TPS2116: Switching from IN2 to IN1 back and forth under load

Hi Martin,

It looks like the PR1 signal is fluctuating above/below VREF (1V typ). PR1 doesn't have any hysteresis - this can make it tricky when working with switching converters and their output ripple. However, you can add a resistor between PR1 and ST to create hysteresis on PR1 - there is a helpful calculator for this at the bottom of this e2e FAQ. I calculated that a 100kOhm resistor between PR1 would ST would make your circuit switch to IN2 when PR1 = 4.2V and switch back to IN1 when PR1 = 4.35V, maybe start with this value and see if that corrects the behavior.

If that doesn't work, can you send a waveform that includes PR1 (with 200mV/div resolution), ST, IN1, and OUT? That would help diagnose where the issue is coming from.

Also, I would recommend putting a 10-20uF capacitor on IN1 in general to support load steps when the device switches back to IN1, but if you find that your problem is solved fine otherwise, then it's not explicitly needed.

Please let me know the result of your tests.

Thanks,

Patrick

Hi Patrick,

the hysteresis is an interesting thing, why is that not mentioned in the datatsheet?

I tried out adding additional 22µF as near as possible to the IN1 terminal, which is as well as near as possible to the PR1 voltage divider.
Additionally, i added the 100k for the hysteresis. Sadly, nothing helped.

I made a few more shots with the oscilloscope and as well add some Logic analyzer pictures for better view on the timing.

As requested, a shot with PR1 (CH1), ST (CH2), IN1 (CH3) and OUT (CH4):

I also modified the 4k7 R2 to be 6k8:

And then, we have one of the switching events zoomed in:

As you can see, PR1 will be completely driven to zero somehow. Is there any circuit inside the TPS which is discharging the PR1 pin from the inside in special situations?
If that is not the case, a small capacitor on PR1 should fix our problem immediately. But if I add 10 or 47nF in parallel to R107, nothing (!) changes at all.

Logic analyzer pictures (50MSamples only, so very small dips are not correctly displayed as seen in the scope pictures):

22µF on IN1 and 100k between ST and PR1:

6k8 instead of 4k7:

Zoomed in one of the events:

Strangely, my third PCB fixed itself again after soldering long enough on PR1, both resistors, ST and all capacitors around, i will dig into this again...

If i find something, if will tell you,
Martin

Hi Martin,

Thanks for the detailed reply. As to the hysteresis not being in the data sheet - it can be inferred from lack of hysteresis specification in the Electrical Characteristics table, but we will add this in the next data sheet revision - thank you!

It looks like there's quite a bit of ringing on PR1 (and IN1) - my first thought is that there's a lot of inductance from TPSM53602 to TPS2116, but yes, a capacitor on IN1 would have solved this. Is there by chance a long and/or thin trace/cable connecting the TPSM53602 VOUT and TPS2116 VIN1?

There's no circuit in the device that would discharge PR1 - it is only an input to a comparator, so we'll need to find what the source of the ringing is.

As a note, I see that TPSM53602 operates in PFM mode (see Section 7.4.2 of TPS53602 data sheet) at your operating conditions, which increases efficiency at low loads but increases ripple and decreases load step performance. And, it just occurred to me that the inductance causing the ringing could be the TPSM53602 inductor itself. Additional (not sure how much, maybe 2-3x?) output capacitance near TPS53602 would help with this.

If not, maybe one of the converter output caps had a soldering issue, or maybe there was a soft short on one of the resistors or TPS2116 pins?

Yes, please let me know if you find anything more or if you have any ideas about what was mentioned above.

Thanks,

Patrick

Hi Patrick,
sorry fort the late reply!

I was working on the problem since on and finally managed to find the solution!
The problem was not the IN1 path but IN2.
IN2 will be generated by another board in our system which is connected via a 15cm cable to the main PCB with the power mux. On our last schematic review, no one noticed, that we do not have a local capacitance in our VIN2 path on the main PCB. In my tests for the PCB, I used a slightly longer wire to supply VIN2.

What is happening here is, that in the moment of switchover, my VIN2 does have a ringing which I did not see before because I did not searched the signals for peaks of <<1µs and especially not on VIN2.

My assumption now is, that the the loaded switchover to VIN1 will drop VIN1 again but release VIN2 the same time. As VIN1 drops, the TPS will do another switch to VIN2 and VIN1 will rise, VIN2 will drop again. The problem is, that with capacitance not directly placed next to the TPS, the voltage drop on VIN2 is nearly to 0. I think the TPS will somehow kind of “restart” because it does not have a proper supply in that moment?
In our case, a small capacitance of 100nF was enough to prevent failure due to constant switching but with a ringing of 2Vpp left on VIN2. I now placed a 10µF ceramic capacitor directly next to the VIN2 pad and now the ringing is down to 0.3Vpp.

Can you share with us how the TPS is supplying itself from both input rails?
I just want to know if we managed to shut down the chip while switching over or if it was only an oscillation.

It is clear that the fault is on our side not having both supplies stable enough to power the TPS or prevent switchover-oscillation (whichever is true), but maybe it will be a good addition to the next datasheet revision to not only mention the stability of the supplies in the last section but include capacitors in the basic and typical application sections?

Thanks for your help and have a nice week,
Martin

Hi Martin,

Thanks for your patience as I have been out of the office this past week due to the US holiday.

Great to hear that you found the solution! I saw that IN2 looked stable and undisturbed in your waveforms so IN2 capacitance didn't look like an issue to me.

The device powers itself from whichever supply is higher - please see Section 7.4 of the data sheet (below) which gives some information about how the device could reset upon switchover.

If the ringing on IN2 creates a slew rate of > 5V/us, then my intuition is that it could cause the device to reset, but I will need to check with the designer to verify this. However, I think It's more likely that the device is just switching rapidly between inputs without a stable resolution as there is a 1ms typical delay before VOUT is fully powered by INx when the device turns on. I will discuss this with the designer and follow up with you on this.

Thanks also for the input on the data sheet - we will add this in the next revision.

Thanks,

Patrick