WFM Stereo Broadcasting Standards - OIRT vs CCIR

[radiolistener]

As you probably know, there are two standards for WFM stereo broadcasting - one applied in the OIRT band (65.9 - 74 MHz) and the other in the CCIR band (87.5 - 108.0 MHz).

OIRT uses 31.25 kHz subcarrier to transmit L-R component, with so called "polar modulation" and partially suppressed subcarrier.

CCIR uses 38 kHz subcarrier to transmit L-R component, with DSB-SC and pilot-tone at 19 kHz in order to recover subcarrier.

I am working on adding OIRT FM stereo decoder support to my receiver. Unfortunately, there is no longer any stereo broadcasting in the OIRT band in my region - only mono transmission remains. I also don’t have access to IQ recordings of actual OIRT stereo broadcasts for testing, so I have to rely on old magazine articles and schematics of stereo decoders.

One point I find confusing is the terminology: in the OIRT stereo composite signal, the modulation is often described as "polar modulation". It is not entirely clear to me what exactly is meant by this.

From what I understand, the stereo composite modulation in both CCIR and OIRT systems is essentially DSB. The main difference is that in CCIR, the 38 kHz subcarrier is completely suppressed and recovered on the receiver side by doubling the transmitted 19 kHz pilot tone, whereas in OIRT stereo-composite there is no pilot tone - instead, the 31.25 kHz subcarrier is only partially suppressed (about -14 dB), allowing it to be recovered directly on the receiver side.

Apart from this subcarrier recovery method, it seems to me that the modulation principle is the same in both systems, so the same DSB demodulator could be used for both standards, with the only difference being the subcarrier recovery mechanism.

Am I understanding this correctly?

If anyone happens to have a recording of an FM broadcast in OIRT stereo mode (65.9 - 74 MHz band), I would greatly appreciate it. Ideally this could be provided as a WAV file containing an IQ IF stream, or alternatively as a direct recording of the composite signal from the FM detector output.

[gf]

Here's what Google Gemini told me:
https://g.co/gemini/share/2583583ca5ad
I can't assess correctness, but it sounds plausible.

[radiolistener]

Here's what Google Gemini told me:
https://g.co/gemini/share/2583583ca5ad
I can't assess correctness, but it sounds plausible.

It seems the Gemini AI response introduces some confusion here. It states:

In the OIRT system, polar modulation is used to modulate the main carrier, not the subcarrier.

which is clearly incorrect. In fact, polar modulation in the OIRT stereo system refers to the modulation of the stereo subcarrier in the composite signal, not the main FM carrier itself.

If I understand correctly, the mathematical difference can be described as follows (Octave code), we have:

Code: [Select]

Fs = 384000;            % sample rate
N = length(L)           % sample count

t = (0:N-1).' / Fs;     % time vector

% Build sum and difference
add_LR = (L + R) / 2;   % sum channel
sub_LR = (L - R) / 2;   % difference channel

Then:
- CCIR uses this:

Code: [Select]

pilot = 19000;          % pilot tone frequency 19 kHz
subcf = pilot*2;        % sub-carrier frequency 38 kHz

pilot = cos(2*pi*pilotf*t);        % pilot tone 19 kHz
c38   = -sin(2 * 2*pi*pilotf*t);   % carrier 2*19=38 kHz

dsbsc = sub_LR .* c38;

% Composite baseband signal (L+R + pilot + DSB-SC)
composite = 0.45*add_LR + 0.1*pilot + 0.45*dsbsc;

While OIRT uses something like this:

Code: [Select]

subcf = 31250;          % sub-carrier frequency 31.25 kHz

c3125 = sign(cos(2*pi*subcf*t));  % squarewave subcarrier

polar = 0.2*c3125 + sub_LR .* c3125;

% Apply bandpass filter with center frequency 31.25 kHz ±15 kHz
polar = filtfilt(bpf_31250, 1, polar); 

% Composite baseband signal (L+R + polar)
composite = 0.45*add_LR + 0.45*polar;

And since it uses squarewave carrier with following BPF, it is actually equals to:

Code: [Select]

c3125 = cos(2*pi*subcf*t);  % sine wave subcarrier

polar = 0.2*c3125 + sub_LR .* c3125;

% Composite baseband signal (L+R + polar)
composite = 0.45*add_LR + 0.45*polar;

which is more easy and clean, because don't involve BPF to remove harmonics added with square wave carrier.

As result, in the OIRT case, the square wave subcarrier generates harmonics, which must then be removed by a band-pass filter to keep the spectrum within ±15 kHz. After filtering, this is essentially equivalent to using a DSB-SC scheme with a partially suppressed carrier. By contrast, the CCIR system is simply DSB-SC with the carrier fully suppressed and reconstructed from the 19 kHz pilot tone.

So in effect, the only difference is that OIRT leaves a residual carrier (at about -14 dB) while CCIR completely suppresses it. Also it means that you can use polar demodulator to decode both - OIRT and CCIR system. The only difference is a subcarrier recovery scheme.

Am I understanding this correctly?


I do not have access to the real OIRT stereo signal recording to verify my understanding of the stereo composite modulation in this system, so I have to rely on fragmented descriptions and circuit diagrams from old radio magazine publications from 1970...

Here is one of the schematics published in Radio magazine, issue No. 3, 1974. According to the circuit, transistor T1 amplifies the signal, T2 recovers the 31.25 kHz subcarrier, then the low-frequency part (L+R) is extracted, while the high-frequency part containing (L–R) is amplified by T3 (L3 is tuned at 31.25 kHz, Q=4.9, R16 is used to decrease Q-factor) and demodulated using a diode bridge. Finally, the L and R channels are formed as:

Code: [Select]

L = (add_LR + sub_LR) / 2;
R = (add_LR - sub_LR) / 2;


[gf]

Here's what Google Gemini told me:
https://g.co/gemini/share/2583583ca5ad
I can't assess correctness, but it sounds plausible.

It seems the Gemini AI response introduces some confusion here. It states:

In the OIRT system, polar modulation is used to modulate the main carrier, not the subcarrier.

which is clearly incorrect. In fact, polar modulation in the OIRT stereo system refers to the modulation of the stereo subcarrier in the composite signal, not the main FM carrier itself.

I really don't known. However, polar modulation makes only sense to me in a context where a single carrier is modulated with two modulating signals. And this applies rather to the main carrier. Since the sub-carrier is only modulated with a single modulating signal (L-R), I see no need for polar modulation for the sub-carrier at all. What would be the 2nd modulating signal?

Of course, instead of polar modulation, L+R and the modulated sub-carrier could also be added, and then FM-modulated onto the main carrier (like CCIR). But then no polar modulation would be in the play at all.

[radiolistener]

Perhaps someone could help by performing a practical test: the idea is to transmit a generated composite signal with a low-power transmitter and check whether OIRT stereo receivers produce correct stereo decoding. The test signal could be prepared either as a baseband composite signal for direct input into an FM modulator, or as a fully FM-modulated IQ recording centered at 0 Hz, which would then only need to be shifted to the desired frequency.

I checked through my old radio receivers, but unfortunately all of them are equipped only with CCIR stereo decoders, so at the moment I have nothing suitable to test with.

If anyone has an old radio receiver with an OIRT stereo decoder, or access to a local OIRT stereo broadcast, I would greatly appreciate it if you could let me know.

[gf]

After challenging Gemini and pointing out contradictions, it changed its mind several times and its most recent conclusion was

Yes, based on all technical sources, pure FM for the main carrier is the correct and only method used in both the OIRT and CCIR standards.

The "polar modulation" that has caused so much confusion does not refer to the modulation of the main carrier. It is a historical term used to describe a specific way that the stereo information is encoded and combined with the mono signal before it ever reaches the main transmitter.

Main Carrier vs. Composite Signal

Think of the process in two steps:

    The Composite Baseband Signal: This is the complete audio signal, which contains all the necessary components for both mono and stereo reception.

        It includes the L+R signal (the mono sound).

        It includes the stereo subcarrier (the L-R signal modulated onto a subcarrier, either at 31.25 kHz for OIRT or 38 kHz for CCIR).

    It might also include other data signals like RDS.

The Main FM Carrier: This is the powerful radio wave that is actually broadcast from the tower. The entire composite baseband signal from the first step is used to frequency-modulate (FM) this main carrier. The main carrier's frequency is shifted up and down according to the instantaneous amplitude of the composite signal.

In both the OIRT and CCIR systems, the main carrier's modulation is strictly FM. The difference between the two systems is in what makes up the composite signal, not in how that composite signal modulates the main carrier.

Based on my sources, the OIRT FM stereo subcarrier uses Amplitude Modulation (AM). Specifically, it uses Double Sideband Reduced Carrier (DSB-RC) AM.

The alternative, and incorrect, claim that it uses FM is a major point of confusion. Here is the reason why the AM explanation is technically sound and the FM explanation is not:

    Synchronization: The primary difference between the OIRT and CCIR standards is how the receiver decodes the stereo signal. The OIRT system's 31.25 kHz subcarrier is intentionally transmitted at a low, "reduced" amplitude (about 14 dB below the main signal). A receiver's phase-locked loop (PLL) uses this partially transmitted carrier to synchronize with and demodulate the stereo signal.

    Technical Impossibility of FM Suppression: The concept of "partially suppressing" a carrier only applies to amplitude-modulated signals. The amplitude of an FM signal is constant; all the information is encoded in its frequency changes. You cannot reduce its amplitude for synchronization without losing the information it's carrying.

The term "polar modulation" is a historical misnomer that has contributed to this confusion. In this context, it describes the OIRT system's method of combining stereo signals in a way that is different from the CCIR system, but it does not accurately describe the subcarrier's actual modulation. The subcarrier itself is still amplitude-modulated.

I don't have direct access to the original, official standards for the OIRT or CCIR systems. These are historical documents not available in my training data, which consists of publicly accessible online text. My information is based on secondary sources, such as technical summaries, forum discussions, and educational texts that describe the standards. This reliance on secondary sources is why my previous explanations contained inconsistencies.

So I think your assumption is correct. It's likely similar to CCIR, but w/o pilot tone, and with AM/DSB-RC instead of AM/DSB-SC for the subcarrier, and the term "polar modulation" is just used to create confusion.

Today, the term 'polar modulation' does not refer to a specific type of modulation, but rather to a particular power amplifier (PA) architecture in which a linear PA is emulated using a highly efficient non-linear PA. The same RF signal that can be generated using I/Q modulation and a linear PA can also be generated using a polar modulation PA, with better efficiency.

[radiolistener]

I came across this discussion on generating the OIRT stereo composite:
https://forums.stereotool.com/viewtopic.php?t=6233

Unfortunately, there are few technical details and no example signals provided, but it seems that the authors of StereoTool have implemented a working version of polar stereo encoding for OIRT. If anyone here uses this tool, could you possibly create a small test wav file with a stereo composite signal in polar stereo format?

The discussion also references a standard that provides some details on deviation and the relative levels of the components in the composite stereo signals for both OIRT and CCIR systems:
https://www.itu.int/dms_pubrec/itu-r/rec/bs/R-REC-BS.450-3-200111-S!!PDF-E.pdf

According to this document, the frequency deviation for OIRT stereo broadcast is 50 kHz instead of 75 kHz which is used in CCIR.
Also it mentioned OIRT stereo-composite proportions:

2.1.2.2 A signal S is produced equal to one half of the difference between signals A and B mentioned above. This signal, S, is pre-emphasized in the same way as signal M. The pre-emphasized signal, S, is used for the amplitude modulation of a sub-carrier at 31.25 kHz; the spectrum of the amplitude-modulated sub-carrier is formed so that the sub-carrier amplitude is reduced by 14 dB and the spectral components of the given modulating signal appear to be transformed as follows:

\$\overline{K}(f) = \frac{1 + j\,6.4 f}{5 + j\,6.4 f}\$

where f is equal to each frequency component (kHz).

2.1.2.3 The stereophonic multiplex signal is the sum of:
– the pre-emphasized signal, M;
– the sideband spectral components which are the product of amplitude-modulated unsuppressed carrier by a pre-emphasized signal S additionally transformed from the law \$\overline{K}(f)\$
– the sub-carrier with the amplitude reduced by 14 dB.

2.1.2.4 The amplitudes of the various components of the stereophonic multiplex signal, referred to the maximum amplitude of that signal (which corresponds to the maximum frequency deviation) are:
– signal M : maximum value 80% (A and B being equal, and in phase);
– signal S : maximum value 80% (A and B being equal but of opposite phase);
– reduced sub-carrier at 31.25 kHz; maximum residual amplitude 20%.

2.1.2.5 The frequency modulation is arranged in such a way that positive values of the multiplex signal correspond to a positive frequency deviation of the main carrier and negative values to negative frequency deviation.

A = left channel
B = right channel
M = (A+B) / 2
S = (A-B) / 2

Both M and S are pre-emphasised with pre-emphasis 50 us (standard pre-emphasis for Europe).
AM modulated signal (which consists of S message) is spectrum shaped with mentioned K(f) law in order to suppress carrier for -14 dB.

From the text it is not entirely clear whether the spectrum of the AM-modulated signal should be corrected according to K(f), or if the more correct approach is simply to combine the full DSB-SC amplitude with the carrier attenuated by -14 dB.

[gf]

From the text it is not entirely clear whether the spectrum of the AM-modulated signal should be corrected according to K(f), or if the more correct approach is simply to combine the full DSB-SC amplitude with the carrier attenuated by -14 dB.

Yes it isn't clear. Interestingly, the transfer function K(f) is a highpass with ~0.8Hz cutoff and 14dB stopband attenuation. If the subcarrier amplitue is suposed to be represented by the DC component of the baseband modulating signal, then this filter - applied to the baseband signal - would provide the -14dB subcarrier attenuation.

[Njk]

If anyone has an old radio receiver with an OIRT stereo decoder, or access to a local OIRT stereo broadcast, I would greatly appreciate it if you could let me know.

Well, we still have 3 working stations of the OIRT band here in St. Petersburg. For one of them, 69.47 MHz, my old receiver indicates a stereo signal (just checked).
https://radiomap.eu/ru/sankt-peterburg
Perhaps you can use a WebSDR connection to view the signal

[radiolistener]

For one of them, 69.47 MHz, my old receiver indicates a stereo signal (just checked).
Perhaps you can use a WebSDR connection to view the signal

Unfortunately websdr doesn't support wideband FM, its limit - max 20 kHz, while OIRT stereo FM requires at least 96.25 kHz bandwidth. And there is no websdr receiver covering 69.47 MHz at that location, only ham bands on HF and VHF...

Could you please record the file with baseband IQ of this station with RTLSDR? You can do it with HDSDR.

It is important to setup LO outside of station bandwidth - use LO=69.3 MHz or LO =69.7 MHz for recording, it helps to avoid DC spur which can affect signal quality. It's better to setup RTLSDR sample rate 960 kHz instead of 2.4 MHz to reduce file size, since I need just +-150 kHz around station frequency. Also please disable AGC and setup good RF gain manually, to avoid unwanted gain variations.

30-60 seconds record will be enough. If possible, please record it when station broadcasting some music or sound which allows to hear stereo effects and distinguish it from mono. If such music composition will allow to check proper left/right channel decoding it will be great.


You can share file on https://transfiles.ru/

[Njk]

The problem is that it's old Soviet receiver, Leningrad RP-015 to be precise. It has no IQ outputs. The model is well known so you can find the schematic diagrams on the internet. There are no patented chips, it's made entirely of discrete components. There is a test point right after the FM detector and before the stereo decoder, marked with "80 mV". I think it would be possible to route the baseband signal from the test point to the input of a PC sound card for recording. But it's not required for a sound card to support the BW above 20 kHz, while all the decoding magic happens at the frequencies of up to 62 kHz. So I'm not sure the recorded file will be useful. Anyway, the stereo decoder in the receiver does not seem a very complex circuit. Perhaps you can get a clue by looking into the diagram.

I don't have an SDR dongle either. In my case, it just does not make sense. It's a big city and the reception is very poor. I can't hear anything on the AM bands no matter of the receiver. While the FM provides only a local content. So I'm not an SDR fan. I can only confirm the existence of the stereo signal in my area

[radiolistener]

It is a pity that you don’t have any SDR hardware available to make a direct recording of the signal. Even a very inexpensive dongle (10–20 USD) can do much more than just listening to radio – it allows spectrum analysis, interference studies from household equipment, debugging and aligning radio circuits across HF, VHF and UHF. I also live in the center of a large city with plenty of noise, but radio is still quite usable here: both HF and VHF broadcasting, amateur stations, satellites, and even frequency standard signals that can be used to calibrate frequency counters, generators, receivers or transceivers.

You are right that the stereo decoder circuitry itself is not extremely complex, but all the phase relationships and details matter a lot, so extracting the exact encoding parameters from schematics alone is unreliable – we really need real-world signals for verification. On your receiver’s diagram you indeed pointed out the correct place for recording the stereo composite: the KT7 80 mV test point. However, as you mentioned, capturing it with a standard sound card is problematic because of the input low-pass filter.

For composite recording an IQ signal is not required – it’s just a mono baseband with about 100 kHz bandwidth. In theory it could be digitized with a sound card at 384 kHz sampling rate (giving 192 kHz bandwidth), but in practice most sound cards have an input LPF around 20 kHz. A better option would be a capture card with 150–200 kHz effective bandwidth, or even a digital oscilloscope with sufficient memory depth. But I suspect you may not have such equipment either.

The most practical solution would be recording the IF with an SDR receiver. Perhaps you could borrow an RTL-SDR from a friend – these dongles are very popular among radio amateurs nowadays because they are inexpensive and versatile. They allow not only receiving nearly any band from longwave to UHF, but also testing and aligning radio equipment.

If you don’t have access to an SDR, you could still try recording the composite with a sound card, just in case. To compensate for the LPF effect, it would be very helpful if you could also record plain white noise from the same input – then I could attempt to restore the frequency response of your sound card and design a correction filter to recover as much of the high-frequency content as possible. Often the LPFs in sound cards are quite weak and do not completely suppress higher frequencies. Of course, this is a rather problematic and challenging approach to try to reconstruct the signal in such a way, but since there are no other available options to obtain an OIRT stereo composite sample, it might still be worth attempting.

It is especially regrettable that, from what you describe, your local station really does broadcast stereo in the OIRT system. According to the schematic, your receiver indeed detects OIRT stereo with the 31.25 kHz subcarrier, so there is little doubt about it. It is a pity that you don’t have the technical means to capture and record this signal. Broadcasts in this system are becoming quite exotic nowadays: much of the old transmitting equipment has aged and gone out of service, and modern replacements for OIRT seem to be very rare. In most cases stations still operating in this band transmit only in mono.

I am very interested in implementing an OIRT stereo encoder/decoder in DSP and experimenting with its pros and cons compared to the CCIR system. Unfortunately, in my region broadcasting in this band is only mono, so I have no chance to capture such signals myself.

Can you help with obtaining such OIRT stereo-composite recording for testing?

[Njk]

How many samples are needed? For me, the simplest option seems to capture the baseband signal with Rigol DS1054Z scope (check the specs. It's the only digitizing scope I have)

[radiolistener]

How many samples are needed? For me, the simplest option seems to capture the baseband signal with Rigol DS1054Z scope (check the specs. It's the only digitizing scope I have)

As many samples as possible

For example, with a Siglent scope that has 14 Mpts of memory, I can capture 14 seconds at a 1 MSa/s rate. To do this, it is necessary to manually disable Roll mode (which is enabled by default below 50 MSa/s) and to enable the 20 MHz low-pass filter to minimize aliasing. I think that 14 seconds should be sufficient to clearly hear the stereo.

There is, of course, a risk that aliases from frequencies above 500 kHz may affect the signal quality, but it is still worth trying.

[Njk]

Ok then. Take the receiver apart, blow the dust off and capture the signal. Will do it on the next week (hopefully). BTW for better anti-aliasing, I can arrange a simple passive LP filter (two res, two caps) with the cutoff frequency of about 100 kHz

[radiolistener]

Ok then. Take the receiver apart, blow the dust off and capture the signal. Will do it on the next week (hopefully). BTW for better anti-aliasing, I can arrange a simple passive LP filter (two res, two caps) with the cutoff frequency of about 100 kHz

No, it is better not to use an LPF at all than to add a simple 100 kHz one, since such a filter can significantly attenuate the amplitude of the encoded L–R signal, given that the OIRT stereo composite has a bandwidth of 100 kHz. If you do apply an LPF, it should have a much higher cutoff, around 300–400 kHz.

I am looking forward to the test sample — it will be interesting to see how the OIRT stereo composite actually looks.

[radiolistener]

I came across an article about polar modulation for stereo signals. It refers to a 31.25 kHz subcarrier, which suggests it is describing the OIRT stereo composite standard. However, what seems odd is that the article claims polar modulation assigns the left and right channels to the corresponding polarities of the subcarrier. This is unusual, since earlier publications indicated that the subcarrier carries the difference signal (L−R), in addition to the mono sum signal (L+R). Moreover, the detection circuit shown appears overly simplistic—essentially just two diodes.

What exactly is being described here? Is this an earlier revision of the standard, or simply a misunderstanding and the circuit itself is incorrect?

[gf]

I haven't checked in detail yet but I think the described asymmetric AM does not fit within the +-15kHz bandwidth around the sub-carrier. And if you remove sub-carrier harmonics and their sidebands, I think you lose the intended asymmetry.

[radiolistener]

Yes, that’s exactly the confusing part. There are many references claiming that OIRT used some kind of “polar modulation”, but I suspect there is some kind of a misinterpretation. From what I can see, the system probably is essentially the same as CCIR stereo: the baseband carries L+R, and the 31.25 kHz subcarrier is DSB-modulated with L−R, only with the subcarrier left partially suppressed (at −14 dB) rather than fully suppressed as in CCIR. 

The question then is: where do all the statements about direct diode decoding come from? The article above even includes a schematic with nothing more than two diode detectors and RC filters. This seems questionable — could such a simple circuit really decode a stereo signal in which the subcarrier is modulated with L−R?

That’s why I’m looking for an actual recording of an OIRT stereo broadcast. With a real composite signal it would be possible to analyze the spectrum and confirm what is really transmitted.

I did find a screenshot of an OIRT station transmitting in stereo — the composite signal spectrum is faintly visible there — but unfortunately that’s the only material I’ve managed to locate so far.

[Njk]

To radiolistener

[gf]

Yes, that’s exactly the confusing part. There are many references claiming that OIRT used some kind of “polar modulation”, but I suspect there is some kind of a misinterpretation. From what I can see, the system probably is essentially the same as CCIR stereo: the baseband carries L+R, and the 31.25 kHz subcarrier is DSB-modulated with L−R, only with the subcarrier left partially suppressed (at −14 dB) rather than fully suppressed as in CCIR. 

The question then is: where do all the statements about direct diode decoding come from? The article above even includes a schematic with nothing more than two diode detectors and RC filters. This seems questionable — could such a simple circuit really decode a stereo signal in which the subcarrier is modulated with L−R?

Doing some numerical experiments, it seems that that positive and negative rectification of a composite signal L + R + subcarrier DSB-modulated with L-R, can separate L and R. The separation was imperfect, though, and I did not get reasonable separation at all with DSB-RC. I also don't know if the experiment generalizes to arbitrary L and R signals, since the rectification is a non-linear operation, so additivity is not granted.

The asymmetric AM-modulation scheme described in the paper generates a composite signal containing a L+R component, the subcarrier DSB-modulated with L-R, and sidebands of subcarrier harmonics (but no subcarrier harmonics themselves). This signal can, of course, be separated to L and R by positive and negative rectification (and subsequent lowpass filtering), but due to the presence of sidebands of subcarrier harmormics, it has a much higher occupied bandwidth. After removing the sidebands of subcarrier harmormics with a lowpass, we would basically end up with a compositive signal L + R + subcarrier DSB-modulated with L-R (but not DSB-RC).


EDIT: Here's a simulation of the asymmetric AM with random L and R random signals, BW-limit to eliminate sidebands of sub-carrier harmonics, and demodulation by positive and negative rectification of the composite signal + lowpass (=> "ideal diode detector", but a much better lowpass). The demodulated signals are decomposed into L, R, DC components and residuals to assess the goodness of separation. With the random L and R the separation is not that bad. Std(resid) could be lower, I don't know yet where exactly it comes from.

Code: [Select]


pkg load signal

fs = 1e6
fsub = 31250
audiobw = 15000
SKIP = 20000
N = 100000 + SKIP

% generate bandwith-limited random L and R signals

L = 2 * rand(1,N) - 1;
R = 2 * rand(1,N) - 1;

[n,Wn,beta,ftype] = kaiserord([0.95*audiobw audiobw],[1 0],[0.0001 0.0001],fs);
hnoise = fir1(n,Wn,ftype,kaiser(n+1,beta),"noscale");

L = filter(hnoise,1,L);
R = filter(hnoise,1,R);
L /= max(abs(L));
R /= max(abs(R));

% L = cos(2*pi*2000*[0:N-1]/fs);
% R = cos(2*pi*3000*[0:N-1]/fs);

% subcarrier
subc = cos(2*pi*fsub*[0:N-1]/fs);

% asymmetric AM
composite = max(0,subc) .* (1 + L) + min(0,subc) .* (1 - R);

% limit BW of composite signal
[n,Wn,beta,ftype] = kaiserord([fsub+audiobw 2*fsub-audiobw],[1 0],[0.0001 0.0001],fs);
hcomp = fir1(n,Wn,ftype,kaiser(n+1,beta),"noscale");
composite = filter(hcomp,1,composite);

% lowpass after rectifier
[n,Wn,beta,ftype] = kaiserord([audiobw fsub-audiobw],[1 0],[0.0001 0.0001],fs);
hbb = fir1(n,Wn,ftype,kaiser(n+1,beta),"noscale");

LR1 = [L; R; ones(1,N)];

% compensate filter delays
delay = (length(hcomp)-1)/2 + (length(hbb)-1)/2

printf("Factor analysis of lowpass filtered composite signal:\n");

x = filter(hbb,1,composite);
lr1 = x(SKIP+1:end) * pinv(circshift(LR1,delay,2)(:,SKIP+1:end));
resid = x(SKIP+1:end) - lr1*circshift(LR1,delay,2)(:,SKIP+1:end);
printf("  mono: %9.6f * L + %9.6f * R + %9.6f + resid, std(resid) = %.6f\n", lr1, std(resid));

printf("Factor analysis of rectified + lowpass filtered composite signal:\n");

x = filter(hbb,1,max(0,composite));
lr1 = x(SKIP+1:end) * pinv(circshift(LR1,delay,2)(:,SKIP+1:end));
resid = x(SKIP+1:end) - lr1*circshift(LR1,delay,2)(:,SKIP+1:end);
printf("  positive: %9.6f * L + %9.6f * R + %9.6f + resid, std(resid) = %.6f\n", lr1, std(resid));

x = filter(hbb,1,min(0,composite));
lr1 = x(SKIP+1:end) * pinv(circshift(LR1,delay,2)(:,SKIP+1:end));
resid = x(SKIP+1:end) - lr1*circshift(LR1,delay,2)(:,SKIP+1:end);
printf("  negative: %9.6f * L + %9.6f * R + %9.6f + resid, std(resid) = %.6f\n", lr1, std(resid));


Example output:

Code: [Select]

fs =  1000000
fsub =  31250
audiobw =  15000
delay =  4012
Factor analysis of lowpass filtered composite signal:
  mono:  0.317287 * L +  0.317287 * R + -0.000000 + resid, std(resid) = 0.000005
Factor analysis of rectified + lowpass filtered composite signal:
  positive:  0.316714 * L +  0.000577 * R +  0.320410 + resid, std(resid) = 0.002654
  negative:  0.000573 * L +  0.316710 * R + -0.320410 + resid, std(resid) = 0.002654

EDIT: My first conclusion:

• The described modulation scheme generates directly a composite signal (including the mono component), not just the modulated subcarrier.
  
• The unfiltered composite signal can be demodulated virtually exactly with two ideal diodes + LPF, but is has too much bandwidth
  
• After limiting the bandwith to 46.25 kHz, demodulation with two ideal diodes is still possible, but it is no longer perfect:
  a) separation between L and R is no longer perfect, but still not too bad.
  b) Besides L, R and DC components, the 0-15kHz spectrum contains some residual components(noise/distortion). Still not sure where it comes from; I guess IMD introduced by the rectification. The residual component is larger when L and R are two sine waves, and smaller when L and R are wideband signals.
  
• In the tube era, these flaws were likely considered acceptable.
  
• While it seems to be a working scheme (with some restrictions), I'm no sure if this is really OIRT. The article does not claim that either; it just talks about "Soviet Stereo". The modulated subcarrier amplitude is relatively high, and the subcarrier itself is not reduced and consumes almost 50% of the composite signal power. But exactly these amplitude ratios, as produced by the described modulator, seem to be necessary in order to make simple diode detection work.
  

[radiolistener]

After limiting the bandwith to 46.25 kHz, demodulation with two ideal diodes is still possible, but it is no longer perfect:
a) separation between L and R is no longer perfect, but still not too bad.
b) Besides L, R and DC components, the 0-15kHz spectrum contains some residual components(noise/distortion). Still not sure where it comes from; I guess IMD introduced by the rectification. The residual component is larger when L and R are two sine waves, and smaller when L and R are wideband signals.

Hm... so, if I understand correctly, it means that the stereo composite can indeed be demodulated using just two diodes, but the result will not be perfect in quality?

[gf]

Hm... so, if I understand correctly, it means that the stereo composite can indeed be demodulated using just two diodes, but the result will not be perfect in quality?

To be more specific: It seems that this particular kind of stereo composite signal can be demodulated/decoded by ideal half-wave rectification + LPF (although the result is not perfect). How well the circuit can do the same job is a different question. This would require a circuit simulation. In particular, I cannot imagine an RC filter with a narrow 15...16.25 kHz transition band and high stop band attenuation at 16.25. But if you have speakers that cannot reproduce the residual ultrasound components in the output, and ears that cannot hear them, you may simply not care...

Btw, it's interesting that the described asymmetric AM method (plus limiting the bandwidth to 31.25+15 kHz) happens to generate a composite signal equivalent to

    (L+R) / pi + subcarrier .* (1 + (L-R) / 2)

IOW, what we get is the mono signal (L+R)/2, scaled by factor 2/pi, plus the sub-carrier, DSB-modulated with (L-R)/2, 100% modulation. No reduced subcarrier. Although it seems to be a feasible method in principle, is this really OIRT Stereo? I don't think so.


EDIT:

... but the result will not be perfect in quality

Attached is an example of the demodulated R channel spectrum, when the original L and R are sine waves with 3kHz and 4kHz. SFDR is about 25dB, SINAD about 21dB. When L and R are wideband signals (random noise), I see a little bit better SINAD of roughly 29 dB.

EDIT: Added audio file of demodulated R channel, with 4 kHz tone.

[radiolistener]

Thanks to Njk, we obtained a test sample of an OIRT stereo composite signal, recorded with a Rigol oscilloscope connected at the test point between the FM demodulator output and the stereo decoder input of the Leningrad RP-015 receiver.

I applied a 50 kHz FIR low-pass filter for decimation and encoded the result into FLAC to reduce file size. The recording contains 24 seconds of music. The song is identified as "Prayer" by Tom Odell at 2:42: https://youtu.be/pTxeOOkTqYo?si=2aeP5-vgOQ81Fkii&t=162

The quality is fairly good, although some frequency distortions are audible. Applying a de-emphasis IIR filter significantly reduces them, but slight artifacts remain, most likely due to the limited linearity of the oscilloscope ADC. The 31.25 kHz pilot carrier is well visible.

The attached images show the spectrum of the original oscilloscope signal and the spectrum after applying the filter and decimation (file included).

Still not tried decoding it yet, only tried to apply a 50 µs IIR de-emphasis filter to the L+R component, which produced good results.

[Njk]

There can be some distortions because of multi-path reception. Very common phenomenon at my location. Tried to mitigate it by adjusting the antenna position but it's of dynamic nature and can happen at any time. BTW, another difference between OIRT and CCIR bands is that the former traditionally uses the horizontal polarization, so a receiving dipole is expected to be placed horizontally, like a TV antenna. While CCIR uses the vertical polarization. But for me, it makes no difference concerning the MPR. It's very annoying and happens with every receiver all the way