The part at 1:03:00 involves a signal that had a large impulsive
component, and was treated with a combination of noise blanking and
noise reduction. Applying pure noise reduction wasn't going to help.
You can actually see from the pure noise reduction example they played,
that they were not comfortable with trying to do pure noise reduction on
a signal that wasn't easily copyable without it.
More importantly, nowhere from the 46 minutes mark do they describe
their algorithms, although, as open source code, it will be possible to
find them.
I would assume that the K3 uses LMS. The spectral one needs similar but
greater processing to that needed in the phase shifting proposal. It
needs greater, in that the it has to compute the parameters by which to
multiply the FFT bins on the fly, whereas the phase shifting case only
needs to compute them when you turn the control knob. Consequently it
is subject to the same performance question: does the K3 hardware
actually have enough processing power to run the algorithm. (I assume
LMS is used because the processing power requirements are rather less.)
I don't know whether it has the power needed.
My description of the general problem of noise reduction actually more
accurately fits the spectral processing model. I believe LMS is trying
to achieve a similar effect, but in a more computationally efficient manner.
Pure morse code doesn't actually need complex noise reduction, as simply
using a narrow filter will do the same thing as a good spectral noise
reduction algorithm would attempt to achieve. It can, though, benefit
from noise blanking. I say morse code, because a literally CW signal
can get perfect (Gaussian) noise reduction by using an infinitesimally
narrow filter. Actually, for a fair test, they should have also shown
the result of applying a very narrow filter to the power-line noise example.
Nothing in the part of the talk I listened to addressed the problem of
distortion caused by the suppression algorithm, in particular the
non-linearities caused by modifying the parameters on the fly.
On 15/01/2019 02:53, Wes Stewart wrote:
https://www.youtube.com/watch?v=FrVDL_-HOds starting at 41 minutes.
Particularly at 1 hour 3 minutes.
On 1/12/2019 9:51 AM, David Woolley wrote:
Do you have a reference for an algorithm that will do this?
______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[email protected]
This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html
Message delivered to [email protected]
