Repository navigation
A release of OOS, an open operating system for the Rigol DHO800/DHO900, with the work of the community in it: the two forks merged (PR #7 by fumingcaps/ichnograph, PR #8 to #13 by W4R/wasimlhr), and many points from the forum thread: every decoder a module and leaving out what you do not use, settings that no longer change each other, STOP over the whole record, Mean and RMS on deep records, a second screen with its own view.
Experimental. It has run on a DHO924S, and DHO804s run earlier versions. Read the disclaimer, the status and the known limits first. Measurements are not calibrated to any standard.
Installing and updating
No build needed: Installation.
- From 0.1.7 or newer: install
oos-0.1.13.oosupdthroughhttp://<scope>:8080/update(or System > Update). The scope restarts into the new build by itself (kexec) and installs it from there. Then switch it off at the mains, wait 30 seconds and switch it on: that start is the trial. - From 0.1.6 or older: install 0.1.7 first, the same way, then this one (details in the 0.1.9 release).
- USB stick (
DHO800_DHO900_OOS_Install.GEL, Storage > Upgrade in the stock system): installs any version directly, and also updates an older OOS build on the card.
What changed
From the forks
- PR #7 (fumingcaps, DHO804 fork): the 2-channel gain lane fix (a DC error of about 7 % of the offset with the channels at different V/div), DMA reads of at most 4 MiB, dots mode keeping the edges of min/max pairs, Min persistence without flicker, STOP that really stops, SINGLE with one frame, the IPv6 link-local fix of the DHCP script. His FPGA functions come along opt-in and off by default (
/oos/system.conf): hardware segments with time stamps, hardware averaging, the hardware DVM, mask and zone tests in hardware, UART/SPI/I2C hardware triggers, AUX OUT, live min/max compression, the ADC core offset trim, and the batch draw pipeline (Acquire > Draw path). So do SSH (Dropbear with OpenSSH's sftp-server), the LAN firewall and IPv6. - PR #8 to #13 (W4R, OOS Plus): SCPI that answers every query and covers much more of the stock command tree, waveform import (CSV, OOS BIN, Tek WFM), an offline mode, many more measurements and measurements on references, SIM, the eye diagram, deskew, Measure > Dock, Cursor B source, the compact bottom bar (Display > Bottom bar), a desktop build. From the second round of PR #8 the fixes are in (SIM and references, math with references, the power page, the eye on a held frame, dock and placement, overlapping labels); its Studio layout is not.
New
- One measurement page for everything: categories on the left, sources and groups, one tap adds or removes a measurement. Lists scroll by bar and finger.
- Every decoder is a module, the core ones too, and so are the measurements and the analyser pages (protocol modules).
- Leave out what you do not use: System > Features switches each analyser page, SIM, Load file and each decoder off; it is then gone from the menus and runs nothing. At build time every optional decoder is a Cargo feature (
OOS_DECODERS="usb dcc",OOS_DECODERS_OFF="avionics"), 260 KB less without them. A setup naming something the build does not have loads the rest. - STOP keeps the whole record and reads it again for every view: zoom out to see all of it, zoom in or move the delay for every sample of that part.
- Display > Monitor view: with HDMI Layout the monitor shows another view than the panel, e.g. the spectrum or four windows, while the panel keeps the controls.
- Short time bases draw every frame they can (Standard draw path, 4 channels at 100 ns/div: 29.5 instead of 3.5 frames a second); Acquire > Draw path chooses PR #7's batch pipeline instead.
- Acquire > ADC cores = Matched takes the offset pattern of the eight interleaved ADC cores off every frame (spur at 156.25 MHz: -48 to -76 dBc on a DHO924S); Acquire warns about aliasing below the ADC's rate.
- Trigger level: no mark with AC or LF reject coupling; with DC a dashed line for 1.5 s after every change.
- Architecture decision records in docs/adr: what was decided, why, and what follows.
Changed
- Mean, RMS, Std dev and Area of deep records (above 1 M points) come from raw reads of the screen window (
meas.deep_raw, on by default): 0.711 Vrms for a 2 Vpp sine from 10 M to 125 M on a DHO924S, about 5 % of the records. Max, Min and Pk-pk come from the min/max pairs. - One setting no longer changes another (eTobey in the forum): on the hardware a trigger type and a decoder protocol set only their own parameters, not the channels, the source, the level or the slope; Auto keeps a memory depth you fixed. New interplay tests check it.
- Frame 0 of a hardware batch of several frames is left out (issue #5,
acq.frame0 = keepkeeps it); the record path no longer splits a batch over two screens (issue #6).
Fixed
- The first record after a change of settings at deep memory could show the older memory of the settings before (twice the RMS right after the depth was changed).
- A deep record measured from its raw window no longer shows "not available" while the measurement runs.
- SCPI over the raw socket answers again when the client closes its side after the command (
echo "*IDN?" | nc <scope> 5555). - Graphics = GPU: buttons and dialog frames covered their area twice; with the measurement page open and HDMI Layout the frame rate rose from 25.3 to 27.4.
Full list: CHANGELOG.md.
Files
| File | What |
|---|---|
oos-0.1.13.img |
boot image, for the second-card install in the README |
oos-0.1.13.oosupd |
update package for a running OOS (0.1.7 or later), signed with the maintainer key (id 4dbd03d9763869dd), which every build trusts |
DHO800_DHO900_OOS_Install.GEL |
install from a USB stick through the stock system's Storage > Upgrade, without opening the scope; keep its name |
SHA256SUMS |
checksums |
linux-6.12.115-base.tar.xz |
kernel source, needed to build the kernel: put it into kernel/src/ |
busybox_1.37.0*, e2fsprogs-1.47.2.tar.xz, wpa_supplicant-2.11.tar.gz, kexec-tools-2.0.32.tar.xz |
sources of the other programs in the image |
dropbear-2026.94.tar.bz2, openssh-10.6p2.tar.gz |
sources of the SSH server and its SFTP subsystem (new in 0.1.13, off by default) |
nftables-1.1.3.tar.xz, libnftnl-1.2.9.tar.xz, libmnl-1.0.5.tar.bz2 |
sources of the LAN firewall (new in 0.1.13, off by default) |
mesa_26.0.8.orig.tar.xz, glibc_2.43*, gcc-16_16-20260322*, libdrm_2.4.131*, zlib_1.3* |
sources of the GPU helper's libraries (Mesa, glibc, libstdc++ and libgcc, libdrm, zlib) |
State
On a DHO924S: this exact package installed from a running OOS through the web page (from the build before it, which had come from 0.1.12 the same way), started by kexec and installed; then switched off and on at the mains: the trial start passed and confirmed itself. The quick tier of the device test on it: 56 of 61 passed; of the five, three two-channel tests did not run as meant because the second channel was wired to the probe compensation output instead of the generator, the generator test with a step of 0.04 V at 2.5 Vpp fails on 0.1.12 too (the generator hardware), and the frame rate case below. Before that the deep-memory measurements at 10 to 125 M ran three times, the rates with and without the raw reads, and the trigger types and decoders round trip on the hardware.
Known: with Graphics = GPU, HDMI Layout and the measurement page open the frame rate is 26 to 27 frames a second, at the test's limit of 27 (the same before these changes).
Bug reports, test results from other models and code are welcome.