Our company had been using AD7656BSTZ-1 for years and there was not any problems until June 2019 when the first popping problem happened, what’s worse, another three cases came out one by one from July 2019 to January 2020, and the popping problem could happens even without any input of any sampling channel. Figure1 is a burst one after a popping with smoke, the four burst chips’ SN are #1524, #1524, 1504, #1428.
We have listed 4 clues as below after schematic checking, but we are not very sure whether or not we have exactly located the root causes, so we’ll deeply appreciate it if we could have ADI’s professional confirmations and explanations.
P.S. Figure3 was plotted according to Figure6 and Figure7, for limited probes to capture all the wave forms in one time.
P.S. Figure4 was plotted according to Figure8 and Figure9, for limited probes to capture all the wave forms in one time.
I think you have clearly identified the change in high voltage supply as the likely candidate for the sensitivity in your new system version (RACK V2). Let me spend some time to think about what might be causing the current flow that is leading to the large differential between AGND and DGND and see if I can't make a recommendation as to how you might protect the device permanently, either through sequence modification or potentially additional protection circuits.
Firstly, thank you very very much for your reply.
Secondly, I'm sorry about hadn't uploaded the GND circuits timely along with the problem descriptions, actually the AGND and DGND are connected with four parallel 10uH inductors, as shown in the appending figure1, by the way, is it risky with an overshoot of 0.3V, even for a transient one ?
Thirdly, one of my friends recently told me that there was a power up sequence recommendation from an ADI engineer's email, and it was recommended as VDD/VSS(±15V)->>AVCC/DVCC(+5VA/+5VD)-->VDRIVE(+3.3V), but there is not a description in the user manuals about this kind of recommendation, so, is it just a recommendation, or a must?
In conclusion, we are looking forward to your professional suggestions and recommendations about the proper using of AD7656, as we've been stuck here for quite a few days because we are not sure about the root cause of the popping problem.
I should add some more details, there was no obvious phenomena except the area near Pin30/Pin31/Pin32 became a hole after the popping, actually, Pin30(VSS, -15V) and Pin31(VDD, +15V) and Pin32(AGND) these three pins are totaly burned broken, as shown in figure3.
Two Schotky diodes are recommended at the VDD and VSS pins in Data Sheet Rev. D. as shown in Figure26, while there was no diodes in our design. As the data sheet described, the two diodes are designed to provide protection for the analog inputs, because diodes will become forward-biased to conduct the current into the substrate when the analog inputs exceeds the VDD and VSS more than 0.3V. But there is no chances for inputs exceeds VDD or VSS in our products because we hadn't load any inputs yet when or before the popping happened.
I'm suspicious that the fact that you only have an AC path between grounds could be the source of your problem. Whatever has changed dynamically by swapping out the PSU from version 1 to version 2 is creating transient current flow that is causing your AGND to DGND voltage to exceed the AMVR by nearly 2V! Under these conditions there is no doubt that there will likely be damage as this is classified as an electrical overstress event.
Given the location of the damage and the likely path through AGND I'm suspicious that the source of the current is through the VDD and or VSS pin(s) to ground and thus the localized damage in that area. If you return one of the devices through our RMA process we can confirm the damage using our internal failure analysis team.
In the mean time I would suggest trying to add a DC path between AGND and DGND by replacing one of the inductors with a small dc resistance (say 10 or 20 ohms) of a suitable power rating to see if you can't get rid of the fast transient through the grounds at start-up.
I agree with you about the advice to add DC paths between AGNDs and DGNDs instead of only one AC path, and I will verify it soon. But it seems unlikely well explained the fact that all the other AGND pins and all DGND pins were OK, but only the AGND at pin32 together with VDD(pin31) and VSS(pin30) were burned broken.
By the way, AGND-DGND exceeds the ABSOLUTE MAXIMUM RATINGs about 400mV, as shown in Figure1 below, not 2V.
As you said RMA proces needs returning a device, so I need to know what exactly the device refers to, the damaged AD7656 chips or our PCB, or both the damaged chips and PCB?
I have another consideration about the ABSOLUTE MAXIMUM RATINGs, as shown in Figure2 below. If it is right to understand the "VDD to AVCC" as "AVCC-0.3 < VDD to AVCC < 16.5V", then we can name the time that doen't meet the requirement as "bad time", so, Rack V2 obviously has a longer bad time because the +15V's rising time lasts too much longer compared with Rack V1, as shown in Figure3 and Figure4, and the longer "bad time" means more heat accumulated and easier to burnout, that's why Rack V2 sometimes comes up popping, Rack V1 doesn not. What do you think of this consideration, is it reasonable?