HB9UF: FM deviation, modulation index and sidebands with an RTL SDR dongle

In this video, I use a Kenwood TM-D700 transceiver to generate a frequency modulated signal that I investigate with gnuradio. In the second part of the video, I discuss some of the theory that helps understanding those observations.

The gnuradio pipeline that I used can be found here:

[YrarcOrg]

I did this on my IC-7100. I didn't use a dongle but a SoftRock IF plugged into my Elecraft K3 for receive with HDSDR software. I also used the "Data In" pin on the mini-DIN connector with the rig set to "9600 baud" mode. It isn't 9600 at all but a wide bandwidth audio input without preemphasis or other equalization. The first thing I tried was setting the audio frequency to 1250 and increasing power gradually. Before I could see the null, the 7100 stopped transmitting due to excessive audio level, above the nominal 100-300mVp-p. The method I ended up using was to set the audio voltage to 300mVp-p and lower the frequency until I saw the null. That worked and showed the maximum deviation to be 1680Hz. Further lowering the frequency revealed a 2nd and third null, and by looking up those modulation indices in Wikipedia, determined that the resulting deviation estimate was very close to 1680Hz as well. I haven't tried the "1200 baud" port yet but it may well show that it is 3Khz. I think the deviation for 9600 is limited due to the bandwidth required to handle audio in the 3-6kHz range. Thanks again for your informative video.

[rdoetjes]

The peek you see on your tuned frequency is a DC component from the SDR. I have it too and also on the HackRF this is also why you usually tune little up or down to get the best result -- It really has boggled me for the past few weeks since I started working with SDR :D

[YrarcOrg]

At 16:20 you are adjusting the voltage while calling it out in kHz. The baseband frequency domain display is growing in amplitude but the frequency is constant. You switched back to volts once you get to the Bessel zero. I liked how you used GNU radio to put everything on one display. It's good for us hams who don't have those $20K mixed domain oscilloscopes with built in frequency analyzers.

[MrCircuitMatt]

+Raymond Doetjes I cannot reply to your comment directly, I hope you'll see this. Yes you are right. Among other things, the LO will mix with itself resulting in a 0 Hz difference product which is DC. The approach you suggest is probably the easiest way to get around it. Thanks for commenting :)

[bennguyen]

So to visualize, would you connect a 1khz sine wave to the WFM or NFM block? In your example, the ampltude of the signal determines the 'deviation'.. and isn't the same as the bandwidth (i.e. Wide vs Narrow)? GRC doesn't seem to allow the a sine signal source to go to the transmit block.. how would you do that?

[randomstring]

Hi there; I can't wait to watch your next video.
It just so happens that in the last few weeks I have been searching info exactly about the very subjects that you are touching in this video. Thank you very much; which wireless protocol did you use to read my mind remotely :-) ????
So, seriously, you are basically saying that given that the Mod. Index is related to the deviation and the modulating freq., if my constrains are deviation and M.I., then I can calculate the max audio freq. I can transmit and receive.
So, given my constrains are Dev. = 5 Khz and max M.I. = 1 (NBFM, right?) I can transmit an audio signal up to 5 Khz (audio), am I correct?
And you mean Deviation 5 Khz total or 5 khz per each side band, so total deviation = 10 Khz (5 LSB + 5 USB)?

[PlasmaHH]

ThumbNet sold some dongles where basically the oscillator was replaced by a quite ok tcxo, which can easily be done on your own, and it is working so much better that I would say it is worth the effort than all the time being lead to nowhere. And yes, the peak is the LO (it is leaking everywhere in these things) and I usually tune center to quite a bit outside of the interested things, as long as I don't need to process MHz of bandwidth. Programs like SDRSharp have a correction for this already built in and can silence that tone.

[davidkrauss]

I tired this, but in the baseband FFT sink window, it shows the Frequency of the tone, NOT the deviation. Any ideas how to fix this?

[SteveWrightNZ]

Someone else made a video of this in 5 minutes.