Running fm95 and attemping rds95

[chodgetts]

Hello,

This is my sound card - which clearly supports 192kHz.

chris@mpxtest:~/rds95/build$ arecord -D hw:2,0 --dump-hw-params /dev/zero
Warning: Some sources (like microphones) may produce inaudible results
with 8-bit sampling. Use '-f' argument to increase resolution
e.g. '-f S16_LE'.
HW Params of device "hw:2,0":

ACCESS: MMAP_INTERLEAVED RW_INTERLEAVED
FORMAT: S16_LE S32_LE
SUBFORMAT: STD MSBITS_MAX
SAMPLE_BITS: [16 32]
FRAME_BITS: [32 64]
CHANNELS: 2
RATE: [44100 192000]
PERIOD_TIME: (83 5944309)
PERIOD_SIZE: [16 262144]
PERIOD_BYTES: [128 2097152]
PERIODS: [2 32]
BUFFER_TIME: (166 11888617)
BUFFER_SIZE: [32 524288]
BUFFER_BYTES: [128 2097152]
TICK_TIME: ALL

arecord: set_params:1387: Sample format non available
Available formats:

• S16_LE
• S32_LE
  chris@mpxtest:~/rds95/build$

No matter what I do when I start rds95 I hear data pulsing - when I set asciig to 0 I see the strings fly by but I never see a RDS sub carrier.

I am using a BW Encore to decode - I can see hte pilot, and it moves when I adjust its level in fm95
I get the L+R / L-R showing up correctly, so I know its producing the MUX, but just cant seem to get RDS working in this - and just wondering what I might be doing wrong :(

[KubaPro010]

Please remember that RDS95 outputs a raw bpsk signal, meaning not on 57 khz and that raw signal goes to fm95 (NOT AS AUDIO), and that puts it on 57 khz (since 57 khz is supposted to be derived from pilot, and fm95 already does stereo so it was easier to do it like this)

Also, its either audio or ascii g in rds95, not both

EDIT: As of 11th July, this is no longer the case.

[KubaPro010]

You have to route the signal to fm95 to encode it via a virtual sink

you can create one in pulseutils (pipewire and pulseaudio) like this pactl load-module module-null-sink sink_name=RDS rate=4750 channels=4

configure fm95 and rds95 to use it

# fm95.conf
...
[devices]
rds=RDS.monitor

# rds95.conf
...
[devices]
rds95=RDS

and it should be on 57 khz

[KubaPro010]

as for rds2

the pactl command already has 4 channels, so no change there but the configs will have to change

# fm95.conf
[fm95]
rds_streams=x

# rds95.conf
[rds95]
streams=x

x should be a number 1-4 and has to be the same in both, otherwise you will NOT DECODE ANY RDS

[chodgetts]

Thank you so much.. that helped a lot...
I am sure I tested them before, but your pactl command might have actually done more for me than all the changes I would have done..

I have the following settings,

chris@mpxtest:~$ cat /etc/fm95/fm95.conf
[fm95]
stereo = 1
rds_streams = 4
preemphasis = 50
calibration = 0
mpx_power = 0.0
mpx_deviation = 75000.0
master_volume = 1.0
audio_preamp = 0.1
deviation = 75000.0
agc_target = 0.0
agc_attack = 0.5
agc_release = 2.0
agc_min = 50.0
agc_max = 100.0
bs412_attack = 10.0 # Slower attack lets short peak transients hit 100%
bs412_release = 25.0 # Slower release stops it from riding the peaks too hard
bs412_max = 1.0
bs412_gate = -40.0
bs412_knee = 3.0
bs412_strenght = 6.0

[devices]
input = fm_audio_input.monitor
output = default
rds = RDS.monitor
mpx = none

[advanced]
lpf_order = 10
lpf_cutoff = 15000.0
stereo_ssb = 0
preemp_unity = 1000.0 # Moved to standard 1kHz to preserve mid/low density
sample_rate = 192000
headroom = 0.05
drive = 0.5 # Increased from 0.5 to drive the peak clippers harder
makeup = 1.0

[volumes]
pilot = 0.09
rds = 0.05
rds_step = 0.0

I am seeing 9% and 5% mod as expected on my testing device, but I am only achiving about 65% total mod when I inject audio, and the audio only lasts about 45 seconds before going silent, even though mpv continues to play or at least indicate audio flowing.

Sorry - there was not example config files included in the package, so I am hoping I am not asking too dumb a question here
And thank your for your project and advise

[KubaPro010]

BS412 activates after 45 seconds (warm-up, because the compresssor measures the loudness over the whole minute) - i had also this issue because the calculation for applying the strenght can output NaN for larger values on start, and since any operation with NaN also results in NaN, the entire mpx output becomes NaN

just decrease your bs412_strenght and you should be fine (applying a higher value when it is running via reloading works fine, but not on startup). i use 3 as the value and it starts up fine

as for 65% modulation, it is not really a real value, as it is without BS412 - don't take it into consideration

[KubaPro010]

i have commited an update to "fix" this - now if the program detects that the gain might be nan (because of strenght) it will just not apply it

it should work after 45 seconds and also the strenght will be applied once the compressor is stable

(update via git pull and just run make again)

[KubaPro010]

also why is your preemp_unity at 1000? pre emphasis makes the treble louder, but it doesnt make the bass quieter, this is so the preemphasis doesn't make the audio too loud (preemp_unity is taken as a reference to where should it be 0 db, essentially if you put the lpf cutoff here, the treble isn't louder, but the bass is quieter) - i use 15000 here

rds_step is also at 0
this is meant to get the amplitude of rds2 subcarriers, so rds2 at 66.5 is rdsrds_step, and rds2 at 71.25 is rdsrds_step*rds_step and so on
you have disable rds2 by doing that

also it does not parse the device names, instead of doing mpx = none just remove the line, because an unknown name is the same as default source device, which is why the output technically works

Sorry - there was not example config files included in the package, so I am hoping I am not asking too dumb a question here

no worries - this is a complicated system meant to replace 100+ euro software and also to run on a raspberry pi so it is expected for there to be confusion

[chodgetts]

Thank you so much for all your work and help - I really appreciate it.

My settings were really just random values while I was trying to get everything working, so I'd definitely be interested to see what you're using if you're willing to share.

I actually work in commercial broadcasting here in New Zealand, but this is a personal project. I know exactly what broadcast processors cost, which is one of the reasons I'm so interested in trying yours out. From what I've seen so far, I'm really impressed.

I'm running everything on an Intel NUC, and I've discovered the built-in sound card is capable of 192 kHz output. It also supports PTP on the Ethernet interface. I don't actually need PTP for this project, but the hardware is phenomenal for the price.

I'm not even sure whether I'm allowed to use RDS2 in New Zealand, and I'm not sure if many car radios here support it yet. That said, if I can get it working and figure out how to send the images, I might end up being the first person in the country to have it running! 😂

It'll only be on a 1-watt Low Power FM service, so it won't reach very far anyway!

I've just started my weekend, and all of my test equipment is back at the office, so it won't be until Monday (my time) before I can confirm any changes to the settings. I'll definitely keep you posted.

[KubaPro010]

as for the configs, everything (including a setup script which converts pipewire into a processing chain) is at this repo

as for rds2, it should already work if you have the streams, but in order to send images, you can put a .script.lua in your rds95 root directory (build/.. - so the repo root) and run make install

here you can write code that rds95 will run on startup - some of with might be ran after start too, this is mine (i changed some rds values from what i actually use though, this station is not real)

-- Custom ODA
i = 0
j = 0
k = 0

function handler()
    i = i + 1
    if i >= 0xffff then
        i = 0
        j = j + 1
    end
    if j >= 0xffff then
        j = 0
        k = k + 1
    end
    if k >= 31 then
        i, j, k = 0, 0, 0
    end
    return true, k, j, i
end

table.insert(hooks.on_state, function ()
    ext.set_oda_handler(ext.register_oda(6, false, 0xffff, 1), handler) - This is also custom ODA

    local station_logo_rft = require("rft").new() -- These two lines are the only things you need for RFT, (you also need the on_state hook thing too)
 and sending logo/images (this is logo - for a generic image you should change 0xff7f to 0xffff)
    station_logo_rft:sendFile(0xFF7F, "/home/user/logo.jpg", 0, true, false)
end)

table.insert(hooks.on_start, function ()
    -- i use this so my settings for the station are loaded
    RDS.pi = 0xFFFF
    RDS.set_ps("RADIOX")
    RDS.pty = 11
    RDS.tp = false
    RDS.ta = false
    RDS.set_lps("RADIO eXtreme")
    RDS.ecc = 0xe1
    RDS.set_grpseq("\x00\x00\x00\x00\x00\x00\x00\x00\x04\x04\x04\x04\x04\x04\x02\x1e\x1e\x1f\x1f\x06\x06\xff\xff\xff") --rds1 group sequence - rds2 doesnt have any thing like this because it is controlled by lua scripting fully
end)

you can use rds spy to test if rft works with its plugin for decoding (available on their website) you can see the logo as received (you should see CRC error checking too)

[KubaPro010]

I have updated rds95 and fm95, and now they send the RDS data over a unix socket, which means you don't need to use the rds sink/source (you have to remove them actually, unrecognized keys which were settings nowcause an error at startup)

EDIT: (this also provides a small benefit, as now the clock which the rds bits are transmitted is actually derivied from the pilot [or rather the pilot is derived from this clock, but that's the same thing] as the clock in rds would only be used for the 57 khz subcarrier itself, but not whats on it) - this can extend rds range on some receivers

[chodgetts]

Hello,

Thank you for that.. this weekend has been the only time I have had any time to set it up.

Just got it running so going to test it today on air - but will change it up after my test..

But it sounds awesome from what I can see and hear - just do not have an RDS decoder so will need to do it on air to test properly :)

[KubaPro010]

I also implemented sync for the rds stream values, meaning you only set the number of streams in the encoder which is easier to do then set both's configs

All communication between rds95 and fm95 should happen in the socket at /etc/fm95/ctl.socket, rds95 sends the bitstream of all the streams as well as telling how many of them there are (fm95 starts always with no rds, but if rds95 is on, it will connected to the ipc, and do its thing, because if there was no data but streams are set then you would see two spikes instead of rds, try it, kill rds95 after running it but leave the fm95 instance running)

this means rds95 and fm95 connect seamlessly, really

EDIT: now i also removed it from the config, and now it should be settable from lua, you can use rdsgen=x command in the ascii udp (netcat 127.0.0.1 5000 -u) or use RDS.set_streams from lua

TL;DR: if you use netcat 127.0.0.1 5000 -u and type in RDSGEN=4, you get 4 RDS streams

i've heard RDS2 is restricted in some countries, but if NZ is not one of them, then use RDS2 (i recommend 2 or 3 streams though, the 4th one is just too quiet to hear it, and if you have a stream no one hears, then simply there might be no data, and receiving the logo might take forever)

[chodgetts]

Hello, thank you for all the work on this..
I have it working really well.
Although I cant seem to hit 100% mod - I only seem to reach about 70% mod without distortion.

Pilot and RDS levles are stable, so its not an output level thing.
When I run the cal tones, I hit 100% mod - this is on a BW Encore decoding the MPX

Just curious, I am running a bunch of pipewire filter chain - and the audio in my view sounds amazing, just a little low overall.

Could you make any suggestions on this one please?

Once again, thank you for your efforts in this, its an outstanding project.

[KubaPro010]

You should be able to have an higher volume by regulating the AGC (Make its its not off by setting agc max to 0)

Also you do have AGC data in your config completly wrong.

here, i'll explain what they actually do:

• agc_target: This selects the target value, theoretically it should be 1 (100%)
• agc_attack: This is the time that the AGC takes to react to a louder level (ms unit)
• agc_release: Same thing, but to a quieter level (ms unit)
• agc_min: This is the minimal multiplier that the AGC can output, if you put 1 here then the agc will do nothing
• agc_max: The max level that the agc can output, if you put 2 (200%) here, the agc will never output more than 2 times the loudness

None of these are in percentages, they are in raw values (e.g. 1% -> 0.01 | 10% -> 0.1 | 100% -> 1) where i believe your mistake was

EDIT: also your audio_preamp was at 0.1, this is not optimal (unless you drive fm95 at 1000% volume), but it does mean the agc, or bs412 (which i think saved you from having deviation measured in mhz, and also maybe the soft clipper at the end) have to work harder

[chodgetts]

thank you for that.
I was just following your configs in.
https://github.com/KubaPro010/scripts/blob/main/configs/fm95.conf

So I will go and amend them tomorrow and see how they go :)

[chodgetts]

sorry I should put this is my currently running config :
[fm95]
stereo=1
calibration=0
agc_target=0.9
agc_attack=0.1
agc_release=0.5
agc_min=0.1
agc_max=3.5
bs412_attack=0.01
bs412_release=1
bs412_max=1.5
bs412_gate=-25
bs412_knee=8
bs412_strenght=3.25
mpx_power=2.9

[advanced]
lpf_order=14
preemp_unity=15500
lpf_cutoff=15000
headroom=0.25
stereo_ssb=0
drive=1.0
makeup=3.2

[devices]
input=AFProcessor.monitor
#input=FM_Audio.monitor
output=alsa_output.pci-0000_00_1b.0.pro-output-0
rds=RDS.monitor

[volumes]
pilot=0.09
rds=0.04
rds_step=0.0

[chodgetts]

Hello,

I am feeling like a bit of an idiot..
I cant seem to get any more than about 50% mod on my MPX pos or neg.
I get a 7.5% Pilot and 4.5% RDS but audio is still only peaking at 60%

chris@mpxtest:~/radio_dev$ cat /etc/fm95/fm95.conf
[fm95]
master_volume = 1
stereo=1
rds_streams=1
audio_preamp=1.00
preemphasis = 50
calibration=0
agc_target=1.0
agc_attack=0.1
agc_release=0.9
agc_min=0.1
agc_max=2.5
bs412_attack=0.02
bs412_release=0.025
bs412_max=2.82
bs412_gate=-25
bs412_knee=3
bs412_strenght=0.8
mpx_power=0.2

[advanced]
lpf_order=14
preemp_unity=15000
lpf_cutoff=15000
headroom=0.05
stereo_ssb=0
drive=1.0
makeup=2.00

[devices]
input=FM_Audio.monitor
output=alsa_output.pci-0000_00_1b.0.pro-output-0
rds=RDS.monitor

[volumes]
pilot=0.075
#rds=0.035
rds_step=0.0

chris@mpxtest:~/radio_dev$ cat /etc/rds95.conf
[rds95]
streams=2
udp_port=5000
tcp_port=1234

[devices]
rds95=RDS

I cant work out what I am doing wrong...
Do I need an external amplifier to get to the transmitter to increase this level ?

[KubaPro010]

Increase mpx_power - it is the reference volume for the bs412 compressor, to something like 3 dbr

If you're not sure what it is, BS412 (as a standard) defines how to measure multiplex power (averaged value over 1 minute) and to refer it in dBr (0 dBr is 19 khz). The reason you're getting low volume because it is compressing your deviation to be by average 19.9 khz in audio

It is a required standard in European countries - including mine, you can't disable it without modifications to the code