Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Digital simulation of FM radio sound (Read 8596 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Digital simulation of FM radio sound

As discussed here, I'm putting together a digital radio listening test.

I want to include a simulation of FM. It must be a digital simulation of good quality FM reception.

So far, I'm simulating FM by adding...

1% harmonic distortion
-60dB mono noise
-50dB mono noise with one channel inverted
-50dB 19kHz tone
some extra noise around this tone
  and shaping all this with the FM de-emphasis curve

I'm also adding -40dB channel cross-talk.

The harmonic distortion is generated by raising the original waveform (range +/- 1) to the power 2, setting the negative half of the waveform (which has been made positive by the squaring) back to negative, and adding 0.02 times this to the original waveform. What's added is approx 1% RMS of the original waveform (typically).

Any comments?

If it's about right, I'll put a sample on-line so people can compare it with their own FM reception.


Currently, I think the distortion sounds too quiet, and the noise sounds (maybe) too quiet, depending on the station.

It would help me if some people with good FM reception could digitally record and analise it for me: Given a very loud signal peaking at 0dB FS, what is the RMS value of the background noise during silence, measured over a bandwidth of 15kHz (i.e. filter out all the stuff above this before measuring). If you can quote the figures for stereo and also for mono, that would be great.

Cheers,
David.
P.S. I've also posted this to the newsgroup alt.radio.digital

Digital simulation of FM radio sound

Reply #1
Don't forget to low-pass the audio somewhere below the pilot frequency.
I don't know what the unrejected SCA sounds like... maybe frequency inversion of the SCA
program?  I don't know what percentage of stations carry SCA either.

Here's some specs for Magnum Dynalab's  "midline"  tuner (perhaps this is like a
"midline" Ferarri):

Audio frequency response (+/- 1dB)    10Hz - 17.5Khz
THD (mono)    0.10%
THD (stereo)    0.18%
Capture ratio    >1.5dB
AM suppression    -70.0dB
Image rejection    75.0dB
Signal to Noise ratio (SNR)    80.0dB
Stereo separation    50.0dB
SCA rejection    -75.0dB
19 & 38 KHz rejection    -70.0dB
Audio outputs* (@100% modulation)    1 volt
Balanced audio outputs    600 ohms

Digital simulation of FM radio sound

Reply #2
In my experience the biggest problem with FM reception is multipath distortion, but I suspect that this would be very difficult to accurately simulate.

Digital simulation of FM radio sound

Reply #3
I forgot to say that I'm bandpass filtering the audio 25Hz-15kHz.

We don't have SCA in the UK - at all.

Your figures (and those from someone on alt.radio.digital) make me think I'm adding too much noise, at least for "good" reception.

I know multipath is a huge problem "on the move" - but in the UK, the BBC state that the FM transmiter network is designed to give everyone good reception "with a directional roof-top aerial" - if you follow this advice, the results are superb. No multipath problems to my ears.

I'll probably carry this on at the Cool Edit user forums http://forum.syntrillium.com/ because I need to discuss actual Cool Edit settings. I'll post exactly what I'm using tomorrow.

Cheers,
David.
EDIT: typos

Digital simulation of FM radio sound

Reply #4
Souldn't signal processing like Optimod be part of the simulation as well? Also question is, if and how DBS and FM get a different treatment with this...
BTW it's forums.syntrillium

Digital simulation of FM radio sound

Reply #5
I've got 5 samples at hand.Total 14 MB
Recorded with analog Yamaha T-560 tuner, double cable on the floor as an antenna, in town.

Mainstream pop : DSP+Compressor+DSP+EQ+DSP+loudness 4 MB (no silence)
National Pop : DSP+EQ 3 MB (micro noises over FM noise)
Associative station, pop : the most silent of all, too quiet 2 MB
Mainstream classical : 3 MB
National classical : quiet, noisy 2 MB

Tell me if you need all, or which one you are interested in.

Digital simulation of FM radio sound

Reply #6
The first one, and the last two please Pio - if that's OK? PM me.


I've posted the question here, describing the process in detail:
http://forums.syntrillium.com/topic.asp?TOPIC_ID=11549

is it just me, or do they really know how to format a forum over there?

Cheers,
David.

Digital simulation of FM radio sound

Reply #7
Quote
Souldn't signal processing like Optimod be part of the simulation as well? Also question is, if and how DBS and FM get a different treatment with this...
BTW it's forums.syntrillium

So it is! Thanks.

After much discussion and thought, I'm not including that kind of processing in all the tests. It is there as a specific test, but to process all "FM" signals with an FM optimod, and all digital signals with a digital optimod (they do have processes targetted at each medium) would make it a test of optimod compressor algorithms as well as broadcast systems - I don't want to confuse the two.

But the effects of hyper-compression will be included as a specific test.

Cheers,
David.

Digital simulation of FM radio sound

Reply #8
I think some of the figures are a little pessimistic.

What is the noise around the pilot tone supposed to represent?

Digital simulation of FM radio sound

Reply #9
Quote
I think some of the figures are a little pessimistic.

What is the noise around the pilot tone supposed to represent?

I don't know why it's there, but in high quality recordings made by myself, and one other person, there's definitely something happening around the pilot tone. I can't decide whether it's noise, or related to the signal itself. I think the latter is more likely, but don't know what it is.

Looking at the spectrum, there's something there, so I'm including it. I realise FM stereo should stop dead at 15kHz, but the real analogue filters in all radios let various things through above this, so I think it's fair to include them.

Does anyone know what's happening around 19kHz?

Cheers,
David.

Digital simulation of FM radio sound

Reply #10


Because of the perfect symmetry, it must be an alias effect due to digital conversion somewhere in the path of the broadcast. Same result recording at 96 kHz.

I got similar symmetrical distortions around treble test tones analogly recorded from the output of my CD player in the frequency response experiment, even when the sines frequency had no relation with the sampling rate. The closer to the Nyquist frequency, the stronger they were.
EDIT : and the symmetry was always around the sine, not the Nyquist Freq.

Digital simulation of FM radio sound

Reply #11
I see that you are seeing exactly the same thing as me.

On your frequency response test, I'm not sure what's happening. If the frequency scale was different, I'd say it was jitter. As it is, the two sidebands either side of the main tone could be caused by amplitude modulation - but I'm really not sure.

You could be right about the source of the nosie around 19k on the FM transmissions, but I can't help thinking that it has something to do with the stereo signal processing. The two ideas (your and mine) are not mutually exclusive.

If anyone knows more, please post it - I don't need it for the simulation, but it would be interesting to know.

Cheers,
David.

Digital simulation of FM radio sound

Reply #12
Quote
On your frequency response test,(...) the two sidebands either side of the main tone could be caused by amplitude modulation

YES !

You're right. I understand now. The CD Player clock was not in sync with the soundcard clock (analog copy) -> a very small beat effect appeared, slower as the frequency approaches an integer divisor of the sampling rate, that show up as two sines in the spectral view, closer to the central peak as the beat gets slower.

 

Digital simulation of FM radio sound

Reply #13
I finally got a proper soundcard, yet the crappy FM radio (brooktree chip) still is not replaced. The soundcard has opened a whole new view on how crappy the tuner is.  Here is a spectrogram of a couple seconds of mono human speech.

http://j7n.sytes.net/www/ha-fm-spectrogram.png

Now, if the 19k signal is amplitude modulated, what secret information could be encoded there in this range? I understand that the tone must be constant and is only used as reference frequency later.

Next I wonder why is there data in the AM sidebands of 38 kHz even if the input signal is aproximately mono. L-R should be about zero then.

Do any other (better) tuner has this auxilary data in the audio output? I've read that you recorded only the pilot tone. Even the information is inaudible, it must consume power in the amplifier. Isn't it extremely dangerous to tweeters to play back this recording?