Different Vout from AD5721R in different packages?

We have a design using the AD5721R DAC for controlling a VGA.

In our former designs, we used the 16-Lead TSSOP package, and experienced that we had to multiply the Vout-value-formula (Datasheet Rev-C, page 25) by 4 to achieve the correct Vout voltage. We use Output range = 0-5V setting M=2 and C=0, N=12.

In our latest revision, we needed to shrink the PCB and chose the CP-16-22 package (Marking is DHN). Now the multiply-by-4 is not necessary (!).

We wonder why??

Has the datasheet change from Rev-B to Rev-C regarding Changes to Transfer Function Section anything to do with this(???):
REVISION HISTORY 1/2018—Rev. B to Rev. C
Changes to Transfer Function Section ......................................... 25
.
.
Please advise.

Tormod, Volue IIoT.

  • 0
    •  Analog Employees 
    •  Super User 
    on Oct 11, 2021 2:08 AM

    Hi, 

    I'm not sure where you got the multiply by 4. Can you give a little more info on how you came up with this? 

    The transfer function change is just to make the formula compatible with both the 12-bit and 16-bit versions. There shouldn't be any difference in the output voltage of different packages. The correct way is to follow the transfer function on the d/s depending on your output range settings. 

    Best regards,

    Ian

  • Hi Ian.

    Sorry about this late reply, I was out of office last week.

    The 'multiply by 4' was found necessary to get the correct Vout using the formula in the datasheet. It was revealed by measuring the Vout voltage while stepping the Vout-register corresponding to 1dB. Unfortunately we did not file this as a possible component error...

    In our new, small design we run the exact same SPI-commands to the 16-Lead LFCSP package, and now the given formula gives the correct Vout-stepping, without the 'multiply by 4' added in our program.

    It is a mystery to us.

    Our first tests with the AD5721R BRUZ was back in Q4 2018. So it is possible that the 16-Lead TSSOP packages is older than that. I can find no date-code on the package. Are there any errata descriptions for old batches?

    Regards Tormod, Volue IIoT.

  • 0
    •  Analog Employees 
    •  Super User 
    on Oct 18, 2021 1:30 PM in reply to Tormod

    Hi Tormod, 

    Are there any errata descriptions for old batches?

    The typo changed from that revision came from the denominator on the transfer function which was set to 65535. It was changed to 2^N to accommodate for the 12-bit part. 

    It should be fine since it's not working as expected on your redesign using the LFCSP.

    What I'm curious about is the TSSOP on your former design. Is the "multiply by 4" still in effect on your old board with the TSSOP part? 

    Could you do a check on your previous code? To output a full-scale of 5V, the input command set should be 0xFFF. If multiply by 4 is necessary, then is the input code you set 0x3FFF?

    Best regards,

    Ian