TPS54336A: Oscillating output

Hi Chris,

It looks like something wrong on your board. Here are my suggestions.

1. Double check if you have any wrong parts on the board. Make sure inductor/caps/resistors are correct values. Especially C3 which should be 0.1uF with correct voltage rating.

2. Re-flow or re-solder TPS54336A to make sure the thermal pad is well soldered on the board. 

3. Re-flow or re-solder the compensation components, then deep clean IC area to make sure no any flux contamination. 

4. You said "you tried lowering the value of C4 by soldering another 12nf across it. Unfortunately IC itself caught fire and I have no idea why". If you add another 12nF cap across C4, it will increase that cap and may cause loop unstable and burn the IC.

5. Measure the switching node waveform to make sure it is correct with 340kHz switching frequency.

Good luck and Thanks,

Nancy

Hi thanks for your reply. 

I did already re-solder 2 boards fully and check / replace the RC parts. But I had 100 boards made, and 15 tested so far have exactly the same fault. The same board has a 5V regulator (another texas part) and no problems with that circuit.

I will check the frequency tomorrow. I don't think I checked C3 value, so will check that tomorrow as well. 

C3 was 1uF, so I changed to 100nF. No change other than "finger on board" no longer works. 

Frequency seems 340KHz.  Can't figure out how to attach the image. 

Starting with a fresh board again. Oddly I have a huge 1volt p-p ripple on the 12V running at 30Khz. When I put my finger on the RC parts, that disappears and everything runs fine.

Changing to a previously known good board, this does not happen at all. The ripple is extremely low in the mV range. 

Hi Chris,

Does the fresh board have 1uF C3? Did you change C3 from 1uF to 0.1uF? I think you may de-solder the IC to see if thermal pad is well connected to GND of board?

Or replace a new IC. This should be manufacture issue related. Also deep clean the board to remove flux. Or swap the IC between good board and bad board to find root cause.

Thanks,

Nancy

All the boards have 1uF. But it seems my older boards have 1uF anyway. I did change to 0.1uF, but as said above, it doesn't help. 

I did wonder about the thermal pad, but that wouldn't make sense as with "finger on board" I can push 4 amps. 

When I mean "finger on board" I am not implying a bad connection ( I have resoldered it several times on a couple of boards now). Is when I touch the RC parts on the COMP pin it starts to behave. I do not have to push down hard at all. If I push down with a plastic pen, nothing changes. Though it is literally down to the conduction of my finger on those parts. Though again as previously mentioned, I have already added much variation onto those values without success. 

The next step is probably to change the IC, but because of parts shortages that is not so simple. I will see if Texas still offer samples for them.

Hi Chris,

Swap the IC between good board and bad board, then you will see the thermal pad connection and identify if it is board issue or IC issue.

Thanks,

Nancy

The problem is I only have one working board and I need it to power other equipment I'm working on, so I would rather not risk breaking that one as well.

I have however removed the IC and re-solder it back on but did not solve anything either.

So "finger on board" equates to 27K across C5. With the 27K there, noise & ripple is under 20mV with a 3amp load. 

Made some progress R3,C4 in the datasheet are 3.74K and 12nf, however using webench it states 36.5K and 330pF. I didn't have those exact values, so 47K and 220pF. I can now push 3amps, but the inductors are still "hissing" and I have about 300mV output noise. Adding the 27K back solves the problem again.

Hi Chris,

Good news. I am not sure about your solution. What is your compensation to solve the problem? R3=47K, C4=220pF, C5//27K? How about your previous board with same compensation? Could you check if you use a wrong part on C5?

Thanks,

Nancy

So just to clarify all points.

The original power supply builds have also got 1uF for C3.  C4,C5,R3, are as per the datasheet In past builds and the current problematic builds.

Adding 27K across C5 on the new builds solves the problem.

Adding 27K across C5 on the previous good builds actually seems to create the problem ?!

I'm starting to think there is something inherently problematical with these TPS chips in general. As I also have used the TPS54628 on the same board. The problem with those is every time I do a batch of 100 boards, I have to constantly adjust the feedback resistors because on one batch I can have it perfectly set at 5 volts. Then on another batch of boards it ends up at like 5.8V. So I have to change all the resistors on them. In fact on the current build of boards I added a trimmer preset to just easily change this because of that problem. I realise specifications & tolerances can have some variation but this seems somewhat ridiculous

I managed to find a scrap board with a chip on it. So I swapped if one of these new boards and it gave 3.5amps easily without shutting down. I still had very large ripple on the output, but I cannot remember if I had changed the capacitors on this board previously or not.

But in anycase, it seems for whatever reason these chips are massively out of spec :( 

Hi Chris,

Thanks for all your information. The support engineer Zhao Ma will be back from China holiday tomorrow. He is an expert for this part and will continue to help you.

Thanks,

Nancy

Hi, Chris 

Some comments: 

1. Could you send me the original schematic? 

And take a summary: What changes you did can make board normal working? 

2. Could you measure some waveforms when issue happen? including inductor current, COMP voltage, and Vout. -->which can help to do the analysis. 

3. What is the SS cap? and what is the total output caps including downstream? 

The schematic is pretty much identical to the datasheet.

1) I change the feedback resistors to output 12V

2) The SS cap is 2.2uF. This value had to be used because the power supply is often connected to equipment which have a lot of bulk capacitance and the power supply would not start up otherwise.

3) 

Adding 27K across C5 on the new builds solves the problem.

Adding 27K across C5 on the previous good builds actually seems to create the problem

4) I did try to upload a image of the COMP voltage. it is basically a sine wave.

5) I don't really have the means to test the inductor current unfortunately.

6) C3 is 1uF. But all the previous builds use this value.I even tried changing to 100nF with no improvement, I even added 10R in series with it with no improvement.

7) I have tried varying the components on the COMP circuit, C4,C5,R3 I think there was some improvement as described somewhere in the above threads. But the the output ripple is extremely large 300mV and the inductors are sounding very "hissy" which they do not ordinarily do.

8) original boards had 100uF output capacitance. The new boards of I think had 47uF due to parts shortages. I have however soldered 100uF ontop for a total of 147uF with no improvement. However if I solder a large electrolytic of 4,700uF there is a noticeable improvement to the point I think it is again operating correctly.

9) Removing the IC and re-soldering it did not change anything. 

10) coating the board with freezer spray did not change anything.

11) changing the IC for a previously known good one solve the apparent overcurrent protection, but the 300mV ripple remained. But I do not remember if any of the components were changed previously on that board. Unfortunately I do not have any other chips to try and unable to source them from anywhere.

As mentioned in point 3, adding 27K across C5  Solves all the problems on the new builds. But if I do that modification on a previous build, it creates the problem. If you can try your own EVM board with 27K across C5, then I would assume you would witness exact same problem. 

I think there have been around 300 builds over the past 4 years Which have been operating without any problems. But I can only really conclude there is some problem with the chips themselves as the new builds all work derived from the same batch. These were obtained from a surplus outlet which I have never had any problems with. In fact they pride themselves on not selling counterfeit or faulty items. While this could indeed being question, the TPS54628 I have had problems with right from the start. 

As previously mentioned, I would constantly have to change the feedback resistors to obtain 5V output on the TPS54628. Otherwise it could be as high as 5.8V. I would expect a little bit of variation may be 0.1,0.2V but even using 1% or better tolerance components did not help. It's why I was forced to use a trimmer on the current build because I did not have time to keep testing and changing the resistors all the time. But this problem has been ongoing for the past 4 years and I ( as far as I can remember) always purchased the parts from Mouser. So in light of the possibility of the TPS54336 Being somehow factory rejects or something, it does not really explain the problems with the other chip either. Unfortunately at this point I can only assume these chips are just very bad in tolerances over production batches or something.

The only other possibility is they are somehow becoming damaged during soldering. All previous batches I soldered myself. The new batch was built at a factory in China as I did not have time to do them all myself. I guess it could be possible they was heat damaged. But I think it will be more likely I would damaged them during soldering as I do them by hand. But again I cannot really test if heat is a factor because of parts shortages. Though I remember when I was prototyping very first boards, the same chip would generally get used multiple times across boards and I do not recall there ever being any odd issues in doing that. So I do not currently think heat damage is a factor. But even if that was the case, I could not really risk using the chips again another build if they are going to be that fragile.

Hi, Chris 

1. It looks you used the recommended schematic, I think it is okay to just adjust FB divider resistors for 12Vout. 

So, could you send me the layout files? screenshot is okay. 

2. "SS cap is 2.2uF", it is too large, maybe you can try 220nF cap. 

3. "Adding 27K across C5", confirm with you, you parallelled a 27k resistor with C5?

If so, it is an error, we cannot do this. 

4. If cannot measure the inductor current. 

Could you measure COMP, Vin, and Vout when issue happening? please measure power on/off, and steady state.   

6. C3=1uF is okay. 

7. The recommended compensation circuit in datasheet is okay, don't need to change it when debugging. 

8. It is weird. 

9. This part have been production for many years, it is robust, i think it is not caused by IC itself,  let's focus on application and debugging. 

How do I send you files, I tried to upload images here before and could not see how to actually do it ? of course I would rather not upload the layout files publicly.

I have uploaded the requested information into a zip file on my server, but I cannot see a way to private message you the download link.

>>3. "Adding 27K across C5", confirm with you, you parallelled a 27k resistor with C5?

Yes in parallel. 

I tried 100nF SS capacitor, no change other than the " power up- power down" restart is happening a lot faster now.

I have also been wondering if the IC somehow thinking it has a light load and entering pulse skipping mode ? Is not very clear if there is anything more to it than that. But I have read some other manufacturers only switch one MOSFET internally to save power on the light loads. I wonder if this is happening, as it would explain the erratic voltage regulation under load. When there is no load, the output is steady with no issues.

Hi, Chris 

You can send it to my email: zhao-ma@ti.com 

please include schematic, layout, and waveforms.