AD7980 Ref Input Driven by AD780

Hello,

Thank you for your post.

When assessing whether a given reference is suitable for hooking up directly to the AD7980 REF pin (or any other precision SAR ADC), one of the key metrics is the reference's load regulation (deviation in output voltage vs. load current). This article provides a great summary of when a reference buffer is recommended: https://www.analog.com/en/analog-dialogue/articles/precision-successive-approximation-adcs.html

For the AD780 specifically, the load regulation spec is 75uV/mA (for sourcing current). The AD7980 REF input pulls typically 330uA at 1 MSPS.

So for the case where VREF = 2.5V and the AD7980 is running at 1 MSPS, the output of the AD780 may deviate by as much as (0.33mA * 75uV/mA) ~ 25uV.

1 LSB of error for VREF = 2.5V for the AD7980 (16-bit, pseudo-diff ADC) would be roughly 38uV, so under typical conditions this error would be less than 25uV. However, the REF current spec provided for the AD7980 is only typical, so it is worth considering if that is enough margin to tolerate for your application.

Note that the REF current for the AD7980 typically scales linearly with sample rate. For example, at 1 MSPS the REF current is typically 330uA, and at 500 kSPS it's reasonable to assume it would be half of that (165 uA).

So the decision to implement a reference buffer with the AD780 + AD7980 will depend on how much error in VREF can be tolerated in order to achieve the gain error specifications of your end design, vs the sample rate the AD7980 will be operating at.

Just as a reminder, be sure to follow our layout guidelines for the REF decoupling cap as written in the Voltage Reference Input section of the AD7980 data sheet. The placement and size of the REF decoupling capacitor also has a direct impact on the accuracy of VREF over the course of AD7980 operation.

Thanks,

Tyler

Dear Tyler,

Thank you for your very informative reply. I went thru the article by Alan Walsh that you referred. It was also very helpful. A couple of points:

1) The figure#2 in the article shows that the dynamic REF current is much worse than the typical specification of 330uA. It shows that Iref could be as large as 2.5mA. In that case the error due to ref deviation would be much larger than one LSB. Is the typical spec an averaged value?

2) I decided to change AD780 with LT6658A-2.5. The reason why I selected AD780 in the first place was because we already had it available in our inventory. LT6658 seems to be much more superior (Load regulation: 0.25µV/mA). If I use LT6658, I won't be needing the ref buffer. Could you please confirm?

Thanks,
Adnan

Hi Adnan,

Glad to hear!

To answer your questions:

1) Yes the REF current specs in our SAR ADC data sheets generally are the average current at a given sample rate. As you noted, the instantaneous current of this pin can be quite a bit larger than that over very short time intervals. This is why we recommend a REF decoupling capacitor to be placed very close to the REF pin on the ADC, with minimal trace impedance between the capacitor and the REF pin; this decoupling capacitor acts as a "charge well" to supply the charge during the peaks in REF current without resulting in much of an instantaneous voltage drop. So generally the load regulation of the voltage reference/reference buffer isn't a factor for those current spikes.

2) It is very likely that the LT6658 voltage reference will be sufficient to drive the AD7980 REF pin directly (without a reference buffer), just based on the fact that it's load regulation spec is quite good, though I will point out that I am not aware that we have ever directly tested this product combination. As mentioned before, just be sure to include the reference decoupling capacitor as recommended in the AD7980 data sheet!

Let me know if you have any further questions!

-Tyler