> I'm still trying to decide how much using an SDR of wider bandwidth might
> help the operation
> of the Linrad noise blanker for EME signals at normal EME elevations. I am
> feeding the same pulse generator to two different SDR's. The first one is
> IQ+, Delta44 and Linrad at 96Khz.
> The second one is SDRPlay RSP1a at 5 Mhz bandwidth 14bit.
> The 96Khz SDR must run at 200 hz pulse rate to show two spikes on the scope
> while the 5Mhz SDR must run at 20Khz pulse rate to show two spikes on the
> scope. The 96Khz SDR resolves to about at 2Khz pulse rate and seems to get
> into trouble at higher repetition rates than that while the 5Mhz SDR seems
> to handle up to about an 80 Khz pulse repetition rate and gets into trouble
> above that. On both of these SDR's the maximum pulse rate shows to be about
> 10 pulses. These two SDR scopes are obviously operating at two very
> different time bases. I would like to change the time base to resolve the
> pulses more fully.
The timf2 oscilloscope runs with the same sampling rate as the signal source.
Each point corresponds to one sample at the input when input is in the form
of I and Q. When the input is in the form of a single channel like the loudspeaker
output from a radio at i.e. 22100 Hz, the timf2 oscilloscope shows the same signal
as pairs of I and Q at 11050 Hz. It is not meaningful to change the time base.
If you want to see more details, just use a magnifying glass or the corresponding
software tool of Windows - or set a lower screen resolution to make pixels
bigger and wider separated.
If 96 kHz sampling needs a pulse repetition rate of 200 Hz for two pulses to
become visible you should need (5000/96)*200 Hz = 10 kHz. Correspondingly,
if you resolve 2 pulses at 2 kHz (48 pixels) you should resolve two pulses
separated by 48 pixels at 5 MHz also if frequency responses are similar
> What is not obvious is how to change the timebase on
> Linrad. I can only change amplitude. Linrad has evolved over time and many
> of the manuals and descriptions are accurate at the time of the writings
> but since the program has evolved some of the newer features can be hard to
> figure out. I am using a 2 Ghz i3 Windows10 laptop computer and it is
> running 80 to 100% CPU at 5 Mhz SDR. I have tried all the usual tricks for
> conserving the CPU. When looking at real antenna noise here it looks like
> the two SDR's are very similar with very similar noises pulses until one
> realizes that the two time bases operate at about a 100 to 1 ratio.
Hmm, 5000/96 is more like a 50 ratio;-)
> means the 5 Mhz SDR is resolving much finer pulses as expected. When
> looking at the 5 Mhz SDR some of the pulses are very close together
> suggesting perhaps something close to a 200 or 300 Khz pulse repetition
Yes. This gives only about 20 pixels per pulse. For the blanker to be
efficient you will the smart blanker with a good calibration to not
loose too much data. That could be challenging since the frequency
response of your antenna could be far from flat - and different in
An improved smart blanker could learn how frequently occuring pulses
look, but it is a non-trivial task to implement that...
> There is also a strong source of ordinary power line noise at a 120
> Hz pulse repetition rate. It's peak power must be very strong because it is
> evident at most antenna headings. This noise is easily resolved and
> blanked. It's the higher pulse rate lower amplitude pulses that remain a
> challenge. If I can resolve the pulses better by using a faster sweep rate,
> I think I can optimize the calibration of Linrad to minimize the width of
> the pulses.
To minimize the width of the pulses you would have to ask for a desired
response that has VERY soft fall-off on both sides. The parabolic slope
with the flat region only about 50%. That would not be optimum because
pulses do not have to be fully resolved when you use the smart blanker
which would know the precise shape of each pulse and subtract the contribution
over as many pixels as required. That would of course require a flat
response of the antenna.
> The power line pulse that shows up at many antenna headings looks a bit
> wider than it should but I suspect that if I am able to increase the
> timebase of the scope I will be able to see that each power line pulse
> actually has several higher frequency pulses bunched together. I would then
> expect my pulse generator by comparison to produce a narrower pulse after
> Linrad is properly calibrated.
The calibrated pulse is symmetric and has oscillations that extend equally
in both directions in time from the peak for pulses that happen to have
the maximum on a sample. The oscillations extend roughly 1/bw on both
sides where bw is the bandwith of the fall-off region. The smart blanker
knows the shape of those oscillations and removes them very efficiently.
You should be able to see that when using the pulse generator.
> I'm starting to suspect what might be an ideal setup for fully blanking
> Urban noise might be a setup as follows: A 4 yagi array of Log periodic
> yagis with a 20 Mhz bandwidth centered at 135 Mhz. A preamp system with a
> 20 Mhz bandwidth. A dual channel SDR with a 20 Mhz bandwidth at 16 or even
> 24 bit resolution. A VERY fast computer. And of course Linrad as a noise
> blanker sending to the network. Linrad as a converter to feed Map65. Then
> finally Map65. Looking at it this way it seems the best SDR hardware has
> not yet been built.
I am sorry, the frequency response of a 4-stack has sidelobes and nulls
that vary with frequency. Pulses will be different depending on direction.
The solution would be adaptive pulse calibration. Certainly possible, but I do
not think I will go in that direction. Currently my interest is in low