LM5168-Q1: LM5168P layout review

Hi Jaroslaw,

Could you please update if the suggestions [given in previous post] has resolved or not? I dont see the modifications in the schematic.

I will look into the layout and get back to you?

Hi
Schematic change from previous thread didn't work - I wrote this in response to that suggestion. Changing injected ripple doesn't resolve issue.

All layouts are actual, we have 3 versions of boards with 3 very similar layouts of DCDC converters modified to fit on different PCB's, but your suggestions in terms of inductor and feedback lines placement are appropriate for all of them and I can agree with them.
It must be taken into account that feedback is shielded with 2 layers of copper and inductor is in shielded version.

I've tested it:
- pulled inductor up so it's further away from feedback
- original switching node is now cut off and connected to ground - so feedback lines cannot be affected by inductor now
It didn't help, DCDC behavior is still the same - sleep doesn't work at lower voltages. DCDC overall stability is working ok, switching looks ok.
Board looks awful because it was covered by isolation layer of plastic - it's hard to remove completely.

You must take into account that we have quite a lot of that PCB's, and around 93% of them are working without issues. Now I'm trying to figure out what's wrong with rest of them. Also, on the beginning we assembled 9 boards for tests (3 of each layout) and all of them were working fine.

One more thing that came out - touching frequency setting resistor Rt (R311) or net that it's connected to (pin-4) also causes DCDC to work ok and brings back sleep function. (touched with needle connected to human body being far away from feedback components). So it makes the same effect that touching feedback - it's the same on modified and not modified boards.

This is holiday in TI India..Kindly expect a delay in response..Will check and get back to you by tomorrow 

Hi Jaroslaw,

1. Even if you move the inductor up but still the TP319 to C317 track is crossing SW node cu so it is prone to SW node noise I think.

2. I think it could be due to soldering issues also, may be the ground is not properly connected. This makes sense to me as you clearly mentioned some of the boards are working fine and some are not and pressing down the components helps.

Regards

Arpita

Hi
1. I cut off inductor pad and all switching area and connected it to ground, you can see soldered area on the photo, so it shouldn't affect feedback that way.

2. Also thought about it, but after removing DCDC converter and placing it again nothing changes, tried it on 2 or 3 PCB's. While removing I can say that they are soldered perfectly and thermal pad is fully connected to PCB.
Replacing IC with different one solves the issue, looking like IC's are the problem here.

Regards

Hi Jaroslaw,

As the ground layer also goes below the inductor (even at the modified setup) 

Due to the eddy current in the ground layer, the inductor value decreases and the loss increase with set-off effect from the line of magnetic force. Signal line other than ground also has the possibility of propagating switching noise caused by eddy current. It is better to avoid wiring directly under inductor. If wiring is unavoidable, I would recommend using shielded inductor. What kind of inductor are you using?

2. As touching external components with needle resolves the issue, the issue very likely due to layout or improper soldering. It seems to be some isue with ground connection as explained above.

I would recommend testing the IC in TI EVM and check if it resolved.

1. Ok, I get this point now. Inductor is shielded, part number is: IFSC3232DBER221M02
2. Also thought about buying LM5168PEVM and modifying it with our parts, probably will proceed with that

Hi

I have an update - bought LM5168PEVM eval board and reconfigured it to be similar as our project, components replaced:
R8 - 62k
R9 - 62k
C4 - 2n2
L1 - 220uH (IFSC3232DBER221M02)
U1 - DCDC converter taken from problematic PCB
rest of components stayed without changes 

Problem with sleep still occurs in the same way, once again touching feedback nodes solves issue.
In my opinion it confirms that layout is not the issue since the problem is identical on TI eval board...

Hi Jaroslaw

Have you checked with any other fresh ICs? Do they show the same behavior? 

Hi

As I wrote at the beginning, replacing IC solves the issue, and only about 7% of assembled PCBs have this problem