Slimming Raspberry Pi OS for a Pi 1 Model B

I had an original Raspberry Pi Model B in a drawer: 512 MB of RAM, a single 700 MHz ARMv6 core, Ethernet, no Wi-Fi, no Bluetooth. The job I had in mind for it was small: run an RTL-SDR dongle headless, around the clock, and record two radio channels to the SD card. It doesn’t need a desktop, a display, a camera, sound, or anyone typing at a console.

Raspberry Pi OS Lite (32-bit, Debian 13 “trixie”) still boots on it. But one image serves every Pi from the Zero to the 5, so out of the box it carries kernels for four CPU generations, firmware for half a dozen Wi-Fi chips, a camera stack with TensorFlow Lite in it, and a fair few services that wake up to look for hardware this board has never had. On a 700 MHz machine that’s boot time, RAM and SD card writes, and every one of those packages gets downloaded again at each upgrade.

So I slimmed it. The result is a script, run once after the first boot, and it’s in full at the end of this post. The rest is how the list in it was built, and the traps along the way.

A word on the timings. The Pi isn’t the only antique here: the SD card is a 16 GB SanDisk SDHC made in August 2009, and it’s probably as slow as you’d expect. A lot of the boot time, and how long the upgrades and purges took, is the card as much as the CPU. On a newer card your seconds will be different. The ones that come from the CPU (Perl starting up at every boot, NetworkManager’s start-up) should still be worth having.

The numbers

Fresh imageSlimmed
Packages640376
Root filesystem used2.1 GB957 MB
Memory the kernel sees426 MB474 MB
Memory available, idle326 MB376 MB
Kernel modules loaded31*12
Boot time (systemd-analyze)1 min 42 s*55 s

* I didn’t measure these on the fresh image. They’re from after the first pass, with the packages gone but NetworkManager, the swap file and the camera drivers still there.

Don’t guess, look

The tempting approach is a remove line of package names you’re fairly sure you don’t need. I started there, and it was mostly wrong. The better way is to boot the fresh card, ssh in, and ask it.

What’s running, and what will run later:

systemctl list-units --type=service --state=running
systemctl list-unit-files --state=enabled
systemctl list-timers --all
systemctl list-sockets
ps -eo pid,rss,comm --sort=-rss
ss -tulpn

Which package each process belongs to:

dpkg -S "$(readlink -f /proc/<pid>/exe)"

What the image chose to install, and what’s big:

apt-mark showmanual
dpkg-query -Wf '${Installed-Size}\t${Package}\n' | sort -n

apt-mark showmanual is the useful one. Everything else on the card is there because something on that list depends on it, so that’s where to look for things to drop.

And then, before removing anything, a dry run of the whole lot:

apt-get -s purge --autoremove <the list>

-s simulates. It needs no root and changes nothing, and it prints every package that would go, including the ones you didn’t name. Read that output. It’s where every trap below showed up.

Let the first boot settle (cloud-init applies the Imager’s settings, then it’s done: cloud-init status says so) and reboot once before you measure anything.

What went

The fresh image ran these, by memory use:

ProcessRSSWhat it’s for here
NetworkManager16.5 MBEthernet with DHCP; replaced, see below
udisksd11.4 MBDesktop disk automounting
polkitd7.0 MBPermissions for udisks and NetworkManager
systemd-timesyncd6.5 MBTime; replaced by chrony
wpa_supplicant5.0 MBWi-Fi, on a board with no Wi-Fi
mpris-proxy2.9 MBBluetooth media controls, from bluez
cron2.6 MBNo jobs
two gettys4.7 MBLogins on HDMI and serial

And the space went on:

  • Wi-Fi and Bluetooth firmware, 244 MB: Atheros 120 MB, MediaTek 63 MB, Realtek 30 MB, Broadcom 20 MB, Libertas 12 MB. All for chips this board hasn’t got.
  • Kernels for other Pis, about 150 MB with the headers. The 32-bit image installs v6, v7 and v8 kernels (the v8 one as arm64, which also means an arm64 package list is downloaded at every update). A Pi 1 runs v6. Check with uname -r before you take any.
  • rpi-eeprom, 31 MB. Firmware updates for the Pi 4 and 5’s boot EEPROM. The Pi 1 has none.
  • rpi-connect-lite, 23 MB. Raspberry Pi’s remote access service.
  • mkvtoolnix, 23 MB, plus 36 MB of ICU for its Qt dependency. A video container tool on a Lite image; I’ve no idea why.
  • cloud-init and its Python libraries, about 30 MB of Babel locale data alone. It’s done its job after the first boot.
  • The camera stack (rpicam-apps, libcamera and TensorFlow Lite), v4l-utils, the GPIO Python libraries, udisks2, NTFS and CIFS support, gdb, and man pages.

The full list is in the script, grouped and commented. Purging it with autoremove takes about 245 packages and frees about 630 MB.

The traps

Each of these was found by the dry run or by the Pi, not by thinking about it.

gnupg has to stay. apt depends on raspbian-archive-keyring, which depends on gnupg. With gnupg on the list, apt refuses the whole removal: “pkgProblemResolver::Resolve generated breaks”. Fair enough.

rsync has to stay. I wanted it gone as a tool I didn’t use. The dry run quietly added raspi-firmware to the purge, which is the package that installs the bootloader files and kernel images. Remove that and the card doesn’t boot. raspi-firmware depends on rsync. To find which name on a long list causes a removal, simulate each one on its own:

for p in <list>; do
apt-get -s purge $p | grep -q '^Purg raspi-firmware' && echo "$p"
done

apt never autoremoves a kernel. Purging the linux-image-rpi-v7 metapackage leaves linux-image-6.18.50+rpt-rpi-v7 installed, deliberately. Then dpkg --remove-architecture arm64 fails, because the versioned v8 kernel is still an arm64 package. The script asks dpkg for every installed kernel, header and kbuild package that isn’t the v6 kernel, at the time it runs, so it keeps working after kernel updates.

The audio packages stay, oddly. I wanted ALSA gone: no sound on a headless recorder. But raspberrypi-sys-mods depends on raspi-config, which depends on alsa-utils. And sys-mods is worth keeping. It’s where Raspberry Pi OS keeps:

  • the journal in RAM rather than on the SD card;
  • the hardware watchdog switched on;
  • its udev rules (vcgencmd access, SD card trim, the CPU governor);
  • its sudo settings.

So ALSA stays, at 9 MB, running nothing, and the sound driver is blacklisted instead.

--no-install-recommends skips more than you’d think. It’s the right flag for a lean install, but it never skips a real dependency (Depends:), only Recommends:. Some of what you need is only recommended:

PackageRecommendsWithout it
cmakegcc, makecmake can’t build anything
gcclibc6-devNo stdio.h; the first #include fails
gitca-certificates, ssh-clienthttps clones fail on certificates; git@ remotes fail outright

build-essential depends on gcc, make and libc6-dev for real, so naming it gets them all. Git LFS downloads over https even from an ssh remote, so if you want LFS files you need ca-certificates too. I didn’t, so neither git-lfs nor ca-certificates is on my list.

Headless, properly

GPU memory. gpu_mem=16 in config.txt gives the ARM 496 MB instead of 448. The firmware then switches to its cut-down build, which has no camera or video codec services. Fine, but their kernel drivers still load and fail at every boot (“failed to open VCHI service”). So they’re blacklisted, and the display, camera and audio lines in config.txt are commented out:

blacklist bcm2835_codec
blacklist bcm2835_isp
blacklist bcm2835_v4l2
blacklist bcm2835_mmal_vchiq
blacklist vc_sm_cma
blacklist snd_bcm2835
blacklist drm

drm is loaded by name by modprobe@drm.service, which a blacklist doesn’t stop, so that unit is masked too.

The DVB-T driver. Linux has its own driver for these dongles, for watching TV, and it grabs the device before librtlsdr can. Blacklisting dvb_usb_rtl28xxu, rtl2832_sdr, rtl2832 and e4000 stops it. The modules are part of the kernel package, so they can’t be uninstalled; a blacklist is the way.

No consoles at all. I’ll never plug in a screen or a serial cable, so the script masks getty@tty1 and serial-getty@ttyAMA0, sets enable_uart=0, and removes the console= entries from cmdline.txt. That’s a commitment. From then on, ssh is the only way in, and if the network won’t come up the fix is to edit the SD card on another machine or re-image it.

Timers

systemctl list-timers on the fresh image, and what I did with each:

TimerWhat it doesVerdict
apt-daily, apt-daily-upgradeRefresh package lists and download updates, dailyMasked: I’ll upgrade when I choose to
man-db, logrotateRebuild the man index; rotate logsGone with their packages
e2scrub_allWeekly ext4 metadata check, LVM onlyMasked: no LVM, so it does nothing
dpkg-db-backupDaily copy of dpkg’s database to /var/backupsKept: it’s the recovery if a power cut spoils the database
fstrimWeekly: tells the SD card which blocks are freeKept: the card supports it, and it helps wear
systemd-tmpfiles-cleanClears out /tmp and friendsKept: part of systemd
rpi-zram-writebackSee swap, belowGone with the swap change

Swap: zram only

Raspberry Pi OS’s swap is cleverer than it looks. The swap device is zram0: a compressed block device in RAM. A page that’s swapped out is compressed with zstd and kept in memory, at about a third of its size. Behind that sits /var/swap, a file attached through a loop device as zram’s writeback device. Once a day a timer marks compressed pages that haven’t been touched and moves them out to the file, which frees the RAM they held.

The cost is at boot. rpi-resize-swap-file sizes the file to match RAM and re-runs mkswap on it every time, whether or not anything changed. It also starts Perl (via ack) to read a number out of /proc/meminfo. On this board that’s 25 seconds of every boot.

A recorder that runs for weeks doesn’t need idle pages paged out to a card it’s also trying not to wear out. So one drop-in file, /etc/rpi/swap.conf.d/50-zram-only.conf:

[Main]
Mechanism=zram

The zram size is still worked out from the RAM at each boot (by rpi-swap‘s generator), so it still follows a change to gpu_mem. There’s no file, no loop device, no writeback timer and no resize. Delete the drop-in and reboot to get the default back.

NetworkManager out, systemd-networkd in

NetworkManager took 32 seconds to start at every boot on this Pi, and it’s the biggest process left. For one Ethernet port on DHCP, systemd-networkd, which is already installed as part of systemd, needs one file, /etc/systemd/network/10-eth0.network:

[Match]
Name=eth0
[Network]
DHCP=yes
[DHCPv4]
ClientIdentifier=mac

ClientIdentifier=mac matters. By default networkd identifies itself to the DHCP server with a DUID, where NetworkManager’s built-in DHCP client uses the MAC address (as far as I can tell), so the router may hand out a different address after the switch. On a headless box with no console, that means hunting for it.

networkd doesn’t write /etc/resolv.conf by itself; that’s systemd-resolved’s job. Rather than run another daemon, the script keeps the DNS servers NetworkManager was given and writes them back as a fixed file once NetworkManager is gone.

The switch is the very last step before the reboot. NetworkManager leaves the interface configured when it stops, so the network (and the ssh session running the script) stays up until the reboot, and networkd takes over at boot. It came straight back at the same address, systemd-networkd started in 4 seconds, and the boot went from 1 min 42 s to 55 s.

One leftover: polkitd was still running, started because networkd asks systemd-hostnamed for the hostname at boot, and hostnamed asks polkit. Nothing depends on polkitd; it’s only there because systemd recommends it. It now goes with NetworkManager, and networkd and hostnamed carry on without it.

Smaller things

No man pages or docs, ever. A file in /etc/dpkg/dpkg.cfg.d/ stops dpkg installing them, for every package from then on, and the script deletes the ones already there (keeping the copyright files):

path-exclude /usr/share/man/*
path-exclude /usr/share/info/*
path-exclude /usr/share/doc/*
path-include /usr/share/doc/*/copyright

Write it before the upgrade, so the upgrade doesn’t put them back.

Order: remove, update, upgrade, install. Remove first, so the upgrade doesn’t download and unpack packages you’re about to delete. That’s a real saving on a 700 MHz CPU. Then update the lists, upgrade, and install last, against the current versions.

DEBIAN_FRONTEND=noninteractive and --force-confold. -y answers apt’s questions but not dpkg’s “keep your modified config file?”, which would stop an unattended upgrade dead.

Old logs and a trim. The script ends by deleting /var/log/*.log and the apt logs, running fstrim -a, and rebooting.

sudo’s cached password is shared. This one confused me. sudo worked over ssh with no password, then later didn’t, and I wrongly blamed the purge. Raspberry Pi OS sets timestamp_type=global (in /etc/sudoers.d/010_global-tty, from sys-mods), so once you’ve typed your password in one session, every session as that user can sudo until it times out. The script was fine. The cached login had simply run out.

Safe to run again

The script uses set -e, so any failure stops it rather than carrying on. It’s written so a second run changes nothing:

  • purging a package that’s already gone only prints a notice, and installing one that’s there does nothing;
  • systemctl mask and enable don’t mind being repeated;
  • the config files are rewritten whole, and the config.txt and cmdline.txt edits only match lines not yet changed;
  • the one step that would fail, removing the arm64 architecture when it’s already gone, is checked first.

The trap for later: a package name that’s no longer in the archive at all stops it with “Unable to locate package”. That’s why the kernel list is worked out when it runs rather than written down.

I ran it four times on the same card while building it, the last three over a Pi that had already been slimmed. It’s a script for a Pi 1 Model B on Ethernet, so check it against your own board before you borrow from it: the kernel names, the firmware and the Wi-Fi removal are all specific to this one.

The script

Run it with sudo bash setup-pi-trixie-v1.sh once the first boot has finished. It reboots at the end.

#!/bin/bash
set -e

if [ $(whoami) != root ] ; then
    echo "Must run using sudo"
    exit 1
fi

# RPi Imager
# Select OS: Raspberry Pi OS (other) ->
#     Raspberry Pi OS Lite (32-bit) Debian Trixie
# Set the hostname (sdr1), the username (sdr) and a password,
# and add your ssh public key.
# This is a RPi v1 Model B with 512MB RAM and 700MHz CPU;
# no WiFi or Bluetooth.
# Always headless: no HDMI, camera or audio. Wired Ethernet with DHCP.
# Run by hand after the first boot has finished
# (cloud-init status: done), with sudo.
# Safe to re-run. It reboots at the end.

export DEBIAN_FRONTEND=noninteractive

echo "---- No man pages or docs from now on ----"
cat > /etc/dpkg/dpkg.cfg.d/01-nodoc <<'EOF'
path-exclude /usr/share/man/*
path-exclude /usr/share/info/*
path-exclude /usr/share/doc/*
path-include /usr/share/doc/*/copyright
EOF

echo "---- Removing unwanted packages ----"
# Survey of the fresh image, 2026-11-21. Purging these with autoremove
# frees about 630 MB.
# Keep: gnupg (raspbian-archive-keyring, and so apt, needs it), rsync
# (raspi-firmware needs it), locales, htop, nano. Keep the audio
# packages (alsa-utils etc., 9 MB, nothing running):
# raspberrypi-sys-mods needs raspi-config, which needs alsa-utils, and
# sys-mods keeps the journal in RAM, enables the watchdog, and holds
# the sudoers and udev settings.
# Radio: no WiFi or Bluetooth on this board
PKG_REMOVE="modemmanager bluez bluez-firmware rfkill"
PKG_REMOVE="$PKG_REMOVE wpasupplicant wireless-tools"
PKG_REMOVE="$PKG_REMOVE firmware-atheros firmware-brcm80211"
PKG_REMOVE="$PKG_REMOVE firmware-libertas firmware-mediatek"
PKG_REMOVE="$PKG_REMOVE firmware-realtek"
# Kernels for other Pi models (this one runs rpi-v6), and kernel
# headers. Named one by one, versioned ones too, because apt never
# autoremoves a kernel.
PKG_REMOVE="$PKG_REMOVE $(dpkg-query -W \
    -f='${db:Status-Abbrev} ${binary:Package}\n' \
    'linux-image-*' 'linux-base-*' 'linux-headers-*' 'linux-kbuild-*' |
    awk '$1 == "ii" &&
         ($2 !~ /rpi-v6$/ || $2 ~ /^linux-(headers|kbuild)-/) {
             print $2
         }')"
# First-boot setup, finished by now
PKG_REMOVE="$PKG_REMOVE cloud-init rpi-cloud-init-mods userconf-pi"
PKG_REMOVE="$PKG_REMOVE libpam-chksshpwd"
# Remote access, file sharing and disks
PKG_REMOVE="$PKG_REMOVE rpi-connect-lite cifs-utils ntfs-3g udisks2"
PKG_REMOVE="$PKG_REMOVE p7zip-full zip unzip"
# Camera, video, display, keyboard and GPIO
PKG_REMOVE="$PKG_REMOVE mkvtoolnix v4l-utils rpicam-apps-lite"
PKG_REMOVE="$PKG_REMOVE kms++-utils fbset console-setup"
PKG_REMOVE="$PKG_REMOVE keyboard-configuration"
PKG_REMOVE="$PKG_REMOVE gpiod python3-gpiozero python3-libgpiod"
PKG_REMOVE="$PKG_REMOVE python3-rpi-lgpio python3-smbus2 python3-spidev"
PKG_REMOVE="$PKG_REMOVE rpi-keyboard-config rpi-keyboard-fw-update"
PKG_REMOVE="$PKG_REMOVE rpi-usb-gadget usb-modeswitch libmtp-runtime"
# Pi 4/5 only
PKG_REMOVE="$PKG_REMOVE rpi-eeprom rpifwcrypto rpi-update"
# Man pages
PKG_REMOVE="$PKG_REMOVE man-db manpages-dev groff-base"
# Tools and services we don't use (NetworkManager does its own DHCP)
PKG_REMOVE="$PKG_REMOVE apt-listchanges debconf-utils dmidecode"
PKG_REMOVE="$PKG_REMOVE pciutils lua5.1 luajit ssh-import-id"
PKG_REMOVE="$PKG_REMOVE isc-dhcp-client isc-dhcp-common logrotate"
PKG_REMOVE="$PKG_REMOVE cron cron-daemon-common gdb"
PKG_REMOVE="$PKG_REMOVE ncdu strace curl bash-completion"
PKG_REMOVE="$PKG_REMOVE vim-tiny vim-common net-tools traceroute"
# Development packages: removed here, and the ones we need are
# installed again below
PKG_REMOVE="$PKG_REMOVE build-essential pkg-config python3-venv"

apt-get -y purge $PKG_REMOVE
apt-get -y autoremove --purge

# The arm64 architecture was only there for the v8 kernel
# (checked, so the script can be re-run)
if dpkg --print-foreign-architectures | grep -qx arm64 ; then
    dpkg --remove-architecture arm64
fi

echo "---- Disabling unwanted timers and services ----"
# apt-daily*: background update downloads.
# e2scrub*: ext4 checks for LVM only.
# getty@tty1, serial-getty@ttyAMA0: logins on HDMI and on the serial
# port; ssh only.
systemctl mask apt-daily.timer apt-daily-upgrade.timer \
    e2scrub_all.timer e2scrub_reap.service
systemctl mask getty@tty1.service serial-getty@ttyAMA0.service

echo "---- Updating packages ----"
apt-get update
apt-get -y -o Dpkg::Options::=--force-confold full-upgrade

echo "---- Adding new packages ----"
# For running NAVTEX and RTL-SDR dongles
# (chrony replaces systemd-timesyncd)
PKG_INSTALL="rtl-sdr chrony"
# For development work
PKG_INSTALL="$PKG_INSTALL build-essential pkgconf cmake"
PKG_INSTALL="$PKG_INSTALL libusb-1.0-0-dev git"

apt-get install --no-install-recommends -y $PKG_INSTALL

echo "---- Removing man pages and docs already installed ----"
find /usr/share/man /usr/share/info -mindepth 1 -delete
find /usr/share/doc -mindepth 1 ! -type d ! -name copyright -delete
find /usr/share/doc -mindepth 1 -type d -empty -delete
apt-get clean

echo "---- Headless boot config ----"
# No display, camera or audio; give the GPU the minimum memory
CONFIG=/boot/firmware/config.txt
sed -i -E \
    -e 's/^(dtparam=audio=on|camera_auto_detect=)/#&/' \
    -e 's/^(display_auto_detect=|dtoverlay=vc4-kms-v3d)/#&/' \
    -e 's/^(max_framebuffers=|disable_fw_kms_setup=)/#&/' \
    $CONFIG
grep -q '^gpu_mem=' $CONFIG ||
    printf '\n[all]\ngpu_mem=16\n' >> $CONFIG
# Camera, video codec and sound drivers: with gpu_mem=16 the firmware
# has none of these services, so they only fail at boot. drm is loaded
# by name, so its loader is masked too.
cat > /etc/modprobe.d/blacklist-headless.conf <<'EOF'
blacklist bcm2835_codec
blacklist bcm2835_isp
blacklist bcm2835_v4l2
blacklist bcm2835_mmal_vchiq
blacklist vc_sm_cma
blacklist snd_bcm2835
blacklist drm
EOF
systemctl mask modprobe@drm.service
# Stop the kernel's DVB-T driver claiming the RTL-SDR dongle, so
# librtlsdr gets it without a detach and re-attach on every start. The
# modules are part of the kernel package, so they can't be uninstalled.
cat > /etc/modprobe.d/rtl-sdr-blacklist.conf <<'EOF'
blacklist dvb_usb_rtl28xxu
blacklist rtl2832_sdr
blacklist rtl2832
blacklist e4000
EOF
# No serial console, so turn the UART off
if grep -q '^enable_uart=' $CONFIG ; then
    sed -i 's/^enable_uart=.*/enable_uart=0/' $CONFIG
else
    printf '[all]\nenable_uart=0\n' >> $CONFIG
fi
# No kernel consoles either (serial or HDMI)
sed -i -E 's/ *console=[^ ]+//g; s/^ +//' /boot/firmware/cmdline.txt

echo "---- Swap in compressed RAM only ----"
# Swap is zram0, compressed swap in RAM. rpi-swap's default (zram+file)
# also attaches /var/swap through a loop device: a timer moves idle
# compressed pages out to it daily, and every boot resizes it and
# re-runs mkswap, about 25 s here. zram alone needs no file; its size
# is still worked out from the RAM at each boot, and nothing swaps to
# the SD card. This file overrides /etc/rpi/swap.conf; delete it and
# reboot to go back to the default, and /var/swap is recreated.
mkdir -p /etc/rpi/swap.conf.d
printf '[Main]\nMechanism=zram\n' \
    > /etc/rpi/swap.conf.d/50-zram-only.conf
# Still in use as zram's backing file until the reboot; the space is
# freed then
rm -f /var/swap

echo "---- Replacing NetworkManager with systemd-networkd ----"
# Done last, just before the reboot: NetworkManager leaves eth0
# configured when it stops, so the network stays up until then.
# networkd is part of systemd.
cat > /etc/systemd/network/10-eth0.network <<'EOF'
[Match]
Name=eth0

[Network]
DHCP=yes

[DHCPv4]
# Identify by MAC address, as NetworkManager did, so the router keeps
# the same address
ClientIdentifier=mac
EOF
systemctl enable systemd-networkd.service
# No resolver daemon: keep the DNS servers NetworkManager was given,
# as a fixed file
RESOLV="$(grep -E '^(search|nameserver)' /etc/resolv.conf || true)"
# polkitd too: kept only because systemd recommends it, and only
# started because networkd's hostname lookup (systemd-hostnamed)
# asks it
apt-get -y purge --autoremove network-manager polkitd
if [ -n "$RESOLV" ] ; then
    rm -f /etc/resolv.conf
    printf '%s\n' "$RESOLV" > /etc/resolv.conf
fi

echo "---- Removing old logs ----"
rm -f /var/log/*.log /var/log/apt/*

echo "---- Trimming the card ----"
sync
fstrim -a -v

echo "---- Done: rebooting ----"
reboot

Passwords dead; long live passwords

Passkeys are here – heralded as the end of passwords. Fully supported on iOS devices and in the latest Windows update.

So the theory is good – we replace passwords with public-private key pairs stored in a password manager (so no remembering anything). The computer fields this (no typing), the private key never leaves the device (can’t be observed in transit) and the public key is worthless to anyone who obtains it (so it doesn’t matter if the site you’re using it on gets hacked). And since a password manager is doing the remembering there’s no cost to having a different key-pair for every login – so even if you obtained the full key pair for a site it’s not useful on other sites or for tracking users across multiple sites – unlike passwords that commonly get re-used. It avoids the more obvious phishing attacks since your browser looks up the passkey based on the website you’re using it on – if you’ve typed it wrong domain name (or been mislead or clicked on a dodgy link) there is no passkey to offer up. The password manager won’t sign the relevant bits without you authenticating to your device e.g. with a biometric that again isn’t shared with the website you’re authenticating with – so its claimed this is two-factor authentication as well.

So does it work? I started experimenting on the iOS device – it has the fullest support, and will store passkeys in Apple’s keychain allowing secure sharing to other iDevices. First problem not that many sites support passkeys. So tried Amazon first – and it seems to just work. I can sign in with a passkey from my iDevice, its little face ID do you want to authenticate thing popped up and it did the rest. A bit more experimenting – you can have multiple passkeys to log into a site (e.g. for different devices). Second problem is Amazon’s UI is in its early stages – you can’t delete a specific passkey, only delete all keys! The third problem is there’s no way to turn off passwords – so despite the benefits of passkeys there’s still a guessable / hackable / leakable credential floating around. Setting it to the longest string of random characters I can dream up and forgetting it seems like a workaround I shouldn’t need.

Then a surprise win. Logging in on a PC Windows pops up a QR code, talks to your device by Bluetooth (which then pops up its biometric ID prompt) and everything just works. So while I could have separate passkeys scattered across devices, keeping them in one family of devices seems easier. Doesn’t seem to work without device internet though.

The Google experience is similar – but again, seemingly no way to disable passwords.

Now I have many Windows boxes – only one has Bluetooth. All the others won’t let the iDevice talk, so time to try Windows Passkeys support. Basically does the same magic – in this case your passkeys are protected by whatever Windows normally protects credentials with (Windows hello, biometrics or if you must a PIN). But it’s not as complete – there’s seemingly no way to export or backup passkeys, nor is there any way to synchronise them across multiple windows boxes unlike e.g. Edge’s password manager. So you end up having a passkey per device in Windows world.

So, er, Microsoft? They claim to support it. But many hours of swearing and you end up running Microsoft authenticator app instead, and I can find nowhere across personal or business Microsoft 365 to offer it a passkey. App, security key, windows hello, linked-to-windows, anything but passkey. Although Windows settings > Passkey seems to have auto-generated a passkey for this purpose. Curious. And even if you setup Microsoft’s authenticator app on Business 365 – it’s a sysadmin fiasco navigating the maze and I was ultimately unable to disable password login.

Paypal is then a disaster; offers to create a passkey, but gets stuck in an infinite loop authenticating you. eBay is more curious – works on iDevice; they’ve utterly defeated doing the QR code thing from a Windows box.

So after all that faff what have we learnt? When it works (which it very much does on iOS and does with limitations on Windows) it’s magic. But I’ve been able to get rid of passwords in zero cases so they linger as security holes. And people’s ability to break it (paypal, ebay), make it nearly impossible to use (Microsoft), or just have crap UI (Amazon) means the world has a long way to go to enable widespread adoption.

While experimenting bigger security holes become obvious. Back in the day closing a web-browser meant all the state vanished. Whereas in the modern world browsers are remarkable good at remaining logged in to sites, and even sync this information across multiple computers. So the chore of remember to log out resurfaces – otherwise no authentication at all.

Nothing much changes if you lose a passkey – like forgetting a password a similar reset procedure is needed. Again, control of your email address is your most valuable credential.

There’s also the usual annoyance – old devices don’t understand all of this, so if you want to be password free you need to turn your old devices into landfill; personally I’m not so happy about this.

So after some experimenting there’s a glimmer of hope that passwords can be consigned to the dustbin of history. It’s sad is this is built on stuff like public-private key pairs which we knew how to use long before the internet existed, and that the tweaks to web protocols to make it handle passing the keys around could likewise have been implemented two decades earlier. I guess the game-changer is smartphones handling the biometrics on-device needed to keep keys secure. But I fear passwords will be with us for a while.

Apollo 13

A theatre production is a team effort and it can bring out the best in people. Strenuous efforts where you sprint for the finish line (opening night) are the norm, followed by the relentless feedback to refine the show performance by performance until that last night where you can relax, sit back and enjoy your work – whilst mentally steeling yourself for one last burst of effort to take the show apart at the end of the night. It can be all consuming, tiring and pushes people beyond what they normally would consider reasonable in say their normal work or home life. With age comes experience, less youthful energy, knowing when to give in for the night and take a break and when pressing on is injurious to the goal. You have a calibration of how hard you can sprint for how long. I’m immensely proud of some of the shows I’ve worked on and to date in life they are the benchmark by which I’ve been able to measure other things.

Until now that is. This blog is written on Day 25 of Apollo 13. After 24 days of extreme effort I am uncalibrated once more as to how hard I, and indeed the people around me, can sprint. My employer is part of the government’s ventillator challenge in response to Covid-19 and our internal project name is Apollo 13. There are some good parallels – think “Houston we have a problem”, running out of a scarce resource, having to invent solutions from limited components and above all a sprint of human endeavour to do something extreme. Our morning team leads meeting is “Mission Control”. Today is day 25 – an enforced day off and midnight-to-midnight down tools for the whole team. What follows are some incoherent ramblings about our situation and some thoughts I need to unpack.

I time the project from its official start on Thursday 19th March. The preceding Friday had seen a call with government and our medical technologists had worked the weekend to demonstrate a rapid prototype. Rapid meaning I was able to view the demonstrator video on the Tuesday. I remember over lunch how impressed I was but also how odd it felt not to be part of the response commenting “if the world needed 5G base-stations to fight a global pandemic I’m sure we’d have responded – shame we can’t contribute.” This had rather changed by the Thursday when the job of software verification was hurtling around the company looking for a team lead and a hasty zoom with a rogues gallery of likely suspects convened on the Friday. The best qualified person could not be unplugged from other commitments. Other equally well qualified people were ruled out as the role is incompatible with supporting young children at home (we’d moved to home working that week). Two minutes is not a long time to ponder a major commitment and my glib response to the brief “we need an awkward bastard” may have swung it in my direction. And that was that. Relieved to say I got the person best qualified onto the team to support me – a vital sanity check.

I spent the next hour extricating myself from ongoing commitments. It was timely in one sense – I’d recently delegated most of Technical Authority role on a long-standing project to several people all of whom hit the ground running and I’m pleased to say have proven me superfluous over recent weeks. The other two projects on my plate were doing final reports that week or the following so some loose ends needed dealing with. When the dust settles on Apollo 13 its team will be commended – but so too should be the people who took on extra things to free up people for the team. Everyone I spoke to said “well duh! this out-prioritises, go do, we’ll cope”. Cambridge Consultants prides itself on responding quickly to new projects – but I’ve never seen a team ramp so quickly. By Friday afternoon we were making inroads into our bit of the problem. Nobody even discussed whether we should work the weekend – we just ploughed on.

I’ve never had a sleepless night questioning whether I should take on a project responsibility before. That’s not meant to sound arrogant. Put simply, nothing has mattered before. The worst a signal processing project can normally go wrong is to miss deadlines or go over budget. Class C medical software (translation “get it wrong and you kill people”) is not my comfort zone. At the time it was believed the UK would run out of ventilators in 14 days. The morals of the situation were terrible. Do nothing – people die. Do something too quickly or carelessly – people die. Do it to perfection and you will be late – people die. There were serious unanswered questions from the MHRA as to what concessions against normal product standards would be allowed – even what specification we were aiming for. The critical path is never a great place to be in such circumstance as software was likely to be. The motivation – do something that could save lives – was clear enough and widely understood. I know I struggled with my composure trying to explain our situation as we brought people into the team. How on earth do you describe what commitment is expected from people in such circumstances? How do you set expectations? With hindsight I needn’t have feared – the team around me share an understanding of the situation and everyone is focussed on the goal. It helped to ponder a conversation from years earlier with Simon Long – ask an ex employee about what they worked on at CC and they’ll pour good humoured scorn on the pointless Bluetooth widgets and other daft things they worked on – while telling you that however dull/miserable/difficult the project the thing they took pride in was the thing that made a difference in life (in this case I think some sort of kidney dialysis machine). Anyway. Seize the moment and press on. No time to ponder.

The first week was a blur. Think get up at 6am with thoughts racing, ingest the overnight changes and test data, plan the day ahead in time to sync with the team at 9am and escalate any issues to other team leads by 9.30am’s Mission Control. Inevitable calls and follow-ups. 12.30 team meeting to absorb the mornings work and reset the direction. Not uncommon to get to 4pm when working from home and think “Crap. Should have a shower now and maybe some breakfast”. Then work on in to the evening and often the night. Oddly it’s not a technical challenge – the pneumatics, mechanics and software in a ventilator are not ground breaking – the challenge is to do it in a context where every day counts and where what you have to aim for depends on when you can achieve it. We are not resource limited – but there is a fundamental limit to how fast you can align people and solidify a design and plug in people and hardware to help contribute.

Last weekend was surreal. On the Friday we had a project team video conference with David Baker – one of the original Apollo 13 mission control people – a fascinating insight into a very famous moment in history and he had some thought provoking things to say about what people are capable of when they push as a team. Also a good motivator going into the weekend – our product would leave 5am Monday for government test. Day 17 Saturday saw the first assembly of a full product prototype – electronics, mechanics, software, casework the lot and it was a hell of a relief to see it go together and spring into life around midnight. My corner of the world was a close run thing – we were trying to clear four software milestones in a weekend for a 4pm Sunday code freeze and 5pm delivery to system test. Milestone 3 (minimal viable product) was looking good so we were free to enhance test coverage and push for the stretch target milestone 4. All good in theory until a late breaking bug in 3 came to light at 4.36pm, the test framework had hit issues due to the push for enhanced tests and 4 hadn’t been tested at all – with the test cycle at >20 minutes. We had nothing at that moment – and a timely prod from the TA “will it be ready by 5” was met with a bold text *Maybe* as the best certainty we could give. My longest and most intense 24 minutes of concentration of the project followed as the test framework came together and both milestone releases went through test in parallel. We delivered 30 seconds ahead of the deadline. And breathe(!) The system test went ahead and were done by 2am with 3 hours to spare.

It’s hard when you’re pushing for a target and the target moves. The global picture is complicated and we’re about plan D – plans A,B,C involving pushing pre-existing design production harder, scaling e.g. formula 1 companies to make the parts that cannot be sourced for existing designs, ramping production of more basic export products for the application and so on. Making a new design from scratch is not the obvious answer but it can be tailored both to the disease and to the enormously complex supply chain issues – basically what is in stock or can be manufactured in volume in the UK. And there is a lot of sense to backing multiple approaches. Our plan now has a different end date because we’re aiming for something more sophisticated – the clinical feedback is changing in response to what is being learnt. The world is awash with “dead simple” ventilator concepts that will keep someone alive for 24 hours; less so the more sophisticated support needed to nurse someone to recovery over 14 days without doing more harm than good. And the early designs that do get approved cannibalise the supply chains for the designs that follow and the designs have to iterate. Another sadness around an end date that moves is that if you’d known it was going to move on day one you would have gone about it differently.

It’s bizarre when people are so focussed to think we don’t have certainty the design will see the light of day. If we are stood down it’s won’t be because we haven’t delivered a viable product in an insane timescale – it will be a competing design getting there first or because the need has receded or because the government’s plans A,B,C have over-delivered. Whilst all these things should be celebrated such an outcome would be gut-wrenchingly hard for such a focussed and committed project team.

I try and avoid the press. The companies working on this all agreed to avoid publicity; it’s not helpful and adds enormous stress to the people involved. Dyson is a dirty word for breaking ranks and talking rubbish on this front. The government and associated press releases linked above were probably a response to this – although putting it in the public domain did help in enabling people to talk to their families and friends about what they were working on. It’s sad but I’ve been very careful writing this to avoid naming names and companies involved or to release details into the public domain that should not be.

I struggle with social media. People spout forth with complete certainty as to what should be done or what mistakes have been made. Such people are idiots. The situation is far more complex and nuanced than many imagine. On no day has it been clear what the right course of action is. A vigorous but tolerant debate is needed not such simplistic and entrenched positions. I’ll avoid a long rant here but people need to think harder. Lockdown has its costs – in mental health alone to say nothing of social, cultural and economic harm it does. The economic worries should not be off topic – without it the supply chain for health services is unviable and no you can’t easily delineate what is “essential”. The people who blame the government for everything simply mark themselves out as people who would blame the government for everything in all situations regardless. I try not to get stressed arguing with such people.

Life has its other oddities. If I’m in work I can pretty much chose an entire empty floor of an office building to go work in. Mostly I’m there to ransack labs for test equipment and to build remote-access lab rigs. The doctor doing the evaluations was isolating so while you can’t meet him the cameras for the video links are operated by his daughters because he’s not isolating from his family. Working with a sub-team of 12 people only one of whom you’ve worked with before and only two of whom you know would be bizarre enough in normal circumstances – now with video conference and home working for added weirdness. Michael Gove has written to us as “dear friends” confirming we’re key workers. And so on….

There’s humour too. I remember the general channel in teams carrying “Rumours are circulating it’s gloriously sunny outside. This is FAKE NEWS.”

The hypochondria gets the better of you. The worry that you might be ill or worse might make a project team ill who have to deliver in about the length of time the illness knocks you out for is not a happy thought. I have no idea how much I cough occasionally in normal life – because I do that when I’m in perfect health. I suspect despite pleading with us we’re not being careful enough on social distancing in labs; people are putting it to the back of their minds.

I am grateful to the people who’ve checked up on me. I’m sorry if I don’t seem to listen when you tell me to take a break. I am managing. Theatre taught me well and I know the warning signs when I am tired and I’m monitoring how well I rest. And it’s not like there’s anything else to do at the moment. The team around me are rotating through rest days and I’m trying to look out for them. After 24 days of 14-16 hour days I was pretty tired by the end of yesterday having only had a few hours break on Thursday. I’m not sure of the wisdom of today’s shutdown given I know the challenge of the week ahead but I’m using it to go back maximally recharged. On the 50th anniversary of the Apollo 13 launch it was nice to sign off for Sunday to a message from David Baker encouraging us in our endeavours. There have been phone calls to friends and a BBQ in the sun and so on today. I have food – a large pile of cow does three roast dinners and several curries so my major food groups are covered.

I am deeply proud of the people I work with; they are doing an amazing job in difficult circumstances. I have faith everyone reading this would do the same in the same circumstances which is why I’m embarrassed people compliment me on what I’m doing. I’m just being me and doing what I do with all the energy I can summon. We’re not out of the woods yet. For me it’s all about the end goal. It could all come off the rails in the week ahead just as quickly as it’s come together so far. Mis-steps are costly when the timescale is in days. We all desperately want to see this in manufacture. We steel ourselves for the coming days.

Updates

I’m glad I wrote the above on the one day off we had (12th April 2020). Now the dust has settled it’s worth sharing some links including Cambridge Consultants’ press release Med tech news Mathworks and a tangential reference about UI design

Synergy

84095489_10214065350422487_2183156941985939456_o.jpgCUTAZZ Synergy here we go. Time is short so will be updating one blog post as the days go by this year… No promises it’s an exciting read but if this stuff interests you then you might be interested to know how this show has evolved over time, how the job of a lighting designer has changed over the last decade and why we do it.

Sunday 9th

Straight into (7 hours of) rehearsal with a bang. For the first time ever we have the first run-through (mostly) in costume which I have eagerly videoed with the GoPro. Then headed home to back up the video to the network and upload requested pieces (plus finale) to share with the company. Video and pieces spreadsheet is now organised to a first pass. Didn’t take the laptop along this year – was easier to make pages of illegible scribbles on paper mainly because you can draw what the movement is if something catches your attention.

But it’s looking good and there are some interesting costumes as well (notes say “OWN COLOUR PALETTE” by a number of them!)

Monday 10th

Today is catchup day. I got ahead in creating a showfile and drawing the basics of the rig over Christmas. However a month of back-to-back 60 hour weeks in work and the occasional skirmish with UK productions’ Shrek drawings put paid to being any further ahead. So have re-digested the entire show e-mail trail to factor out the TODO list and per-piece notes, fettled the rig plan and poked the venue for some clarifications on lenses.

The rig has gained a new positon (the stalls bars underneath the circle) which I am hoping will do low-level sidelight better than the 4 PAR Cans I normally use. Plus centre-line highlight. Oh and some speakears! Most of the missing bits of focus done in the visualiser (e.g. new zooms for frontwash and specials). Agonised over backlight gobos, realised that 3-5 working days meant buying two more of them today, spent out of own pocket (budget already spent and time is short etc).

Huge amount of time sunk into the lx bits and pieces – plan has all fixtures addresses and on hopefully the right universes, standard rig is in a different colour so you can see what has moved, masking and focus points all labelled. Legend and key wizards suitably tamed. There are enough gobo holders, DMX leads and powercon link throughs.

Risk assessment, method statements, CDM paperwork, schedule and plans are ready to email to the venue tomorrow morning.

All computers now on the plot desk; all updates applied.

And the curve-ball? Next week some anti-social bastards from Extinction Rebellion are attempting to close the roads all around the venue – which will be a mess for my hire order arriving, taking anything by vehicle to the get-in, the company rehearsing, and who knows what happens for the audience if they haven’t been moved on by Friday.

Long day but still feeling behind. Much to do tomorrow…

Tuesday 11th

I ususally start the day by clearing the correspondence to keep things ticking along on email (plus some time to drink my morning coffee) and today is no different. More videos to share with the company. RobH and I have a pretty slick system re keeping in sync with sound and we run from our version of the truth in run-throughs to flag any problems. So as of Sunday we were chasing five pieces as of last night that’s three.

Risk assessment, method statement, schedule, lx plan and section and masking and focus points plan are all copied to the venue. This prompted the need to draw the colour on the plan for the non-colour-changing fixtures and checking I have the colour is now on today’s TODO list.

Lunchtime says the morning has vanished but desk has been setup from scratch and visualiser is talking to it. Mission for the afternoon is get all the stupid control channels that never move at sensible HOME values (lots of user manuals to check), get the colour palettes in and time-code working so that who knows might get to actually plot some stuff this evening. Then we’re on track.

More of Tuesday then disappears to Hyper Procrastination. 32 All Fade Mark cue in 32 labelled cue lists with time code linked, 32 Intensity block assert blackouts. More venue comms. Visualiser mostly playing ball. Magic sheet taking shape. Colour palettes tamed. Better handling of colour palettes between show files could be the subject of a whole blog on its own.

Untitled.png

WednesThursday 12-13th

Well, I always said there’s no time to blog when you’re getting on with the job. And there’s always more work hiding than there should be. Some quality time spent looking over all the videos and pondering costume and colour palette vs run order. And noting all the ones that can’t be plotted because sound / costume / choreography / whatever – more emails resulted.

And some actual pieces plotted. It’s always slow at first but it’s nice to get going and ideas tend to lead to more ideas. I tend to start by ignoring the clock and seeing where the process takes me – I know now I need to speed up but experience says that happens. Friday is heads down get on with it.

Wyg.png

Friday 14th

Friday was heads down and get some numbers plotted. Started with the ones that have detailed notes (and hence fewer degrees of freedom – both within the piece and with making the pieces either side look different). Then some competition pieces as the sounds is tentatively confirmed, who knows may even match what’s in the video, and we don’t get to see those on Sunday.

Saturday 15th

Postie has delivered gobos and frost. Need to get at least 7 pieces plotted today. Will break to buy sandwiches for rehearsal tomorrow and to print plans for Wednesday.

Stop press! Purple strobe excitement.

Cannot label the plan as to which way around the movers go. Or at least the user manual has no information as to which way tilt goes based on the DMX setting and the orientation of the unit.

Wierd desk bug. Channel 513 is 513 in Park. Unless it’s patched whereupon it becomes 2/1.

Just done a complicated piece – work through it 5 seconds at a time tracking the movement and poking in the focus positions. Then got to the 20 second bit they hadn’t choreographed(!) Fail.

Effects engine many bugs what’s new. Especially where the Stop/Back button is concerned.

Sunday 16th

Today is freeze point; have videoed all pieces and not expecting any new information to come to light between now and the tech/dress. Uncertainties will have to be dealt with as best they can. Need to go over some of Saturday’s plotting as some choreography has changed since last week. But the task ahead is clear (if sizeable). Might have tea first while the video copies across though.

Monday 17th

The inability of the technology to just make me a proper orange colour continues to be upsetting.

Past half way in terms of pieces but don’t quite have a complete first half yet.

Nyaarrrgh. Was all going so swiftly this morning then there’s the one that won’t plot. I mean I want to be in the forest with little red riding hood, just where the shadows are purple and the sun shining through the trees is cyan. Still, naming cues always amuses. TopTwinkleReprise. And StrobeCeaseAndDesist!

Nodding buckets ate my plot time. 20/32 of the way there.

Tuesday 18th

To recall a conversation from last year “Time to channel your inner Britney”. There was some debate as to whether everyone has one of these, even me? Either way, lots of flashy stuff needs sorting out today so it’s heads down to get mostly the second half done. And find the glow tape, lx plans, print bar tapes, make sandwiches and generally get everything ready for being in the venue tomorrow.

Eeek! Just needed the user manual for the lighting desk. Danger. Although I can be forgiven for not guessing that [Shift][Forward-slash] means open-subgroup-bracket.

Right 26/32 means it must be tea time and time to make danger yellow sandwiches for tomorrow. The good news is that the hires have been delivered to the venue. The bad news is the magic smoke has escaped from the venue’s architectural lighting controller so it looks like white worker floods 100% houselights OFF until spares arrive from ETC. Should make for an exciting focus and plot tweak (!) More reasons to get more plotted tonight.

28/32 done but the remaining four are biggies. But my pile of stuff is ready to go to the venue (colour, frost, gobos, bar tapes, plans, tools, sustenance, glow tape). Have typed up my focus palette notes for all the mover cues. Will attempt to eyeball the focus in the morning so it’s fresh in my mind what all 100 fixtures do.

Wednesday 19th

Back from venue. Rig is in and focussed. Always amazes me that the human brain can marshall what over a hundred different lights do. Masking and spot marks are done. Colour palettes fettled (orange fail still makes my cry). Focus and beam positions for the movers mostly straightened out (have found new bits of tension wire grid that get in the way – must draw these on a plan). Here is somewhere where exporting pictures of all of them from the visualizer would have helped but time ran out on that one. Here’s us firing things up for the first time (prior to top masking):

IMG_20200219_144306.jpg

New stalls position for low level sidelight seems to work well as does the extra units I’ve poked under the prosc. to deal with the very front of stage. Rob seems to have sound under control including new and improved over-stage speakers. Many thanks to Anna and Michael for coming along to help rig and focus. Delighted we ran to time today.

Show file has come back home and the visualiser updated to match the venue patch. This is only the 14 conventionals – the rest of the rig being LED/intelligent/etc it was all correct in advance. We’re not a 100% LED rig yet.

Home stretch now – four big numbers remain to plot.

So three done; am giving in on the right side of midnight will do the last one tomorrow. Experience says fewer mistakes that way!

Thursday 20th AM

Social media is convinced I have an event to go to this weekend. Wonder what that might be. Right. 11.15am let’s call that plotted. Time for fish finger sandwiches to power me through the rest of the day. We have three hours to look at it in the venue – check what real light looks like on real people, none of this visualizer nonsense. Main plan from now is to go over the notes – about says there are 536 lx cues to look at so without a system it’d be a long afternoon! And for whoever asked me what the plot desk look like:

GOPR2893.JPG

Programming wing is on the left and by default cue sheet left monitor magic sheet right monitor. Next monitor along is everything that doesn’t have anywhere else to go – so show spreadsheet of doom, rehearsal notes, and spectral view of the sound for setting any timecode cues. Right hand monitor is dedicated to rehearsal video. The tiny laptop does the automation of sound and timecode. The visualiser is normally locked to auditorium view and projected onto the wall. The network storage box that keeps all the information in sync is hiding. And yes with four keyboards and three mice on one desk there is comedic potential for typing into the wrong window! Oh, and the remote for my camera is on the smartphone.

Thursday PM

Fun and games at the venue because the architectural lighting (house lights, workers etc) very not working. Mostly ignored this and got on with looking over the lighting plot. Managed to get through all 32 cue lists by 6pm as scheduled.

IMG_1055.JPG

New magic for this year we ran the dress rehearsal with the technical point in the auditorium. I don’t have the words to describe how much better it is to watch from an audience perspective. In fact its probably the first time I’ve ever been able to watch the whole show from out front. In the less busy pieces totally lost in the moment. Unusually for me I’m mostly happy with how it looks – the moving lights are a good addition, the better colours make good skin tones, the cues and timing are there, it’s looking suitably sculptural.

IMG_1054.JPG

Home now – it’s nearly midnight, video has just finished copying and the show file from the lighting desk is backed up. All set to do the tweaks tomorrow. Not a huge list – mostly need to poke the units that cover the very front of stage and poke a few moving light focus positions now the cast have performed on the right sized stage. Then we should be set for the opening night.

Friday AM

First ETC support case of the week – taking my F915FC colour palettes and [CopyTo] F415FC neither errors nor does anything at all….

If I get time I should file the logs from the venue’s Ion as well – lost use of encoder selects during the plot yesterday.

First half notes done. It’s always quite slow going after the dress – a lot of seeking around the video and very very carefully adjusting focus positions. At this point the focus is slightly off in the visualiser as it has been fixed for the venue (which isn’t perfectly square and true and the lights aren’t exactly 90 degrees to the bar etc) so it’s easier to make small relative corrections. But it’s no hardship to watch the show back 🙂 And it will help later because it’ll be easier to remember what I’m looking out for as the show runs live.

ETC are mailing my colour palettes bug to their developers so have turned around some more details on e-mail. Lunchtime!

Friday PM / Saturday

Is all a blur….

So the odd thing about CUTAZZ is having only a single tech/dress in the venue. A typical show has a separate tech (so change anything) and a dress (whereupon run the first night exactly like this). With only one opportunity to rehearse in the venue the choice was to force errors out into the open. So spots are as narrow as they can be so that if they miss it’s obvious, and the fill under the prosc arch is out unless it’s definitely needed.  Without forcing these errors the show is over-lit – the lights under the prosc broaden the space and take the audience eye out of the stage space thanks to the inevitable spill on the prosc, and the mover spots are too big with no safe way to reduce them. Which is all fine when you can spend Friday fixing this offline in the cases where it needs fixing. Until the desk bug bites – and your colour palettes for the lights under the prosc don’t match the rest of the rig. And that’s when Friday turns into a needless pile of stress. Manually colour matching over 20 palettes between fixtures eats venue time and with other fixes you’re then heading into a first night with a bunch of colour palettes you’ve never seen light the cast and which are changed since you looked over the lighting plot. Not ideal.

Friday is a come down for lighting. The contrast is palpable between watching the dress out front, with an excellent audience eye view and a screaming crowd cheering the show on and the control box experience of being insulated from the crowd with a poor view from the wrong angle with mainly your reflection in the blue-lit box window to look at. Which is a shame as the company had a great night and the audience loved it. Timecode failure on piece 23 was a pain and gets filed as a bug report to ETC.

Notes. An offline plot gets you 90% there. The Thursday afternoon gets you another 9% of the way. There’s still another 0.9% to have from the first night – the odd escape out of the front of the stage lighting and the odd timing tweak. Having given in post show Friday for a much needed beer and a round of drinks to say thank you to the crew who helped there’s an hour or two on Saturday to make the changes.

Some hurried phone-calls on Saturday morning and Rob has some instrumentation to observe what timecode is up to and whether the issue from Friday is the lighting desk or the sending laptop that does the music playback. You don’t really want to run your timecode through this but hey:

IMG_1065.JPG

The bug shows itself in testing  pre-show (it was the desk) but not during either of the subsequent performances. Matinee is a battle with tiredness but the show is looking good and only three notes to fix before the evening. Pub tea intervenes.

Saturday evening is sheer joy. Nothing to adjust as no next performance the experience of running lights is psychic control. Stare out of the box window and get lost in the show. Every time you have that thought they’re about to escape the light or a cue is needed here you look down and the desk has already started running what is needed. It’s nice when you have a show under control to that point.

Then all too soon it’s over. A swift get-out follows – gel and gobos packed up, movers back in their flight cases, cables coiled and taped.  Car loaded, stuff extricated, and that’s it for another year. Thanks are due to Rob for being utterly on top of making it sound fabulous, to Rachel for being totally on top of the stage space and to Michael and Anna for making Wednesday go more smoothly than it ever has before. It’s a luxury to have a 100% trust in the people around you and I hope CUTAZZ realise how fortunate they are to have such competent people help them. On a personal level it helps me relax and enjoy what I am working on. And I think we all appreciate what a privilege it is to work with such a talented and energetic company.

Sunday is for sleeping and for tidying …

Decade

I’ve gone grey as my lights got more colourful; here’s hoping they won’t fall out of the grid. So how does the last decade measure up and what does the next one hold? It’s a story of shows, of friends, of venues and of producing companies as much as it is a story of technological and process evolution.

IMG_7907_p.jpg

Dance is still my passion. No longer for Cambridge Contemporary Dance after the company re-founded but its back catalogue of 2007-2014 is one to be proud of. Nor for the student dance society (variously Manx Gnat, King’s Contemporary Dance, CU Contemporary Dance Workshop and CU Dance Society). A tough decision after 18 years working with them but it was time to move on and the show was losing its way (and eventually lost its slot). But it was a delight to discover CUTAZZ in 2011 and a decade on their enthusiasm is undimmed and it is sheer joy to work on their shows.

16832338_10158380936645151_8124646396730363281_n.jpg
2017 CUTAZZ Pulse

I abandoned paid freelance work in the last decade. While I miss the exposure to other designers, lighting desks, toys and technology on large corporate budgets the money was never good at persuading me to be in London for a 4am call! It’s a different contract with amateur shows. At one extreme you have no need to commit anything for the producing companies that get it wrong; at the other you can expend blood, toil, tears and sweat far beyond what would be economic for the projects you love.

10068_1451977721681403_845304785_n.jpg
My last design for CUCDW Fantasy in 2014

My connections with my student contemporaries have dwindled. I’m delighted to see so many go on to be famous and successful but I miss the development shows at Battersea Arts Centre, the ADC and the Edinburgh fringe. No touring, no trips to London for Resolution at The Place and no taking over Her Majesty’s for Channel 4 on a Sunday afternoon this decade.

The decade gone by was blessed with friends. I am lucky to know so many talented people in Cambridge and for the reciprocal commitments that arise with such a full calendar of amateur shows.

Then there is the drinking club with a theatre problem  – or is that the other way around? Despite a decade of designing in Cambridge it took until 2007 for the Penguin Club to recruit me and back in the noughties its shows were lit by one or two usual suspects. So much has changed over the decade and now the club counts a double-digit number of lighting designers working most years and still more opportunities than Penguins. I was even persuaded to serve time on its committee – at least I now have the T-shirt.

IMG_0298.JPG
Priscilla Queen of the Desert in tech rehearsal, 2017

Co-designing was the biggest change. Spam High Res-44.jpgThere are a lot of facets to a good lighting design – the concepts, style and look, spotting cues in rehearsal, the rig to support the design and implementing and focussing it in the venue, programming, tweaking and operating live, practicals and tuning things as a show beds in. Having overlapping but different skill sets gives a huge boost to the productivity and to the quality of the end result. Co-designing Hot Mikado with Rob Loxley was a “Duh! Why haven’t I co-designed before” moment and it also got me back into lighting musicals. It’s been a productive working relationship and there are many shows to be proud of – Secret Garden for its tight focus enabling the stylised use of colour on such a confined stage, Avenue Q for its outstanding cast and for being hilarious fun, Spamalot for heralding the return of hired sets and for fitting in the venue and Priscilla Queen of the Desert for its sheer scale and ambition to name but a few. There were funny moments and frustrations too – I should tease Rob for vastly increasing his collection of incandescent lightbulbs over the decade though rarely in time for the first night! Although that’s a bit like my collection of dratted Eagle gobos I suppose.

578051_10100281537613530_413869243_n.jpg
2012 Hot Mikado

10482598_10155598639075655_3454746245907716830_n.jpg
2015 Secret Garden

The venues and companies have changed too. Back in 2009 the received wisdom was that the ADC (with more lights than seats in the auditorium) was a fantastically equipped venue and the amateur shows in there were first rate. The Mumford redeveloped right at the start of the decade. It made minor changes to the dimming and invested in a large pile of Source 4 and ADB Fresnels that served it well – the rig was modern and practical and well suited to the space. When the Leys Great Hall re-opened in 2014 they spent the bulk of the money on the infrastructure and left the rig alone. LED’s time was coming and they’ve augmented the rig with new LED fixtures every year since – standard rig is entirely LED and very flexible for it. It’s a great space and a welcome addition to the options for Cambridge drama. The ADC redeveloped the building in multiple phases but until the most recent phase left the lx rig alone. I often think the ADC has suffered from its starting point of being well equipped as it has been overtaken by venues that were less equipped but spent the money more recently. It’s Patt 243s are 59 years old this year! If it retired its conventionals faster it would have a more suitable and flexible LED rig – and supporting the conventionals with a brand new 120 ways of dimming right at the end of the decade was surely bonkers.

1609596_10100860014032150_2004310778248200524_n.jpg
2014 Pied Pipers Guys and Dolls in the sewer scene

Cambridge has a great pool of performing talent but I worry that the producing companies have not kept pace with the changes to technology and the production process. And the less said about the show budgets the better. The ADC ticket pricing is a contributory factor – it’s depressing to see amateur shows elsewhere sell for twice the price and the results are clear to see on stage. Cambridge Theatre Company deserve a mention here for bucking the trend – for its ambition and high production values and for proving you can make this work on decent budgets outside the ADC. I don’t wish to be negative about the other companies – I would love nothing more than to see Festival Players and Pied Pipers explore other venues, CaOS do more modern shows like Sister Act, and maybe another town group return to the Arts theatre.

13557757_10101753932173030_5924279104154717971_n.jpg
CTC Spamalot at the ADC in 2016

 

Spam High Res-5.jpg
Less of this?

Technology has been continually evolving over the decade. A decade has gone by where I haven’t drawn a paper lighting plan and haven’t used a Strand lighting desk. For the big shows they were all pre-plotted and all have rehearsals captured on video (latterly in HD). Most were pre-visualised. At the start of the decade my shows were mostly lit with conventionals – today they are a rarity. ETC have made huge progress – back at the start of the decade my rant “Ramblings of a grumpy Strand user” was the top hit in Google if you searched  “ETC Ion Strand“. But continual improvements to the software, good customer support, retirement of the wretched Ion face panel layout, timecode and a competent offline mean the rant has deservedly slipped to at least sixth. The iPad, released in 2010, is now a default control surface for wandering around the venue with. Programming operates at a higher level – fewer numbers, less writing fixture profiles, more labels and reference data. User interfaces have improved and you can pretty much organise your control surface how you like.

IMG_0297.JPG
More like this?

Programming wings are commonplace. Interoperability with MIDI, Ethernet, media servers and show control has all improved. A lot more can be done outside of the venue. I can now pre-visualise in gratuitous 4K UHD!

 

Predictions are hard – especially about the future. What for the decade ahead? Despite the leaps and bounds it has made control of lighting has much room to improve. Augment 3D is one to watch – as is the general potential of AR when combined with an understanding of the underlying geometry of your production space. I suspect my gel guillotine will gather dust for the decade ahead. While I don’t know what will replace it DMX as a form of control is daft and running up against limits. Who knows someone might even come up with a wireless control system you can trust – although I’d put power and data over the same cable as an alternative contender given lights are always going to need power. I hope and pray the world gets better at colour space management and colour calibration. Visualisers need to become more accessible and more integrated with control systems. Better interchange of data (control and optical performance) between manufacturers, control systems and visualisers.

IMG_0281.JPG
Plotting at home

What do I regret? Mostly that I focussed so hard on what I was doing I rarely had time to capture it on camera. My eternal thanks to Duncan Grisby, Claude Schneider, Rob Loxley, Peter Buncombe and many more for capturing some of what I’ve illuminated. Perhaps I don’t regret failing to buy a van load of Vari*Lite VL2000s. Perhaps.

What of the decade ahead? I should probably light some drama. I plan to raise my game, work harder and strive for better shows. Mainly though I plan to spend more time in the moment.

CUTAZZ - Ben Griggs (66 of 77).jpg
Gratuituous anti-gravity. Because dance.

10985362_10155371486705151_6742987823874206459_n.jpg
Not lighting people is also fun.

SC dress-90.jpg
Beams, choreography, dance sidelight all still pretty.

Happy New Year!

Compromise

“Design is the art of compromise.” is an uncompromising statement. It forces us to confront the negative concepts we have of compromise and elevates that to an art form. Seriously? Nobody likes to compromise. Dictionaries talk of expediently accepting standards below desirable and of settling disputes by making concessions.

And yet nothing is perfect so compromises are inevitable. If I move the masking off stage I accept the show won’t mask from seat A1, and I’ve traded that for better views for many more people or bigger dance numbers or whatever. That’s a good compromise – an optimal trade-off given an unmoveable constraint (the shape of the building).

There are many more subtle compromises. Almost every light you focus is a compromise between where the wanted light is going and what the beam flare is spilling onto that it shouldn’t. Time is finite so there’s always a time tradeoff – more time spent plotting the finale or more time spent making the more mundane bits slightly better. The rig is finite – should there be more lights dedicated to specials that are used briefly or a better general cover that will be used throughout the show? Making a given dance number flashy is more effective if you don’t make all the other numbers flashy, but in keeping your powder dry the other numbers are somehow not as pleasing. These are all more interesting compromises because they don’t have obvious solutions. They require skill, judgement and experience to get right, and “right” is a matter of opinion – it’s art not science. Bring it on – it’s why designing things is interesting!

Compromises can mount up though. For every idea, light or cue that makes it into a production there are probably several ideas that don’t – each show is a tradgedy of discarded ideas and unexplored potential. These ideas go silently to their graves, known only to the designer, who is in danger of measuring the show against what might have been. I’m often critical of my own work and of the productions I’m involved in. And kindly rational souls tell me “don’t be silly – given the circumstances you were working under it was the best that could be achieved”.

This is where it gets ugly though because it forces you to evaluate the circumstances and how you feel about them. If a production company says to me “there’s no budget this year because we’re saving everything for our big centenary show next year” and I understood the rationale at the outset and it still seemed worth doing, fair enough. I chose to accept the constraints. The bad ones are the ones where the circumstances that lead to compromise were avoidable or unnecessary. The shows where the budget is wrong simply because the people who set it were out of touch, uninterested, on auto-pilot or just “doing the same as last year”. The ones where despite knowing when the show was many months in advance there’s still no useful run-through until three days to go. The ones where these failure modes were pointed out when accepting the project and yet the errors are still repeated as a sort of slow-motion car crash. Doing your best “given the circumstances” is less consoling if the circumstances are infuriating. The gap between what you produce and what might have been grows so it’s less rewarding – but the effort that goes in goes up since you’re also working around the problems. Less for more.

The pernicious long term effect of all this is it forces you to choose your shows more carefully, to take a chance on fewer things, to negotiate harder up front, to force choices and budget before accepting the project. I don’t want to travel too far down that road. Perhaps its time for a long break from shows…

Evita-t-minus-6-weeks

Week t-minus-6 has been and gone. Rehearsal schedule update has moved the first stagger later to t-minus-11-days. CAST Software have sent me updated SpectraCyc 200, Roxie and ColorSource PAR files which I have incorporated into the model. Essentially we are stalled. No change re set but there may be conversations about that early next week. I’m going to spend some of the bank holiday playing trying to get organised on RDM and other new desk features. From tomorrow it’s t-minus-5 weeks.

Evita-t-minus-8,7weeks

Evita87 because, well, Evita is the next show and this blog covers 8 and 7 weeks to go…

We’ve been naughty lampies. Back at the beginning we said “we don’t do set design”. We said this because techies with CAD tools and a good understanding of what’s in the stock pile are great at drawing stuff on plans that’s easy to make. It definitely skews things towards “functional” rather than “artistic”; it leads to “structural” more than it leads to “coherent visual look for the show”. It is not our area of expertise – people called “set designers” are widely believed to be better at it. However, the production team doesn’t have one of those, so….

We met with the director to go through some initial thoughts on lighting with 8 weeks to go. The discussion broadened into the set since a rough idea of the space we’re lighting is kinda important. And yes, naughty us, we drew some plans.  Iterated them at 7 weeks to go. And so we turned our attention to what we could work out on the lampie front given what we know already re set. Which is?

It’s not giving too much away to say there’s a raised section at the back of stage and space for dancing in front of it. The upstage part of the raised section may also be home to the band, who we might (TBD) hide behind a cloth at various points. So what is the new ADC lighting rig and remodelled auditorium good for?

Basics. Frontwash. Tentatively assuming we want a ton of sidelight (this being a musical) the frontwash had might as well be straight on. So let’s play with the new Source Four Lustr2’s and see what happens. Immediately we don’t quite understand the bridges. At least not the combination. Bridge 1 naturally does what the old bar zero did quite well – if you want a second layer wash starting upstage of the safety curtain it will do fine – in 26 degree the angles are about right US-DS although ideally you want a bit wider to smoothly wash left-to-right:L2-26.PNGWhat is difficult is that Bridge 2 is so shallow that anything you put on it to light the forestage massively overlaps the next layer. Here’s two layers of 26 degree units in section – with layer two tweaked upstage to meet the front of the deck. What you see is that L1 also reaches the front of the deck.26 both bridges.PNGA bit more experimenting says that to span the width of the forestage right to the corners in Lustr says we need three 36 degree units, or five 19 degree units, from Bridge 2. The 19s work better in terms of overlap. There is no plan to use Lustr as frontwash – no reason we would waste decent colour-changing units on frontwash – however, given the lens pile is restricted (8 of 19, 26, 36 degree) then it is useful to know that eight 19 degree units (5 making layer 1 wash from bridge 2 and 3 forming layer 2 from bridge 1) would work if they were leftover. So three RJ614 will do the whole stage from bridge 2, and reach the corners, and they zoom so we can get the angles right, and having set them up to cover the width they will stretch back to the deck as well. They won’t be desperately bright, but we don’t need them to be, and it means we will run them closer to full brightness and get a better colour temperature from them. This may change pending rehearsals if we need to be more selective with the frontwash but it’ll do for now. We did try a zillion other things with the Lustr – too many to list here – but nothing seemed to quite work for us. Fans from either side didn’t seem to help.

Important stuff. Sidelight. Pairing a 26 and 36 degree Lustr to do the sidelight seems obvious (that’s my ballet brain talking) and seems to work. So the 8-of-each-lenses do forestage, downstage, midstage, and maybe deck. Hard to say without a sidemasking plan. And with the question hanging over the cloth it’s possible we need to split things US/DS of the cloth to cover the band deck. Hiring lenses for the nominally leftover 19 degree units is an option.sidelx.PNGA pile of Lustrs doing sidelight suggests a new option – can the ColorSource PAR (normally a shoe in for sidelight) do backlight? With the round lenses there are issues. You end up with only three units to cover the width – and a non-sensible overlap – but upstage-downstage this barndoor-free unit will dribble all over the front of the auditorium – and not be desperately punchy at the same time. The oblong lens suggests four on a bar and three bars will cover stage nicely and with a bit more punch.

One thing we’re fairly sure of is that the show will want one or potentially several followspots. Having seen them used from bridge two in recent shows they look really good – no daft beam the length of the auditorium lighting up the haze/dust, and about the right steepness. However, there’s a catch. We don’t yet know if we’ll need to followspot on the deck – an issue because you can’t. From the Lx Box a spot can light a 6′ person stood on 6′ deck (the prosc. height being 12’5″) and this gets worse the further upstage you get. The new bridges make this slightly worse. But from the bridges it’s basically not possible to spot someone on deck – they’re similar in height to the lx box spot, closer to the stage, and from bridge 2 bridge 1 gets in the way as well. Perhaps we throw some moving lights at the problem.

Some experimenting

So for what we know we have some sketchy thoughts to cover off the basics in terms of washing the stage. It leaves a ton of TBDs – mainly around having pencilled in all of a given type of unit and then needing more to deal with the deck or band. For Lustr we might hire lenses, for ColorSource it’s not an easy hire option (a different type of unit being colour matching pain at least). And the new rig? Partially working for us (it’s progress that all bar three of the units we’ve mentioned are colour-changing) and partially a source of worry (bridge 2 angles are shallow, the lens pile is restrictive, would be great if there were some punchy LED fresnels with barndoors). Who knows we might rip up everything I’ve written here as more detail emerges – and experience says the first time you light a new rig or venue is always the hardest.

It’s a start. I’ve done all the procrastination things. There’s a wyg model with 27 blank scenes, a load of side-masking and top-masking drawn just waiting to be dropped into place, and some tentative lx bars.  The rest is pending on blocking, run-throughts and set.

As I write this it’s Sunday – 6 weeks from now we’re in the get-in. More news as the situation develops…

Redevelopment

So the ADC has done (another!) redevelopment – this time replacing the ceiling and adding air handling in the auditorium. The lampie interest is of course the remodelled ceiling and lighting bridges. This led to quite a lot of updates to the Wyg model and stage plan. It transpires that while the 3D CAD needs to know about a load of stuff it didn’t previously – ceiling shape for example – on the actual stage plan this stuff is better omitted for clarity. But for you, dear reader…

So what’s new? Well in plan view you have two lighting bridges with lx bars on the front and handrail on the rear edges. The red lines show the ceiling – the centre third of which is flat with the edges sloping to meet the top of the auditorium walls. It also slopes the meet the bridges, and the shape between the front bridge and the proscenium is a bit more complicated:

Plan.PNG

In section view it makes a bit more sense – again the red is the ceiling height in the centre, the blue the top of the auditorium walls:Section.PNG

Astute readers will have spotted an issue that has bitten a few people – the ceiling ends up lower than the lx bars at the edges. Hence a whole new plan (“Front Elevation”) has been added to the PDF drawings – with the relevant bit shown below. The other thing to beware is that there are quite a lot of vertical supports to the lx bar – you can’t just rig anywhere. Hopefully these are visible enough as circles on the bar in plan view.

Elevation.PNG

How does it compare to what went before? Well before I consigned the old CAD layers to the dustbin of history I first made them all pink and plotted a few comparisons (old is pink). So the houselight bars move a bit and the new bridges span the position of the old side bars. I’ve shown the 614 on the bridge about as far as it can be rigged towards the edge – whatever you do with your front-light it’ll definitely be more flat and less side-lit in the new world.

OldNewPlan.PNG

What of the angles? Well here’s the section:OldNewSection.PNG

Bridge one is farther from stage than the old bar zero position – welcome for less steep spots on the forestage – perhaps less so for layer 2 frontwash which will end up shallower. Bridge two is even shallower than the old Dome position (also shown). I’ve yet to play with this in detail – but this probably matters more now the layer 1 frontwash isn’t from the sides as the spill will go further upstage and not vanish off into the side-masking. Also shown is the old and (much larger) new DSM desk.

I’ll write more about frontwash when I’ve had time to experiment.

Finally, if you like playing spot-the-difference, here’s a link to three previous versions of the ADC.