TPS26600-02EVM: Not shutting down under short circuit condition

Hi Neal,

Thanks for reaching out. I will get back by tomorrrow EOD on this.

Regards

Kunal Goel

Hi Kunal,

I look forward to your response. While I am waiting, I today setup a similar test to the one I reported yesterday, but I lowered the input voltage to 36V, and used a brand new EVM board. I still had the additional upstream capacitance in the circuit (440 uF in aluminum electrolytic and ~ 6.6 uF of MLCC). The only changes to the EVM board from it's new in box configuration were to remove J4 so that mode is floating, and J5 jumper was moved to position 5-6 for a 1A current limit. Otherwise, it took a very long time to cause the EFuse to fail (more than an hour of consistently trying), and many times the device locked into a fault state until it was reset (S1 button press). But the important thing is that this is setup basically the way the EVM is setup with the exception of mode and current limit, and the part has failed with a short from V_IN to V_OUT. I'm wondering if there's any way to protect the part from failing in this way?

Thanks,

Neal

Hi Neal,

1. Output voltage behavior is looking to be due to the way short is done. It will have many arcs and bounces. 

2. Reason for fail could be VIN abs max violation or VOUT abs max violation.  Looks like you are using long wires at input and output. So may be present TVS and schottky not clamping within limits

3. We should probe VIN also. Also lets observe VOUT at µs scale and see if it is going negative when short is applied.

4. We can also probe output current.

5. How much cap is at output?

Regards

Kunal Goel

Hi Kunal,

Today I took another EFuse evaluation module kit, I set it up with a 36V power supply, removed jumper J4 so that mode is floating, and J5 jumper was moved to position 5-6 for a 1A current limit. I did NOT include any additional upstream capacitance. I repeated the testing I was doing yesterday, and again was able to cause the EFuse to fail after about 10 minutes of repeatedly generating lots of arcs and bounces. While doing so I captured VIN, VOUT, and FLT on a scope. I did not capture current as we unfortunately do not have access to a high speed current clamp. Unfortunately the fatal arc was not captured on the scope, but I was able to observe and capture many of the arcs and bounces. In some instances Vout got as low as -3V, but I suspect the input voltage is where the problem may be. Whenever there was a particularly loud/large arc, the input voltage had a large decaying ring to it. In the worst case I measured a VIN of ~ 61V. I assume what's happening here is when there's a loud/large arc, a lot of current is flowing through the short, and that's causing a large voltage spike when the EFuse opens. What I am unclear on is how to protect/prevent this from occurring in a system that is operating off of a 48V battery that at full charge can be as high as ~54.6 V? I was able to kill the EVM board with a 36 V nominal power supply. It seems like this problem will only get worse as the nominal voltage increases?

I've attached 2 scope captures, one of the 61V V_IN swing, and one of the -3V V_OUT swing. I've also attached one image of the EVM, wire in, wire out, and the handle of a C-clamp (that's otherwise isolated/foating). We're generating the arcs and bounces by rubbing frayed ends of the ouput wires on the handle of the C-clamp, as we've found that's a good way to make lots of arcs and bounces.

You'll see that the test setup is quite straightforward, and doesn't use much other than a standard eval board, power supply and some wire. If you would like to reproduce the test I think it will be pretty easy for you to do so. I am going to have trouble conducting more tests as I used out last EVM today, and replacement EFuse parts are out of stock from all authorized distributors.

I'll also try to answer your questions:

1. I agree that our test method will generate many arcs and bounces. I anticipate that the EFuse should be able to protect the system in the event of any type of short circuit downstream, including ones that create many arcs and bounces.
2. The wires in this test are between 1-2 feet in length on both the input and the output, why would this prevent the protection diodes from operating to protect the EFuse?
3. Done, the latest images are on the shorter time scale and show VIN, VOUT, and FLTb.
4. Can't probe output current because we don't have a high speed current probe.
5. The only output capacitance in the system is that which is present in the default configuration of the EVM, namely 1 uF MLCC (C3).

Thanks,

Neal

Hi Neal,

1. Can you share image of failed device?

2. I believe reason for fail is more because of output abs max violation.  It is going -3V and is way beyond -0.3V abs max spec. Input of 61V is still below abs max spec of 62V but above recommended of 60V .  The level of voltage going positive and negative depends on wire length and so you will need to choose TVS and schottky accordingly for your application. 

3. One solution is to use high output cap . That will cause VOUT to go less negative.

Regards

Kunal Goel

Hi Kunal,

1. Please find attached 2 images of the EVM I failed yesterday, along with a few of the failed chips we've removed from prior EVM's that have failed in the same manner. 
2. I would not be surprised if at the point of failure the VIN exceeded 62V. In my response yesterday I was trying to say that we observed transients on the 36 V Vin rail when the EFuse opens up that varied from only a few V pp, up to the 41.4 Vpp shown in the scope capture in my post yesterday. It seems entirely possible to me that the Vin ripple may have exceeded 62 V at the time of the failure. Never the less, we can chase down the output voltage going negative to see if that can resolve the problem.
3. Okay, should this be a large MLCC or aluminum electrolytic cap? Is there a value that you recommend we start testing with?

And a question from me:

4. Is there a way to get TI to send us some more sample parts to test with, it seems like the 26600PWP* is out of stock at all authorized distributors.

Thanks,

Neal

Hi Neal,

1. From scope shot you shared, VIN looks to be going till 61V. 

2. You can start with 47uF on EVM by putting J3 in position. You can also put cap at input side also to limit VIN peak.

3. I see some stock on ti.com. https://www.ti.com/product/TPS2660?login-check=true#order-quality You can directly order.

Regards

Kunal Goel

Hi Kunal,

I have ordered 10 more pieces of the EFuse part, it seems like that's the limit per customer on a monthly basis. So hopefully we can get to the bottom of this before I go through these next 10 parts, or we can find another way to get some more samples.

We are replacing an EFuse on an EVM currently, and I will retry with the additional output capacitance. You'll recall from my earlier testing that we had been testing with ~ 440 uF electrolytic and 6 uF MLCC upstream in our earlier tests (I removed this recently to try to make it so that the circuit looks as similar as possible to the default EVM circuit to make it easy for TI to reproduce the failure). I can reinstall the upstream capacitance, and the 47 uF MLCC on board downstream capacitance jumper J3, and spend some time today seeing if I can still cause the same sort of failures we've seen previously.

Thanks,

Neal

Hi Neal,

Sounds good. We can start with high cap on output side also like 400uF or so to be on safer side. If that much cap is causing VOUT drop to be limited to safe spec and not causing fail then we can decrease cap and see at what value fail starts happening. That way you can decide minimum needed cap for your system. Aim is to prevent VIN and VOUT abs max violation. What is the max operating voltage in your application? I think we need to choose correct TVS and schottky based on your application.

Regards

Kunal Goel

Hi Kunal,

We're operating a 48V nominal battery as our supply, so up to 54.6V on a full charge.

Thanks,

Neal

Hi Kunal,

I started with the 47 uF aluminum electrolytic capacitor that is in the EVM design. I still had the output dipping negative when the EFuse opened on a short circuit event. The lowest I observed was -2.5 V. However, I have been unable to cause an EVM EFuse to fail with the 47 uF output cap connected in the EVM. To be clear, the -2.5 V is OUT wrt GND, but looking at the datasheet for the part, this doesn't seem to be sufficient information to know whether or not we're violating the absolute max limits, as the datasheet specifies a limit for OUT wrt RTN, and RTN wrt GND. Since I don't have access to a differential probe, I suppose I will need to measure return and out at the same time wrt GND and see if they combine to result in a violation of the OUT wrt RTN max rating of -0.3 to + 60 V. I'm not in the lab today, but can try to capture this data later this week.

Additionally, if we are not concerned about reverse input polarity, it seems like we could tie RTN and GND together. Based on the challenges we're already having with this device in trying to prevent it from failing short when a downstream short circuit event occurs, is it advisable to connect RTN and GND together?

Thanks,

Neal

Hi Neal,

During normal case internally RTN is shorted to GND. We have some margin over that -0.3V spec. Up to -0.6V it should be fine. 

Connecting RTN and GND may not help here.

Regards

Kunal Goel

Hi Kunal,

Is it possible to message you about our specific application in a less public forum?

Thanks,

Neal

Yes we can use e2e messaging.

Regards

Kunal Goel

Hi Kunal,

I have built some prototype boards, which I am testing, and one of the tests that I ran was checking the reverse standoff voltage of the downstream Diodes Inc. B260A-13-F Schottky diode. I noticed that the diode didn't start to conduct until somewhere above 70V reverse voltage. This is confirmed by figure 2 of the datasheet, which shows the knee occurs at > 120% of the rated peak reverse voltage, or > 72 V in the case of the 60V reverse working voltage diode. I'm wondering if this is intentional, as 72 V would exceed the max ratings of the EFuse, so a positive transient above the EFuse max ratings may not get shunted via this diode? Is it recommended to put a smaller reverse breakdown voltage Schottky in order to limit over voltage conditions?

Thanks,

Neal 

Hi Neal,

Are you planning to place this at eFuse output or input? For input of eFuse please use a TVS diode. I think they are better as compared to schottky for clamping as they are designed for transients clamping. For output side of eFuse use a schottky diode. 

Regards

Kunal Goel