Have a few TLC5947 off the shelf boards with the same circuitry as he Adafruit to kick off work with the TLC5947. These boards have a header on each side where all signal (Vdd, Gnd, Latch, clk) are connected together except for Data In/Out of-course, blank is pulled down with an R to ground. These board have an onboard regulator of 5V to the TLC5947 while the connected input voltage is directly supplied to the LED outputs. Interestingly, the boards work when supplied with 3.3 or 5V as well. An RPI 3 is used to control the board using bit-bang I/O and the boards are fed with an exterrnal power supply (78xx regulators) independently from the RPI, RPI and the boards share a common ground. The R is set for 15mA CC.
When the boards are in parallel (all boards get Data In from the same wire form RPI) this works all well and stable at 3.3, 5V and 9V input voltages. For all transactions, all lights all have the same color repeatedly tested in a loop for several hours for many different colors settings.
In cascading (serial) changing the DATA Out to from one to another board's DIN,it works fine and stable as well for hours at 3.3V. However changing the voltage to 5V or 9V to the board, all LED outputs immediately shut down (no LED signal). This happens already with 2 boards and currently using 3 LEDs per board. when removing the DIO to DIN wire, the first board continues to work normally. Swapping the boards has the same effect. Lowering down the clock speed to very slow frequency has no effect either.
It does not look like this is related parasitic capacitance . Somehow connecting anything to DIO causes this problem. Is there something in particular to consider when cascading ?