• FM Stereo Radio Principles

    From G8MNY@21:5/101 to TECH on Sun Jul 21 05:52:53 2019
    R:190721/0848Z 9657@N7HPX.#BOI.ID.USA.NOAM BPQ6.0.18
    R:190721/0839Z 33513@N3HYM.MD.USA.NOAM BPQ6.0.18
    R:190721/0844Z @:GB7CIP.#32.GBR.EURO #:13065 [Caterham Surrey GBR]

    From: G8MNY@GB7CIP.#32.GBR.EURO
    To : TECH@WW

    By G8MNY (Updated Jul 19)
    (8 Bit ASCII graphics use code page 437 or 850, Terminal Font)

    Being involved with 30 short term radio stations in UK (RSLs) since 1994 & now a perminant station, here is part of a talk I do on it, on the technical side.

    AUDIO
    There are 3 main parameters to Audio quality.

    FREQUENCY RANGE
    Here is the approximate frequency plot for some audio sources. Note that the
    old disk system was not really limited & special equipment can do Quad audio
    with high frequency sub-carriers!

    │Sub Sonic│<- - - - -Young Human ear response - - - - - - ->│Ultra Sonic >
    │I BASS MID TREBLE
    │N <- - - - - - - - - - -Disk H i F i- - - - - - - - - - - >
    │F < - - - - - - - - -C.D. H i F i- - - - - - - - - - >
    │R <- - - - - - - Broadcast HiFi - - - - - - - - >
    │A < - - - - - Original Cassette- - - - - ->
    │ <- - - - - ─ A.M. Radio - - - - - ->
    │ F E E L I N G < - - COMMS - - ->
    └───┬────┬──┬───┬────┬───────┬───┬────┬───────┬───┬────┬──┬─┬──┬>Frequency
    DC 10 20 30 50 100 300 500 1k 3 5 10 15 20 30kHz

    CDs are the best source most people are familiar with, its frequency range is
    limited by the 44kHz sampling rate & the requirement to filter off all the
    frequencies higher than 22kHz to stop aliasing mixes caused by the sampling
    rate.

    Broadcast FM is limited to 15kHz to stop problems with the stereo pilot
    used. But 15kHz represent a good compromise for the upper limit which is why
    it was chosen.

    Broadcast AM (EU) uses 9kHz chanel spacing, so in theory 4.5kHz should be the
    upper limit, but in practice 6kHz is the limit (-40dB @ 9kHz) to make it sound
    a bit better.

    Comms Audio is the smallest bandwidth that can easily be understood, but not
    having any treble there is confusion over F & S, B P E G D, M N letter sounds!

    I have not included Digital Broadcast, as the quality is quite variable, from
    near CD quality, right down to phone call quality, dependent on the data rate
    assigned for a particular programme/ch.

    SIGNAL TO NOISE RATIO
    This is the measure of unwanted noises below the wanted sound..
    e.g. Hiss & Hum, or windage/engine noise, Neighbours/street noises etc.

    0dB┤ Noisy Conversation

    10 ┤ Poor Comms, NORMAL CAR
    20 ┤ Fair Comms
    30 ┤ VERY QUIET CAR AM RADIO
    40 ┤ Typical Cassette Tape, Living rooms
    50 ┤ Reel-Reel tape, Dolby Cassette
    60 ┤ New Vinyl Record FM RADIO/TV SOUND
    70 ┤ Mini disk (unmasked noise)
    80 ┤ Dat tape. DAB
    90 ┤ Perfect Digital CD, apparent Minidisk
    100 ┤
    110 ┤
    120 ┤ Ear Threshold Signal : Silence to Pain ratio

    With Digital sources there is also "Quantization Noise/Distortion", which is
    due to the sample step size, & is a set No of dBs below any sound level.

    HARMONIC DISTORTION
    This is the amount of unwanted signals generated in harmonics of the wanted
    signal, in the audio pass band of interest. It is usually very dependent on
    the level, except for digital systems where it is a mathematical design
    feature. Note the ear generates these too!

    It is measured as a % of the signal, so 10% = -20dB in harmonics.
    % dB
    10 ┼ -20 Cheap AM Radio at high Volume, Comms Audio
    5 ┼ -26 Film Optical Sound
    3 ┼ -30 Cassette & Reel Tapes
    1 ┼ -40 Quality Valve Amps AM BROADCAST
    .5 ┼ -46 High Quality Disk, FM BROADCAST
    .3 ┼ -50 Most Loudspeakers (not loud)

    .1 ┼ -60 Most AF Amps, DAB, NICAM
    0.05┼ -66 Good modern AF Amps.
    0.03┼ -70 Most Digital AF sources.

    Sometimes the above parameters are joined together in a "Signal In Noise And
    Distortion" (SINAD) rating for measured RF signal level of a Rx.

    FM EMPHASIS
    The FM Radio system suffers noise level that rises with frequency. To mask this the treble is lifted on Tx & cut on Rx, this substantially reduces the hiss, & top end harmonic distortion, but at the loss of treble peak loudness!
    e.g. with a time constant of 50uS (75uS USA) 15kHz is lifted by 14dB, that is 14dB less peak loudness at 15kHz, or only 20% of max level after de-emphasis.

    +14dB_│ TX Response _ │ Rx Response │Overall Response│
    │ ,/│ 0dB_│ ________ 0dB_│ ____________ │_100%
    │ ,/ │ / Flat `\ │ / `. | │
    0dB_│ ________./ 6dB/O │ 6dB/O`\ │/ `. │ │`.Level
    │ /Flat Lift 14dB_│ Cut `\ │ `. │
    └──┬──────┬──────┬─ └──┬─────┬──────┬─ └──┬─────┬─────┬─┴ 0%
    30Hz 1kHz 15kHz 30Hz 1kHz 15kHz 30Hz 1kHz 15kHz

    15kHz WALL FILTER
    As the treble is lifted & there is a requirement the audio does not interfere with the 19kHz stereo pilot tone, & higher frequencies of the stereo system, so a matched pair of sharp audio cut off filters are needed. The filter time delay MUST be the same, as the stereo image positioning you hear is all about treble timings.

    Other audio tailoring may reduce the subsonic audio as well.

    0dB_│ _____________________________________________
    -3dB-│ .-' '-.
    │ /' ▌
    │ ,/' ▐
    -70dB_│ /' V'-.__
    └┬────────┬────────┬────────┬────────┬────────┬────────┬──┬──┬─────
    10 30 100 300 1k 3k 10k15k 19k Hz

    LIMITING
    As FM must not be over deviated to keep the bandwidth down, a limiter is used, this is unlike a simple clipper used on comms Tx that lets the signal distort.

    Broadcast limiters have fast attack to cope with the spikiest peak, & with several decay time constants, mask the limiter's breathing effects. Complex limiters may also treat the treble separately with separate faster ti conqt`nts,A4he DBabld aonTEnt 7iL♀ be a mor proI♫%fd Pat lf t02e-

    e-phaSised qigA. ip♣ahp th♣♀%&♫☼``php♣☻GA N♀aeApa`AB♦♀ A☺
    D♣♫♣@@A☻♦SB TECH @ WW
    FM Stereo Radio Principles
    By G8MNY (Updated Jul 19)
    (8 Bit ASCII graphics use code page 437 or 850, Terminal Font)

    Being involved with 30 short term radio stations in UK (RSLs) since 1994 & now a perminant station, here is part of a talk I do on it, on the technical side.

    AUDIO
    There are 3 main parameters to Audio quality.

    FREQUENCY RANGE
    Here is the approximate frequency plot for some audio sources. Note that the
    old disk system was not really limited & special equipment can do Quad audio
    with high frequency sub-carriers!

    │Sub Sonic│<- - - - -Young Human ear response - - - - - - ->│Ultra Sonic >
    │I BASS MID TREBLE
    │N <- - - - - - - - - - -Disk H i F i- - - - - - - - - - - >
    │F < - - - - - - - - -C.D. H i F i- - - - - - - - - - >
    │R <- - - - - - - Broadcast HiFi - - - - - - - - >
    │A < - - - - - Original Cassette- - - - - ->
    │ <- - - - - ─ A.M. Radio - - - - - ->
    │ F E E L I N G < - - COMMS - - ->
    └───┬────┬──┬───┬────┬───────┬───┬────┬───────┬───┬────┬──┬─┬──┬>Frequency
    DC 10 20 30 50 100 300 500 1k 3 5 10 15 20 30kHz

    CDs are the best source most people are familiar with, its frequency range is
    limited by the 44kHz sampling rate & the requirement to filter off all the
    frequencies higher than 22kHz to stop aliasing mixes caused by the sampling
    rate.

    Broadcast FM is limited to 15kHz to stop problems with the stereo pilot
    used. But 15kHz represent a good compromise for the upper limit which is why
    it was chosen.

    Broadcast AM (EU) uses 9kHz chanel spacing, so in theory 4.5kHz should be the
    upper limit, but in practice 6kHz is the limit (-40dB @ 9kHz) to make it sound
    a bit better.

    Comms Audio is the smallest bandwidth that can easily be understood, but not
    having any treble there is confusion over F & S, B P E G D, M N letter sounds!

    I have not included Digital Broadcast, as the quality is quite variable, from
    near CD quality, right down to phone call quality, dependent on the data rate
    assigned for a particular programme/ch.

    SIGNAL TO NOISE RATIO
    This is the measure of unwanted noises below the wanted sound..
    e.g. Hiss & Hum, or windage/engine noise, Neighbours/street noises etc.

    0dB┤ Noisy Conversation
    10 ┤ Poor Comms, NORMAL CAR
    20 ┤ Fair Comms
    30 ┤ VERY QUIET CAR AM RADIO
    40 ┤ Typical Cassette Tape, Living rooms
    50 ┤ Reel-Reel tape, Dolby Cassette
    60 ┤ New Vinyl Record FM RADIO/TV SOUND
    70 ┤ Mini disk (unmasked noise)
    80 ┤ Dat tape. DAB
    90 ┤ Perfect Digital CD, apparent Minidisk
    100 ┤
    110 ┤
    120 ┤ Ear Threshold Signal : Silence to Pain ratio

    With Digital sources there is also "Quantization Noise/Distortion", which is
    due to the sample step size, & is a set No of dBs below any sound level.

    HARMONIC DISTORTION
    This is the amount of unwanted signals generated in harmonics of the wanted
    signal, in the audio pass band of interest. It is usually very dependent on
    the level, except for digital systems where it is a mathematical design
    feature. Note the ear generates these too!

    It is measured as a % of the signal, so 10% = -20dB in harmonics.
    % dB
    10 ┼ -20 Cheap AM Radio at high Volume, Comms Audio
    5 ┼ -26 Film Optical Sound
    3 ┼ -30 Cassette & Reel Tapes
    1 ┼ -40 Quality Valve Amps AM BROADCAST
    .5 ┼ -46 High Quality Disk, FM BROADCAST
    .3 ┼ -50 Most Loudspeakers (not loud)
    .1 ┼ -60 Most AF Amps, DAB, NICAM
    0.05┼ -66 Good modern AF Amps.
    0.03┼ -70 Most Digital AF sources.

    Sometimes the above parameters are joined together in a "Signal In Noise And
    Distortion" (SINAD) rating for measured RF signal level of a Rx.

    FM EMPHASIS
    The FM Radio system suffers noise level that rises with frequency. To mask this the treble is lifted on Tx & cut on Rx, this substantially reduces the hiss, & top end harmonic distortion, but at the loss of treble peak loudness!
    e.g. with a time constant of 50uS (75uS USA) 15kHz is lifted by 14dB, that is 14dB less peak loudness at 15kHz, or only 20% of max level after de-emphasis.

    +14dB_│ TX Response _ │ Rx Response │Overall Response│
    │ ,/│ 0dB_│ ________ 0dB_│ ____________ │_100%
    │ ,/ │ / Flat `\ │ / `. | │
    0dB_│ ________./ 6dB/O │ 6dB/O`\ │/ `. │ │`.Level
    │ /Flat Lift 14dB_│ Cut `\ │ `. │
    └──┬──────┬──────┬─ └──┬─────┬──────┬─ └──┬─────┬─────┬─┴ 0%
    30Hz 1kHz 15kHz 30Hz 1kHz 15kHz 30Hz 1kHz 15kHz

    15kHz WALL FILTER
    As the treble is lifted & there is a requirement the audio does not interfere with the 19kHz stereo pilot tone, & higher frequencies of the stereo system, so a matched pair of sharp audio cut off filters are needed. The filter time delay MUST be the same, as the stereo image positioning you hear is all about treble timings.

    Other audio tailoring may reduce the subsonic audio as well.

    0dB_│ _____________________________________________
    -3dB-│ .-' '-.
    │ /' ▌
    │ ,/' ▐
    -70dB_│ /' V'-.__
    └┬────────┬────────┬────────┬────────┬────────┬────────┬──┬──┬─────
    10 30 100 300 1k 3k 10k15k 19k Hz

    LIMITING
    As FM must not be over deviated to keep the bandwidth down, a limiter is used, this is unlike a simple clipper used on comms Tx that lets the signal distort.

    Broadcast limiters have fast attack to cope with the spikiest peak, & with several decay time constants, mask the limiter's breathing effects. Complex limiters may also treat the treble separately with separate faster time constants, as the treble content will be a more prominent part of the pre- emphasised signal. Any peaks then left not properly gain limited (hardly any) then get hard clipped.

    To maintain the stereo image both left & right gains must be tracked together!

    The result is a signal that has it's ± peak value accurately limited, but sounds perfect! With a good limiter you should not be able to tell the difference between live studio feed & off air with limiting of around 12-20dB.

    The peak values result in the actual FM deviation, which ensures the correct overall Tx bandwidth.

    │ │ Peaky pulse
    +│ ││ The result is a signal that has
    │_.│ │ /'\ it's ± peak value accurately
    0┼─────┼──┴──┼──────. limited, but sounds perfect!
    │ '|,^| │ With a good limiter you should
    -│ U\/ not be able to tell the
    difference between a live
    Multiple CR studio feed & off air even with
    0dB│-. Recovery . high limiting levels of around
    │ │ ..-'\/\ _ _.' 12 - 24dB used for AM.
    GAIN │ │|' `' \|
    -20dB│ ' Fast attack
    └───┬────┬────┬────┬───>Time
    2 4 8 10s

    As the limiting process must have fast attack times to handle all the peaks one half cycle will Rx a different compression factor to the other half cycle, this results in some low frequencies down to DC being added to the signal.

    A scope X-Y plot of Stereo (Pilot tones filtered off)
    ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐
    │ │ │ ┬ │ │ · │ │ · │ │ ┐ │ │ │
    │ │ │ │ │ │ \ │ │ / │ │ / │ │ ▒▒▒▒▒ │
    │ ├─────┤ │ │ │ │ │ \ │ │ / │ │ / │ │ ▒▒▒▒▒ │
    │ │ │ │ │ │ \ │ │ / │ │ / │ │ ▒▒▒▒▒ │
    │ │ │ ┴ │ │ · │ │ · │ │ └ │ │ │ └─────────┘ └─────────┘ └─────────┘ └─────────┘ └─────────┘ └─────────┘
    Limited Mono Anti-phase 2D Stereo Hard Limited Stereo fills
    Left only Right @ 45° (behind U) Pattern ±75kHz Box

    If you set up a scope (or PC) to display broadcast audio you can see the stations that run hard limiting, as they end up with a tightly defined box filled all the time. This is not that they have clipped audio with distortion, but quickly acting AGCs do that do not generate AF harmonics. The result is very LOUD audio, each audio frequency can still have 40dB or so dynamic range, but the overall modulation is 100% nearly all the time (each millisecond). Such stations can be a strain on the hearing to listen to!

    BAD LF RESPONSE AFTER LIMITER
    It is also important that there is no phase distortion between the limiter & the Tx over the frequencies to be transmitted. If there is the limited signal can actually get larger....

    No Phase error @ 20Hz +110kHz_│ ._ With Poor LF Phase
    +100%_│ _ __________ _ _ _ _ _ 140%_│_ _│ `-._ _ _ _ _ _ _ _│ _ +75kHz│ │ │ 100% │ │ `-._ │
    │ │ │ │ │ │ │
    │ │ │ │ │ │ │
    │ │ │ │-._│ │ _│ -100%_│___│ _ _ _ _ _│__________│ _ -100%_│_ _ _ _ _ _ _│ _.-'_ _ -75kHz│ -140%_│ │_.-'
    Audio with large 20Hz -110kHz│ '
    content limited to just fit -3dB @ 20Hz
    inside this 20Hz envelope. Tx Peaks now seriously over modulated

    The same problem occurs in an AM Tx, where the poor LF phase response on high level Modulation Transformers cause unexpected hard clipping on certain AF envelope waveforms. (not the AF's LF content!)

    In practice an FM Tx will have another hard limiter (clipper) to protect it from accidental over modulation.

    STEREO MULTIPLEX
    The system used for all stereo radio broadcasts is known as the Zenith-GE Pilot Tone System (so-called after the names of the two companies who devised it).
    It has been designed to be fully compatible with Mono FM radio Rx & without too much increase in bandwidth.

    The MUX signal can be made with a 38kHz Double Side Band exciter fed with a LR difference signal, & added to the mono signal. But modern linear switching electronics, means the simpler methord is now used. That is to take a sample of the left & right channels every 38kHz. So a switch samples the left or right channels @76kHz, to keep the Rx's switch in step a locked 19kHz sinewave pilot tone is sent at -20dB (10%) below peak level. The phase of the pilot tone is critical to good channel seperation.

    76kHz
    Left ──o Switch
    \,_______________________┌─────┐ ┌────────┐ ┌─────┐
    Right──o /|\ │ Add ├──>─┤ 53kHz ├─>──┤ Add ├──> to FM Tx
    │ ┌──┐ ┌───────┐ └──┬──┘ │Low Pass│ └──┬──┘
    76kHz >──┴───┤÷4├───┤ 19kHz ├─────┤-20dB │ Filter │ │-30dB
    └──┘ │Filter │ │ └────────┘ ┌────┴────┐
    └───────┘ └─────────────────>─┤RDS 57kHz│
    PLL │ UNIT │
    └─────────┘

    _ Left Right MUX _
    / \ ───┐ ┌───── ─ ─ ─/─ ┐\ ┌ ─ ─ ─ ─
    ▐ ▌ │ │ ▐ │ ▌ │
    │ │ │ │ │ │ │ │ │
    ▐ ▌ │ │ │ ▐ │ ▌
    \_/ └───┘ └ ─ ─ ─ ─\┘_ /
    Output after switch
    (Expanded for diagram)

    The low pass filter @ 53kHz used, must have low phase shift (group delay) so that the timings of the stereo samples are not affected. In some designs digital tricks are used to null out the 2 & 3rd order harmonics of the switching process (76kHz & 114kHz), so a less aggressive "lower distorting" low pass filter can be used.

    RDS (Radio Data System, similar to ARI system in Germany)
    This is data phase modulated on to a 57kHz carrier that is phase locked to the 19kHz, so it actually reduces the overall modulation & is added at 2-3% to the MUX output. The data is QPSK @ 1187.5Hz (76kHz/64) which only occupies about 2kHz bandwidth (seen as 2 data Carriers). It contains many features, not many supported by stations, the common ones are...

    PS = Station Service Name
    PI = Tx ID code, up to 256 stations in a network
    AF = Alternative Frequency list (radio searches this on weak signal)
    CT = Time
    TA & TP = Traffic Flags, allows radio to change AF source/level.
    PTY= Programme Type 16 types, e.g. Jazz, News, Pop. etc.
    RT = Radio Text, e.g. current song title

    MUX BASEBAND SPECTRUM
    __ __ __
    0dB_│ _____...---''' │ P │ ```---..._____ _____...---''' │
    ││ │ I │ │ │ │
    ││ │ L │ Difference │ │ Difference │
    ││ L+R │ O │ Signal │ │ Signal │
    -20dB_││ MONO │ T │ DSB L-R │ │ DSB L-R │ RDS -30dB_││ │ │ │ │ │ │
    └┼────────────────┼───┼───┼────────────────┴┬┴────────────────┼───╨──
    30Hz 15 19 23kHz 38kHz 53 57kHz

    To achieve good channel separation linear frequency & phase response is needed between the Tx coder & the Rx decoder. Reduced levels or phase shift @ 38kHz make the channel separation poor (tending to MONO), & increased HF gain widens the channel separation.

    The increase in baseband bandwidth from 15kHz for mono to 53kHz for stereo causes about 20dB loss in overall signal to noise ratio on an FM system. As it adds in, not just the noise from 3x the bandwidth (9dB), but the very poor signal to noise, of the higher frequency stereo difference signal that you Rx on FM systems. The resultant "noise" in the stereo image appears as noise from behind you....

    LEFT RIGHT
    o^o
    [ ]
    \_/

    NOISE

    TX SPECTRUM
    The Bessel Functions shows the FM sideband harmonic levels, for any particular modulation index. For mono the modulation index is peak Dev/Mod 75kHz/15kHz = 5 but this analysis is less useful for real very complex signals.

    This is where Carson's rule for minimum bandwidth is needed, can give clearer indication.
    Bandwidth = 2x Peak Dev + 2x Highest Mod Freq

    This gives the bandwidth of sidebands needed for NO distortion. But it does not take into account, that the levels of the highest modulation frequency are only 3% (-30dB) of the peak deviation, with resultantly weak sidebands.

    Deviation
    Mod <-2x 75kHz -> Mod 0dB_│ _ │_100%
    <57.5>┌────────────┐<57.5> │ / \ │
    ______│ │______ -20dB-│ │ │ │─10%
    │ │ _│ /' `\ │_
    │ │ -40dB │ ./' `\. │ 1% ───┼─────────────┼────────────┼─── ─┴─┴────┬─────┬────┴─┴─
    < - - - - - 265kHz - - - - > -60dB 100kHz 0.1%
    Theoretical Full Bandwidth More Typical bandwidth
    of RDS Stereo Broadcast signal as seen on spectrum analyser
    for ZERO distortion. under heavy modulation.

    TX RF Harmonics & Mixes
    These should all be > -60dBc, so added filters are normal. On multiple Tx sites there is a risk of PA mixing, where RF from a nearby Tx can be Rx at the Tx PA at enough strength to cause a Mix. A narrow resonant channel filter or directional coupler (Circulator/isolator) in the Tx feed can protect the Tx from these signals.

    \│/ TYPICAL TX SITE LINE UP
    │ ┌──────────┐ ┌──┐ ┌──┐ ┌─────┐ ┌──────┐ ┌──────────────┐
    └─><─┤CIRCULATOR├─┤PA├─┤FM├─┤ RDS ├─┤STEREO├─┤PRE EMPHASISED├─o-o─<L STUDIO
    Rx+Tx│OR FILTER │ │ │ │Tx│ │CODER│ │CODER ├─┤STEREO LIMITER├─o-o─<R FEED Signal└──────────┘ └┬─┘ └─┬┘ └──┬──┘ └──┬───┘ └────┬─────────┘ │ │
    test point \│/ \│/ Data Mono \│/ Backup
    Power Mod flags Levels Source

    Also see my buls on "FM Deviation Calibration", "RF Directional Coupler",
    "AM Broadcast principles" & "1W @ 531kHz MW system".


    Why Don't U send an interesting bul?

    73 de John G8MNY @ GB7CIP


    --- Mystic BBS v1.12 A43 (Linux/32)
    * Origin: HAMRADIO telnet lu9dce.dynu.com (21:5/101)