Dunno why but the frequencies are way off... The space freq should be 999 Hz and my RIGOL shows 775 Hz! Mark should be 2376 and it's 1852 Hz (so the same 1.29 factor). For now the timing is done with Arduino's delayMicroseconds(), that should have been precise to ± 5%... (I'm pausing for 20 µs between each DAC ouput update)
Anyway, I plan to use SAMD21 TC5 for timing, so I don't need to investigate any further this discrepancy. Stay tuned!
The illustrated 15ms interval is the maximum tolerable delay of desynchronization between voice and image, according to this article .
The 2 sine pulse train is generated on the DAC output of SAMD21 (with Direct Digital Synthesis rather than PWM) and is triggered by the appearance of a 1 PPS pulse from the GPS.
Note that the oscilloscope (low) sampling frequency is deliberately set to 50 kSa/s to get near to the audio recording frequency in cinema (48 kHz). Fitting of the sin pulse will be done on the desktop with available python libraries.
I've set up a gitlab repository for the firmware, ppssync, and cabled the prototype. Below is an artistically blurred picture of it:
The external antenna is for testing purpose only, the GPS is quite capable to fix indoor with its ceramic antenna. Here's the hookup guide between the Trinket M0 and the GPS board:
The trinket provides the 3.3 V power (from onboard regulator) to the GPS, greatly simplifying the set up! Thanks, Adafruit!
measure some input impedances (cam and audio recorder)
implement reference voltage for true audio AC with LM358
test mic level output
test line level output
test small speaker for acoustic coupling
Maybe I shouldn't reveal I've been researching this project for ten years! It would cast a shadow over my capacity to deliver... 8-) But you wouldn't have access to my treasure trove of related links...
The first traces of my journey are links kept on Delicious, now Pinboard: they are tagged smpte and date back to 2011. I later diversified my classification and introduced year after year more nuanced tags, here are direct links to them: (this is stratigraphic information paleontology!)
From now on, maybe the project will stall again... I'm back to school this January after a sabbatical semester (ducking down during COVID: I just can't stand doing zoom classes), we'll see.
Raymond Lutz
Dunno why but the frequencies are way off... The space freq should be 999 Hz and my RIGOL shows 775 Hz! Mark should be 2376 and it's 1852 Hz (so the same 1.29 factor). For now the timing is done with Arduino's delayMicroseconds(), that should have been precise to 
