Why is the output voltage of the ADR3650A outside the expected range?

Hi Sebastien,

Thanks for reaching out. To clarify, are the parts that are sitting outside the specified range always the same parts, and are they always outside the range? Are they always >5.004V, or are you seeing units at <4.996V as well?

You also mentioned having difficulty reproducing this--are you saying these previously outside-of-range parts end up back in spec when they are powered-on again?

Hi, 

Yes, the components that do not meet the specifications are always the same (once identified). And when they fall outside the range (>5.004V), each one has its own specific value, for example: one is always at 5.021V, another at 5.014V, and so on. When these components are not faulty, they read 4.999V. No component has fallen below the range <4.996V. The problem always occurs above 5.004V.
The problem occurs occasionally after the power supply is switched on. When the problem occurs on certain ADR3650s, the Vref voltages remain stable at their incorrect value. If the power is switched off and on again, the Vref voltage may reset correctly (4.999V).
Please note that our product is equipped with 25 circuits containing an ADC and the ADR3650 Vref. Most of the time, the power supplies start up correctly. The problem occurs approximately once in ten cases and affects 10% of the ICs installed. (I can move the ADC circuits to my main board; the problem follows the ADC circuit and is not specific to the placement of the ADC circuit within the product).

Hi Sebastien,

Thanks for the additional info. Have you been able to reproduce this with a faster power supply start up? 700ms does seem slow. Are you using the fast-startup functionality on the LT3045? The PGFB pin can be configured with Rpg1/Rpg2 to greatly increase supply startup times.

Hello,
Yes, reducing the Vin rise time of the ARD3650 using an LT3045 was also one of the ideas I wanted to test.
Unfortunately, the LT3045’s “fast start-up” function is not implemented on my PCB. The PGFB pin is connected directly to Vout. So I carried out the test in stages:
1) I changed the LT3045’s Cset from 4.7uF to 330nF (without enabling fast start-up). The rise time went from around 620ms to 44ms. I still had the problem of an incorrect Vref voltage on the ADR3650.

2) I then modified my circuit to implement the LT3045’s fast start-up function. The rise time dropped to 12 ms. So far, I haven’t managed to reproduce the Vref issue, but I’m continuing to run tests to ensure the problem has indeed gone away. I’ll keep testing over the next few days and will keep you updated.

Do you have any maximum power supply rise time specifications for the ADR3650?

Thank you

Hi Sebastien,

We don't have specifications for the ADR3650 rise time, but generally 700ms is on the very long side of startup times for references.

Did you see the frequency of incorrect voltages change at all with the 44ms rise time? And given the fast-startup function has seemed to address the issue (let us know otherwise), what specifically did you change--just the two feedback resistors from PGFB to OUT?

Hi,
I can’t say whether, with a rise time reduced to 44 ms, the frequency of Vref failures is as high as it is at 620 ms. I haven’t carried out that test. I experienced it three or four times during the test, and even once is unacceptable in our application. However, with a rise time of 12 ms, I haven’t yet been able to reproduce the problem. It looks very promising so far...
And yes, I’ve only implemented the ‘RPG’ resistors between the Vout and PGFB pins of the LT3045 to enable the fast start-up function up to 90% of Vout.