Software Defined Radio with HackRF by Michael Ossmann, Lesson 7: Complex Numbers in DSP

[@CraigHesling]

This explanation is legitimately too good. Thank you! It should be noted that the previous material is critical to unlock the understanding in this one.

[@joelmiller246]

Great python example! The application to a problem definitely cleared it up for me. Appreciated, Thanks

[@andreasclaesson3436]

Awesome, this is one of the best explanations to a complex (pun intended) concept I've seen!

[@GretaKnauer1980]

Excellent insights Michael, I'm getting a lot out of these these lessons

[@mattpopovich]

One thing I think would be useful up to this point would be an explanation of how we go from the amplitude vs time plot (top at 43:51) vs the IQ plot (bottom). Maybe you touch on that in future lessons... I'll keep watching!

[@firehdtablet490]

You need to watch lesson 6 to understand that.

[@tuathadedanaan1424]

I love watching your video's.

[@GreatScottGadgets]

Thank you.

[@Panakotta000]

Quick question regarding IQ and FM for determining the angle. You mentioned calculating the angle of the first and then the second sample. Could we also instead subtract the latter sample from the former sample and just get the angle from that? Instead of two lookups and one subtract, we have two subtracts and one lookup. 1. Is that even allowed? 2. How is the performance difference?

[@sebaschtl9710]

thank you.

[@GreatScottGadgets]

You're welcome

[@pakkiemelk]

Thank you for the excellent videos. I really enjoyed this and the previous video about complex numbers and why they are useful in DSP. There is one thing I don't understand after experimenting with complex numbers and real numbers in Gnu Radio Companion, I don't see any difference in CPU load when working with complex numbers over real numbers. The way I tested it was I use the flow graph of lesson 2 (signal source, throttle and FFT sink (or frequency sink)), while using real numbers. I keep my eye on htop looking at my CPU load and increase the sample rate. When the CPU gets loaded to lets say 50% I stop the process and repeat at the same sample rate with use of complex numbers. There may be some difference in CPU load, but not significantly, at least not that I can see in htop. Why is this? I even tried with a noise source instead of a signal source, to make it harder for the FFT sink to compute, but no difference. I would expect a better performance with the use of complex numbers.

[@GreatScottGadgets]

That is an interesting question. We suggest reaching out to the GNU Radio folks: www.gnuradio.org/

[@bfkmnemonic]

It seems to brik when you use equal or opposing forces? ((90, 1), (270, 1)) ((0, 1), (360, 1))

[@etwasanderes2]

Opposing winds breaks because the total is 0 (no wind), to which the concept of direction doesn't really apply (which means that the logarithm breaks, as you can not compute the logarithm of 0)