Heatmap with unwanted vertical lines

Leif Asbrink leif at sm5bsz.com
Sun Aug 2 03:25:22 UTC 2015


Hi John,

Heatmap steps the frequency of the dongle and places spectra with
a bandwidth of about 2 MHz side by side. Depending on hardware and how
you configure it you might have a center spur that repeats.

You cover 118 to 137 MHz which is nearly a factor of 10 above the
bandwidth of the dongle so you loose 90% of the data on each frequency
with an associated higer noise floor (5 dB perhaps) that makes weak signals
more difficult to find.

It seems you have 10 sub-bands (based on the regions of low loise level.)

I have not followed this thread so I do not know what tuner you are using.
The image does not give any hint.

All the dongles have many things that can be controlled by software.
As a first step I suggest you use a "normal" SDR software that uses a single 
center frequency. SDRsharp, Linrad, HDSDR or whatever. Look at the spectrum 
and play with available settings.

Regards 

Leif










> Update:
> 
> Termination:  My dongle has 75 ohm antenna connection so I used 75 ohm 
> terminator.  The lines remained.  The lines are much diminished or it 
> could be due to fact "background" signal is very prevalent.  I have 
> attached this one image for reference.  I was expecting to see no signal 
> at all (or no single where the lines are not).
> 
> Bandpass filter: ave not done this.  I have a broadcast FM notch filter 
> I could try.
> 
> RF Chokes:  Tried these on 8" usb cables feeding this RTL dongle, 
> another RTL dongle and an 802.11 dongle.  No apparent change in lines.
> 
> Enclosure:  The RTL dongles were already in aluminum enclosures. They 
> are right next to an Odroid single board commuter that has no case.  All 
> of this is in a tin box.
> 
> Linrad:  Have not tried this.
> 
> All my debugging was air band.   When I switched to 450-470Mhz the 
> problem goes away.  Since this is my range of interest I will not be 
> working on debugging this anymore.
> 
> I can provide other images if someone is interested themselves.
> 
> Thanks again,
> 
> John
> 
> On 07/15/2015 08:03 AM, Nikolay Dimitrov wrote:
> > Hi John,
> >
> > Put a 50- or 75-ohm termination on the rtl-sdr antenna connector and
> > redo the plot, to see whether the beat-frequencies are generated inside
> > or outside your dongle.
> >
> > Next you can put a bandpass filter in front of your rtl-sdr dongle, in
> > order to reduce the out-of-band signals that probably overload your
> > front-end. In practice, we shouldn't be using any RF device without
> > input and output bandpass filters.
> >
> > Next, you can also try putting an rf choke/ferrite (a common-mode
> > transformer) on the USB cable, in order to reduce the noise coming from
> > the USB-host and through the cable.
> >
> > Next, putting the dongle inside a metallic enclosure will help
> > screening the RF circuits, and will allow it to receive signals only
> > through the input connector (and preferably through an input bandpass
> > filter). You can create an effective "poor man's enclosure" by cutting
> > and soldering pieces of double-sided PCB.
> >
> > Finally, you can test your dongle with Linrad with its patched version
> > of librtlsdr. Linrad uses a different gain distribution and there's a
> > big chance that it can satisfy your needs. You can do similar
> > experiments by reducing the RF gain and AGC on rtl_power and see
> > whether it influences positively your measurements. Please try these
> > and share your experience.
> >
> > Regards,
> > Nikolay
> >
> >
> > On 07/15/2015 02:52 PM, John wrote:
> >> Hello,
> >>
> >> When I use heatmap.py with output from rtl_power I get regularly spaced
> >> vertical lines that do not appear to be related to any signal. They
> >> look they like repeat at the dongle bandwidth (2048000Hz in this case).
> >> The crop option for rtl_power reduces the presence but I am not sure f
> >> that is intended by that option.  Even at -c of 70% they are still there
> >> (see attachment).
> >>
> >> Is this because of small bin width?  If I use a larger bin (32k) they
> >> are still there.  In this case there is no frequency legend along top so
> >> can't compare if they happen more often.
> >>
> >> Are these lines expected?  Can they be removed?
> >>
> >> John
> 



More information about the osmocom-sdr mailing list