LM5175: Output catastrophically not working.

Hello Diana,

What is the minimum input voltage that you want to support with this design? At the moment the converter starts at 4.7V.

You need to update the compensation to fit your requirements. At the moment, the converter is unstable for voltages below a certain level.

Please use the quickstart calculator for this and set the voltage on the Bode Plot to the minimum input voltage: https://www.ti.com/tool/LM5175QUICKSTART-CALC

Best regards,
Brigitte

Good morning Brigitte,

Thank you very much for your feedback. I changed the com network as you suggested. I changed the res to 7.5KOm, and the parallel capacitor to 120pF. I had this configuration working before for me on a different design with this supply that works from 10V an avobe. I still got the same results. I also changed the EN resistors to match the previous design. 

Based on what I have read in the forum I am going to take measurements for SS, SW1, and VOut simultaneously, as well as CS, FB, COMP, SS simultaneously. I want to see what is faulting first.

I will double-verify that the board was fabricated properly, and get back with additional clues later today. 

Regads,

Diana

Vin = 21V with updated Com Network:

Vin = 18V

Vin = 14V, this one is the closest it gets to working:

At 15 V is the only"sweet spot" at times: Vout was not connected:

Vin = 19V Sw1 start osc

same, just zoomed out

VCC on the new design (not working-CHannel 1) and the old design (working, channel 2)

Hello Diana,

Would you please share which components you changed from the old to the new design?

Did you measure above problems with load connected or without load? If the load was connected, please check the behavior without load.

The last pictures on the failing board seem to show hiccup mode. Which would be correct as the output voltage is too low.

How many boards do you have that show the same behavior?

What happens if you take one of the new boards and change it back to the old design? Does it then work or are there still issues?

Best regards,
Brigitte

Good morning Brigitte,

Thank you very much for your feedback. 

These are the changes from the old to the new board (for increased current and environment temperature):

Most FETS.

COMP, which I already changed back since 3/15 (and will stick with old version)

SS cap from 47nF to 22 nF. I will change that one next.

Cslope I have 82pF now, I haf 100 pF before.

The diode is a big chage the one in paralle with LDRV FETS.

Sense resistor changed from 5m to 7m.

All other control signals are wired similarly, through the same Caps, resistors , and diodes.

For all captures the power supply was not loaded.

I will try your experiments. Thank you very much.

I will get back to you with more results later.

Regards,

Diana.

Hello Diana,

Please remove the diodes D6 and D8 for testing if they cause some problems as well. And please replace the gate drive resistors with 0 ohm ones for testing as well. It is possible that too big gate resistors cause shoot-through.

Please check the current sense signal and the COMP as close as possible to the IC as well.

Best regards,
Brigitte

Hello Brigitte,

 I ran the experiments you suggested, besides increasing input capacitance closer to the actual part, and I gained some range towards the lower side.

I have solid Vin=14V and Vin =11V performance, then as low as 11 Vin and the other values up to 17V is noisy, but it kind of works. From 18V input and forward it goes into drastic hiccup mode, almost not drawing current.

I wil post the images later, have to transfer them later tonight.

Diana

COMP at 14 Vin (400 mA) :

COMP at 12 V (current oscillates between 80 mA and 400 mA)

COMP at 19 V (little current drawn 80 mA)

Hello Diana,

Thank you very much, so I wonder if the circuit is creating noise issues causing the device sensing wrongly. Could you please check the current sense signal from CS to CSG?

And please summarize which tests you did and which result you got with these tests. This might give hints to the direction we need to go.

Best regards,
Brigitte

This is a differential measurement across the capacitor between these two pins. The cap is located very close to the part, as well as the sense resistor on the two boards pictured below.

Both boards were measured at 14 V input. The first one is one of my older boards, the second one is the new board. None of them is loaded at the output. 

For the newer board (second picture), the sense resistor is connected to a bigger plane that feeds the source of the two transistors. That may be causing some noise issues, I guess.

Also, for input voltages above 17 Vin, the picture looks similar, just intermittent, with periods of lower noise (more like the good board).

Finally, I also noticed the current on my new board (bad) is closer to 400 mA for the better working range (more stable Vout) of Vin (14 V to 17V), however, the current on the good board, under the same conditions, is closer to 27 mA (unloaded)

Thanks,

Diana

Hello Diana,

400mA input current at no load is way too much, even for CCM.

Can you please check if the differential pair on CS and CSG is really routed differentially? In addition, I notice that you are not using a separate analog GND for the sensing parts, so possibly the root cause of the issues is the layout.

I think it would be an interesting experiment to take an EVM (or one of your existing working boards) and change it 1by1 to your new design and check if and when it gets into trouble.

Best regards,
Brigitte