cren - "Why keep the MCP1703?"
I wanted the reworked mini to get as close to Maple-mini performance as practical. If the ADC noise on my design is inferior, I'd prefer to know its because of the PCB, and not be left wondering whether its the analogue supply, or the PCB.
"How much Vripple is on the LM1117-3.3? Does the difference in Vripple make a huge amount of difference?"
I am assuming LeafLabs did some measurements, and it makes a difference.
However I think its worth designing and making a board without it, to get some facts.
"Could you ditch the MCP1703 for just a LC filter and use that for AV+ instead of just Vcc directly?"
That is certainly worth trying.
On my larger LQFP64 board I had ferrite beads on both ground and 'Vin' into the analogue supply, but I removed the 'Vin' ferrite because I assumed the regulator would do a better job anyway.
Edit: I'd like to 'ditch' the MCP1703 for several reasons. It might be possible to reduce the header spacing of the 1206-based wide mini from 0.8" to 0.7" if I can use fewer the parts. The MCP1703 is another piece of active circuitry to test (and fail), and it's a relatively expensive part (though that only really matters for very low cost parts like STM32F051 or bottom end STM32F1xx). The MCP1703's function is to improve the ADC behaviour, but much of the time, the bottom 3 or 4 bits of a sensors ADC reading is noise.
"Lastly regarding analog, why is L1 even there?"
The intent is to segregate analogue ground, and reduce the noise on ground reaching the ADCs.
"Why is D2 there?"
I've just realised, the diodes have different names on the two boards. I'll go fix that later today to reduce confusion.
If you're looking at the mini, then D2 is on Vusb. In that case it is simply there to protect the Host USB socket from a higher voltage applied to Vin. This allows the user to have the board in a circuit, with external power, and then plug in USB to program it. This seems to be a common use case.
If you're looking at the wide-mini (mini-w), then D2 is on Vin. In that case there are two reasons:
a) To protect the board from reversed polarity external power. AFAICT neither regulator has reverse polarity protection.
b) To protect the external power (Vin) from Vusb. Without that diode, the external power supply would suffer about +4.6V when USB is plugged in.
"I don't think it would be of any use and actually an annoyance like the original Maple Mini r2 where you're wanting to get Vusb to power an LCD, but can't."
I think it does serve a function in both cases. Maybe you could build the board, and replace the diode with a 0 ohm resistor.
I have been thinking that it might be useful to have a 'raw' voltage connection. It could either be after the polyfuse, or after the two diodes. After the diodes is useful because that voltage is available regardless of the power source. Before the Vusb diode, and after the polyfuse sounds like the voltage you are describing.
A place to put an alternative voltage would be 'av+' or 'av-', and leave the other pins unchanged.
However, exposing a different voltage on a header pin makes the board incompatible with Maple-mini, and I am trying to avoid that.
"At a later stage, would it be possible to revamp the board to use 0603 components to obtain a board similar size to the current mini?"
Is it critical to use 0603? If it is, then use Siy's mini48. My aim is to design boards which are easy enough to make that my old doddery body, and newbies, can assemble them :-)
However, the board I showed in the first article Improving Maple Mini (part 1) is identical physical pin spacing, 2 x 20 0.1" pitch header pins 0.6" apart. It is also signal and voltage identical to the Maple-mini. It uses 0805 parts, which are hopefully even easier to handle than 0603. I hope to come back and 'finish' that one later this week. I am planning on getting both designs made at the same time to reduce cost.
"Either way, liking the look of the board :D Great work everyone.
Thank you very much for the encouragement, and thank you for the helpful feedback and questions.
"I would like to see STM32F303 components in the future, but for now, getting something working the same would be good enough."
I had thought ST had 'given up' on LQFP48 for high-end parts, but the availability of LQFP48 STM32F303's revived my interest, and hence these designs. I wish they'd do a 28pin DIL, or 20pin SOIC part :-)
So, YES, the further goal is to make STM32F303 boards!
However, as you say, getting something working would be good enough for now :-)