LM5085: Double switching pulse

Hello Norbert 

Would you please send the link of your WEBENCH design ?

-EL

Dear Eric,

here ist the link: webench.ti.com/.../SDP.cgi

Dear Eric, the only component that is different to the design is Cout with 10uF (Tantal).

Can the Cvcc cause this problem? I used a X7R 470nF Ceramic. I think in one application i have seen a polarized one?

Hello Norbert 

The link doesn't work....COUT affects the switching stability. VCC capacitor doesn't matter 

-EL

https://webench.ti.com/appinfo/webench/scripts/SDP.cgi?ID=AA5922D684B3DD66

it works on my pc..

Do you have an email?

Hello Norbert 

The link works now. Would you please share the schematic and PCB layout too?

-EL

Hello Norbert

C7//R13 should be directly connected to the positive connection of the current sense resistor. It is possible that the "short-on" pulse is caused by a noise injection to thi spoint. 

-EL

Dear Eric, in the first post i forgot to mention, that i had a 10R resistor populated on R14. Now i changed this in 0R and the double pulse moved in the middle of the signal. (as you can see below). I changed the circuit as you suggested, but there is no change in behaviour. You can also see an oszi-picture from the gate signal.

Hello Norbert

I guess you made a wrong connection. You have to connect the other side of RADJ 

Based on your experiment using 10R, I believe the issue is caused by a noise injection between ADJ and ISEN pin. 

I haven't tried, but I guess it is valuable experiment to check if any improvement after populating 100pF between ADJ and ISEN. 

-EL

Hello Eric,

the connection is correct. The unpopulated pads in front are Vin. Now the wire goes directly to Vin and the other side of R13,C7 goes to Pin1 (Adj).

I tried your suggestion with the 100pF - nothing changed. This is the oszi with populated 100pF:

I did some more measurement:

This is the output ripple:

and this is the signal at the current input:

Now i populated the ADJ-Resistor with 4.7K and left the 1n away, since my LED's are controlling their current by themselves. This means, i do not need the current control of the LM5085. (I used it only for the voltage range 18V to 60V). And now, at a voltage of 22V the spike is away and switching frequency is correct (160KHz), with 18V it is only 65Khz, which is also wrong...

22V:

18V:

Any suggestions?

Kind regards

Norbert

Hello Eric,

my connection is correct. The wire connects R13/C7 directly to +C4 and R18. I think a much shorter connection is not possible. As i use an aluminium pcb, i cannot put a groundplane to the bottom side.

The improvement is only at a voltage of 22V and up. From 18V to 22V, the failure is present, even with the 4K7 at ADJ. So, this circuit is not usable for me.

I think i have to set up a new layout, but i will not use the LM5085 again. What do you think, will the LMR38020 do a better job for this application?

Thanks for your help,

kind regards

Norbert

Dear Eric,

i did a new design with the LMR38020 which works very fine.

But now customer did emc surge test and the LMR38020 died. Customer told us, that they measured transients with 120V at the ic.

Do you have any suggestions to improve my input filter, so it withstands the emc tests? I think this is not the first application used in a range from 19V to 60V, that has to be emc test?

19V-60V-V02.pdf