CCES SS_App with SigmaStudio vs. uController with Exported files, behaviour and properties

I have with great help from the ezone support made a microcontroller application running at 192khz with USB/I2S input and external DAC I2S output. The application is a crossover filter for active speakers.

We use a ADSP-SC589_EZ-Board with a BRKOUT extension board, 4 channel external I2S DAC's for output, and 6 channels for input, analog, SPDIF and I2S from a USB interface.

We now observe that there is a "clear" difference in performance / audial quality when the SigmaStudio project is run with the default SS_App with the adapters ICE1000 and SigmaStudio interface connected, versus the autonomous booted/flashed microcontroller application with the system files exported from SigmaStudio and all cables disconnected.

I realise that this is a very elusive problem, difficult to quantify. We cannot easily see any measureable differences such as distortion or frequency response, but we all agree that the audial impression is that the SS_App with SigmaStudio sounds better than the booted/flashed microcontroller application. Better clarity, better "soundstage", less granular, more "life".

The default SS_App and the microcontroller application are definitely two different programs, that may handle the data and buffering differently.

The only programmed I/O we use in the microcontroller code is an interrupt controlled sensing of three GPIO pins to determine which input that should be used, analog, SPDIF or I2S. Volume control is handled by the DAC's so it is not in the code. 

In order to solve this, where should we look? What can we modify? What are the main differences between the two modes of operation?

Any qualified suggestion is greatly appreciated.