TPS55340: okj

What happened to applications engineers who used to support a part?  

Hi Robert,

Sorry for the delay. Can you provide the schematic that you are using, as well as any key waveforms that you have? Ideally, it would be good to have VIN, VOUT, COMP, SS.

Regards,

Richard

Can you tell me how to attach a Jpeg and a PDF from my PC hard drive to this thread. I sent them in an Email but apparently that didn't get noticed? I dont see an insert file path in this UI. 

1700-67 v157.pdf

The schematic appears to have been stepped on6560.1700-67 v157.pdf

I took a picture of the power supply part of the schematic

Finally what a pain. 

Hi Robert, 

Thank you for providing the image. Can you clarify what trace is what on this plot- this will help make the image easier to understand

Regards,

Richard

The yellow trace is Vin its about 5 volts in this situation.  Its 4 batteries in series. The Green trace is Vout. The design is set for a Vout of 15v. as you can see when I turn on the power supply it over shoots to close to 20 volts.  We wait a bit before switching the load on and you can see when we do this the distortion on Vout and Vin. Vin is dropped to below 3volts. you can also see when we turn on the switcher with no load the Vin is also upset. The load is an Ultrasonic sensor that draws about 50ma but has an inrush of about 250ma for a short duration. This 250ma is from charging capacitors in the power supply of the Sensor.  Sorry but I have written this description several times. I initially sent them in Emails and didn't realize they were no reply Emails. I have been frustrated as this support request is several weeks old. I had trouble posting in this forum initially. I talked to a customer support person and also shared all this information. That person could only tell me to post in this forum which I thought i did.  Can you see my schematic?  I cant seem to post the pdf here. Can I email it somewhere so you can see it if you cant see it here?  

Hi Robert,

Thanks for providing information on the traces. The PDF is uploaded. I apologize in advance, the team is undergoing some changes so responses are delayed. 

Regards,

Richard

Hi Richard,  Sorry this is not directed at you personally. I have received Zero input on this issue. What ever is going on at TI I cant say. but what I can say is that to date customer service is nonexistent. This is despite many interactions with a customer support person who told me as soon as I posted my issue here I would get prompt attention.  What if I had 1000 parts in manufacturing?  

Hi Robert,

I'm sorry for the delay. 

You may try two things to debug this:

The first is increasing your Rcomp value to 2.5kOhm in order to speed up your transient response. 

The second is to do an ABA test to see if you can replicate this issue which will tell us if this is specifically with the IC or with the circuit.

Please let me know how the two methods work. 

Regards,

Richard

Not sure what an ABA test is. I have several examples showing over shoot and poor transient response. They all show this and and as my battery voltage drops the transients eventually confuse my processor before the battery voltage is low enough for us to put the processor in conserve power mode. Once the processor is confused it gets stuck in a mode where it draws 120ma and the system dies. normally we draw 100ua in conserve power mode

I have tried 976 and 1.74k for Rcomp and both act the same. Web bench gave me those values for .5amp and 1 amp load currents (You could check). Pretty sure 2.5k will do little but I will try it. This brings up why does Web bench give those compensation values? it even simulates power up and load transients which look like what I wanted.  My feeling is the math in Web bench for this chip is wrong. I have not done the math by hand so perhaps I should. The capacitor value and the other resistor all work together to tune this part. It would seem in order to figure this out empirically you would need to vary all three to really understand this. these are all 0603 parts, super fun to vary and everyone has a complete set of 0603 resistor and caps to play with.   The soft start is suppose to dampen the inrush too. You can see its not doing that either.  Something is not correct here. Since web bench takes seconds to generate this design and I don't have access to the algorithms behind the curtain. It would be very nice if someone at TI would take a few minutes to look. I have been waiting weeks for help. I have a small company with a design in limbo because we used your tools to design in your chip and it doesn't work right.  Perhaps its a layout issue? I tried to follow the recommended layout. Its interesting you have not asked about that yet. I would dump this part and might have too for lack of support. It means we need to spin another board and have another 10 built which costs 1000's of dollars as well as the schedule impacts.  It would seem like you could get someone's attention at TI

Hi Robert,

Webench model development is based off of the EVM and experimental data. I would suggest to use Webench as a baseline, and then provide the design file for review here if you have questions. 

The reason I asked you to perform an ABA test is to check if your TPS55340 unit has been damaged once the battery has fallen below 2.9V. In order to perform this test, let's define the unit with the confused tuning as unit A. I would like you to take this bad unit on a working board with a working unit, and swap the two units to see if the issue carries over to the working board. If the issue carries over, then unit A is at fault. 

Then, taking the working TPS55340 IC on the working board, which we define as unit B, place it on the board showcasing the issue. If the board works, then we have further proof that your design is working as expected. If the board doesn't work, then we can conclude that there is something wrong with the board. 

If you have questions about the layout, please provide the design files and I can take a look. 

Regards,

Richard

Hi Richard, I don't currently have anything that works that much better. Besides that removing a SM chip (or two) and swapping them between PCB assemblies is problematic. In my experience taking off one that is suspected as bad and replacing it with a new unused part is much less risky in terms of actually learning if the part I removed was different or bad. In this case the part I used is a leaded part but it has a large ground pad underneath. I have no way to de-solder that  pad other then a hot air removal tool which risks de-soldering all the 0603 passives in the general area. I could maybe send out this assembly to a more capable surface mount rework shop if I have too but that takes a while

Again I have a supply that starts and runs. It sources current. All of them do that. Ones that appeared to work better still show over shoot and large load transient disturbances. This is why I though tuning was the issue. I have several assemblies that never saw low voltage conditions and it still shows less then expected performance in terms of overshoot and transient response. When looking at them as a group they just appear to be tuned wrong. 

This board was designed in fusion. I am looking at it right now and wonder how to show it to you. Its a two layer PCB with a bottom layer that's all ground. 

It occurs to me that TI some times makes demo pcb supplies. Perhaps one exists in this case and I could get one and apply the same load to it and see how that one works. If the demo supply operates with no over shoot and no transient response then I would have to lean tward my layout causing the  overshoot and poor transient response. 

My requests for inspecting the tuning part of web bench seem to have been brushed aside. It seems like the simplest answer but one that would require Ti to participate. I have used that tool to design many switch mode supplies and everyone always worked. In fact every Product the company I work for has a web bench switcher. I think there are 5 or 6 in production.  I come from having to do all that long hand so when it became available I liked the time it saved. I never have altered tuning from a web bench design in the past. The tool sources not only the schematic but also a parts list. If I was asked I would have guessed that all the formulas for calculating the values where just like the ones in the data sheet. They are all there in the data sheet. 

I am going to go back to my roots and use the data sheet to figure out what those parts are suppose to be using the formulas provided. If I get much different values I will try those as removing a 0603 part is within my tool set. 

Hi Robert,

I understand your situation better now. If you still see the overshoot and large load transient disturbances, then this may be an issue with the board design and layout. In this case, we do have an available TPS55340 BST EVM: https://www.ti.com/tool/TPS55340EVM-017

Let me send you a friend request so that we can discuss how to get you this EVM to you. 

Regards,

Richard 

If you can send me a evaluation board that would be great. I work for Senix Corporation.  10516 Rte 116 #300, Hinesburg, VT 05461

Send it to my attention. 

Hi Robert,

Will do. I've sent you a friend request, so if you could also provide an e-mail, that would be great!

Regards,

Richard

 I got no friend request. My Email is n my TI profile but here it is again gifford@gmavt.net

Well its nice you think this is resolved. Its not. TI basically gave me no help other then basically telling me I am on my own to figure out why web bench gives compensation components that result in such poor performance. Thanks for not telling me to reset my PC and trying again. What a joke.  

Sorry that Webench isn't working the best for you. I've gotten in contact with someone on the sales team, with you cc'd to see if we can give you a sampled EVM. Were you able to receive it?