Does anyone know whether there is a specification for the required accuracy of the 19 kHz stereo pilot tone frequency?

I was trying to figure out whether it is possible to use the same crystal to provide a reference for a PLL and for a stereo-coder. Using an 8 MHz crystal for example, it's possible to get 38095 / 19047 kHz tones (divide by 210 then by 2) which is around 0.25% (or 2500 ppm) off frequency.

I guess I could try using this and see whether most receivers will work OK, however I wondered if anyone had seen a standard which specifies what is actually/theoretically needed.

According to "Technical Standards and Requirements for FM Broadcasting Transmitters" from 2005:

A.2.2.2.1 The frequency of the pilot signal shall be 19,000 ±2 Hz.

The spec. is +/- 2Hz, but many commercial coders don't manage this. One Bext coder I have here at the moment is 17 Hz high. Practically, it doesn't seem to matter much, but if the spec says 2Hz.....

I've used a 10 MHz reference (with GPS steering from a "Navman" satellite receiver) with a couple of PLLs to give 19 kHz and 57 kHz at very high accuracy. It's also easy enough to divide the 10 MHz to a sensible reference frequency for a transmitter PLL, so yes - it's possible and is done in some commercial gear.

My other favourite frequency accurate off-air source (198kHz longwave) is being closed down, so I'm going to have to rebuild some of my gear to use GPS receiver modules.

Albert H wrote: ↑Tue Feb 04, 2025 4:05 am My other favourite frequency accurate off-air source (198kHz longwave) is being closed down, so I'm going to have to rebuild some of my gear to use GPS receiver modules.

Or make a low power 198khz rig and GPS lock that

Thank you @1608cc and @Albert H - 2 Hz is more of a challenge without adding the complexity of an additional PLL. Time for some creative thinking on division ratios to try and get a suitable reference frequency and the necessary 19 & 38 kHz tones.

It turns out it's quite easy... With a 4864 kHz crystal, you can divide by 4864 (4096 + 512 + 256) (just 3 bits on a counter) to get a 1 kHz reference. With a divide-by-100 pre-scalar you can then operate the PLL for the FM synthesiser. Equally, you can divide by 128 to get 38 kHz and 256 to get 19 kHz. Luckily I have a bag of 4864 kHz crystals so can give it a try. Divide by 100 is more complicated but something like an MC12080 (div-by-10) followed by any-old-CMOS divider would do the job. I still have a couple of SP8629's kicking around but not enough to be useful.

Consider also the 8,664MHz, 4,332MHz and 456kHz resonators, which are quite easy to get online. One guy from polish forum used 11,0592MHz with 8bit DAC that gives - as he said - 1,13ppm off from ideal.

Years ago, I made a run of stereo coders (mixer type with MC1496 for the DSSC modulator) using 455kHz IF filters for the reference. I found that if they were lightly loaded, it was quite easy to get a nice, stable 456kHz out of them.

A couple of CMOS divider ICs later, I had 19 and 38 kHz. I divided first by 6 (with a 4017) to give 76 kHz, but with an assymmetrical squarewave - almost a pulsewave. This fed both halves of a 4013 to give 50 : 50 M/S 38 kHz and 19 kHz. I used simple LC filters to make the squarewaves into pretty good sinewaves, and found that I had to tweak the inductor cores to get the 19 and 38 kHz in phase.

The downside of this type of stereo coder was that the 1496 couldn't handle "line-level" signals, so I had to attenuate the difference signal to prevent distortion in the analogue mixer, and then amplify it back up to match the sum signal. This, of course, gave a noise penalty - the "Side" (modulated difference) component was about 16dB noisier than the "Mono" path, leading to an overal degradation in the noise performance. I used the quietest op-amps I could find (NE5532 or LM833), but still wasn't entirely happy with the result.

Soon afterwards, I moved to oversampled, switching coders, and got much better noise performance. I still used the 455 kHz resonators or filters for the reference, though.

i’ve missed this talk of resonators, sounds abit like something mechanical I guess the 455 ones are what you find in old am/fm radios, you could make your own version of a clock opera.