It works! It works!

on a messy desk, a small USB midi keyboard is connected to a Korg NTS-1 mini synthesizer via a small micro-controller board that acts as a USB host for the Akai keyboard, converting USB MIDI to traditional MIDI for the Korg
Akai LPK25 keyboard has USB MIDI out, but the Korg NTS-1 only has regular MIDI in. The little board in the middle acts as a USB host for the Akai and MIDI source for the Korg

This is great: gdsports/midiuartusbh: MIDI DIN to MIDI USB Host Converter allows your USB MIDI instruments to act as traditional MIDI controllers. It uses a Adafruit Trinket M0 to act as the USB host and MIDI output.

I modified gdsports’ design very slightly:

  1. Instead of using a 74AHCT125 Logic level converter and driver, I used a FET-based SparkFun Logic Level Converter
  2. Instead of a 5-pin DIN socket, I used a 3.5 mm stereo socket.

And it works!

breadboard showing Trinket M0 microcontroller board, logic level shifter, audio socket breakout and two resistors
Breadboard layout for MIDI-standard 3.5 mm output (Korg). The resistors are both 220 ohm, and the boards need 5 V power

Mystery lockdown phone message

“Just a test call. Time to stay home. Stay safe and stay home.”

This message from an unknown caller has sat on our landline answering machine since 2020 or 2021. No idea who or what sent it. All I know is it came in just before noon on a Tuesday morning. The entirely synthesized voice makes me think it’s a junk call, but there’s no scam attached. Just this message, slightly eerie, quite inexplicable.

The Quite Rubbish Clock, mk.2

this is bad and I should feel bad

In early 2013, I must’ve been left unsupervised for too long since I made The Quite Rubbish Clock:

It still isn’t human readable …

Written in (Owen Wilson voice) kind of an obsolete vernacular and running on hardware that’s now best described as “quaint”, it was still absurdly popular at the time. Raspberry Pis were still pretty new, and people were looking for different things to do with them.

I happened across the JASchilz/uQR: QR Code Generator for MicroPython the other day, and remembered I had some tiny OLED screens that were about the same resolution as the old Nokia I’d used in 2013. I wondered: could I …?

small microcontroller board with USB C cable attached and an OLED screen on top. The OLED is displaying a QR code which reads '172731'
OLED Shield on a LOLIN S2 Mini: very smol indeed

The board is a LOLIN S2 Mini with a OLED 0.66 Inch Shield on top, all running MicroPython. One limitation I found in the MicroPython QR library was that it was very picky about input formats, so it only displays the time as HHMMSS with no separators.

Source, of course:

# -*- coding: utf-8 -*-
# yes, the Quite Rubbish Clock rides again ...
# scruss, 2022-06-30
# MicroPython on Lolin S2 Mini with 64 x 48 OLED display
# uses uQR from https://github.com/JASchilz/uQR
# - which has problems detecting times with colons

from machine import Pin, I2C, RTC
import s2mini  # on Lolin ESP32-S2 Mini
import ssd1306
from uQR import QRCode

WIDTH = 64  # screen size
HEIGHT = 48
SIZE = 8  # text size
r = RTC()

# set up and clear screen
i2c = I2C(0, scl=Pin(s2mini.I2C_SCL), sda=Pin(s2mini.I2C_SDA))
oled = ssd1306.SSD1306_I2C(WIDTH, HEIGHT, i2c)
oled.fill(0)


def snazz():
    marquee = [
        "   **",
        "   **",
        "   **",
        "   **",
        "   **",
        "********",
        " ******",
        "  ****",
        "   **",
        " quite",
        "rubbish",
        " clock",
        "  mk.2",
        "<scruss>",
        " >2022<"
    ]
    for s in marquee:
        oled.scroll(0, -SIZE)  # scroll up one text line
        oled.fill_rect(0, HEIGHT-SIZE, WIDTH,
                       SIZE, 0)  # blank last line
        oled.text("%-8s" % s, 0, HEIGHT-SIZE)  # write text
        oled.show()
        time.sleep(0.25)
    time.sleep(5)
    oled.fill(1)
    oled.show()


snazz()  # tedious crowd-pleasing intro

qr = QRCode()
while True:
    qr.add_data("%02d%02d%02d" % r.datetime()[4:7])
    qr.border = 1  # default border too big to fit small screen
    m = qr.get_matrix()
    oled.fill(1)
    for y in range(len(m)):
        for x in range(len(m[0])):
            # plot a double-sized QR code, centred, inverted
            oled.fill_rect(9 + 2*x, 1 + 2*y, 2, 2, not m[y][x])
    oled.show()
    time.sleep(0.05)
    qr.clear()

If your output is glitchy, you might need to put the following in boot.py:

import machine
machine.freq(240000000)

This increases the ESP32-S2’s frequency from 160 to 240 MHz.

Canada’s longest running radio program is wrong!

Rob Cruickshank noted the other day:

Naturally, I had to verify this. So I tuned to the WWV 10 MHz time signal on my amateur rig, tuned a portable radio to CBC Radio 1 FM, which broadcasts on 99.1 MHz in Toronto and recorded them together:

Noisy recording of two radios. The WWV 13:00:00 EDT tone is at roughly 9 seconds, and the NRC tone broadcast by CBC is at 19 seconds

Yup: Rob’s right – CBC is broadcasting the NRC 13:00:00 signal at 13:00:10, which for time nerds might as well be the change from Julian to the Gregorian calendar.

Waveform from Audacity showing a low tone at 9s recorded from WWV, and CBC's long beep coming at 19 seconds
Annotated waveform: the CBC long beep is ten seconds after the WWV tone

This recording was made directly from the airwaves. There should be effectively no difference between the signal broadcast times, but here we are with the “National Research Council official time signal” going out at a very wrong time indeed.

Update, October 2023: Well, CBC has noticed, and rather than trying to fix it, they’re going to end it: The end of the long dash: CBC stops broadcasting official time signal | CBC News

Adding RGB LEDs to an illuminated arcade button

Somewhat painterly view of the button doing its thing. The weird clunking sound is my camera’s continuous focus. For a clearer but more flickery view, see here

Following on from a customer query at Elmwood Electronics, I can confirm that one can install install addressable RGB LEDs/NeoPixels inside one of these large buttons. It’s not the easiest build, so whether one should attempt this is another matter entirely.

You’ll need:

  • Large Arcade Button with LED – 60 mm White (tall version) – this is larger and more domed than the flat-top one that Adafruit sells
  • RGB LEDs – I used a generic 8 LED ring, but anything not too tall and under 45 mm in diameter should fit. Either a 7 X WS2812 5050 RGB LED Ring or Adafruit’s NeoPixel Ring – 12 x WS2812 5050 RGB LED with Integrated Drivers could also work
  • Thin (and I mean thin: I used 28 AWG) Silicone Cover Stranded-Core Wire in several colours. You’ll want to cut this quite long at first, as you have to ease it through some tiny holes in the button assembly. If you solder connectors on the end, you won’t be able to disassemble or install the button without cutting them off. Do I speak from experience here? You betcha!
  • The usual soldering/hot gluing/bending/prying/grabbing/cutting tools you already know and love. In addition, you might consider a non-marring spudger and a pair of small(ish) arterial forceps (aka hemostats, aka Kelly forceps, aka fishing hook removal pliers)

I’m not going to cover soldering the wires to the LED PCB in any depth here. You’ll need three wires: 5 V power, Ground and Data. Even though the LEDs I used need 5 V power, they are quite happy with 3.3 V logic on the data line. They need more than 3.3 V power to light, though.

a large arcade machine style button on the left: it has a clear domed top and a threaded base. On the right is the combined microswitch and LED holder that fits into the button base
The button in two pieces, as you might expect to receive it
the top of the button disassembled into its main parts: bezel ring at top left, threaded lock ring at bottom, and main button mechanism. The mechanism is upside down, so the return spring and button actuators can be seen inside the threaded shaft
Main parts of the button top, once you’ve removed the lock ring
Close up of inside the shaft: return spring and its retainer tabs, and button actuators can be seen
First step is to ease the spring out without bending it too much or breaking the retainer tabs
Close up of inside the shaft: the tips of a pair of forceps have eased the top of the spring past its retainers
I used small forceps to ease the spring out. Once you get it started, it unscrews easily from behind the retainers
Close up of inside the shaft: the button actuators have been pushed down the shaft, allowing the top of the button to be pulled out
Now the spring is out the way, you can squeeze in the actuator tabs and push them down the shaft to liberate the button top
button top components arranged: black threaded button base on left, return spring in the middle, and domed clear top with white underside and white actuators sticking down
The button top disassembled
clear button top attached to its white underside. A blunt metal tool (spudger) is pointed at the push-fit join between the two parts
Carefully lever off the clear top with a blunt tool like a spudger. Now would have been a great time to clean dust and other wee bits off your workspace, as they’ll surely end up inside the button, looking nasty
clear button top separated from its white base. A translucent white diffuser is inside the clear top. The white base has a hollow centre and a circular cavity
The button top opened up. The cavity is about 45 mm in diameter and only a few millimetres deep
The microswitch with the LED holder attached on top. The blade of a blunt metal spudger is inserted under a plastic tab that holds the LED holder onto the switch
Removing the LED holder from the microswitch is done by levering open (gently) the plastic tab that clamps the holder onto the switch.
the LED holder at left, and the bare microswitch. The LED holder has an LED in a white plastic retainer, and below it two spade contacts. The switch has three spade connectors: Com(mon) on the base, and "NO 3" (Normally Open) and "NC 2" on the right side. Normal operation connects COM and NO
LED holder and microswitch separated. For normal button operation, the contacts NO and COM become connected when the button is pressed. The spade contacts on the LED holder look like they should come out, and they will (soon)
LED holder disassembled into two parts. The black LED holder base is on the left, with the two conenctor clips slightly blurry at top. On the right is the LED in its white support, pulled out of the holder base
Pull the LED out from the holder, and you’ll see the metal clips that held it in place. These clips have to come out: I found the pushing them in slightly while pulling down on the spade connector eased them out eventually
White button top underside with an 8 RGB LED ring hot glued into it. Three thin insulated wires (from top: yellow (data), red (5 V) and black (GND)) are previously soldered behind the LED board, and are secured against strain with a large deposit of hot glue
LED ring hot glued into place. Make sure that the wires are properly secured, as you don’t want to take this apart again
threaded button base with clear top fitted, seen from underneath. The white button actuators have been pushed back into place, and the three coloured wires are feeding through the hole in the shaft. The return spring is outside the wires, and is being fitted around the retainers inside the shaft
Fit the clear button top back inside the base, feeding the wires through the shaft. Fitting the return spring back in is a bit more chaotic than getting it out. I ended up jamming it in with forceps, and it seemed to sort out okay despite that
underside of the button shaft, with microswitch attached to LED holder. The wires coming from the LED ring inside the button top have been fed through the small cavities where the original LED holder clips/contacts have been removed. The red/black power wires are on the side towards us, while the yellow data wire is behind the microswitch
The really fiddly bit: feeding the wires through the tiny gaps where the LED holder clips/contacts used to be. Even using thin (28 AWG) silicone covered wire, all three wires couldn’t fit down one side. Make sure the wires are pulled gently through, and aren’t snagged anywhere
Fully reassembled button, with microswitch installed into its bayonet connector in the threaded shaft, and the button actuator lined up with the microswitch lever on the left. The yellow data wire is in front of the microswitch at bottom
Finished! Make sure that the switch actuates properly by lining up the LED holder in the bayonets inside the shaft. Of course, you’ll have wanted to install the button in your project before doing this assembly, as you’ll have to feed those pesky wires back through again if you haven’t …

More Magic Designer Nonsense

This is what I simulated earlier – except drawn on a real Magic Designer. Something’s off with what I modelled …
All of these are drawn with the Magic Designer angle set to 57, which puts the crank discs exactly in phase. The blue circle in the middle is an exactly image of a crank disc, if perhaps a very dull plot.

Slightly imperfect Hoot-Nanny/Magic Designer simulation

round figure with three interlaced 6-fold curves picked out in red, green and blue
I even emulated the locating notches at the edge of the paper …

Simulated (and not quite right yet) output from a “HOOT-NANNY” or Magic Designer, a proto-Spirograph toy that drew six-sided curves on round paper sheets. It was made by Howard B. Jones and Co. of Chicago, IL and first sold in 1929. The company’s better known for producing Jones Plugs and Sockets, sometimes known as Cinch-Jones connectors. The “HOOT-NANNY” name was dropped when production moved to the Northern Signal Company of Saukville, WI.

My eBay-acquired Magic Designer is quite beaten up, and doesn’t always produce accurate results. Here’s how one should look, from the instruction pamphlet:

diagram of the Magic Designer toy: central turntable holding paper is rotated by crank at bottom right. On the left are two crank discs (upper and lower) each with a crank pin that can fit into holes along two arms. These arms are joined at a pivot, and in the centre of this pivot is a pencil.

The upper crank disc can be moved in an arc relative to the lower disc. This is controlled by a locking shift lever on the right
I can’t shake the feeling this was originally something like an artillery ranging tool or suchlike

As far as I’ve been able to work out, the parameters of the machine are:

  • central turntable is 6″ in diameter, with 192 gear teeth around the edge;
  • each crank disc is 1″ diameter (32 teeth), with the crank pin at 3/8″ radius;
  • the shift lever has a 10-70 degree scale, which corresponds to moving the upper crank disc between 30-90 degrees of arc from the lower (fixed) crank disc;
  • the pencil arms have 18 holes labelled A to R, at 1/4″ spacing from 5.75 to 1.5″. The perpendicular distance from the pivot holes to the pencil is 5/16″. This small offset makes very little difference to the overall arm length.

If we model the toy with a fixed turntable:

  • the crank pins describe epitrochoids around the edge of the paper;
  • the pencil point traces the intersection of two circles of radius the lengths of the pencil arms, each centred on a crank pin.

Here’s a very simple model in Python that emits a hard-coded (but editable) pattern in HP-GL: Slightly imperfect Python simulation of the “HOOT-NANNY” (or Magic Designer) drawing toy (static local copy: hootnanny.zip). It doesn’t do anything with the fixed circle studs (yet)

A new old calendar

simulated dot-matrix printer on ruled paper layout of an ASCII-art Snoopy in WW1 Air Ace regalia yelling "Curse you Red Baron!" at his imaginary nemesis, and a 2022 calendar laid out on the facing page
not seen the likes of this since ’78, I’ll be bound

It’s unlikely anyone wanted a faux-lineprinter ASCII art calendar for 2022, but you’re getting one anyway. You can print this yourself:

If you want to make your own, here’s a script: snoopycal.sh

Credits:

  • Original “WW1 Fighter Pilot” Snoopy ASCII art from “SNOOPY.BA” for the DEC PDP-8, written by Mr Kay R. Fisher of DEC some time before July 1973. It’s referred to in the first printing of the “101 Basic Computer Games” book, which was published in 1973.
  • ncal, banner: their respective authors
  • pstext an ascii to PS filter by Dan Judd, usenet comp.lang.postscript, December 1989. I had to really mess around with the output of this program to use a custom font and add the music ruling, but it produces cleaner PostScript than the giant messes that enscript and a2ps have become
  • Font: mnicmp, by me. Based on the DecWriter II font.
  • iffy scripting, scribbly lines: also me.

I (U+1F494, BROKEN HEART) UTF-8

Something has gone very wrong with the database encoding on this blog after a recent update, so all my lovely UTF-8 characters have gone mojibake.

Trying to find ways to fix it. It may have to be manual. Remember, kids: have backups before letting WordPress upgrade!

Here’s the Python equivalent of what I think the database has done:

bytes("I ???? UTF-8", encoding='utf-8').decode(encoding='cp1252')
'I 💔 UTF-8'

Quite why my hosting thought a character encoding from last century was appropriate, I’ll never know.

Update, November 2023: kinda-sort fixed the backend, but the encoding is still weird — can we…?

Autumn in Canada: PicoMite version

more leaves

So I ported autumn in canada from OpenProcessing to PicoMite BASIC on the Raspberry Pi Pico:

a small black screen images with text in the centre: autumn in canada scruss, 2021-11 just watch ...
no leaves
a small black screen images with text in the centre: autumn in canada scruss, 2021-11 just watch ... with one red and one orange maple leaf sitting on top of it
a couple of leaves
a small black screen images with text in the centre: autumn in canada scruss, 2021-11 just watch ... with four red/yellow/orange maple leaves sitting on top of it
more leaves
a small black screen images with text in the centre: autumn in canada scruss, 2021-11 just watch ... with sixteen simulated fallen maple leaves mostly covering it
plenty of leaves
a small black screen image completely covered with many simulated fallen maple leaves
far too many leaves

The biggest thing that tripped me up was that PicoMite BASIC starts arrays at 0. OPTION BASE 1 fixes that oversight. It would have been nice to have OpenProcessing’s HSV colour space, and an editor that could handle lines longer than 80 characters that didn’t threaten to bomb out if you hit the End key, but it’ll serve.

Source below:

' autumn in canada
' scruss, 2021-11
' a take on my https://openprocessing.org/sketch/995420 for picomite

OPTION base 1
RANDOMIZE TIMER
' *** initialize polar coords of leaf polygon and colour array
DIM leaf_rad(24), leaf_ang(24), px%(24), py%(24)
FOR i=1 TO 24
    READ leaf_rad(i)
NEXT i
FOR i=1 TO 24
    READ x
    leaf_ang(i)=RAD(x)
NEXT i

DIM integer c%(8)
FOR i=1 TO 8
    READ r%, g%, b%
    c%(i)=RGB(r%,g%,b%)
NEXT i

' *** set up some limits
min_scale%=INT(MIN(MM.HRES, MM.VRES)/8)
max_scale%=INT(MIN(MM.HRES, MM.VRES)/6)
min_angle=-30
max_angle=30
min_x%=min_scale%
min_y%=min_x%
max_x%=MM.HRES - min_x%
max_y%=MM.VRES - min_y%

CLS
TEXT MM.HRES/2, INT(MM.VRES/3), "autumn in canada", "CM"
TEXT MM.HRES/2, INT(MM.VRES/2), "scruss, 2021-11", "CM"
TEXT MM.HRES/2, INT(2*MM.VRES/3), "just watch ...", "CM"

kt%=0
DO
    cx% = min_x% + INT(RND * (max_x% - min_x%))
    cy% = min_y% + INT(RND * (max_y% - min_y%))
    angle = min_angle + RND * (max_angle - min_angle)
    sc% = min_scale% + INT(RND * (max_scale% - min_scale%))
    col% = 1 + INT(RND * 7)
    leaf cx%, cy%, sc%, angle, c%(7), c%(col%)
    kt% = kt% + 1
LOOP UNTIL kt% >= 1024

END

SUB leaf x%, y%, scale%, angle, outline%, fill%
    FOR i=1 TO 24
        px%(i) = INT(x% + scale% * leaf_rad(i) * COS(RAD(angle)+leaf_ang(i)))
        py%(i) = INT(y% - scale% * leaf_rad(i) * SIN(RAD(angle)+leaf_ang(i)))
    NEXT i
    POLYGON 24, px%(), py%(), outline%, fill%
END SUB

' radii
DATA 0.536, 0.744, 0.608, 0.850, 0.719
DATA 0.836, 0.565, 0.589, 0.211, 0.660, 0.515
DATA 0.801, 0.515, 0.660, 0.211, 0.589, 0.565
DATA 0.836, 0.719, 0.850, 0.608, 0.744, 0.536, 1.000
' angles
DATA 270.000, 307.249, 312.110, 353.267, 356.540
DATA 16.530, 18.774, 33.215, 3.497, 60.659, 72.514
DATA 90.000, 107.486, 119.341, 176.503, 146.785, 161.226
DATA 163.470, 183.460, 186.733, 227.890, 232.751, 270.000, 270.000
' leaf colours
DATA 255,0,0, 255,36,0, 255,72,0, 255,109,0
DATA 255,145,0, 255,182,0, 255,218,0, 255,255,0

You could probably use AUTOSAVE and paste the text into the PicoMite REPL. I used an ILI9341 SPI TFT LCD Touch Panel with my Raspberry Pi Pico along with some rather messy breadboard wiring.

Fun fact: the maple leaf polygon points are derived from the official definition of the flag of Canada.

Raspberry Pi Zero 2 W: slides and thermals

2 out of 4 cores burning, 32-bit mode: time to overheat = basically never

Slides from last night’s talk:

It’s impossible to have a Raspberry Pi Zero overheat unless you overclock it. That’s why you don’t get any cases for it with fans or heat sinks. The quad-core Raspberry Pi Zero 2 W, though, has the potential to do so. Here are some numbers:

  • Used official case with lid fitted: increases SoC temperature +3 °C over free air
  • Test – CPUBurn: https://github.com/pmylund/cpuburn
  • Tested 4, 3 and 2 cores burning in 32-bit and 64-bit modes: time from idle to throttling (80 °C) measured
  • GPU overheat not tested.
line graph of cpu temperature against time. Temperature rises sharply from about 47 degrees C to 82 degrees C in around four minutes
All 4 cores burning, 64-bit mode: time to overheat = under 3½ minutes
line graph of cpu temperature against time. Temperature rises sharply from about 47 degrees C to 82 degrees C in just over four minutes
All 4 cores burning, 32-bit mode: time to overheat = just over 4 minutes
line graph of cpu temperature against time. Temperature rises moderately from about 47 degrees C to 81 degrees C in around seven minutes
3 out of 4 cores burning, 64-bit mode: time to overheat = just over 7 minutes
line graph of cpu temperature against time. Temperature rises slowly from about 47 degrees C to 81 degrees C in around ten minutes
3 out of 4 cores burning, 32-bit mode: time to overheat = 9½ minutes
line graph of cpu temperature against time. Temperature rises very slowly, reach 70 degrees C in 40 minutes and then only rising very slightly to about 73 degrees C in the entire run time of 3 hours 20 minutes
2 out of 4 cores burning, 32-bit mode: time to overheat = basically never

Unless you’re doing things that might indicate you should be using a bigger computer, a Raspberry Pi Zero 2 W won’t overheat and doesn’t need any form of cooling. If you’re overclocking, well … it’s your choice to have cooling equipment worth more than the computer it’s trying to cool.

Raspberry Pi Zero 2 W: initial performance

Running A Pi Pie Chart turned out some useful performance numbers. It’s almost, but not quite, a Raspberry Pi 3B in a Raspberry Pi Zero form factor.

32-bit mode

Running stock Raspberry Pi OS with desktop, compiled with stock options:

pie chart comparing multi-thread numeric performance of Raspberry Pi Zero 2 W: slightly faster than a Raspberry Pi 2B
multi-thread results
pie chart comparing single-thread numeric performance of Raspberry Pi Zero 2 W: slightly faster than a Raspberry Pi 2B
single-thread results
time ./pichart-openmp -t "Zero 2W, OpenMP"
pichart -- Raspberry Pi Performance OPENMP version 36

Prime Sieve          P=14630843 Workers=4 Sec=2.18676 Mops=427.266
Merge Sort           N=16777216 Workers=8 Sec=1.9341 Mops=208.186
Fourier Transform    N=4194304 Workers=8 Sec=3.10982 Mflops=148.36
Lorenz 96            N=32768 K=16384 Workers=4 Sec=4.56845 Mflops=705.102

The Zero 2W, OpenMP has Raspberry Pi ratio=8.72113
Making pie charts...done.

real	8m20.245s
user	15m27.197s
sys	0m3.752s

-----------------------------

time ./pichart-serial -t "Zero 2W, Serial"
pichart -- Raspberry Pi Performance Serial version 36

Prime Sieve          P=14630843 Workers=1 Sec=8.77047 Mops=106.531
Merge Sort           N=16777216 Workers=2 Sec=7.02049 Mops=57.354
Fourier Transform    N=4194304 Workers=2 Sec=8.58785 Mflops=53.724
Lorenz 96            N=32768 K=16384 Workers=1 Sec=17.1408 Mflops=187.927

The Zero 2W, Serial has Raspberry Pi ratio=2.48852
Making pie charts...done.

real	7m50.524s
user	7m48.854s
sys	0m1.370s

64-bit

Running stock/beta 64-bit Raspberry Pi OS with desktop. Curiously, these ran out of memory (at least, in oom-kill‘s opinion) with the desktop running, so I had to run from console. This also meant it was harder to capture the program run times.

The firmware required to run in this mode should be in the official distribution by now.

pie chart comparing 64 bit multi-thread numeric performance of Raspberry Pi Zero 2 W: slightly faster than a Raspberry Pi 2B
multi-thread, 64 bit: no, I can’t explain why Lorenz is better than a 3B+
pie chart comparing 64 bit single-thread numeric performance of Raspberry Pi Zero 2 W: slightly faster than a Raspberry Pi 2B
single thread, again with the bump in Lorenz performance
pichart -- Raspberry Pi Performance OPENMP version 36

Prime Sieve          P=14630843 Workers=4 Sec=1.78173 Mops=524.395
Merge Sort           N=16777216 Workers=8 Sec=1.83854 Mops=219.007
Fourier Transform    N=4194304 Workers=4 Sec=2.83797 Mflops=162.572
Lorenz 96            N=32768 K=16384 Workers=4 Sec=2.66808 Mflops=1207.32

The Zero2W-64bit has Raspberry Pi ratio=10.8802
Making pie charts...done.

-------------------------

pichart -- Raspberry Pi Performance Serial version 36

Prime Sieve          P=14630843 Workers=1 Sec=7.06226 Mops=132.299
Merge Sort           N=16777216 Workers=2 Sec=6.75762 Mops=59.5851
Fourier Transform    N=4194304 Workers=2 Sec=7.73993 Mflops=59.6095
Lorenz 96            N=32768 K=16384 Workers=1 Sec=9.00538 Mflops=357.7

The Zero2W-64bit has Raspberry Pi ratio=3.19724
Making pie charts...done.

The main reason for the Raspberry Pi Zero 2 W appearing slower than the 3B and 3B+ is likely that it uses LPDDR2 memory instead of LPDDR3. 64-bit mode provides is a useful performance increase, offset by increased memory use. I found desktop apps to be almost unusably swappy in 64-bit mode, but there might be some tweaking I can do to avoid this.

Unlike the single core Raspberry Pi Zero, the Raspberry Pi Zero 2 W can be made to go into thermal throttling if you’re really, really determined. Like “3 or more cores running flat-out“-determined. In my testing, two cores at 100% (as you might get in emulation) won’t put it into thermal throttling, even in the snug official case closed up tight. More on this later.

(And a great big raspberry blown at Make, who leaked the Raspberry Pi Zero 2 W release a couple of days ago. Not classy.)

Modding an Adafruit PIR for 3.3 volts

green circuit board covered in surface mount components. A grey wire has been soldered to the output pin of the SOT-89 package 7133-1 voltage regulator
slightly dodgy soldering of a grey jumper wire to the Vout pin of the PIR’s voltage regulator

Consider the Adafruit PIR (motion) sensor (aka PIR Motion Sensor, if you’re in Canada). Simple, reliable device, but only runs from a 5 V supply. Yes, there are smaller PIRs that run off 3.3 V, but if this is what you have, you need to do some soldering. Annoyingly, the sensor on the board is a 3.3 V part, but the carrier was designed in Olden Tymes when King 5 V ruled.

You can try powering it from 3.3 V, but it’ll go all off on its own randomly as its own power supply won’t be supplying enough voltage. There are a couple of sites on how to modify these PIRs that either describe old kit you can’t get any more, or do it completely wrongly. Just one post on the Adafruit support forum gets it right.

One way of doing this is to provide 3.3 V directly to the output pin of the voltage regulator, and ignore the 5 V power line entirely. The regulator’s a SOT89-3 part that looks a bit like this:

71xx-1 SOT-89 package outline, with three pins at the bottom and one large ground tab (connected to centre pin, but not visible) at the top
wee leggy thing

In the photo above, it’s flipped over. Whichever way it’s oriented, we want to put power directly into the Vout pin. There may be easier points to solder this to than a tiny surface mount pin (almost definitely one of the capacitors) but this has held for me.

How to use it in MicroPython? Like the TTP223 capacitive touch sensors I looked at before, a PIR gives a simple off/on output, so you can use something like this:

from machine import Pin
from time import sleep_ms

pir = Pin(21, Pin.IN)

while True:
    print("[", pir.value(), "]")
    sleep_ms(1000)

value() will return 1 if there’s movement, 0 if not. There are trigger time and sensitivity potentiometers to fiddle with on the board if you need to tweak the output.

line graph showing output signal going from 0 to 1, back down to 0 and ending at one over a period of about 20 seconds
Thonny plotter output showing a couple of movement detections. High output (on my device) stays up for about 4 seconds, so you can be pretty leisurely about polling PIRs

Just remember: don’t connect the 5 V power line if you make this mod. I’m not responsible for any smoke emitted if you do — but I can always sell you a replacement …