TPS92691-Q1: Switching transistor and diode keep buring

Sorry, the forum glitched. Consider this to be a continuation of my explanation. Here is the screenshot.

As you can see, prior to switching, the voltage across the inductor is 0 (as would be expected). When Q1 turns on, the voltage across the inductor increases (which would be consistent with a discharge, again as expected). So as far as I can tell, the inductor is pumping 9 A into Q1. The problem with this is that Vis is to high, and switching shuts off for 35 us.

We replaced C7 with a 10 nF capacitor to slow the rise time at the IS pin. It is at this point that we noticed the strangest thing happening.

If we kept Viadj at 0, everything was fine. There was no switching and the rails were stable, but as soon as we increased Viadj, even to 10 mV, our 12 V rail, coming directly from the power supply oscillate between about 10 V and 4 V. I'm sure UVLO was kicking in at 4 V, resetting. It was repeating this at a frequency of about 100 Hz.

When we disconnected the PCB and measured for continuity, it showed as though 12 V was shorted to GND and the node connecting the inductor to D1 and Q1. However, if we set Viadj to 0, the 12 V rail was fine. I will note that we set the current limit of our benchtop power supply to 1.4 A, which is far higher than the threshold limit of our laser diode.

At this point, I'm at a bit of a loss.

Hi Tuan,

I'll give this a try. Just to clarify, once the TPS92691 has started up, we can drop the current down to 0?

Hi Tuan,

I just gave this a try, and no luck. We applied 1 V to IADJ before powering it. I will note that we replaced C7 with a 1 nF capacitor again. We are still overshooting IS, just like the first screenshot I submitted.

Thank you,

-Alexander

Hi Tuan,

It's a similar design for a different product. The application 5 months ago was a very high power circuit, 10 A, if I recall correctly. This one is much lower. It peaks at 2 A. Most of the pins are operating as one would expect. Pins 1 through 8 all appear to be functioning correctly.

I'll ask our tech to review the solders.

The design isn't 100% the same. It has different optimization. The main constraint on this design was size, so I couldn't use a DDPAK transistor and diode, so they are different. I nonetheless took away a lot of lessons from the pervious design.

Our tech is doing a node by node check right now.

Ultimately, L1 is discharging to rapidly into R10. That doesn't really sound like a soldering issue.

Hi Tuan,

We checked all the solders and we did not find an issue.

Let me clarify our observations:

When we look at GATE and DDRV, we see them both turn on, and there are several irregular pulses on GATE before DDRV turns off. Our going theory is that when Q1 turns on, L1 discharges into R10, the voltage across which is > 525 mV, so it triggers a fault and a 35 us timer. This should shut down both GATE and DDRV simultaneously, but there are 3-5 pulses at GATE before DDRV turns off.

Hi Tuan,

We are not using the same inductor. The current rating for the inductor is 2 A. Right now, we just want to get any current through the Laser Diode. We set Iadj low, to get only a ~200 mA, but we're still getting overcurrent on IS. I could reduce the value of R10 (the thought crossed my mind), but then we're just increasing the current limit, which we don't want.

Also, we would have to decrease it by quite a bit. If we're hitting 4 V, and we can't exceed 400 mV, then we'd need to make IS 24 mO, which just doesn't seem right.

Here is the link for the inductor we're using: MGAH0603330M-10 Laird-Signal Integrity Products | Inductors, Coils, Chokes | DigiKey

Or an alternate question would be, "why are we putting 15 A through the switch"?

But also, the GATE pulse width is about 90 ns, and IS should be disabled for 150 ns.

Hello Alexander,

Tuan is out on vacation. Please expect delays on responses. 

Best,

Pedro Arango Ramirez

Hi Pedro,

Conveniently, I was on vacation as well. Sadly, time has not resolved the issue.

Hi Tuan,

I'll look into this today.

D1 was soldered backwards. It now seems to be working.