• State Resolved
  • Locked Locked
  • Replies 3 replies
  • Subscribers 46 subscribers
  • Views 578 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

ADS1610 accuracy and output format

Rick
Prodigy 20 points
Other Parts Discussed in Thread: ADS1610, DAC8820, ADS1605, AMC1200

Hi All,

We are planning to use an ADS1610 16-bit 10MSPS ADC converter in an application where it is connected back-to-back to a DAC8820, such that we can use a digital interface to galvanically isolate the two circuits.

However, after doing some preliminairy research, I found that the datasheet explicitly mentions a constant gain-error of 3.8%, with a variable error of +-0.3%FS, whereas similar parts (e.g. ~ADS1605) and competitive parts (ADI's AD7626 for instance) seem much more accurate.
Is there any particular reason why the part exhibits such a large constant error? Shouldn't this kind of constant errors be trimmed or removed in the factory already? And is there any way to minimize the effect of this constant error in the transfer from ADC to DAC? For instance, would it be an option to adjust the gain of the preceding ADC-driver stage in order to get the constant gain error to 0?

Another question I have relates to the output format, which is in 2's complement. Since the ADC is used in back-to-back configuration with the DAC8820, the output format needs to be unsigned, i.e. the output code should increase linearly between 0x0000 and 0xFFFF for -Vref to +Vref. Is there any common solution for changing this output format? I was thinking about inverting the sign (MSB) bit, but i'm not sure whether this results in the correct data format and does not create additional timing problems with one of the data pins having an inverter in the path. 

Thanks,

Rick 

  • 0
    TI__Mastermind 23235 points

    The increased gain error is due to the design of the part to account for the higher speed of operation.  Various design decisions (architectures, layout, etc) had to be adjusted versus other designs.  Additionally, this device was targeted more at AC applications, so the DC performance was not as critical as for other designs.

    Regarding two's compliment, there is no good way other than the standard method that inverts all the bits and adds 1 (if the MSB = 1); this gets you the absolute value of the negative value.  If you don't need the entire range of the converter, you can offset the input signal such that it only uses the upper half of the range (only the positive values) and design the rest of your circuit to compensate/remove the DC offset.

  • 0 in reply to Greg Hupp
    TI__Mastermind 23235 points

    One additional comment, have you looked a devices like the AMC1200?  This part has isolation built in and is analog in and out.  It might be an simpler solution than you are using currently.  Just another option

  • 0 in reply to Greg Hupp
    Prodigy 20 points

    Thanks Gregg.

    Actually I had a look at this kind of devices, and although they offer a substantial isolation strength, it is still not enough for our application in which we need to isolate from voltages up to 15kV. Moreover, the bandwidth we are looking for is more towards 1 MHz.

    So I think we'll just try to integrate a small size FPGA or microcontroller on the board, just to be able to correct for any gain- and offset errors, and to get the output data in the desired format.