jks

So it appears the Debian project and Beaglebone.org have stopped making updates to the Debian 8 (Jessie) release. For the Beagle, Debian 8.11 seems to be the last release. It has been run by a number of customers for a long time and also by us. So it seems safe now for Kiwi owners generally to do an upgrade and get the benefit of whatever security patches there have been between 8.5 and 8.11.

To do this login as root using ssh/PuTTY and use the following commands. (this can not be done from the Kiwi admin console tab)
If you haven't logged in this way in a while remember that your Debian root password has likely been changed to either the Kiwi admin password or the Kiwi's serial number if no Kiwi admin password has been set. The Debian root password has not been touched if you had set it to something other than the default of having no password.
See this post for details: http://forum.kiwisdr.com/discussion/1801/v1-354-security-improvements-debian-root-password-no-longer-unset-blank

cdp
mst      (stops the Kiwi server)
cd
pkup     (shell alias for "apt-get -y install debian-archive-keyring; apt-get update")
pkug     (shell alias for "apt-get -y dist-upgrade")

(lots of output, takes roughly 10 minutes)

reboot

A few minutes after the reboot the Kiwi server should be responding again. If you login to the Beagle and type the "dog" command you should see:

Debian 8.11
BeagleBoard.org Debian Image 2016-05-13
Linux kiwisdr 4.4.9-ti-r25 #1 SMP Thu May 5 23:08:13 UTC 2016 armv7l GNU/Linux

There is still the open question of how the Kiwi distribution, and thousands of Kiwi customers, will migrate to the current Debian 9 (Stretch) and/or Debian 10 (Buster) releases which are available for the Beagle. This needs time for research as there are many issues involved.

The DRM mistuning detection is extremely good. Screenshot below is Radio Marti, received in PA, mistuned by +1 kHz (!) It locks and decodes fine. It also probably helps that they are running a very conservative configuration.

The DRM spec has a little more information about the frequency pilot cells, section 8.4.2, page 121, http://kiwisdr.com/files/DRM/DRM.spec.v4.1.2.pdf

From the CHANGE_LOG file:

v1.394  May 31, 2020
    Noise blankers and filters:
        Added the spectral noise reduction filter by Frank, DD4WH and Michael, DL2FW, via the
        Teensy-ConvolutionSDR and UHSDR projects. See: github.com/df8oe/UHSDR/wiki/Noise-reduction
        This noise filter works really, really well. Especially for voice signals.
        A good alternative to the LMS filters.

    Note that the Wild blanker and all the noise filters (wdsp, LMS, spectral) don't currently work
    with the modes: IQ, SAS and DRM (i.e. the 2-channel IQ modes). This will be fixed at some point.

This filter is on the main control panel, audio tab, "noise > filter" menu. Click the "more" button to the right of the menu to adjust the filter parameters, although the filter works well with the defaults. In some cases you might have to increase the filter gain to match the filter-off level, especially during A/B comparisons. Try it with a weak station on the AM broadcast band or ham band SSB signal.

Note that some noise filters always require their sections be enabled via the "more" panel. Otherwise simply selecting their menu entry does nothing. In particular the wdsp and LMS filters since they have independent denoiser and autonotch sections that are enabled separately (or both simultaneously).

From the CHANGE_LOG file:

v1.392,393  May 28, 2020
    Noise blankers and filters:
        Added the Michael Wild noise blanker. Not a lot of operating experience with this yet.
            It seems not to work on some noise sources compared to the standard blanker.
            A good test source is 100 kHz Loran-C in AM mode using a UK Kiwi.
            Increase the gate time (standard blanker) or impulse samples (Wild blanker) to
            account for the longer Loran-C pulses compared to the usual impulse noise.

        Fixed wdsp filter: the control sliders had no effect previously!
            Now it works quite well. You have to adjust the gain carefully to prevent overload.

        Added test pulse gain and pulse width controls. And also whether the pulse is
            applied before or after the passband filtering. The Wild blanker seems only
            effective on post-filter pulses. Probably because the filter is turning the
            pre-filter step function pulses into nicer band-limited pulses.

From the CHANGE_LOG file:

v1.391  May 23, 2020
    User interface changes:
        SAL, SAU mode passband image will only show single sideband (just like LSB, USB).
        Ctrl/alt-click on mode button toggles modes backwards (most useful with SAM modes).
        Same works with keyboard shortcuts, e.g. ctrl/alt-A toggles SAM modes backwards.
    
    Noise blankers and filters:
        Reworked the audio tab user interface to support multiple blanker/filter algorithms.
        New algorithms planned but not yet available are greyed-out in the menus.
        From Warren Pratt's wdsp package added the variable-leak LMS algorithm.
    
    Merged Howard Su's RPi changes.
        Note: RPi is not officially supported by KiwiSDR. Ask for help on the KiwiSDR forum:
            http://forum.kiwisdr.com/discussion/1986/port-kiwisdr-to-raspberrypi-3b
    
    Bug fixes:
        Audio recordings made in SAS (synchronous AM stereo) mode now work.
        Fixed problem with ITU region 2 Kiwis and shift-clicking waterfall in 75/80m band segment.
        from lazywalker: fixed problems with frequency memory and reload value when transverter offset used.

From the CHANGE_LOG file:

v1.391  May 23, 2020
    User interface changes:
        SAL, SAU mode passband image will only show single sideband (just like LSB, USB).
        Ctrl/alt-click on mode button toggles modes backwards (most useful with SAM modes).
        Same works with keyboard shortcuts, e.g. ctrl/alt-A toggles SAM modes backwards.
    
    Noise blankers and filters:
        Reworked the audio tab user interface to support multiple blanker/filter algorithms.
        New algorithms planned but not yet available are greyed-out in the menus.
        From Warren Pratt's wdsp package added the variable-leak LMS algorithm.
    
    Merged Howard Su's RPi changes.
        Note: RPi is not officially supported by KiwiSDR. Ask for help on the KiwiSDR forum:
            http://forum.kiwisdr.com/discussion/1986/port-kiwisdr-to-raspberrypi-3b
    
    Bug fixes:
        Audio recordings made in SAS (synchronous AM stereo) mode now work.
        Fixed problem with ITU region 2 Kiwis and shift-clicking waterfall in 75/80m band segment.
        from lazywalker: fixed problems with frequency memory and reload value when transverter offset used.

Aluminum enclosures: Mouser now has 40+ in stock.

"/?ext=drm" since DRM is both a mode and an extension. Use e.g. "drm,3965" to select one of the preset frequencies.

I can't remember why DRM can't be used as a mode selection in the URL. Some problem that was too difficult to fix no doubt.

When the aluminum enclosure was being developed I specifically asked Seeed to use a ball bearing fan (i.e. sleeveless). But I'm pretty sure now they didn't given the observed failure rates. I should ask again. Note that we receive no royalty on enclosure sales in an effort to keep the cost low. It's all Seeed's deal.

Alternative fans and ideas are mentioned in these posts:
http://forum.kiwisdr.com/discussion/comment/8667#Comment_8667
http://forum.kiwisdr.com/discussion/comment/7102

A simplistic but useful way of thinking about this issue is this: Suppose the waterfall is slowed to draw one line every ten minutes. During those ten minutes it could be sampling at ~21 Hz (full rate) and integrating those samples so the coherent QRSS signal is gaining a huge SNR advantage over the random noise. But the waterfall doesn't do that currently. It would just sit idle for ten minutes then take one sample.

The integrate extension does a little better. It does integrate every time its waterfall output wraps around top-to-bottom. If the integration period (time to draw everything top-to-bottom) is short (e.g. 1 sec for time stations) then this is useful. But it isn't designed to work any better with long periods. So it's no good for QRSS signals.

On the wish-list for a long time is a high-resolution FFT extension. One that could for example be used to resolve the millihertz differences of AM BCB stations (a technique already used very effectively by AM BCB DX'ers). Such an extension would naturally do long-period averaging in addition to high frequency resolution.