TPS7A03: Vout ripple

Hi Roy,

Would you be able to provide the schematic used here?

What values of Cout and Cin are used? And for load transient, what low current, high current and slew rates are used?

Also, in the failure case, there are some additional output deviations that have fast edges (circled in picture below). Would you be able to provide zoomed in scopeshot of one of them such that the frequency could be determined?

Thanks

Bobby

Hi Bobby,

I think we may not the issue to be more complicated. Because we had done the ABA swap action to see the issue was following by suspected chip.

For suspected chip, we need to add 8pcs 2.2uF Cout to have small ripple. 

From our side, I would like to know 

1. Why the suspected chip having the higher transient ripple?

2. Does the chip is ok for the kind ripple?

Regards,

Roy

Hi Roy,

Thanks for the information.

What package do you have for TPS7A03? WCSP (YCH), DQN, DBV?

Also, what voltage do have on Vin and what is the headroom?

When you add 8x 2.2uF, does the high frequency ringing stop or does it continue to happen?

It is possible the failure part might be damaged. Would you be able measure the quiescent current of the failure part and a passing part and compare them. If Iq is out of spec, that is a good sign the part is damaged.

Regards,

Bobby

Hi Bobby,

1. We used YCH.

2. Vin = 3.3V, what does headroom mean here?

3. We would like to lower the ripple and so we add the Cout.

4. I will measure Iq, thank you.

Regards,

Roy

Hi Roy,

For 2., headroom is the difference between Vin and Vout. What was the output of the device?

For 3., after putting the caps, the high frequency ringing that showed in my first reply, does that happen?

Regards,

Bobby

Hi Bobby,

I'm covering same customer as Roy in their Canada site, I wanted to layer in some details to help:

(old news:) Customer has had a systems test fail that indicates higher ripple/noise on the 1.8V LDO powering our PPG AFE. Historically we don’t think we’ve had this kind of failure but currently it is about 2% on some builds. For more flavor, the left image has a triangle-shaped ripple about 2x higher at the point where the AFE (load) finishes sampling and the load transitions from ~10mA -> ~1uA. While the AFE is sampling (positive edge), there is more high frequency noise.

My first thought was capacitor derating caused us to go below the 0.5uF needed for stability. The datasheet indicates >0.5uF is required for stability. It’s possible that the 1uF caps we are using (Murata and Kyocera 6.3V, X5R 0201) are derating too close to that limit. In the past, we haven’t noticed any problems with these values.

BUT, customer's Taiwan engineers tried increasing the amount of output capacitance and found that the abnormal ripple reduced to normal values with 4.7uF but not with 2.2uF which is not what I would expect if the LDO was within spec.

Wondering if you can confirm a few things for us.

1. Is the output waveform we are seeing indicative of instability due to low output capacitance? Is there additional testing I can do with the units when I receive them here to confirm this?
2. In the past, TI has provided output load transient waveforms similar to the datasheet with modified parameters. An example is attached for 1uF + 10uF Cout and a load transient from 10uF -> 10mA. We had initially tested this part with 10uF Cout but found that the transient response was too slow, and reverted to 1uF for the fastest response. Is it possible to run a similar simulation with 1uF and 4.7uF to look at the difference in the settling time? TPS7A02 transients - Garmin.pdf

Hi Cameron,

Thanks for the information.

Just to know, do you see the 'failure' behavior for multiple devices or just one device?

Like it was mentioned in previous replies, it seems that there are some additional oscillations in the 'failure' case. Do you know for the given scopeshots, what low current, high current and slew rates are used for the load transients, and what is the output of the device?

This could help us to get to the cause of the issue.

Best,

Bobby

Thanks for the help Bobby!

This is on multiple devices. I'm unsure the sample size, but it is ~2% failure rate. 

Bobby said:

Do you know for the given scopeshots, what low current, high current and slew rates are used for the load transients, and what is the output of the device?

Vin = 3.8V, Vout = 1.8v

Ilow = 1uA Ihigh = 10mA. 

For slew, the load current I’d say it’s on the order of 50us for the rise time. The absolute values here aren’t as important as understanding if the change in output capacitance results in a significant change in the transient characteristics of the part.

thanks! Cameron 

Hi Cameron,

Usually when there is an instability, the output deviation doesn't change that much. You would see more ringing as the LDO output is settling if it was just an instability issue. Here, the failing part output goes much higher before it starts to recover. Also, the high frequency ringing before the large output deviation looks strange. 

So these looks like a damaged part, than a good part that has less stability margin.

Would you be able to measure the Iq on a failed device and a good device and compare them? Or you can submit the failed device(s) to TI for Failure Analysis (FA).

Also, if possible, would you be able to provide a scopeshot which shows both current and output voltage for these cases.

Regards,

Bobby

Hey Bobby,

Here's what we see for IQ 5vin and 3vin:

The customer has an open action to 

• Measure the ripple when no load (ok vs suspected)

Hi Bobby,

Add the test result for no load ripple between ok and suspected socket. The ripple was similar (13mV vs 11mV) Even the suspected chip had lower ripple, but it may be the measurement difference.

So currently we are focus on why ok and suspected chip had the different transient response? Who judge the BW in LDO? May you help explain more for the issue?

Regards,

Roy

Hi Roy and Cameron,

With different data points you have provided, we can see

1. Setup Issue: since you did swap the device in the same circuit, and one device is working fine, there are likely no issue with the circuit setup

2. Operating Conditions: Since the operating conditions you have used is within the recommended spec of the device, and there are devices that works as intended, this is not the issue

3. Cap Stability: You are using caps that are better than the minimum recommended condition. Usually, if there are stability issues, there would be much more ringing before the part settles. So most likely, stability is not the issue here

4. Iq: usually if the part is damaged, the Iq of the failed device would be out of spec. But in this case they are within the spec.

I also did consult other people who are knowledgeable in this area and they are suspecting this is a damaged part.

Would you be able to submit failing parts to our quality org to have FA done on them?

This way we could have some more information about what could be wrong with those parts.

Regards,

Bobby

Hey Bobby! I'm moving this to email thread with Eric and Stephen looped in as well. I will have failing boards by the end of week. 

Bad:  

Good: 

Bad with 4.7uF on VOUT: