We currently have an application where five TLC5940 chips are cascaded for a total of 80 channels of control. The chips are driven by a Teensy 3.1 microcontroller using hardware SPI. A series of 160 distinct patterns are being displayed in rapid succession; we are aiming for a refresh rate of 4800 Hz. The accuracy of the patterns reproduced is crucial to our application and unfortunately, we already noise and artifacts in the patterns with a refresh rate of around 2400 Hz. In fact after continued testing, there seems to be a direct correlation between the amount of inaccuracies in the patterns and the refresh rate above 1200 Hz. Below 1200 Hz, the patterns are reproduced perfectly. It seems as if the distortions of the patterns are happening in a repeating yet not systematic way. Using the minimum target refresh rate, a SCLK frequency of around 9 MHz is needed, which should be under the 30 MHz max as specified by the datasheet. Currently the circuit is built on protoboards: one for the LEDs and drivers and one for the microcontroller. Between the two is a run of five wires about 1.5 feet long, which are the wires needed for serial data transfer. All five wires are twisted into a braid.
Where could the source of inaccuracies be? The noise introduced in the cable run between the two protoboards? Or have we reached the limitation of the TLC5940?
Thank you for your time!