ADS1174: Lockup

Hello Gene,

What you are describing is likely a latch-up issue due to some type of Electrical Over Stress Event, which exceeds one or more of the ABS MAX values in the datasheet.  The 100mA momentary current can occur when one of the inputs exceeds the ABS Max voltage, causing an input current to flow through the internal ESD protection diodes.  When the current exceeds 100mA, this can cause a parasitic SCR to turn-on.

The ABS MAX specs can be exceeded due to numerous reasons.

1. An external input amplifier powered from higher voltage rails over-ranges, which exceeds the input voltage ratings of the ADC.

2. The Analog and Digital ground pins see a voltage transient difference of more than +/-0.3V.  If you use separate analog and digital ground planes on your board, any transient currents flowing can cause a potential voltage to develop, which can easily exceed the +/-0.3V rating and cause the device to latch-up.  This is the most common cause of a latch-up in the ADS1174.  We typically recommend to tie all ground pins on the ADS1174 to the same ground plane directly beneath the device to improve performance.

3. Transients on one or more of the power supply rails.

If you can share the schematic showing the ADS1174, along with the reference and input amplifiers, I will take a closer look.

Also, take a look at this e2e post regarding grounding of the device.

https://e2e.ti.com/support/data-converters/f/73/p/641966/2545203

Regards,
Keith Nicholas
Precision ADC Applications

I can share the schematic, but only in private.

This board designed by another engineer who left the company - so I have to be the janitor and clean things up :D

Yes, the ground planes are separated. They are on the same layer, but do not connect at the chip itself. I haven't seen any long-term potential differences at the ground pins but it might be a transient thing.

I did read through that other post. Seems this part does have some known problems with noise. At the moment, I'm probing around to see if the part is exceeding any max voltages.

How important is the power sequencing? This board has some voltage regulator, and some comparators that sequence things - I wonder if noise on one of the comparators momentarily shuts down a power rail to the ADC chip, which would then have the digital signals exceeding the power rails.

If power sequencing is not as important as the data sheet implies, then I can eliminate this circuitry.

Hi Gene,

The power sequencing is important to properly reset the device.  However, after the supplies have stabilized and data first appear, you can pulse the SYNC pin low to force a reset.  In this case, you would not need to observe the power up recommendations in the datasheet.

Please note that the above comments assumes the ABS MAX input voltages are not exceeded.  If one of the supplies shuts down and there is an input voltage still present, then this would exceed the ABS MAX specs.  When the ABS Max voltage is exceeded, there will be current flowing into the inputs through the input protection diodes.  If the current peaks above 100mA, then you could get a latch-up condition.

As long as the input voltages do not exceed the supply voltages by more than +/-300mV, then you could eliminate this circuitry and rely on the /SYNC pulse to properly reset the device after a power-up.

Since the grounds are separated, this is the most likely cause of the latch-up condition.  I suggest directly connecting all grounds on the ADS1174 to the analog ground plane and run some tests to see if this improves the operation of the device.  Under normal operation, it is not likely you would see any significant difference in voltage between the two grounds.  However, any transients caused by an ESD strike, or some other high current/voltage transient that results in a transient current running through one or both of the ground planes could create a high voltage potential do develop.

Regards,
Keith

Hi Keith,

Some further news...I remain suspicious of the power sequence circuit on this board. The ADC connects to a microprocessor which is powered up slightly earlier than the ADS1174. Although from your explanation, this might be a problem, I have yet to see the chip latch up from power on. In fact, it often will happily run, taking readings for a long time. It surely seems that the problem only occurs when the micro communicates off board. Even then, it's not all of the time - very random.

The power sequence circuit uses a couple of comparators, as I mentioned, and they have no hysteresis and filtering (I'll look at the layout later as it might also be not sufficient). I suspect that the off-board communication is created some sort of noise on the pcb that the comparators pick up. Without any filtering or hysteresis, it would not take much to trip either of these. That would shut down all power to the chip, except for 1.8V, I believe. Yet, the micro would continue to run. So, that case could easily exceed the max levels you speak of.

This afternoon, I bypassed the power sequence circuit and proceeded to make a number of readings. I'm not ready to declare victory yet, but the board worked correctly the entire time.

Once I send you the schematic, I think you will see that this is at least one potential problem.

regards,

gene

Hi Gene,

Based on your observations, I do not disagree with your assertion that the power-up circuit needs cleaned up.  The ADC is certainly not going to behave well if the IO supply is removed and the MCU is still driving the inputs.

Regards,

Keith

Problem is not yet solved. 

The 1st board behaves okay with the power sequencing bypassed.

A 2nd board goes into latch-up frequently. I bypassed the power sequencing again, just like the 1st board. On this board, the symptom found is the ADC readings lockup with a steady but random values (the 1st board always returned 0). Additionally, the chip gets really hot.

You recommended connecting the ground planes together at the chip - so I'll give that a try. Is a single #30 AWG wire sufficient? If not, I'll have to come up with something better.

FYI, the PCB separates the AGND and DGND, but they exist on the same layer. I noticed some moats in the planes, with some traces crossing the moat. 

Hi Gene,

30AWG should work if very short in length.  Hopefully your supply bypass capacitors for the analog and digital supplies are very close to the ADC.  You could start by shorting the ground returns of these capacitors together.  If this does not help, then jumpers directly between the AGND and DGND pins are the next best option.  Since you see such a high occurrence of latch-up, this should show improvement.

Best case is to redo the board with a solid ground connection beneath the ADC.  Take a look at this post that discusses this in more detail.

https://e2e.ti.com/support/data-converters/f/73/t/755516

Regards,
Keith

My technician used #26 AWG to connect 2 of the AGND pins to DGND. So far, so good. I will keep testing.

The engineer who did this board sectioned AGND and DGND as 2 partial planes on the same layer. The distance from AGND to DGND is roughly 10", if you can believe it.

Although I don't approve of the layout, in his defense, the data sheet is extremely ambivalent about this. From the pin description:

"Analog ground; connect to DGND using a single plane"

From the 'Application Information' section:

" Ground Plane: A single ground plane connecting both AGND and DGND pins can be used. If separate digital and analog grounds are used, connect the grounds together at the converter, or at the power entry point of the printed circuit board (PCB)."

Did more than one person write this data sheet? The data sheet gives 3 methods to handle ground planes. The engineer who made this board used the "or at the power entry point" option which is clearly the worst option in retrospect, but is okay per the data sheet.

I spoke with another engineer here who uses the 8 port version of this chip. He has no power sequencing, uses a single plane for all grounds, and incidentally frequently exceeds the power rails when driving the analog in. He further stated that it was unavoidable, so used current limiting series resistors. He reports no latchup in years of using the part.

Based on all of the forum entries, the grounding is the biggest culprit. That was your earliest point to me - and I think you may be right. So then, why is the data sheet so ambivalent on this crucial issue?

gene

Hi Gene,

It appears that the recommendation to connect grounds at the power entry point on the board should have been removed.  The ADS1278, which is the 24b version of this device, and follows all of the same board layout requirements, does not include this statement.

I will bring this up to the owner of the ADS1174 datasheet so that it can be corrected on a future revision.

Please let me know if connecting the grounds together fixes the latch-up issues with your board.

Thanks!

Keith

Thank you, Keith.

Today, I used the circuit board that failed regularly. With the fixes we spoke of (connected the AGND and DGND with #26 AWG, and bypassing all the power sequencing, eliminate the power series resistors, etc.,  I was unable to get the chip to lock up.

Tomorrow, my tech will make the modification permanent on 2 PCB's. I'll work with both and see if the situation has improved. If so, I have about 6 more PCB's that I can make the changes to and try them as well. I expect all of the ADS1174 is from the same lot, but at least 8 PCB's will give me some level of "statistical okay-ness"  (TM).

I will post the results in a few days.

gene