I have a Texas Instruments DAC8568 in their BOOST breakout board package. The DAC8568 is an 8 channel, 16bit DAC with SPI interface. The BOOST package has headers to connect it to my raspberry pi, and it has LEDs connected to the output voltage so you can easily check to see if your code is doing what you think it does. Links to the BOOST package and datasheet of the DAC8568 are in my python code below.

I have the package wired to the raspberry Pi with the 3.3V supply, the 5V supply (needed for LEDs), and ground. The DACs SCLK goes to Pi SCLK, DAC /SYNC (which is really chip select) goes to Pi CE1, DAC /LDAC goes to Pi Gnd, and DAC MOSI goes to Pi MOSI. I do not wire the DACs /CLR, but I can physically hook it to ground to reset the chip if I need to.

I believe my wiring is good, because I can light the LEDs with either a python script or from the terminal using: sudo echo -ne "\xXX\xXX\xXX\xXX" > /dev/spidev0.1

I learned the terminal trick from this video: https://www.youtube.com/watch?v=iwzXh2V1SP4

My problem though is that the LEDs are not lighting as I would expect them to according to the data sheet. I should be lighting A, but instead I light B. I should light B but instead I light D, etc. I have tried to make sense of it all and can dim the LEDs and turn on new ones, but never in the way I would really expect it to work according to the datasheet.

Below is my python script. In the comments I mentioned where in the datasheet I am looking for the bits to send. I am very new to working with analog components and am not a EE, so maybe I am not doing the timing correctly, or making some other silly error. Perhaps someone can look at the datasheet and see my error without having to actually have the chip in hand. Thanks for the help!

# -*- coding: utf-8 -*-
"""
Created on Sat Jul  8 16:33:05 2017@author: pifor texas instruments BOOST DAC8568
for BOOST schematic showing LEDs http://www.ti.com/tool/boost-dac8568
for DAC8568 datasheet:  http://www.ti.com/product/dac8568
"""import spidev
import timespi = spidev.SpiDev()  #create spi object
spi.open(0,1)  #open spi port 0, device (CS) 1
#spi.bits_per_word = 8  does not seem to matter
#spi.max_speed_hz = 50000000  #does not seem to matter#you have to power the DAC, you can write to the buffer and later power on if you like
power_up = spi.xfer2([0x04, 0x00, 0x00, 0xFF]) #p.37 Table11 in datasheet: powers all DACSvoltage_write = spi.xfer2([0x00, 0x0F, 0xFF, 0xFF ]) #p.35 Table11 in datasheet supposed write A--but lights B
voltage_write = spi.xfer2([0x00, 0x1F, 0xFF, 0xFF ]) #supposed write B--but lights D
voltage_write = spi.xfer2([0x00, 0x2F, 0xFF, 0xFF ]) #supposed write C--but lights F
voltage_write = spi.xfer2([0x00, 0x3F, 0xFF, 0xFF ]) #supposed write D--but lights H
voltage_write = spi.xfer2([0x00, 0x4F, 0xFF, 0xFF ]) #supposed write E--but does nothingspi.close() 

Note for future readers, the power up needs to power on the internal reference which is

power_up = spi.xfer2([0x08, 0x00, 0x00, 0xFF]) #(p.37 datasheet

Comment: the bits are shifted. ... how ... compensate for the shift or eliminate the shift?

This could be the case, if the SPIDIV.mode is not in sync with the DAC.

DAC Datasheet Page 6/7:
This input is the frame synchronization signal for the input data.
When SYNC goes low, it enables the input shiftregister, and data are sampled on subsequent SYNC falling clock edges.
The DAC output updates following the 32nd clock.

Reference: Clock polarity and phase

According to the above and the Timing Diagram I come to the conclusion that SPDIV.mode == 2 is the right.

1. Check the actual SPDIV.mode
2. Change to SPDIV.mode = 2

I can confirm your used Values by reading Table 11 Page 35.
Write to Input Register - DAC Channel X
My Example set Feature Bits = 0

            3         2         110987654321098765432109876543210RXXXCCCCAAAADDDDDDDDDDDDDDDDFFFF
A = 32-bit[00000000000011111111111111110000]:0xffff0 ('0x00', '0x0f', '0xff', '0xf0')3         2         110987654321098765432109876543210RXXXCCCCAAAADDDDDDDDDDDDDDDDFFFF
B = 32-bit[00000000000111111111111111110000]:0x1ffff0 ('0x00', '0x1f', '0xff', '0xf0')

Page 33: DB31(MSB) is the first bit that is loaded into the DAC shift register and must be always set to '0'.

The wireing seems straight forward and simple, but worth to doublecheck.

Code Snippet from testing:

def writeDAC(command, address, data, feature=0x0):address = ord(address) - ord('A')b1 = commandb2 = address << 4 | data >> 12          # 4 address Bits and 4 MSB data Bitsb3 = data >> 4                          # middle 8 Bits of datab4 = 0xF0 & (data << 4) >> 8 | feature  # 4 data Bits and feature Bitsvoltage_write = spi.xfer2([b1, b2, b3, b4])# Usage:
# Write Command=0 Channel=B Data=0xFFFF Default Features=0x0
writeDAC(0, 'B', 0xFFFF)