STM32F217 Pinout

[marcot]

Posted on May 13, 2011 at 20:22

STM32F217 Pinout

[frankfrank9]

Posted on May 17, 2011 at 14:34

You're right, the pin muxing system, while better than nothing, is still quite limited. Particularly for the analog signals. There is no muxing at all for the two DACs for example - their pin numbers are fixed. I think the pin muxing is more for the digital signals than for the analog signals. I know on our board we had to give up a feature because the pin muxing simply wouldn't allow it. It's nice to have the flexibility we have, but of course we always want / need more.

[Tesla DeLorean]

Posted on May 17, 2011 at 14:34

For analogue you have pin drivers that are specifically designed/characterized.

The muxing occurs behind the pin drivers, you don't move the pins/drivers. Yes, its impossible to use all the peripherals together, you have to strategically map out the functions you want and how to escape them from the part. Which peripheral instantiation, which mapping, or remapping. Some vendors have fitting tools, I don't think ST does, but anyhow it is a useful due diligence task to perform or model. Most people aren't coming close to utilizing half the peripherals, and most can't deal with high ball count, fine pitch BGA's either. Most of these, and similar class products have limited muxing options, you are not going to find parts with huge cross-bar switches capable of routing any signal to any pin. Such structures have problems with speed, size and cross-talk. If you need more control, consider an Cortex M3 with an FPGA fabric. You don't have to use USART#1 for USART#2 to work. You can remap timers off the pins if you just need them to generate internal timing

[marcot]

Posted on May 17, 2011 at 14:34

What breaks my heart is that even with the 176 pins case, many functions are mutually exclusive.

It is a shame that the design team worked so hard a producing these marvelous functions just to have their use so limited by the lack of ''distribution'' of the functionalities on the pins.  All of that after spending manu pages on the datasheet explaining how they have these great multiplexers to avoid this problem. Somebody didn't think this through properly.

[marcot]

Posted on May 17, 2011 at 14:34

I understand that the analog pins are special but at least bring enough out to make it useful and don't use the same pins for the three A/D converters.  This makes them mutually exclusive as far as the channel count is concerned.  I need to add external multiplexers to obtain the channel count needed when the chip has already all the mux needed internally.  They are just assigned to the same set of pins making them useless.  What a waste.

I am aware that there are usually some compromise to be made.  The full crosspoint switches might be an over-kill but ST's ''pin layout'' people have been greedy in this case. I have been spending a day re-configuring my IOs under various combinations and still cannot get everything I need out.  This is with the 176 pins case.  While many of the lower pin numbers are over-loaded, the higher numbers are under used.  It would have been easy to reproduce the lower pins crowded functions onto the higher pin to give a choice to those of us who are ready to pay more for a full set of functions in a larger case.  To be fair, they did it a little with some incomplete timers but they should have been more methodic about it.  It looks more like an aftertought than a real resource planning effort.  Since the F200 is a second generation product, they should have seen this weakness in the previous generation and correct it. Since I am at the evaluation stage for the processor to be used on my new product line, I am going to look somewhere else because while the this chip is impressive on the functionality list, its implementation is very limited.  It presently does not meet my requirements. Thank you for your help.

[Tesla DeLorean]

Posted on May 17, 2011 at 14:34

No doubt, and I agree with you to a point. A better pin muxer would help escape the required peripherals, but complicate the configuration significantly.

But the challenge is to make a die that fits in multiple foot-prints, and can be manufactured in enough volume to be viable in the market as a general purpose, low cost device. It's not a unique problem. And it's compounded by backward compatibility, and bond out issues. People with specific needs and volumes use SoC type designs, ST will no doubt entertain such requests if you are buying 10-100's of millions of parts. Or might steer you to the Spear parts with a gate-array fabric for customization.