2 questions about duplex devices and automatic sample rate conversion

[aubrey-nureva]

Hi, I'm trying to build a better understanding about what's going on with my duplex device after noticing this line in the example for duplex devices:

For this reason, miniaudio forces you to use the same sample rate for both capture and playback. If internally the native sample rates differ, miniaudio will perform the sample rate conversion for you automatically.

So for example, if I have a mic with a sample rate of 16kHz and a playback device with a sample rate 48kHz paired as my duplex device, I have two questions:

1. Am I correct in thinking that the input is in this case "upsampled" to match the playback device's 48kHz? Or is something else done for this conversion?

2. Has this conversion already completed by the time we enter the callback? As in the example, I can see we check the format and number of channels here, but not the sample rate... Does this mean that pInput has already been resampled by this point?

void data_callback(ma_device* pDevice, void* pOutput, const void* pInput, ma_uint32 frameCount)
{
    MA_ASSERT(pDevice->capture.format == pDevice->playback.format);
    MA_ASSERT(pDevice->capture.channels == pDevice->playback.channels);

    /* In this example the format and channel count are the same for both input and output which means we can just memcpy(). */
    MA_COPY_MEMORY(pOutput, pInput, frameCount * ma_get_bytes_per_frame(pDevice->capture.format, pDevice->capture.channels));
}

Thank you in advance!

[mackron]

1. When you specify an explicit sample rate in the device config, that'll be the rate that's used and both the capture and playback side's will be resampled to that if necessary. Otherwise, if you leave the sample rate set to 0, it'll use the capture device's native sample rate, meaning the playback device will be resampled to match the capture side's sample rate.

2. Yes, if resampling is required on the capture side, pInput will already have been resampled by the time it hits the callback. You can assume pOutput is the same sample rate. If the playback side requires resampling, it'll be done internally by miniaudio after the callback returns.

The data flows like this:

capture > resample > callback > resample > playback

Of course, the resampling step only happens if it's necessary.