Make AudioOutputSPDIF3 sync to AudioInputSPDIF3 to avoid glitches #446
Conversation
AudioOutputSPDIF3 can use the Teensy audio clock at 44.1kHz, OR, if an AudioInputSPDIF3 object (which derives its clock from the incoming audio stream) is instantiated, it can and must be synchronised to that. This approach allows a glitch-free system using SPDIF I/O only. The design GUI has been modified to allow this combination to pass conflict checks; a better implementation may be required.
…0u5e/Audio into feature/sync-SPDIF-IO2
|
Doesn't do what I thought! Looking into it... |
Was just passing audio through, which was not what was wanted at all! This update feeds the Rx clock back through SAI1 MCLK[3], which seems to be the valid way to do it; SAI2 or SAI3 would also work. As yet untested is the concept that this will sync the entire audio system to the S/PDIF Rx clock, so it becomes possible to use the audio shield too. More clock trickery is probably needed...
Also copy across code which removes need for a dummy I²S object
I²S on SAI1 will now work with S/PDIF input as master clock
…0u5e/Audio into feature/sync-SPDIF-IO2
…0u5e/Audio into feature/sync-SPDIF-IO2
...all unsuccessful
Add offset to clock used by output_spdif3.cpp, which changes the interference rate It looks like it's slipping 1 sample every so often, so if the SPDIF and PLL4 clocks are e.g. 77Hz different then clicks occur at 77Hz But why?!
Note the comment above the assignment of fs in AudioOutputSPDIF3::config_spdif3() (lines 231-236). For some reason the internal (PLL4-derived) clock has an audible effect, even if the system sample rate is supposedly being derived from the S/PDIF input. The effect seems fairly minor, with the occasional missing or extra samples rather than huge glitches, but it's not hi-fi, and I can't figure out the root cause!
|
OK, so this now achieves its original aim of syncing AudioOutputSPDIF3 to the AudioInputSPDIF3 sample rate, avoiding the overhead of resampling. It also adds clocking of I2S1 / SAI1 from the S/PDIF input sample rate, tested at 32, 44.1, 44.117, 48 and 96kHz - the update rate changes to match, so anything frequency-related will be more-or-less wrong, depending on how different the S/PDIF rate is from AUDIO_SAMPLE_RATE_EXACT. However, for some reason there's an extra or dropped sample on I2S every so often, despite the shared clock source. I can't find this to fix it... I've toggled an output pin in both the I2S and SPDIF ISRs, and they're definitely locked together. On this basis, I've updated the Design Tool to permit use of I2S with AudioInputSPDIF3 without flagging incompatibility, and put some comments in the info pane relating to "minor degradation" of the I2S audio quality. |
Although AudioInputSPDIF3 is rightly documented as being incompatible with most other audio I/O, it's possible to use its sample clock to drive the AudioOutputSPDIF3 object, making a workable S/PDIF pass-through possible without the need to resample,
In addition to the required code changes, the design GUI has been updated so it doesn't show a conflict between the two SPDIF3 objects, and the info pane text has been edited to give a little more information on usable combinations of I/O objects with these two.