Hi all,

There are benchmarks added to the novasdr program.
Here is an example:

time ./novasdr-server benchmark cpu-fft-complex 1024 1024000
Run fft_benchmark for: iterations=1024 fft_size=1024000 is_real=false accelerator=None ...

run ./novasdr-server benchmark --help to see all benches you can do.

It can help you decide what fft_size is best for your system and if it's faster via GPU or non-accelerated.

It slipped in a few weeks ago

root@Websdr:/opt/novasdr# ./novasdr-server benchmark --help
Run benchmark

Usage: novasdr-server benchmark [OPTIONS] <KIND> [ITERATIONS] [FFTSIZE]

Arguments:
  <KIND>        [possible values: cpu-fft-complex, cpu-fft-real, cl-fft-complex, cl-fft-real, vk-fft-complex, vk-fft-real, ssb]
  [ITERATIONS] 
  [FFTSIZE]     

The 'time' command will time how long very test takes.

All info to be found here.

https://github.com/Steven9101/NovaSDR/issues/22

As my usercount kept growing, but no users where online, or just a few.

I suspected it to be wrong as my router confirmed the traffic upload to be nothing, while Phil had 6mbit up....

Just need to chime in here.  I was using an earlier version of NovaSDR.  I like it.  But I also watched my outbound bandwidth, and it was high compared to Pieter's original WebSDR software.  I also watched it crash often and the demands it put on the GPU.  I have not built a new machine yet with the latest NovaSDR yet, but I will!  I did build out a quick deployment of UberSDR.  Ridiculously simple with a Docker container.  And it just works.  AND mostly importantly, my outbound bandwidth is a fraction of what it was.  I'm using a little Dell machine - it's new but nothing fancy.  No GPU.  Just CPU.  So... I think in my personal silly opinion that there's probably an amazing happy medium somewhere.  You guys might know I run a pretty popular USA northeast online sdr (we really should stop using the term "websdr"!) with a big listener base.  My listeners are a mixed group.  Some are advanced, but some just want easy.  I've heard pros and cons from both sides.  I wish I had more time to help develop with you guys, but I just don't have time.  But I think a collaboration would be amazing.  Each system has some really great positives.  Thank you for doing all this great work.  I'm a fan of both camps!!!

The FT8 Decodes window is not movable. I want to be able to move this window around the screen, since it covers the waterfall.

And that's great. Phil's program has been running on our Websdr for a long time. It's a wonderful piece of software.

Exciting!

Hi all,

I just updated my Nova-SDR server to the latest code, as I do from time to time.
But I also read the issues and pull requests....then a message appeared.

For me important enough to inform you all. Link is here:

https://github.com/Steven9101/NovaSDR/issues/19

It reads:

QuoteHi, I just discovered your project. Are you using fast convolution to do multichannel reception? If so, that's what I've been doing in github.com/ka9q/ka9q-radio. My project is basically a downconverter engine optimized for lots (hundreds) of simultaneous channels on a single front end, producing multicast IP streams that can be fed to decoders, web servers, etc. I don't require any GPUs; I use FFTW on the host computer. It is not restricted to blocksizes that are powers of 2. I typically use a 20 ms frame time with 20% overlap and no windowing. Demodulators are built in for analog AM/SAM/FM/SSB/CW/IQ and wideband FM with stereo decoding. Spectra are generated either from a downconverter output (for small resolution bandwidths) or the A/D input stream (for large resolution bandwidths) with a choice of windows, overlaps and averaging levels. The output sample rate can be any multiple of 200 Hz and audio can be encoded with Opus or in any of several PCM formats.

ka9q-radio is being widely used for FT4/FT8/WSPR skimming (AI6VN's WSPRdaemon project uses it), HamSCI WWV carrier doppler observations, HFDL decoding, FM repeater archiving, etc. It's also being used to feed ham bands to some web sdrs, eg, at the Northern Utah SDR site. There's a web front end (ka9q-web) but is still fairly early; my emphasis has been on the downconverter engine itself. If you are interested in any joint work, let me know.

73, Phil Karn, KA9Q
karn@ka9q.net
San Diego CA

I for one would welcome Phil to join the project. As his KA9Q-Radio project on github is a game-changer.
Also, making a solid connection between the 2 projects will open many stuff people want to use/have.
Like automatic WSPR reporting, KA9Q does this, regardless of the frontend.
As he doesn't use GPU, and we do, it can rocket both projects to new heights.

This is the reason I send out a message to all of you, as getting Phil on the team will benefit all of us.

Bas ON5HB - Heppen Wideband Websdr.

Sorry mates, I can't stop laughing, UberSDR has a Soapy driver.....see what you can do with it.
Checkout TOP RX

You cannot view this attachment.

     

How do you do this?

Clone their github:

git clone https://github.com/madpsy/ka9q_ubersdr/tree/main

Then go into the clients directory, soapy subdir, then:

mkdir build
cmake ..
make
make install
Now Soapy will list a lot of UberSDR, with './nova-server configure' you add as many as you like

Safari issue.

I'm using the RX888 receiver, so I won't be able to use Soapy. Am I wrong?
I noticed that the waterfall jumps noticeably in Safari. In other browsers, there's no noticeable effect.
There is a GPU accelerator, but it is not used.