Dear all,
I am currently testing NovaSDR 2.0.0 with a RTL-SDR V4 dongle and and having some issues with playback stopping randomly when demodulating FM signals. It looks like other users are having the same issue [1,2]. The behavior can be reproduced for example by selecting FM for both VFOs and switching between them. In the debug panel, the audio delay drops to 0 and playback is stopped while the waterfall is still working. The behavior differs when NovaSDR is compiled and run under Ubuntu 24.04 or 22.04.
On Ubuntu 24.04 the playback completely stops. On Ubuntu 22.04 the stream seems to be re-initialized (audio delay still goes down to 0 and goes up again). As the current version of NovaSDR (spectrumserver and frontend) does not compile on a vanilla Ubuntu 22.04 Jammy, my current workaround is to compile the frontend on Ubuntu 24.04 while compiling and running the spectrumserver under Ubuntu 22.04 using G++-13 installed via PPA. However, this is quite cumbersome and since Jammy is reaching its end-of-life in the not too far future, it would be great, if this bug would be fixed timely.
73+55,
13RF7373
[1] https://www.phantomsdr.fun/index.php?topic=171.0
[2] https://www.phantomsdr.fun/index.php?topic=178.msg994
It would help if you post the toml you use.
Quote from: Bas ON5HB on 26 Oct 2025, 14:04 It would help if you post the toml you use.
Dear Bas,
please find attached my config file.
Best regards,
Christian
P.S.: In the meanwhile I figured out how to use a newer nodejs/npm version under Ubuntu 22.04, so I can compile both components under Ubuntu 22.04 now.
# NovaSDR Configuration — RTL-SDR (curated)
# Device: RTL-SDR via rtl_sdr
# Capture format: u8 (8-bit unsigned IQ)
#
# Run example (WBFM @ 100.9 MHz, 2.048 Msps):
# rtl_sdr -g 48 -f 100900000 -s 2048000 - | ./build/spectrumserver --config config.example.rtlsdr.toml
#
# Notes:
# - Conservative defaults chosen to start reliably on typical hardware.
# - Keep waterfall zstd + FLAC audio; these do not require extra codecs.
# - Avoid "opus" and "av1" unless NovaSDR was compiled with those codecs.
# - If CPU/memory is tight, reduce input.fft_size, then input.sps.
[server]
port = 80
host = "0.0.0.0"
html_root = "frontend/dist/"
otherusers = 1
threads = 1
[websdr]
register_online = false
name = "NovaSDR (RTL-SDR)"
antenna = "2,75m Boomerang"
grid_locator = "JN59"
hostname = "novasdr"
operator = "13RF7373"
email = "webmaster@novasdr"
callsign_lookup_url = "https://www.qrz.com/db/"
chat_enabled = true
[limits]
# Multi-user caps; adjust to your host capacity
audio = 100
waterfall = 200
events = 200
[input]
# Core parameters
sps = 2048000 # 2.048 Msps
frequency = 27415000 # Center frequency (Hz)
signal = "iq" # RTL-SDR outputs IQ
# Performance and quality
#fft_size = 131072 # Reduce if OOM or high CPU (e.g., 65536)
#fft_size = 262144 # Reduce if OOM or high CPU (e.g., 65536)
fft_size = 524288 # Reduce if OOM or high CPU (e.g., 65536)
fft_threads = 1
brightness_offset = -18 # try small steps like -6, -4, 0, +4
audio_sps = 12000
waterfall_size = 2048
waterfall_compression = "zstd" # Leave as "zstd" unless AV1 compiled
audio_compression = "flac" # Leave as "flac" unless Opus compiled
accelerator = "opencl" # "none" | "opencl" | "cuda" | "mkl" (only if compiled)
smeter_offset = -24
[input.driver]
name = "stdin"
format = "u8" # rtl_sdr outputs u8 IQ
[input.defaults]
frequency = 26925000
modulation = "FM" # AM | SAM | FM | WBFM | USB | LSB
Toml seems ok, what hardware do you use for OpenCL?
It is a Radeon RX 6700 XT. But it also happens with CPU processing.
Are you running a desktop at the same time? As we saw other problems reported with using Radeon+OpenCL+Desktop.
Therefor most of us use Intel with internal GPU. Gen6 or newer series.
Quote from: Bas ON5HB on 30 Oct 2025, 14:49 Are you running a desktop at the same time? As we saw other problems reported with using Radeon+OpenCL+Desktop.
Yes. But how could that interfere also with the purely CPU-based processing without OpenCL?
CPU based isn't fast enough. I tried on an AMD Ryzen 8 core, it could not do what an old GPU can.
FM is a very difficult mode for SDR as it's not audio-based.
I just tested it under Debian stable (trixie): same behavior as under Ubuntu 24.04 - playback stops in FM mode after switching frequency.
We use Ubuntu Jammy 22.04 not 24.04 as we had more problems.
I tried to recreate your playback problem but don't have it.
http://websdr.heppen.be see if mine has it too for you.
Could be your browser, I use Firefox.
BTW, Jammy LTS will not end before Apr 2027, and total support stops Apr 2034, long time to go and work out the problems.
There is no need to use the lastest version, as this is Linux not Windows ;D
Quote from: Bas ON5HB on 05 Nov 2025, 15:41 We use Ubuntu Jammy 22.04 not 24.04 as we had more problems.
I tried to recreate your playback problem but don't have it.
http://websdr.heppen.be see if mine has it too for you.
Could be your browser, I use Firefox.
I use both, Firefox and Chrome, both with the same behavior. And the bug does not occur with your instance. That is Jammy, right?
Quote from: Bas ON5HB on 05 Nov 2025, 15:41 BTW, Jammy LTS will not end before Apr 2027, and total support stops Apr 2034, long time to go and work out the problems.
There is no need to use the lastest version, as this is Linux not Windows ;D
That period is only for certain flavors of Ubuntu and for certain components like main and restricted. If you are running Kubuntu like I do, then all your desktop packages are already out of support (3 years, ended in April 2025). Same holds for the universe and multiverse components.
I do not run the desktop. I can start it but it's disabled by Grub-starting-entry.
Strange you mention desktop packages being out of date, this is not the case with LTS versions.
Only the standard version has this.
But why run desktop-packages on a server at all? Server distro's are always LTS.
I run Jammy. My Apt-sources:
deb http://archive.ubuntu.com/ubuntu jammy main restricted universe multiverse
deb http://archive.ubuntu.com/ubuntu jammy-updates main restricted universe multiverse
deb http://archive.ubuntu.com/ubuntu jammy-backports main restricted universe multiverse
deb http://security.ubuntu.com/ubuntu/ jammy-security main restricted universe multiverse
I suggest you do the same.
Quote from: Bas ON5HB on 07 Nov 2025, 17:34 I do not run the desktop. I can start it but it's disabled by Grub-starting-entry.
Strange you mention desktop packages being out of date, this is not the case with LTS versions.
Only the standard version has this.
But why run desktop-packages on a server at all? Server distro's are always LTS.
The release notes explicitly states that most flavors have only 3 years of maintainance:
"Maintenance updates will be provided for 5 years for Ubuntu Desktop, Ubuntu Server, Ubuntu Cloud and Ubuntu Core. All the remaining flavours will be supported for 3 years. Additional security support is available with ESM (Extended Security Maintenance)."
[https://web.archive.org/web/20250513154743/https://fridge.ubuntu.com/2022/04/21/ubuntu-22-04-lts-jammy-jellyfish-released/]
5 years maintenance sounds great on paper. In practice, most maintenance, bugfixing, and development is actually done by the upstream maintainers i.e., the developers of all the software projects that are bundled as packages. They are mostly volunteers and not paid for fixing bugs in old versions. They are fixing it in their development branches that are then published as release versions. The Ubuntu staff will backport some of those fixes, if the bugs are really critical, but most fixes are brought in into the next Ubuntu releases. That is why I would always recommend to upgrade your LTS system within a year of release (to the first minor update version) no matter which flavor you use.
Since you added universe and multiverse as apt sources, you might want to have look at some reddit threads like this one, that discuss this topic:
https://www.reddit.com/r/debian/comments/1i8qm0v/how_to_explain_my_users_the_difference_between/
Regarding the question why I use desktop packages: for convenience. I'm running NovaSDR only locally for myself for testing. Since there are no binary package and since I have to compile it by myself, I need a complete development toolchain and an IDE or editor for making changes. And since Ubuntu Server and Ubuntu Desktop and all the desktop flavors share the same packages, they actually only differ by the packages that are pre-selected during installation.
Sorry but I fail to see your point, Linux is not Windows and it really doesn't matter.
You can add packages yourself if wanted. Typical many packages do not go out of date.
Windows needs constant updates, Linux does not.
But I'm starting to see your problem, you probably installed the compilers from the Ubuntu package sources.
You may want to uninstall those and read this page to install the ones we use:
https://github.com/ny4qphil/PhantomSDR-Plus/blob/main/INSTALL.md
And this may help too:
https://www.phantomsdr.fun/index.php?topic=166.msg931#msg931
I do not know if Nova works with the new compilers yet, I have not heard it does.
And please stop the Ubuntu is old etc stuff, we all know this, none of us cares about it as we run it as standalone server and only this package.
Someday it will be updated and run under a newer version, but the developper has no time to look at this yet.
Ergo, it works in Jammy ;D
Quote from: Bas ON5HB on 08 Nov 2025, 16:03 I do not know if Nova works with the new compilers yet, I have not heard it does.
It does (mostly). It compiles with gcc-13 (tested under Ubuntu 24.04 LTS "Noble") and also with gcc-14 (tested under Debian Stable "Trixie"). The only difference that I noticed is the runtime behavior described in this thread.
The GCC compiler doesn't matter, it's the frontend-compiler that gives troubles and fails.
Beware, the audio-processing doesn't happen in the Spectrumserver but at the client-side.