postbag topics

[ag123]

I have what is probably a lifetime supply of F103 blue and black pills. That is not saying much really, as so far in my lifetime I've only used about 1.5 of them. Truth be known, for the sort of things I do, lowly AVR based Nanos or Pro Micros are all I need.

Some time ago I got a couple of F303CC black pills, thinking that a bit more RAM, the FPU, DACs and OPAMPs must surely be useful for the small increase in cost. Then I spied F411CE some-shade-of-black pills and I just had to get a couple of those. They have provision for a SPI flash chip to be added - my SMD soldering skills are practically non-existent, but on a good day I could maybe manage to do that. So perhaps you could say I have switched to the F411, but I have yet to find a real use for them (or the F303s).

switching to f401 and maybe f411 is partly to overcome the 20k sram hard limit. as the 'bigger' stm32f103 e.g. stm32f103rb or re series after all cost more.
i figured that f401 / f411 pill boards happen to be convenient and that 64k sram is plenty more compared to 20k on stm32f103c8. and in addition stm32f401 has an fpu and 'ART' accelerator (on chip cache). app runs quite a bit faster on stm32f401 vs stm32f103c8
some benchmarks like the adafruit ili9341 lcd graphics test would give an idea of how much 'better' it is
https://github.com/ag88/Adafruit_ILI934 ... nd-results
it doesn't look 'a lot' but that every different benchmark shows an improvement, and running various other sketches feels faster as well

one of those things i ran out of sram for is when using stm32f103c8 pill boards as a 'web server', connected across to esp8266 (esp-01) module
e.g. https://www.instructables.com/Getting-S ... 66-ESP-01/
the idea that you can show a web, say the led and you can 'blink' the led clicking the button on the web is still an attractive idea.
and obviously there are more, e.g. a 'web' oscilloscope using stm32's adc, the ideas go on
the trouble is all these 'web servers' are extremely ram hungry and no matter how much ram there is i'd guess it'd never be enough.
just that 64k still feels much breather vs 20k sram

for the f303 i've got one piece gathering dust as well, is its 5 msps x 4 adc giving a max sample rate of 20 msps quad interleaved.
then it is also one of those chips that has in addition comparators and op amps built into it. very much a 'analog' stuff chip.
in the same way i've not played much with that

one of those postbag stuff that may be worth considering is these spi psram
https://lcsc.com/product-detail/RAM_Lyo ... 61881.html
https://www.aliexpress.com/item/1005001958882492.html
https://www.aliexpress.com/item/4000164056594.html
i've one of these similar one soldered on the f401cc black pill board at the spi1 pads below
they kind of give an extra 8 MB of 'ram storage' (not quite memory) for these tiny mcus

[mrburnette]

...
one of those things i ran out of sram for is when using stm32f103c8 pill boards as a 'web server', connected across to esp8266 (esp-01) module
e.g. https://www.instructables.com/Getting-S ... 66-ESP-01/
the idea that you can show a web, say the led and you can 'blink' the led clicking the button on the web is still an attractive idea.
and obviously there are more, e.g. a 'web' oscilloscope using stm32's adc, the ideas go on
the trouble is all these 'web servers' are extremely ram hungry and no matter how much ram there is i'd guess it'd never be enough.

Déjà vu

Were not similar words uttered by users of the early IBM computers in the 1980's?

The problem is that we are in 2021 and are now saying this about microcontrollers! This should not happen.

I just had this exact conversation with a PhD friend working with robotics. The issue is building and rebuilding and repurposing hardware to "save time" on reworking existing code! We're friends and I'm sure he will eventually get over me calling him foolish.

One must use prudent engineering principles to review project requirements and size/select the appropriate hardware. "Scope-creep" happens even when the programmer and customer are the same. There comes a point of diminishing return when one must apply good judgement and move to the next discrete level. This may mean more capable hardware and often software refactoring or a major rewrite.

There are reasons for large, powerful 32-bit microcontrollers with a MByte of RAM, but blinking LEDs is not one of them. Sometimes a Raspberry Pi running a commercial tool such as NodeRED makes the microprocessor implementation easy.


Good capable hardware is readily available and at low to reasonable prices. Think before selecting "cheap."

[ag123]

the rp2040 should fix that given it has 264KB sram

[BennehBoy]

I decided that I'd improve visibility for my in car multigauge system by moving to an SSD1309 instead of an SSD1306.

2.42" vs 0.96"

~20GBP vs 3GBP, but a lot more visible.

The size does make the lack of resolution a bit apparent though, still 128x64 pixels.

I may yet move to some form of colour TFT, but this is literally a drop in replacement for the 1306, so a quick hardware update.

[mrburnette]

Yippie

Got my new ESP32-C3 DevKit order of 3 of the little suckers. Now, to get familiar (again) with the IDF ...

ESP32-C3_DevKit.jpg (91.52 KiB) Viewed 5134 times

[mrburnette]

I do not purchase new notebooks ... like expensive cars, they depreciate too quickly. My development notebook is an older i7 HP EliteBook and it came with a 5400 RPM commodity drive - I replaced that immediately with a (at that time) fast 480GB SSD ... that was years ago.

The SSD has been good, no errors have been reported in a couple of years of compiles, downloads, links, and un-Zipping. While the notebook is still reasonably peppy, I saw a great deal on
1TB Hard Drive - 7200 RPM SATA 6Gb/s 64MB Cache 2.5 Inch - WD10SPSX

I love it and it is outperforming the SSD. (Especially with writes over existing non-empty directories; such as un-zips.)

WD_Black_1T.jpg (52.59 KiB) Viewed 5056 times

The older SSD has become a backup drive in an $11 USB 3 enclosure.

Ext_SansDisk_SSD.jpg (87.79 KiB) Viewed 5056 times

[BennehBoy]

So had a bit of a diversion recently as I was gifted both a Commodore 64 (breadbin version), and an Amiga 500.

Have had fun recapping both, and fixing keyboard issues.

Lot's of things arrived in the post, but one that I thought might be of interest to those in this forum, is a PCB for a 'Greaseweezle'.

See here -> https://github.com/keirf/Greaseweazle

Anyhow, this is a flux level floppy disk tool that uses either an F1 or F7 microcontroller.

I've built 2 using bluepills that were kicking about - please forgive the horrible soldering!





This turns the bluepill into a shugart capable controller and any 3.5" floppy drive can be connected (best to power the floppy from an ATX/AT PSU since the pill can't provide 12v for any drives requiring it).

They work really well, and I've used them to burn several Amiga Disk Format (ADF) snapshots downloaded from the internet onto floppy disks - the first one being a full diagnostic kit for the 500's hardware - which incidentally helped diagnose some guru meditations caused by a faulty trapdoor expansion 512KB of ram.

Fun & games - literally

[fpiSTM]

please forgive the horrible soldering!

I can't

[BennehBoy]

Ha,

Well to be fair the bluepill headers were soldered on a number of years back when I first got back into electronics. The soldering on the greaseweezle PCB is 'better' and I at least cleaned off the flux residue with IPA. TBH I'm not too careful with this cheaper stuff, the C64 & A500 had much more care taken over the replacement capacitors.

My iron is also a bit limiting, it's a cheapo 'Maplin' temp controlled one with pretty crappy tips, I'm currently dithering over buying a Hakko FM203 (so I can get some hot tweezers too!).

That's my excuse anyway

[fpiSTM]