V3.2 Release Notes

Links to kits, all release notes 💾
User avatar
Simon G4ELI
Posts: 2037
Joined: Thu Aug 06, 2020 7:27 am
Location: Mawnan Smith
Contact:

V3.2 Release Notes

#1

Unread post by Simon G4ELI »

V3.2 is available as a public bets, download beta kits here.

Main Features
  • Updated dot product code with highly optimised assembler (machine code). This is part of an ongoing project to reduce the CPU load with high bandwidths.
  • Added Continuum display: Ribbon Bar, View, Spectrum, Continuum.
July 25th 2022

SDRplay API 3.0.10

With the introduction of the 3.0.10 API some users of the RSP1 and possibly other SDRplay radios are having problems with SDR Console. The logic used in SDR Console is to open a device by matching the saved serial number against the serial number of the available SDRplay devices. For an unknown reason the serial number for the RSP1 may change over time, I have logfiles from a user showing this.

Something else I've noticed is that the RSP1 now has a unique SN, prior to 3.0.10 it was always 1.

To overcome this problem the serial number is now only used if there are more than one SDRplay devices of the same hardware type. So, if you only have one RSP1 / RSP2 / RSPdx etc. then it will be opened irrespective of serial number.

I've added extensive diagnostics in the logfile to help with any future workarounds.

Broadcast-FM, PCM Display

Fix fatal crash when in B-FM with audio muted and the PCM audio display selected. Caused by an optimisation which was bad.

Playback Navigator

Fixed / finished the auto-scroll for non-default layout; playback time moves from bottom to top.
Improved performance when building a big waterfall, depending on your system this may not be noticeable.
Added an auto-scroll option so the waterfall scrolls to keep the current playback time visible. Note: still got a small issue to resolve...
Options window swapped the direction Up and Down labels.

HL2 PTT

The HL2 PTT fix which Max has already checked,

External Radio

Extended the bandwidth when connected to the external radio's IF Output from 200 kHz to 1 MHz.

ELAD S3

Working with ELAD I am now shipping firmware v1.9 (it's in the FX3 dll). This may correct the problem where the S3 has to be power cycled to get it working properly.

14:03:24.731: Radio Elad-FDM-S3: Open S3> Load USB firmware (Pass 1 of 3), status = 0
14:03:24.731: Radio Elad-FDM-S3: Open S3> Address = 1721-0628-SNSK0C93
14:03:24.731: Radio Elad-FDM-S3: Open S3> Firmware = 1.9
14:03:24.731: Radio Elad-FDM-S3: Open S3> Serial = SK0C93


13 July 2022

AFEDRI radio definitions can now be added / edited manually.

Playback Navigator - replaced the Layout dropdown in the Playback Navigator with a Menu button and options window, also added 0.1, 0.2 and 0.5s resolution.
Playback Navigator options.jpg
Increased the filter length in the fractional resampler (used in decimation) to provide better support when listening close to a very strong signal - typically 85dB stronger than the station you want to decode. This is usually only noticeable on VHF and even then, you're likely to be suffering from splatter or phase noise.

Celestrak.com replaced with celestrak.org (satellite tracking TLE downloads).

More RAM allocated to data recording cache.

Minor RDS tweaks, these only affect very weak signals on an empty band.

Fixed almost irrelevant RDS decoding bug where the first station received would not show the decoded text immediately.

Minor performance improvements.

Band II FM:
  • More intelligent loading of US PI Codes,
  • Added support for Canada (big thanks to William Hepburn).
Notes:
  • Some PI codes are allocated to multiple US stations which are far apart, so at some stage I'll add the callsign for the station closest to the user.
  • Will look at adding Mexico support but apparently discipline on Band II is not great in Mexico.
Fixed fatal error in the Matrix caused by the performance change in the June 26th kit.

Removed unwanted artifact in the AM/SSB squelch window (not always visible).

Text in spectrum and waterfall now drawn with grey scale anti aliasing instead of clear type.

Increased #Memories from 20 to 40.

Minor performance improvement, optimised MFC <=> Direct2D drawing interface (remove SendMessage overhead).

FFT Overlap now shown correctly in waterfall statistics.

24 June 2022

When the monitor display state changes (example: switched on / off due to inactivity) Console was not properly refreshed. I found out how to do this, supported for Windows 8 and above according to the documentation, it also works with Windows 7.

The recording data cache now works again!

20 June 2022

Fixed a transmit frequency bug whereby the transmit frequency was not correctly synchronising as the receive frequency was changed.

Added support for the Delta offset option. Use Shift + Mouse left button, Shift + Mouse scroll wheel to adjust.

16 June 2022

Reduced unnecessary updates on the status bar.
Wide FM output level reduced by 10% to 90% of theoretical peak. Was too high for some external demodulators, also may have been clipping internally

14 June 2022
Real Men Write Machine Code
Writing low-level stuff to squeeze a CPUs performance is fun, haven't really done this in anger for 30 years. Written my own SSE / AVX / AVX2 code for DotProduct, managed to beat Intel's performance library by a factor of two.

This is needed when the 160 Msps / 600 Msps SDRs appear.

09 June 2022

Been reading Understanding Digital Signal Processing: Lyons, Richard: 9780137027415: Books (only 954 pages) and have been learning, understanding, and optimising some decimation code. This is a magnificent book, well worth the money, everyone should have a copy.

03 June 2022

This kit refreshes the display when starting and changing monitor. Seems that a recent W11 (and W10 maybe) update has changed part of the Windows display manager.

Tune drive level is reset to 0 if either:

More than 10 seconds has elapsed since last Tune, or
The transmit frequency has changed.

24 May 2022 06:33

NR4 settings will now revert to the default, future changes are saved. This is because the currently saved defaults are for the old design.
Voice Keyer now saves the file selection between sessions.
DX Cluster will now continually try to reconnect if the selected server is unavailable, for example I must take mine offline for a few minutes every month for OS updates.

21 May 2022 19:49

Added customisable font size for the Signal History / Continuum grid annotation.

20 May 2022 20:51

Added option to display the Signal History / Continuum signal level at the cursor position.

12 May 2022

New NR4.

lucianodato/libspecbleach: C library for audio noise reduction (github.com). The code will be continually updated, so I’ll be updating NR4. Note that the Github code uses FFTW which I have replaced with Intel’s IPP library. The default values should be OK, I’ll add more documentation later.

06 May 2022

This kit has the first version of a Continuum display:

Ribbon bar, Spectrum, Continuum Mode (shares the display code with Signal History). At present the continuum is computed from the frequency and bandwidth of the main display. The first second of data is ignored, it will take a few seconds before the display is updated.

04 May 2022

Added support for WinRadio WR-G69DDCe (Excelsior Ultra) Professional HF/VHF/UHF/SHF scanning receiver, 8 kHz - 8000 MHz
Simon Brown :shock:
www.sdr-radio.com

Do not send me direct e-mail, thank-you!

rainstar
Posts: 60
Joined: Sat Dec 19, 2020 4:06 am

Re: V3.2 Release Notes

#2

Unread post by rainstar »

Oh yea! FANTASTIC release Simon! I am PUMPED for the future of SDR.
Nothing gets me hot like optimized decimation code, too- I am absolutely going to play with the fractional decimator a bunch.
This is a feature that a lot or practically most SDR software simply do not have and only goes further to show that yours is the best of all.

https://github.com/haoheliu/voicefixer
Have you played with this? Warning- requires graphics card.
I have already tried porting it to intel using the tools, but it uses some custom code that won't allow the inference model to be
converted to openvino for use in applications. Supposedly there are newer and better network architectures coming along, so
eventually we should see some fantastic results.

What it can do:
It can take a amateur radio SSB 5khz or less signal and reconstitute it to a rich, full 44khz voice(a bit artificial but much better sounding)
It can remove some noise.

User avatar
NO6B
Posts: 52
Joined: Wed Aug 19, 2020 7:06 am

Re: V3.2 Release Notes

#3

Unread post by NO6B »

Same issue as always: defective squelch on NFM. Have to turn it up so high to keep it closed on a dead channel that 20 dB quieting signals won't keep it open.

Post Reply