LTC2157 Alternate Bit Polarity

Hello,

Ignoring the randomizer for a moment,  ABP mode inverts all the odd bits of which DB13 (or MSB) is one of those bits that get inverted.   It appears that  you state that this is indeed occuring in the transmitted digital output.   On the receive capture side, one would need to invert all the odd bits again (including MSB) to restore the correct digital value representing the sampled waveform.  So it is unclear what you mean by "we have to invert the bit that is already being inverted" based on previous statement.

Both ABP and randomizer have the same objective in trying to decorrelate the digital data pattern (during transmission) so that any digital noise that is inadvertently fed back to the ADC input appears more noise like.   Considering the 14-bit ADC's SNR of 70 dBFS (vs theoretical 86 dBFS), it would seem that the randomizer itself would be more than sufficient (vs both features enabled) since the LSB will be toggling quite frequently being dominated by noise floor of the ADC.

Regards

Here is a screen shot from your software to show you want I mean:  

Hello,

Referring to table below.............what digital output mode or combination of modes appear not to be working with regard to ABP, RAND, and TWOSCOMP?  For instance........are you suggesting that when ABP, RAND are enabled with Offset Binary output that the output appears to be 2's complement?  If true...........does this hold if only ABP or RAND is selected with Offset Binary?

We are in (or at least we think we are in) 2s compliment mode and in ABP.  If we are only using randomizer the output is as expected. 

We are in (or at least we think we are in) 2s compliment mode and in ABP.  If we are only using randomizer the output is as expected. 

With 2's complement and ABP, I believe a double inversion of the MSB takes place relative to offset binary since MSB is inverted to generate 2's complement and ABP inverts the odd  DB13 (MSB) resulting in additional inversion.  

That said...................if randomizer is working as expected and has similar goal of spreading digital pattern noise, I would suggest using randomizer only.   I do not work for ADI so I do not have ability to debug actual silicon to determine if flaw exists in the IC itself or the FPGA/software that should be reversing the decoding that is used on the IC.

Interesting.  I've seen articles where using both ABP and randomizer yield the best results, especially with low level signals.

Figure 5. Noise floor with alternate bit polarity and data output randomizer enabled.

I would like to know if there is a problem with the IC or something just no documented in the datasheet. 

Hello,

Someone from the ADI factory would need to comment if issue exists on IC or documentation issue.

Assuming that one is able to enable ABP and Randomizer independently and simultaneously..................one can do an experiment to determine what configuration provides the best SFDR performance for a amplitude sweep of a single tone sine wave from -1 dBFS down to -70 dBFS.