diff --git a/src/freedv_api.c b/src/freedv_api.c index 5cedcf1a..b56d6a02 100644 --- a/src/freedv_api.c +++ b/src/freedv_api.c @@ -57,22 +57,28 @@ #include "ofdm_internal.h" #include "varicode.h" -#define VERSION \ - 15 /* The API version number. The first version \ - is 10. Increment if the API changes in a \ - way that would require changes by the API \ - user. */ +/* The API version number. The first version is 10. Increment if the API + changes in a way that would require changes by the API user. */ + +#define VERSION 15 + /* - * Version 10 Initial version August 2, 2015. - * Version 11 September 2015 - * Added: freedv_zero_total_bit_errors(), freedv_get_sync() - * Changed all input and output sample rates to 8000 sps. Rates - * for FREEDV_MODE_700 and 700B were 7500. Version 12 August 2018 Added OFDM - * configuration switch structure Version 13 November 2019 Removed 700 and - * 700B modes Version 14 May 2020 Number of returned speech samples can vary, - * use freedv_get_n_max_speech_samples() to allocate buffers. Version 15 - * December 2022 Removing rarely used DPSK support which is not needed given - * fast fading modes + Version 10 Initial version August 2, 2015. + + Version 11 September 2015 + Added: freedv_zero_total_bit_errors(), freedv_get_sync() + Changed all input and output sample rates to 8000 sps. Rates + for FREEDV_MODE_700 and 700B were 7500. + + Version 12 August 2018 Added OFDM configuration switch structure + + Version 13 November 2019 Removed 700 and 700B modes + + Version 14 May 2020 Number of returned speech samples can vary, use + freedv_get_n_max_speech_samples() to allocate buffers. + + Version 15 December 2022 Removing rarely used DPSK support which is not + needed given fast fading modes */ char *ofdm_statemode[] = {"search", "trial", "synced"}; @@ -877,13 +883,17 @@ static void codec2_decode_upacked(struct freedv *f, short speech_out[], with the input samples. We may squelch based on SNR. Need to handle various codecs, and varying number of codec frames per modem -frame Squelch audio if test frames are being sent Determine how many speech -samples to return, which will vary if in sync/out of sync Work with real and -complex inputs (complex wrapper) Attenuate audio on pass through Deal with 700D -first frame burble, and different sync states from OFDM modes like 700D Output -no samples if squelched, we assume it's OK for the audio sink to run dry A FIFO -required on output to smooth sample flow to audio sink Don't decode when we are -sending test frames + frame. + Squelch audio if test frames are being sent. + Determine how many speech samples to return, which will vary if in + sync/out of sync Work with real and complex inputs (complex wrapper) + Attenuate audio on pass through. + Deal with 700D first frame burble, and different sync states from OFDM + modes like 700D. + Output no samples if squelched, we assume it's OK for the audio sink to run + dry. + A FIFO is required on output to smooth sample flow to audio sink. + Don't decode when we are sending test frames \*---------------------------------------------------------------------------*/