Re: Correlation Estimator block threshold settings in Absolute/Dynamic mode
There is definitely a weird logic with this block. Let me explain: Suppose your training sequence is c(t) with energy E=sum_t |c(t)|^2. The training c(.) is exactly what the parameter "symbols" is in the correlate block. Suppose your signal is some scaled version of the training sequence eg, y(t) = a c(t). You would expect that the block correlates the incoming signal y(t) with the given training c(t) and declares a peak when the correlation is above a certain threshold, A. ie, something like sum^t y(s) c^*(s) > A <==> a sum^t |c(s)|^2 > A <==> sum^t |c(s)|^2 > A/a Now since the full correlation, sum_s |c(s)|^2 equals E, you would expect that the quantity A/a should scale with E (ie, it should be a fraction of E), ie, A/a = fraction E <==> A = a fraction E. This means that if one wants to build a correlator that works with an arbitrary scaling "a" (suppose a genie gave you that at the Rx), the only thing you should have to do is to put as the parameter "threshold" in the block the quantity "a fraction". This would make the whole system transparent: no matter what the scaling is, your estimator would work exactly the same. HOWEVER, as it turns out the "correct" scaling that works for this block is to set "threshold= a^2 fraction". This is really weird and I do not see any intuition behind that type of scaling... Please see attached minimal working example and notice the squaring of scaling inside the block that makes it transparent to scaling the input signal. However if you change this to anything else, the system breaks down (in the sense that its behavior changes with scaling the input signal). best Achilleas test.grc Description: Binary data
Re: Correlation Estimator block threshold settings in Absolute/Dynamic mode
How close to 1 did you use thresholds for the relative threshold method? For the relative threshold method I use thresholds between 1-exp(-5) .. 1-exp(-8). Cheers. Christoph On Wed, Apr 1, 2020 at 5:58 PM Fabian Missbrenner wrote: > > Hi everyone, > > I am currently developing a wireless transceiver and try to get the > transmission detection running at the receiving end. To help with > detection and synchronization, a known word is prepended to every > message. I now want to use the Correlation Estimator block to detect > the presence of the known word on a noisy, bursty channel. The block > computes the (magnitude squared of the) correlation of the incoming > stream of complex baseband samples with the known word: > https://github.com/gnuradio/gnuradio/blob/master/gr-digital/lib/corr_est_cc_impl.cc#L219 > > Attached you can find the Flowgraph I am using for testing purposes. > > With Threshold Method set to Absolute, once the correlation exceeds a > user-defineable share of the auto-correlation peak of the known word > (Threshold parameter: between 0 and 1), > https://github.com/gnuradio/gnuradio/blob/master/gr-digital/lib/corr_est_cc_impl.cc#L178, > the known word is considered detected and a tag is added to the stream. > > Unfortunately, the block does not quite behave like expected. The > threshold needs be set to a value that is off by a factor of 1000 in > my case: > cf. PDF attachment, slide 1. > With a Threshold of 0.8/1000=800u, detection works reliably. But I > don't understand where this factor of 1000 comes from. > > When using an AGC3 block before the Correlation Estimator (yes, I made > sure to correlate against the normalized known word now), there is no > correction by factor 1000 necessary anymore: > cf. PDF attachment, slide 2. > Here, with a threshold of 0.5, detection works reliably. > > With the Dynamic mode, the semantic of the Threshold parameter > changes: Now it means a probability of false alarm: > https://github.com/gnuradio/gnuradio/blob/master/gr-digital/lib/corr_est_cc_impl.cc#L175. > Unfortunately, regardless of configuring a Threshold close to 0 or > close to 1, there are many false-positive detections: > cf. PDF attachment, slides 3 and 4. > > I'm not quite sure if I am missing some important point about using > the block or if the described oddities are bugs. Maybe you can help me > out? I also saw a discussion about this block in a Github issue > mentioning its lack of documentation: > https://github.com/gnuradio/gnuradio/issues/1207#issuecomment-279573149 > > Best regards > Fabian Missbrenner
Re: Correlation Estimator block threshold settings in Absolute/Dynamic mode
The magnitude of the correlation depends on the magnitude of your signal. The AGC normalizes the signal amplitude before the correlator, so you see expected behavior. The correlation estimator block expects the input to have a normalized amplitude. Nick On Wed, Apr 1, 2020 at 8:58 AM Fabian Missbrenner < fabian.missbren...@gmail.com> wrote: > Hi everyone, > > I am currently developing a wireless transceiver and try to get the > transmission detection running at the receiving end. To help with > detection and synchronization, a known word is prepended to every > message. I now want to use the Correlation Estimator block to detect > the presence of the known word on a noisy, bursty channel. The block > computes the (magnitude squared of the) correlation of the incoming > stream of complex baseband samples with the known word: > > https://github.com/gnuradio/gnuradio/blob/master/gr-digital/lib/corr_est_cc_impl.cc#L219 > > Attached you can find the Flowgraph I am using for testing purposes. > > With Threshold Method set to Absolute, once the correlation exceeds a > user-defineable share of the auto-correlation peak of the known word > (Threshold parameter: between 0 and 1), > > https://github.com/gnuradio/gnuradio/blob/master/gr-digital/lib/corr_est_cc_impl.cc#L178 > , > the known word is considered detected and a tag is added to the stream. > > Unfortunately, the block does not quite behave like expected. The > threshold needs be set to a value that is off by a factor of 1000 in > my case: > cf. PDF attachment, slide 1. > With a Threshold of 0.8/1000=800u, detection works reliably. But I > don't understand where this factor of 1000 comes from. > > When using an AGC3 block before the Correlation Estimator (yes, I made > sure to correlate against the normalized known word now), there is no > correction by factor 1000 necessary anymore: > cf. PDF attachment, slide 2. > Here, with a threshold of 0.5, detection works reliably. > > With the Dynamic mode, the semantic of the Threshold parameter > changes: Now it means a probability of false alarm: > > https://github.com/gnuradio/gnuradio/blob/master/gr-digital/lib/corr_est_cc_impl.cc#L175 > . > Unfortunately, regardless of configuring a Threshold close to 0 or > close to 1, there are many false-positive detections: > cf. PDF attachment, slides 3 and 4. > > I'm not quite sure if I am missing some important point about using > the block or if the described oddities are bugs. Maybe you can help me > out? I also saw a discussion about this block in a Github issue > mentioning its lack of documentation: > https://github.com/gnuradio/gnuradio/issues/1207#issuecomment-279573149 > > Best regards > Fabian Missbrenner >
European GNU Radio Days: contribution submission
The initial deadline for contributing to the European GNU Radio Days conference https://gnuradio-eu-20.sciencesconf.org/ is supposed to be today. In these uncertain times, some might be hesitant to spend time summarizing the result of their work: since no official guideline has been issued for events planned for end of June, we are still planning on holding the event *assuming* some contributions are provided. At worst, me might consider virtualizing the conference, which we hope to avoid, but at least that might provide some framework in which to share results. Feel free to submit at https://gnuradio-eu-20.sciencesconf.org/registration in case you have completed work to present (or let me know if you plan on doing so). Best wishes, Jean-Michel
