LeafLabs Garden

A place to share, learn, and grow...

1. 

   dreamcat4
   Member

   Soon the STM32F103RGT6 (1Mb Flash 64 LQFP) will be coming up to volume. Which is great since that aught to be compatible with RET6 design.

   However after this we don't seem to get much more significant improvements from ST on the STM32 Series 1 range. They will be (are already) promoting their STM32 Series 2 products and the first of those shall start to become available. The F103 is already a very good microprocessor. What I am really excited about for these STM32 F-2 chips is:

   They will be the first ones to feature builtin 10/100 Ethernet! Which is currently the most compelling reason I can think of for adding STM32 F2x support into Maple IDE.

   STM32F207 or STM32F217. Specifically the STM32F207VGT6 or STM32F217VGT6.

   http://www.emcu.it/STM32/STM32F2xx/STM32F2xx.html

   Also noteworthy: F207 requires a move away from 64LQFP optimum size. These ethernet variants are all 100 LQFP minimum.

   To find on ST website:

   http://www.st.com/internet/mcu/subclass/1169.jsp?WT.ac=micro_stm32f2_011210

   For the filter options select:
   LQFP100, LQFP64, 72Mhz, 120Mhz, Flash size 512,768,1024

   Posted 4 years ago #
2. 

   poslathian
   LeafLabs

   oh yes, we are very excited about the F2xx series. The great thing about these chips is that while the specs are being upgraded and new peripherals added, its not a completely new line. Its still good old stm32. So we will be transitioning smoothly to the new chips when they come out. Dont hold you breath, though, it takes quite a while for these sorts of chips to hit the price point where hobby hardware can really take advantage.

   But as you correctly anticipated, the Maple line has a lot of steam left in it thanks to the F2 series!

   Posted 4 years ago #
3. 

   gbulmer
   Moderator

   According to ST, the Cortex-M4 will be (mostly) pin compatible too.
   That runs even faster and adds hardware floating point. Sampling summer.

   Posted 4 years ago #
4. 

   dreamcat4
   Member

   ETHERNET:

   The consideration about the 100 LQFP minimum informs future PCB design for Maple (or compatible) boards. It will make sense to design the ethernet circuit onto the same board. Which can go unpopulated if not needed.

   Sorry - I appear to have missed out the STM32F107 in all of this. It probably wasn't a big seller because ST hasnt made one with more memory. What a shame. There was some discussion about it elsewhere on the forums. A part with fewer memory is currently available - STM32F107VCT6.

   This is great news because it means that ST isn't really stopping us from starting development for the necessary board + networking software library. The only thing that may go wrong here is if ST change their minds and don't release those expected MCU variants. If we design a new 100 LQFP board, it can work for the F-2 series. But we are expecting ST to come through and deliver those specific F-2 variant(s) in quantity. What are other people's views on such expectation?

   For getting started early on networking library, a 64 LQFP chip (the STM32F107RCT6) is also currently available. Being 64 LQFP means its pin-compatible to fit onto an existing Maple Rev 5 board. However I don't think anyone has yet checked / confirmed that those specific pins for Ethernet are broken out and represented on Maple Rev 5.

   http://forums.leaflabs.com/topic.php?id=560

   Elsewhere are parallel software efforts for Arduino ethernet shield (Wiznet 5100 chip).

   http://forums.leaflabs.com/topic.php?id=244

   Supporting the Wiznet 5100 chip are not what we are intending ^^ above. But they may have included LwIP stack or other reusable API code in their efforts.

   Posted 4 years ago #
5. 

   gbulmer
   Moderator

   dreamcat4 -
   

   Sorry - I appear to have missed out the STM32F107 in all of this. It probably wasn't a big seller because ST hasnt made one with more memory. What a shame.

   Maybe the timing of the move to STM32F2 influenced them.

   The STM32F207, which includes Ethernet, has part numbers which range from 256K to 1MB Flash:
   http://www.st.com/internet/com/TECHNICAL_RESOURCES/TECHNICAL_LITERATURE/DATASHEET/CD00237391.pdf

   I agree, if you want Ethernet in the future, then it has to be at least LQFP100; as you say the STM32F207 has no LQFP64 part listed yet.

   I believe Adam was working on LwIP.

   There is an application note An3102 from ST on using LwIP on the STM32F107:
   http://www.st.com/internet/com/TECHNICAL_RESOURCES/TECHNICAL_LITERATURE/APPLICATION_NOTE/CD00255062.pdf

   Posted 4 years ago #
6. 

   dreamcat4
   Member

   F205's are now available from digikey.

   http://search.digikey.com/scripts/DkSearch/dksus.dll?vendor=0&keywords=STM32F205RGT6

   http://search.digikey.com/scripts/DkSearch/dksus.dll?vendor=0&keywords=STM32F205VGT6

   If you look on the digikey page, it says VGT6 has 132Kb ram. Is that correct? I thought they had 60Kb.

   There are also some cheaper F205s with less onboard memory (512,256, and 128Kb flash) too. Depending how fast digikey sells their initial stocks will probably dictate how many they try to re-order next time. The launch price (for the 1Mb part) is around $15-16 dollars per unit. Quantities of 100 units or more makes it about $10 dollars.

   Arrow will also be stocking 1Mb VGT6. At a lower single-unit (1xQTY) price of only $10 (compared to $16 at Digikey).

   http://components.arrow.com/part/detail/49663162S8891564N7728

   However those are not available quite yet. Arrow says "lead time 7 weeks" to receive 440 of the VGT6. So presumably that figure reflects when we can see a second batch of 205s and perhaps Digikey may get some more around that time too. That date works out to be around the 16th August.

   Posted 4 years ago #
7. 

   gbulmer
   Moderator

   dreamcat4 - Well spotted, I wondered when they might appear.

   132 kBytes looks right:
   http://www.st.com/internet/mcu/product/245091.jsp

   It is 128 kB RAM plus 4 kB battery-powered backup RAM
   It is even sneakier because the 128 kB RAM is in two independent blocks of 112 kB and 16 kB. I believe it is possible to have a DMA transfer to one block while the processor simultaneously uses the other block, with no stall.

   At $10 for 1 MByte Flash, and 128kByte RAM + 4kByte backup is pretty impressive for an STM32F103, but that STM32F205VG is bonkers: 14 timers, external memory interface, USB OTG HS, CAN, I2S, ...
   It's a significant step up from a STM32F105, even ignoring the 72MHz to 120MHz speed-up.

   Posted 4 years ago #
8. 

   dreamcat4
   Member

   The good news is that an F2 RGT6 chip could work on a Maple RET6. And similarly, Maple Native.

   The bad news is that ST arent releasing any 48 pin LQFP chip. So nothing for the Maple mini. This is a shame because I always thought the Mini was a much improved design over all the other ones :(

   Posted 4 years ago #
9. 

   gbulmer
   Moderator

   An STM32F2xx is not a drop-in replacement for an STM32F1xx.

   The STM32F205xx STM32F207xx datasheet:
   http://www.st.com/internet/com/TECHNICAL_RESOURCES/TECHNICAL_LITERATURE/DATASHEET/CD00237391.pdf

   Section 2.1 "Full compatibility throughout the family" says:
   "The STM32F205xx and STM32F207xx devices maintain a close compatibility with the whole STM32F10xxx family. All functional pins are pin-to-pin compatible. The STM32F205xx and STM32F207xx, however, are not drop-in replacements for the STM32F10xxx devices: the two families do not have the same power scheme, and so their power pins are different."

   So the existing STM32F103 PCB's won't work :-(

   Posted 4 years ago #
10. 

    dreamcat4
    Member

    Yes... and the next line reads:

    "Nonetheless, transition from the STM32F10xxx to the STM32F20x family remains simple as only a few pins are impacted."

    Posted 4 years ago #
11. 

    becker
    Member

    They changed a few of the +3V power pins to instead be capacitor connections to the internal power layers. It's pretty easy to change an existing circuit board layout to work with the new chips.

    I was hoping to take a knife to one of the boards we designed and modify the power connections so that we could try out one of the new chips. But the power and ground planes are very well connected to the pads with a "copper pour". There is enough area that any cut would likely bridge with solder. My guess is that most other boards are similar and will need a re-spin.

    I'm looking forward to CAN bus being simultaneously usable with USB. I'm not expecting that Ethernet will be generally useful. The memory is just too small to do more than simple UDP reporting.

    Posted 4 years ago #
12. 

    gbulmer
    Moderator

    dreamcat4 - I apologise if I was unclear.

    You wrote

    The good news is that an F2 RGT6 chip could work on a Maple RET6. And similarly, Maple Native.

    according to the section of the STM32F205xx/STM32F207xx datasheet THAT IS NOT TRUE

    As becker has explained, several power pins are connected to the internal power system, and need extra decoupling capacitors, which are not on the PCB for Maple rev5.

    Posted 4 years ago #
13. 

    dreamcat4
    Member

    Its true that an F2 chip could work on a RET6 board, with only a slight modification. So I believe I'm absolutely correct in saying that it COULD WORK. According to the next page in the ST Manual it shows the 64 LQFP package has only one decoupling capacitor. I'm not claiming that you wouldn't need to modify the board slightly - you would.

    Posted 4 years ago #
14. 

    gbulmer
    Moderator

    dreamcat4 -

    Its true that an F2 chip could work on a RET6 board, with only a slight modification.

    Thanks for the clarification about modifying the board. I interpreted your previous comments as using an unmodified board. This is clearer.

    According to the next page in the ST Manual it shows the 64 LQFP package has only one decoupling capacitor.

    The datasheet for the STM32F205xx STM32F207xx:
    http://www.st.com/internet/com/TECHNICAL_RESOURCES/TECHNICAL_LITERATURE/DATASHEET/CD00237391.pdf
    in section 2.1 "Full compatibility throughout the family", Figure 3. "Compatible board design: LQFP64", shows TWO decoupling capacitors added to the 64pin LQFP STM32F20xxx, and two breaks in the Vss tracks.

    The two pins, 31 and 47, change from Vss_1 and Vss_2 to Vcap_1 and Vcap_2 respectively.

    The decoupling capacitors are inserted by breaking the direct connection from pin 31 and pin 47 to Vss, which is ground, not as becker says to +3V.
    Most PCBs for processors which run at 20MHz+ have a 'copper pour' ground plane which is more difficult to modify than signal tracks.

    On a 64 pin LQFP STM32F103, the neighbouring Vdd pins 32 and 48, are normally connected to pin 31 and 47 via a decoupling capacitor which are (recommended to be) very near to both Vdd & Vss pins.
    My interpretation of section 5.1.6 "Power supply scheme" is those Vdd pins will still need to remain connected to their decoupling capacitors

    So the modification needs to leave the existing (STM32F103) decouping capacitor connected to Vdd and the ground connection, so there isn't much room to manoeuvre.

    On Maple Rev 5, pin 31 is connected to Vss (ground) using a via UNDER the chip, so the chip would need to be removed from the PCB to sever the track, or the via might be removed by very carefully drilling out (it isn't a blind via) but this can be tricky as a 'wisp' of copper may be left to short between layers. Pin 47 is a bit easier as the connection to Vss is a short visible track to a via.

    Once the tracks are broken two new, extra, decoupling capacitors need to be connected to the pins and Vss (ground).

    Having said all this, I am willing to believe that you are sufficiently skilled to be able to make the modifications. I don't count those as slight modifications on an assembled board, but maybe you do. I'll be interested to see how you do it.

    I am posting this to ensure that other folks realise it is NOT likely to be straightforward on a PCB which is not designed to support the modification.

    It would be relatively straightforward to design a PCB to take an STM32F20xxx, and make it backwardly compatible with an STM32F103xx. I might do that.

    Posted 4 years ago #
15. 

    becker
    Member

    I'll second that, there are few but very difficult modifications.

    I looked at our board design. compared it to the new power layout, and decided that it would be very difficult to modify with a knife and wire. And further, that any board not designed to support both chips would be similar to ours, and similarly difficult to modify.

    It's only a matter of a few minutes to modify the board layout in Eagle. Except that will lead to a few hours of other changes I've wanted to make.

    Posted 4 years ago #

RSS feed for this topic

Reply »

You must log in to post.