It depends how the host application (Audacity is sending the request to the sampling) . First the device needs to enuerate and show that it can support multiple sampling, which you can do using the descriptors as mentioned in the UAC spec. Next, the host application can select one of the supported sampling rate, else you can provide a default selection. Now whatever the host application select, needs to be conveyed to the device (please check if Audacity supports this and how does it inform the USB device). It can send some setup/class specific request to the device. In the device firmware, you need to check for this event and accordingly create DMA channel/buffers. Please make sure about the command send from the Audacity on changing/selecting one of the sampling rate.