TMP392: Start-up "Blip" of TMP392

Hi Cody,

Do you have a schematic that you can share that details the connections of the TMP392?

Also, for clarification— what is represented by the green, yellow and purple signals? 

Jalen

Cody,

Additionally, which channel is showing the behavior?

Jalen

Hi Jalen, 

Both channels are, somewhat randomly, with Channel A doing it more frequently than channel B.

In the scope shot above, the green is the output of the temp sensor. The pink is the schmitt trigger inverting buffer following the temp sensor. The yellow trace is the power supply that powers the 3.3V switcher (It is mislabeled in the screen capture below- disregard that.) that runs the temp sensors, the pull up resistors, and the inverting buffer. Here is a screen capture of the schematic.

I should also mention that the pull-up resistors on the output of the temp senors have been moved all the way down to 2.2k from their shown values and it did not fix the issue. We had also tried pull-downs on the gate of Q1.9, and pull-up caps on R10.9 and R11.9, in hopes it was only transient- to no avail.

Cody,

Thank you for providing the additional information. I don't see anything wrong in the schematic so I will have to replicate your setup in the lab with our EVM.

I will update you in the next couple business days with the results I've found.

Jalen

Hi Jalen,

Thanks again for your help.

There is brief mention of the "Power reset time" in the electrical characteristics table, as shown below:




Can you please ask the design team today if there is any known behavior in which the output may go high during that 10ms? Unfortunately, we'll have to make a decision on whether or not to pull this circuit by tomorrow morning, so appreciate your quick response.

Thanks,
Cody

Cody,

Checking with design on your Power reset time question.

I will reply by end of day with their feedback.

Jalen

Thanks so much Jalen- I appreciate your help.

Hello Cody

We understand the criticality of getting clarification on the TMP392 device behavior. I have a few questions as we get the answers over to you

1. Are the resistor values used have the 1% tolerances as required by the data sheet. If you can send us the part numbers for the resistor that will be great

2. You mentioned that there is a 1-2 ms blip, however looking at the scope snapshot and time base it looks like more than 11-12 ms. Please correct me if I am reading the time base on the scope snapshot incorrectly.

3. At what temperature are you powering up the device?

4. Is this an issue on every board you have or is this a one-off.

Having said that we have not received a report on any such issue so far. Is it possible for you to remove the schmitt trigger and op-amp buffer and check the behavior?

Cody,

In addition to Amit's questions, I am making an assumption that the scope shot you provided is from the OUTA pin? Can you show a scope shot of the OUTB channel as well?

Jalen

Cody,

Using the TMP392EVM I placed the same resistor components as in your schematic and captured Channel A. I do not see the blip that you described in your scope shots.

So that we can help further, can you confirm the debug steps that Amit proposed?

Jalen

Amit, Jalen,

We removed the temp sensor from the design until we can understand it's expected behavior (or we find another solution). 

To answer your questions:

1. Yes- 1% resistors. Specific part numbers will not be supportable from a long-term sourcing perspective, so assume any generic popcorn device.
2. The blips are spurious, and not always the same width. Sometimes they don't happen at all.
3. The device is being powered up at room temp.
4. Issue on every board.

"... we have not received a report on any such issue so far.."

Can I assume that this means that the spurious toggling of the outputs is confirmed as unexpected behavior? If so, we can dig in a little deeper. (If not, I'd like to not spend any more time on it and just find another solution.)

To ask another way- What does the spec "power reset time" pertain to? Is this simply how long the temp sensor will take to trigger after being powered up, or does this mean that the outputs are invalid until 10ms after power-up?

Thanks in advance for your help,
Cody

Hello Cody

We checked with our design and the only condition this could happen is if the resistor are incorrectly decoded during power up. That is where the specs for the resistors other than 1% are also important, i.e TCR of 100 ppm/C and LTD of 0.2%. If you can share the part numbers, we can source the same and check it on the EVM to reproduce the failure and root cause the same.

The spec on power reset time is where the device spends time to decode the resistor and get the first temperature readout for comparison against the set threshold. During the time of power reset, the output are help inactive. Only, when the first stage of decoding and temp comparison is complete, does the output toggle.

Thanks Amit, this is useful information. 

The resistors used are: CRCW0402332KFKED and CRCW040214K0FKED. As you can see they are 1% and 100ppm/°C.

I think that, when the device is first starting up, it is trying to get it's decode in a really noisy environment- since simultaneously the PFC next to it is coming to life.)

It also is on the top layer, which is capacitively coupled to the PFC boost swiching node via an internal plane, so that could be making for a rough go of the initial decode. It seems to work very well when powered from an external supply, so that seems like it could be the culprit. 

I'll circle back when we have a chance to do a bit more testing. Thanks again Amit.

Hello Cody

Thanks for posting back with the resistor part numbers. I agree that the specs meet up as expected by TMP392. At the same time your assessment is correct. In a noisy environment, any external noise could cause an issue for resistor decoding. The good news so far is that the device does reattempt to acquire the resistor value. 

Let me check with Jalen as to what can be done to reduce the coupling of noise.

Thanks Amit.

We have a redesign coming up of the board anyway, so we'll work on a new layout that should be able to fix it. We'll also be moving to a single channel device. If you or Jalen know of a temp sensor with an integrated shutdown pin, we could just hold the unit in shutdown until we get everything up and running. We can do it in logic of course, but having the pin would be nice if you know of one like that.

When we implement the new single channel device, we'll pass some layout files over for review. Always good to get a 2nd (or 3rd) pair of eyes on it to see if you guys see any issues. We'll have to work that through email instead of here on e2e, to protect the client's confidentiality.

Thanks,

Cody

Hi Cody

I have sent you a PM to reach out directly. Having a shutdown pin defeats the purpose of the thermal switch which is supposed to be autonomous. Let's say that the MCU pin got disconnected or failed. It may lead the Temp Switch to be always ion shutdown, causing it to float the line high, and never give an over temp warning via the OUTA and OUTB.

Having a dual output is rather useful. You can set them to the same trip point. This will allow you to check the output from both pins before making a decision. In case one of them fails, the other OUT pin provides redundancy.

Can you consider a load switch to power up the TMP392 once the PFC is stabilized?