Problem with TPS6102x

Hi,

I found one TPS61020 EVM available with me and I tried using the feedback componenet values given above (R3=2.2M, R4=240k). The EVM is working properly with this values.

I also tried touching the IC (specially the feedback node) with my finger, but still it is working fine.

So I don't think there is a problem with the part. I would strongly recommend to check the layout of the board.

For any integrated switch DC/DC converters, layout is very imporatant. TI engineers have given proper guidelines of the layout in TPS61020 datasheet and the layout of the EVM board is also given in the TPS61020 EVM user's guide. I would suggest comparing the your board layout with the EVM layout.

I hope this helps.

Thanks.

Hi there,

Thanks for trying this out.  The problem is that I can make the EVM go nuts just like my own design.  If I put my finger on the feedback components of the EVM, I can see the voltage at the output swing all over the place.  Were you attempting this with a load on the output or just testing the module itself?  I tested the EVM with no load and with a load in my design.  It didn't seem to matter.  The results were the same.

The real problems seem to begin with that parallel capacitance and whether it is something that should not be done and customers should be warned about it.   Before that, the voltages were just wrong.  After the cap, the input was shorted out and the chip was toasted.  I bet if you put a small cap across R3 your board would be dead but I don't recommend it unless you can get that chip removed and replaced easily.

I have paid much attention to the layout guidelines.  The supply is solid the first time the parts are placed but after that you can't mess with them at all (which is troubling for something being prototyped and tested).

Hi,

I connected an 18pF capacitor across R3.

With load of 110mA connected at the output, I touched the IC. The output drops to around 3.5V at that moment but comes up again when I release my finger from the IC. Even with no load condition, the same result I got. 

In both the cases, IC works fine.

Hmmn,  I don't know what to say.  The moment the voltage dips on any board I have here, that's it, it's gone.  It never recovers after that.

Thanks for the second test setup as well.  It's nice to read some positive results.

You are correct in that a boost converter is sensitive to stray capacitance and inductance.  When you place your finger on the IC, you increase the stray capacitance, cause the ringing on the SW node to increase above the abs max rating of the part and destroy itself.  The app note at the link below explains the problem and gives a potential solution using a snubber circuit. 

http://www.ti.com/litv/pdf/slva255

I have never tried adding the feedforward capacitor (Cff) to the EVM but will try it in the lab.  The datasheet says you only need Cff if R4, and therefore R3, are significantly smaller than the recommended values.  Unfortunately, a significant resistor value change is not quantified.  Including that capacitor introduces a small signal pole and zero in the control loop that might cause the part to be unstable, oscillate and destroy itself.      From a large signal perspective, the capacitor feeds through any switching spikes directly to the FB pin.  You might be able to reduce those spikes by changing the bulk output capacitor from tantalum to ceramic.  One of my colleagues, Scot Lester, wrote an application showing how to carefully size the inductor so that you can use the part with ceramic output capacitors.  Method/Example 2 gives a LiIon to 5V circuit that might work for you. 

http://focus.ti.com/lit/an/slva264/slva264.pdf

Thanks Jeff,

I've never seen that slva264 document before.  I have already switched from tantalum to ceramic.  I have some audio circuits attached to this supply and there was a bad ringing that could be heard which is gone now with the use of ceramics.  To be honest, finding tantalums anywhere near 30 milliohms is impossible, especially for cost considerations.  The ones I have are 250 milliohms and some other ones are over an ohm.  I am also assuming that most ceramics are not going to be as low as 2 milliohms as that document suggests and so they should be stable (which I have proven on my board but am trying to consider longevity now).

So basically, I would like to keep the resistor values I have chosen (2.2M, 240K), not use any parallel capacitance at all and things should be fine as long as I don't ever touch it and give strict instructions for testing/repair that it cannot be touched when live.  I haven't seen any clarification as to a limit on how big the resistors can get.  Is 2.2Mohm fine?  I'm using a 6.8uH inductor with a 1.7A current rating and 99mohm DCR.  I want to be able to get about 900mA out of the 61029 (has bigger FET than 61020) at the extreme edge of current drain with a 3.6V input.  Normal current usage will be much less of course.

One other thing: slva264 shows a 6.3V ceramic capacitor on the output of a 5V supply.  Do you think that is enough margin between output voltage and part rating?  It seems like that could cause a 15% drop in capacitance right there.  These parts are big and expensive so I thought I would ask because the 6.3V parts are cheaper but might degrade too much for anything other than lab testing.

Large feedback resistors cause more noise to couple into the high impedance FB pin.  You are not much higher than the 200k recommendation, and assuming you have a tight board layout with those resistors close to the FB pin, there shouldn't be a problem.

And yes, a 6.3V ceramic on a 5V supply is risky.  I recommend a at least a 10V rated ceramic capacitor.

Hi,

I have one other question about this chip.

When the enable pin is low, is it safe to assume the output is isolated from the input?  Can I attach another 5V supply on to the output for other testing or will it get back to the input through some internal diodes or anything like that?

Thanks again.

Hi,

When EN = Low, the output is totally isolated from the input.

I tested this by connecting a 5V supply at the output (and EN = LOW), No voltage gets back to the input.

I hope this answers the question.

Thanks.