100 has got to stay but,
if noting else, you can get a $10 ethernet adaptor to plug into the USB to go.

I will eventually write an essay on why I want double precision floating point (which will probably, only, convince me to use it). Most processors that do single precision floating point in hardware will do double in less than twice the time (seems like it should be 4X but its not).
The double precision floating point requirement can be reposed as single precision hardware floating point at 15-20 mega macs.

Might be time to consider 100 pin chips as its starting to add up.

Prototyping of a 4 layer board, I'm ready to contribute.
But, I admit, its a "rule of thumb" preference on my part.
I would do it just to be on the safe side but know better than me.

I've got an F4 Discovery that I could start testing ports or helping in some way.
I'd like to get your result at any rate.
Also, I have a few extra F4 discoveries if know anyone that needs one.

I like the carrier board idea for all the Arduino/maple compatibility stuff.

I built a micro in the 70s (although I couldn't afford a $400 6800 back then) and did that kind of hand assembly too but, I mean hand soldered by someone that can solder SMD.

Rod, let me post what I got so far while I look at your stuff.

Posted 3 years ago #
Rod
Member

If I need casual Ethernet, this is my plan: WIZ820io from Mouser. An 80Mhz SPI peripheral the size of the Ethernet connector. Known to work well. Even when talking to Microsoft. When I need serious Ethernet, the plan is the TCP stack on the STM32F415.

You can say what you like about how far behind ST is, but I'm here because nothing else was remotely close. And I specifically mean TI. Admittedly I was looking at low power, but its an awesome set of chips. Just a truly terrible low level software layer. And I so hate all the expensive programming dongles and proprietary C compilers piled up on my desk.

Posted 3 years ago #
gbulmer
Moderator

MediumKahuna -

... guys I know that, know say, "ST stands for yesterday's technology, tomorrow". ...

I'm curious what it is they recommend?

AFAIK, ST had the first Cortex-M4 silicon (that I could buy).
Freescale were shipping Kinetis development boards for a long time, but I couldn't find any chips for sale.

AFAIK, no one uses more advanced process than ST:
http://techreport.com/discussions.x/19154

I'm interested in robots, audio, and getting beginners to use MCUs, so my focus may be utterly different to you.

Don't assume I am only into ST,
I'm interested in other Cortex-M MCUs, especially NXP LPC43xx.
I like the idea of a 2nd processor to offload work, but I think that combo will be hard to program and debug.
I very much like the look of the 'State Configuration Timer'. I like the sound of the 'Serial General Purpose I/O (SGPIO) interface', but there is nothing in the user manual about it!?
But in many peripherals, the the LPC43xx seems no better than the old STM32F103 parts.
LPC43xx's two 10-bit ADCs are slower than old STM32F103, and its single 10-bit DAC is less useful to me than the STM32F103's stereo 12-bit DACs.
I'm also interested in robots. The STM32F103 are in good shape compared to the alternatives. For example the LPC43xx seems to have a single quadrature decoder, where the STM32F103 has 6 timers which can be used to decode and track quadrature.
I can't find any 1MByte Flash parts for LPC43xx or Kinetis either, which is a bit worrying (I want huge look-up tables).

Cypress PSoC 5, are exciting, but (e.g. CY8C5468AXI) are over $20 each, so a tad expensive.

So, IMHO, at the moment ST look in good shape on Cortex-M4, and the Cortex-M3, especially in 64pin packages (I don't need lots of pins), seems as good, or better, than any alternatives for my use cases.

Posted 3 years ago #
gbulmer
Moderator

Rod -

... And I so hate all the expensive programming dongles and proprietary C compilers piled up on my desk. ...

The price of JTAG seems ridiculous to me. I have an Atmel JTAG which lists at $337 at Mouser!

The STLINK/V2 as a stand-alone JTAG and SWD debugger is under 19 GBP+tax. Much more my price bracket. I could afford a class-room set of those.

I'm very pleased by the on-board STLINK/V2 SWD-only debugger on the STM32F4Discovery. It works perfectly with texane/stlink for both upload and with GDB. I'm aiming to put a SWD header on my boards in future.

And GCC is free! Thank goodness for GNU GCC; IMHO, we'd be in a deep hole without it.

Posted 3 years ago #
gbulmer
Moderator

MediumKahuna -

... hand assembly too but, I mean hand soldered by someone that can solder SMD.

Ooops :-)

My interest is DIY. I try to use 0805 or larger parts, and all SMD on one side to make *hand assembly* easier.
I might eventually do a 1206+ board because the local school children are introduced to 'toaster oven' SMD using 1206 parts. It might be a hoot for them to make an STM32F4 board :-)

Some of the 'whiz-people' ('kids' is ageist :-) can do 0603, and 0402, but I find 0603 tiring. Even Chris prefers a microscope for 0402, and he gets tired, which IMHO puts it beyond most people.

Posted 3 years ago #
MediumKahuna
Member

There's some great stuff already there. Maybe I don't need any more.
Time to refine the requirements:
This is what I would want. Of course I don't know exactly what I want and I know I'll need to compromise toward main stream and toward anybody that works on it.
Anyway, I trust you guys to straighten me out.

First is to get a reasonable ways down my hobby project list, before I die, in as fun a manner as possible.
Two (inferred by One). Make the coolest MCU block I can, that works for those projects and has Leafness.
Leafness pretty much settles the processor issue to STMF4 (I think).
Three, form factor.
Most projects are built by plunking the MCU block on to a proto board with some sensor breakouts, actuator drivers etc or stacked up. The only high speed interfaces are on the mcu block and all the pins under 2MHz (Well?).
Pin spacing is dictated by reasonable availability of proto boards and, I think, it would be nice to have a breadboard-able option (come on, what's more fun than a breadboard). Maybe you'd have to leave out a row of pins to breadboard.
But, if it has to sit next to the bread board, its not the end of the world. I've got a RET6 screwed down (stainless steal) next to a bread board under my elbow right now and its plenty of fun.
The proto board is a must though. I've never seen protoboard that isn't 0.1" spacing, but I've never looked either.
The MCU block gets used everywhere. Retrofitting electronic speedometers into vintage mopeds can get tight. Small size is a wild card. You might not need it but, if you do, there is no substitute. I'm thinking, at least 40 pins but probably not more than 60.
I have one other board I want to make, a motor controller, that I would like to be an extension (off one side) to the MCU Block. This works best if the Block is 2ish inches long and as skinny as possible (an inch would be nice).
To not have some plan for Maple or at least Uno adaption would be sheer arrogance (and not friendly). Only Microsoft would do something like that (and those jerks would do it on purpose). On the contrary, we should at least rough out some carrier/adapter concept that's so slick that somebody else will do it.
I think, put the compatibility stuff (maybe even fidi usb) on the carrier/adapter and blaze a new trail in Mini pinouts. I haven't looked that hard but I don't think there is standard Mini pin out or form factor begging to be followed.
I think the form factor/pin out needs careful consideration. I'm going off to do what Siy told me then, learn Eagle, bring up the existing boards side by side and try to figure out how to organize suggestions (some kind of trade off diagram or something.
Four, Floating point, RAM and other luxuries.
For a significant portion of my illustrious career I was engaged in embedded software development (not at all recently). I routinely stuffed 200lb of poop into 5lb bags often, with great vigor. I'll go heads up with anybody in bag stuffing and do it with somewhat intelligible code. I might have almost enjoyed the challenge sometimes but not any more. Now I want to code with impunity. I don't want to digress and muck things up with work arounds. I want my code to look just like its pure essential object oriented concept. I want to collect garbage in real time. I want to wake up in the morning with an idea and see it walk and talk by noon. I want to free minds including my own. I want to soar...
Well, anyway, the bottom line is, I don't want to stuff any more poop.
I don't want some high zoot processor so I can do do 20X zootness with it. I want a 20X zoot processor so I do my everyday stuff with ease and abandon.
There are substantial, engineering reasons for hardware floating point but, fortunately for you, I'm out of wind.
Five, Ethernet.
I want to use telnet for development. USB stinks for data collection. Ethernet provides essential precise system timing. And no more than 1000 other reasons.

Posted 3 years ago #

RSS feed for this topic

12 3 Next »

Reply »

You must log in to post.