Strange conversion results from AD7609


I'm working on a PCB using a AD7609. The read out of the ADC is done via a AT90CAN128 8-bit AVR using the parallel interface of the AD7609.

For a first test all inputs where connected to AGND. I've collected 969 samples with 1 sample/s and plotted them in a histogram.

The x-axis represents the conversion result (using the two's complement for interpretation of the 18 bit), y-axis shows the occurrence of each conversion result.

This picture shows the histogram for channel 2. For the other 6 channels it looks the same.

One can clearly see 7 discrete peaks. There is also an 8th peak at 0x20000 (-131072) which is not visible due to the y-axis.

After power up there is a 1 us long reset pulse applied to the ADC.

I'm using the same schematic as figure 41 on page 24 of the datasheet except of the pull-down/pull-up resistors at RANGE, STBY, ...

Are these pull-up/down resistors mandatory? Could the absence of these resistors damage the ADC resulting in such strange conversions?'

Best regards,