Re: [Discuss-gnuradio] FM pre-emphasis and de-emphasis

[Richard Lamont]

On 08/01/18 21:59, Müller, Marcus (CEL) wrote:

> please don't reply to existing emails on the Mailing List to start a
> new discussion thread

Sorry about that.

> Anyway: If I remember correctly, they work pretty OK – but it's hard to
> mimic analog filters in digital domain, so OK might be different for
> people that care more about the actual acoustic properties of the
> transmission than me!

Fair enough. I've just started playing with grc and did the usual FM
broadcast RX thing as my Hello World - and, as a former broadcast
engineer with pretty good ears - I immediately noticed that there was
far too much 'top' in the the audio. I have been using the FM
de-emphasis block that is shipped in the Ubuntu gnuradio 3.7.9 package -
I have added any others yet.

> So, I'm pretty sure you're not being thick – it's pretty possible that
> GNU Radio contains a mediocre implementation. Have you seen the source
> code? If you want to, we can guide you through that, and if you have
> recommendations, we'll be happy to cooperate with you to implement a
> better de/emph.

I haven't seen the source code and wouldn't be able to do anything with
it if I had. I'm an analog guy and have quite enough trouble just
pointing and clicking in grc. I can't code and the maths required to
calculate filter coefficients would make my head explode. That's why I'm
using grc in the first place!

If the block shipped with vanilla gnuradio isn't doing its job, then
surely it would be best to fix that bug.

When working correctly, the amplitude/frequency response should be
within, ideally, 0.1 dB of the values shown. I've shown values for both
75 and 50 microsecond time constants.

Freq (Hz)     Gain dB (75us)     Gain dB (50us)

30             0.00                0.00
300           -0.09               -0.04
500           -0.23               -0.11
1000          -0.87               -0.41
3000          -4.77               -2.76
5000          -8.16               -5.40
8000         -11.82               -8.64
10000        -13.65              -10.36
12500        -15.52              -12.15
15000        -17.07              -13.65

For pre-emphasis, obviously the gains should be positive.

This .grc demonstrates the problem - the amplitude/frequency response is
essentially flat:

http://www.lamont.me.uk/misc/deemph_test.grc


-- 
Richard Lamont
<address@hidden>