Alternative BluePill sized F4

racemaniac
Sun Oct 15, 2017 11:00 am
In this thread we’re designing an F405 board that is pin compatible with a bluepill (and with some nice extra features).

But for myself, i was wondering: if i get a more powerful chip, what would i want? one of the main things would be SDIO (but those pins aren’t broken out on the bluepill clone). For fabricating, <5cm length would be ideal (but the bluepill is slightly longer due to its 20 pin rows), and the huge boot header & the 32Khz crystal are nice to have, but not necessary.

This is my first attempt at something slightly smaller than a bluepill (4.85cm long), the pins are further apart (on a breadboard, you’ll have 1 row of pins free on 1 side, 2 on the other), so i could fit a micro sd card slot in between the pins. No 32Khz crystal, but the pins are available, and the capacitors for it can be fitted (so just take one of the cylidrical ones, and put it on the pins). And except for the pins used for SDIO (PA8, PC8-12, PD2), all pins are broken out (PA8 was chosen for the card detection pin, as PA9-12 are USB, so most of the higher 8 PA ports is already in use for an important peripheral).
Besides that the features would be the same as the other one (dual power supply so the analog rail is separated, and if we figure out a good way: USB OTG support) :

BPF4SDFront.jpg
BPF4SDFront.jpg (72.98 KiB) Viewed 887 times

Pito
Sun Oct 15, 2017 11:47 am
1. use an 800mA voltage regulator for Vdd
2. do use more vias (heat, grounding)
3. wait on stm32H743/753 in 100pin flatpack

racemaniac
Sun Oct 15, 2017 11:56 am
[Pito – Sun Oct 15, 2017 11:47 am] –
1. use an 800mA voltage regulator for Vdd
2. do use more vias (heat, grounding)
3. wait on stm32H743/753 in 100pin flatpack

you don’t think 600mA is enough?
and you’re expecting heat issues?? (i’ve only got experience with the 72/84 Mhz microcontrollers, and heat doesn’t seem to be one of their issues ^^). or do you mean heat at the voltage regulators (i’ll have to check their datasheet on how hot they can get)


Pito
Sun Oct 15, 2017 12:00 pm
Heat around the voltage regulators. You have got a large gnd plane beneath them, but you do not have any vias for a thermal connection to that plane (the heatsink). Calculate the temperature of the Vdd chip (look for the topic I posted few weeks back). That sot23 package will smoke with 5V input and >=200mA output.

racemaniac
Sun Oct 15, 2017 12:12 pm
indeed good idea :)

bobc
Sun Oct 15, 2017 12:20 pm
[Pito – Sun Oct 15, 2017 11:47 am] –
3. wait on stm32H743/753 in 100pin flatpack

Oh yes, that would be awesome :D

Haven’t looked, maybe there is some pin compatibility? Even if not, experience with F4 should help with H7 design.


Pito
Sun Oct 15, 2017 12:26 pm
[racemaniac – Sun Oct 15, 2017 12:12 pm] –
indeed good idea :)

Do use a larger package with Vdd regulator (ie SOT-223 or Dpack – a package with a heatsink fin) – with large copper around/beneath the heatsink fin. Use a lot of vias to thermally connect the top/bottom copper.
Read the chapter 10.1. carefully:
http://www.ti.com/lit/ds/symlink/lm1117.pdf


racemaniac
Sun Oct 15, 2017 1:15 pm
btw, i’ve done some reading about the absolute maximum power conditions i can expect:
the stmf4 is rated to draw max 240mA, and sd card writing full speed probably max 100mA
The f4 running full speed with all peripherals will require 100mA (and then has 140mA extra to drive things with on its pins). and then +10mA for the 2 leds.

350mA from that [email protected] would mean it heats up to about 100°C at ambient temperature, which would be well within its max rating (160° i think according to the datasheet), and is an absolute max ratings for everything on the board (but that’s then of course not calculating the various modules people could connect to the board…).
I’d consider changing it to something else if i found something with better thermal properties and an equally low voltage drop. Moving to an LM1117 (or an AMS1117) would mean it not being able to properly run of a lithium battery without first boosting the voltage >_<…

but it’s an interesting point. it’s a difficult design tradeoff to make… would 2 of those ap2112’s in parallel work (i would say no as one of both might end up doing all the work and burning out if it has small differences to the other one). anyone got experience with that?

i’ve just made a testsetup here with a maple mini with a ap2112 doing 270mA with 4.6V in, and it’s handling it graciously so far ^^


Squonk42
Sun Oct 15, 2017 5:42 pm
(4.6 – 3.3) * 0.27 = 0.35W
(5.0 – 3.3) * 0.35 = 0.6W. Assuming a SOT25 with 184°C/W for tja, this is 0.6 * 184 = 110°C above ta, so you fry your chip if ta > 40°C. Not so unlikely in temperate countries behind a car windshield… Anyway, going to these extreme temperatures will decrease substantially your chip lifetime, beside being dangerously hot.

Putting LDOs is parallel is calling for disaster.

Why don’t you use the bigger AP2114 in a larger package like DPack or D2Pack? It is rated at up to 1A with lower dropout voltage, same Iq, same PSRR and better thermal resistance, down to 90 or even 73°C/W for these packages?

Is it your final layout? If so, there is something really strange above the reset button (top side): why are there vias on both the 2-signal group and the 3-signal group? It is completely useless.

I also suggest moving all the passive components on the top side (except for the SD connector), such that in the case this latter is not mounted, the bottom side is flush. I also suggest turning the SD Card connector 180°, so that the card exit faces the short PCB edge.

For silkscreen text, every character with height below 32 mils (0.8 mm) won’t be printed correctly with standard PCB manufacturing quality.


racemaniac
Sun Oct 15, 2017 6:46 pm
[Squonk42 – Sun Oct 15, 2017 5:42 pm] –
(4.6 – 3.3) * 0.27 = 0.35W
(5.0 – 3.3) * 0.35 = 0.6W. Assuming a SOT25 with 184°C/W for tja, this is 0.6 * 184 = 110°C above ta, so you fry your chip if ta > 40°C. Not so unlikely in temperate countries behind a car windshield… Anyway, going to these extreme temperatures will decrease substantially your chip lifetime, beside being dangerously hot.

Putting LDOs is parallel is calling for disaster.

Why don’t you use the bigger AP2114 in a larger package like DPack or D2Pack? It is rated at up to 1A with lower dropout voltage, same Iq, same PSRR and better thermal resistance, down to 90 or even 73°C/W for these packages?

Is it your final layout? If so, there is something really strange above the reset button (top side): why are there vias on both the 2-signal group and the 3-signal group? It is completely useless.

I also suggest moving all the passive components on the top side (except for the SD connector), such that in the case this latter is not mounted, the bottom side is flush. I also suggest turning the SD Card connector 180°, so that the card exit faces the short PCB edge.

For silkscreen text, every character with height below 32 mils (0.8 mm) won’t be printed correctly with standard PCB manufacturing quality.

the ap2114 indeed seems great, but is a lot more expensive, but indeed maybe worth it :)
gonna have a look at it :).
there is only 1 seller of it on aliexpress, about 0.7$ a piece (10 times more expensive than the 2112, but he’s only selling per 5), but it looks like a great regulator indeed.

about the sd card slot, i put it like that on purpose but indeed, other orientations are also possible.

i think most silkscreen is 0.8-1mm, i usually go for dirtypcb where they specify the silkscreen lines have to be at least 0.15mm thick, and KiCad can go to 0.7mm with that line thickness.


Squonk42
Sun Oct 15, 2017 7:41 pm
You may use the now obsolete IRU1206-33CY, also 1A, $11.15 / 50 PCE:
https://www.aliexpress.com/item/sku/32485042169.html

What about the strange traces I mentioned in my earlier post?


racemaniac
Mon Oct 16, 2017 7:59 am
The top & bottom plane are VCC & GND, so it’s where i switch between layers to allow the fill to pass by those traces
and this is not the final layout. it was a first go at seeing of something like that would work :). and also to get some feedback of what you guys think :).

Ollie
Mon Oct 16, 2017 4:55 pm
I like this idea. There are several cases where I would like to have something between blue pill (103C8) and blue pad (407VE). When and how I can buy one. Initially, the price is not critical for me.

racemaniac
Mon Oct 16, 2017 6:03 pm
[Ollie – Mon Oct 16, 2017 4:55 pm] –
I like this idea. There are several cases where I would like to have something between blue pill (103C8) and blue pad (407VE). When and how I can buy one. Initially, the price is not critical for me.

Same here, as i’m learning how to design boards, it’s fun to see what would be possible in the range of boards i’d actually like. i love the bluepill/maple mini size, and i’m wondering if we can’t pack more power into that size. But i’m still pretty new at this, so maybe after building & using those boards, i’ll discover why they’re not being made ^^’.
i’m also wondering what other’s capabilities are on this forum, if it just ends with a nice design for this, will people have them produced at one of the boardhouses & assemble a few themselves? (it’s easier than i used to think :).)
it’s a fun road to go down, making your own boards, and seeing the f4 bluepill thread, i’m wondering where this will lead :).

and i also like the feedback on the design (like the voltage regulator)
aliexpress can be *very* frustrating in finding good parts >_<…


victor_pv
Tue Oct 17, 2017 4:52 pm
Racemaniac, this article from Linear can provide some information on using several LDOs to share the load:
http://www.linear.com/solutions/5759

About voltage regulators, I just noticed this one in Mouser, listed as new:
LDL1117S33R

Only 250uA quiescent current, and 350mV dropout

Still the AP2114 Squonk pointed too seems like the best for a drop in 1117 replacement.


Pito
Fri Oct 20, 2017 8:00 pm
The problem with the original design is not the power dissipated itself, but the package type which has got a large thermal resistance so it will smoke at Iout >200mA. A sot-223 or sot-89 or dpack is a better choice, but still the power must be dissipated somewhere.

If you chase a small size and large currents and less heat then the only solution is a switcher.. The switcher is more complex and adds noise which is not easy to filter.


Ollie
Sat Nov 04, 2017 3:16 am
There was a reference to 100 pin STMH743 as the next step after F4. Here are some reference points of prices:
1. STM32H743ZI Nucleo board
– I got mine today and paid $23 at Mouser
– After the purchase, I have seen the prices as low as $19
2. STM32H743ZI chip (144 pin)
– Directly from ST for $8.90 (in lots of 10,000)

Leave a Reply

Your email address will not be published. Required fields are marked *