Automating WIA Broadcasts

[VK2OMD]

Interesting comments Owen..

I have two test sets, so I think I will just use one of those to set levels I have only set it by ear at this stage, but it sounds good.. I did have a look on the deviation scope, and it looked pretty good there too.. It wont take a lot to generate the 1k tone and set it up your way, so I might try that and see what the results are..

Dion.

If you want to play...

I have dropped the level 4dB in a test tone file I made back then to suit the defaut normalisation by mp3gain. The file is at owenduffy.net/pub/tt2.mp3 .

The description follows:

Test files

A digitally synthesised test tone file is useful in setup of the FM side, it should be played from the same source as the broadcast sound file. The test tone file contains the following:

- 1kHz sine wave at -16.4dBfs (peak) for adjustment of the tx level using Bessel first carrier null (2.4kHz dev);
- 500Hz, 1kHz and 2kHz sine waves at -24dBfs for testing for gross slope across the passband; and
- silence between tones for observing the background noise level.

For operators with access to a deviation meter, the 1kHz test tone at -20dB should be used to set 1kHz peak deviation. Having set the audio level, the passband flatness can be checked using the other tones, the 500 Hz -20dB tone should cause 500 Hz deviation and the 2kHz tone should cause 2kHz deviation.

[VK7DB]

I generated a 1kHz tone at -16.4dBfs and measured the deviation. It was 1.2khz, so 6dB lower than yours?

I played your test file, adjusted level to 2.4kHz deviation for 1kHz AF. The other measurements are ok, maybe a little too much bottom end, it measured 600Hz deviation at 500Hz AF. 2kHz measured 2kHz.

I then played the normalised WIA file for today and I thought it sounded a bit loud. This might be better for noisy mobile signals though..

I have rearranged the TX audio circuit in the PRM8010 by bypassing the first mic filter stage and the CTCSS filter.. I adjusted the gain of the mic amp so I can run the output of the Pi set at 100% output.. Seems to work ok, and I dont think I will try to improve it.

Dion.

[VK2OMD]

I generated a 1kHz tone at -16.4dBfs and measured the deviation. It was 1.2khz, so 6dB lower than yours?

I played your test file, adjusted level to 2.4kHz deviation for 1kHz AF. The other measurements are ok, maybe a little too much bottom end, it measured 600Hz deviation at 500Hz AF. 2kHz measured 2kHz.

I then played the normalised WIA file for today and I thought it sounded a bit loud. This might be better for noisy mobile signals though..

I have rearranged the TX audio circuit in the PRM8010 by bypassing the first mic filter stage and the CTCSS filter.. I adjusted the gain of the mic amp so I can run the output of the Pi set at 100% output.. Seems to work ok, and I dont think I will try to improve it.

Dion.

Dion, all noted.

The pre-emphasis is never ideal... and the end to end combination is part of a fraud to make SINAD look better.

At the end of the day, when you apply a 6dB/octave preemphasis over the entire voice band (or as good as), then you have PM... as witnessed by your measurement that for constant modulating voltage, frequency deviation is proportional to modulating frequency... still everyone calls it FM!

Re the loudness, I did a theoretical workup, and it sounded good... but the mobiles wanted more level which as I said, is a compromise between loudness and acceptable distortion. I think they are an important demographic, people at home can listen on their computer!

Get some reports... I will be interested in your final solution.

Owen

[VK7DB]

I think I worked out why mine sounded louder. When I was playing earlier today, running mp3gain on the files with defaults caused a drop in level.. I added some switches, seems I forgot about those.

Code: Select all

/usr/bin/mp3gain -r -m +3 -c wianews.mp3 tasnews.mp3

I will try again without the switches, I'm sure it will be closer to where it should be..

Dion.

[VK2OMD]

I think I worked out why mine sounded louder. When I was playing earlier today, running mp3gain on the files with defaults caused a drop in level.. I added some switches, seems I forgot about those.

Code: Select all

/usr/bin/mp3gain -r -m +3 -c wianews.mp3 tasnews.mp3

I will try again without the switches, I'm sure it will be closer to where it should be..

Dion.

Ok, you have normalised 3dB higher than defaults.

I have done a series of tests on mp3gain with the cal file, and they confirm the filter weighting and spot loudness determination if I can call it that. It seems that one loud section sets the overall loudness statistic.

I normalised this week's broadcast file using defaults, and performed a modulator lineup using the cal file in my earlier post, then did a drive test.

I am confident that the procedure gives good results, I was particularly interested in observing the signal when it was dropping in and out and comparing the apparent loundess of a very noisy signal with a full quieting signal, and readability under high speed vehicle noise. The FM receiver will usually produce about the same noise open mute with no signal than a well modulated signal, and that occurred. There are a couple of lay bak low level talkers on the file, but loudness for modulating purposes is really set by the louder talkers like the intro. This is an indication that the 'tracks' that combrise the broadcast could do with better levelling... perhaps your solution for levelling various broadcast source would be a good one for the individual contributions.

So, I think if you normalise to defaults, and follow the modulator calibration procedure you should have a close to optimal solution Dion.

Owen

[VK2OMD]

Dion,

I know you are using a deviation meter to calibrate the modulator.

My original article explained also how to calibrate the modulator using Bessel zeros.

The method derives from knowledge of the spectrum of an FM signal. When the modulation index (m=deviation/freq) is 2.4, the carrier amplitude is zero. My cal file is set up with a 1kHz sine wave of such level (relative to the normalised broadcast file using mp3gain) that we want 2.4kHz deviation, and so m=2.4.

This can be done be ear listening on an SSB receiver to the carrier beat note, and increasing modulation level until the carrier drops out the first time.

I have demonstrated that in the following video, and have a spectrum analyser showing the receiver output visually, but the dropout is very plain to the ear.



Once the modulation level is set as above, the broadcast file can be played and the level should be appropriate.

Owen

[VK7DB]

This photo demonstrates the same thing.. It did this last night, but its easier to just use the deviation meter.



I have redone the mp3gain thing on the files and now it sounds pretty good.. A broadcast repeat is scheduled for tonight, so I will ask for feedback on level and quality. I might even go mobile myself and see how it sounds.

Dion.

[VK2OMD]

This photo demonstrates the same thing.. It did this last night, but its easier to just use the deviation meter.



I have redone the mp3gain thing on the files and now it sounds pretty good.. A broadcast repeat is scheduled for tonight, so I will ask for feedback on level and quality. I might even go mobile myself and see how it sounds.

Dion.

Good, the carrier has dropped out and the low order sidebands are clearly visible.

Whilst a spectrum analyser might be the tool of choice in setting up a microwave link or whatever, the technique can be applied to a land mobile transmitter using an SSB receiver and listenting to the carrier null by ear.

Measurement is about using what is on hand to achieve an outcome, and the accuracy of the technique I described is dominated by the accuracy of the 1kHz sine wave (freq and low distortion).

Ok, I will be interested to hear of the outcome... especially the summary of reports from mobiles.

Owen

[VK7DB]

All went well, but our audience is small.. 4 reports, all stations. I went mobile, took a variable attenuator to get some noise on the signal and had a careful listen.. I'm pretty happy with how it is now, readability was good in poor S/N conditions. The others commented that the level and quality was good, but said there was nothing wrong with how I ran it Sunday morning..

I'm happy enough to leave it as it is now, but I think I could improve it more with an external DAC on it. I noticed an improvement on my EchoLink machine with a USB DAC on it rather than the onboard sound.

I have written a simple script that sends me an email half hour before start time, just to let me know its still alive. It also emails me the get file logs so that I know its done its job.

The next mod will be to use the GPIO to switch power to the radio with a high side FET or something.. I can put it on solar power and forget about it then..

Dion.

[VK7DB]

Code: Select all

root@broadcast:~# uptime
 12:17:38 up 31 days,  2:00,  1 user,  load average: 0.01, 0.04, 0.05

Reliability? No problems.

[VK2JDH]

All went well, but our audience is small Dion.

Well done. My Pi pulls the weather, tide and other info from the BOM amd other places and puts it out on APRS.

[VK7DB]

Interesting.. Hows that done? I have a Pi doing a tri-band APRS igate with APRX at the moment..

[VK7DB]

I had to reboot my broadcast Pi to cancel a broadcast on a repeater that was dying, it was the quickest way to kill it remotely from my phone.. I have lost the uptime now.. I have another Pi running APRX and its been going for about the same time.

Code: Select all

root@APRX:~# uptime
 20:12:09 up 55 days,  1:42,  1 user,  load average: 0.04, 0.07, 0.06

Code: Select all

root@broadcast:~# uptime
 20:20:12 up 13 days, 10:51,  1 user,  load average: 0.08, 0.06, 0.12

I have no reason to think that I should be concerned about the reliability of these..

Dion.