LM5085 Input Voltage Gap Problem

Hi Seth,

Can you post a schematic for your design?

I am going to contact the applications team for this device.

Regards,

Marc

Thanks Marc.

I will wait for the schematic from Seth.

https://e2e.ti.com/cfs-file.ashx/__key/communityserver-discussions-components-files/196/7178.4237.LM5085WEBENCH.pdf<a href="/cfs-file.ashx/__key/communityserver-discussions-components-files/196/3858.7674.TEK0005.JPG"><img alt=" " src="/resized-image.ashx/__size/550x0/__key/communityserver-discussions-components-files/196/5025.7674.TEK0005.JPG" border="0" /></a><a href="/cfs-file.ashx/__key/communityserver-discussions-components-files/196/8640.0042.TEK0004.JPG"><img alt=" " src="/resized-image.ashx/__size/550x0/__key/communityserver-discussions-components-files/196/3463.0042.TEK0004.JPG" border="0" /></a>

I've uploaded the schematic from Webench and three scope plots.  The first scope plot was with an input slightly over 20VDC.  I did verify that the incoming power supply was smooth and did not have any ripple. Between the voltages of 18V-36V, the supply is unstable.  As you can see by the first scope plot it starts injecting a second pulse.  This effect continues with more pulses until about 36V and the device just pops back in to stability.  VERY odd....

The last two scope plots are at 15V and 40V.  It's weird that the mid-voltages are when the supply goes unstable. 

Any thoughts?  I'm scratching my head.  I've designed numerous switching supplies and have never had this problem.  Could the IC's (LM5085) be bad?

Thanks,

Seth

Also, I've found that if I touch the device at the point where the transistor/inductor/diode all come together it will sometimes pop back into stability.

Also, the schematic from the datasheet has a some more filtering that doesn't show up in the WEBENCH schematic.  See attached.

8228.5810.LM5085DATASHEET.pdf

Looks like the schematic didn't post properly.

6710.1263.LM5085WEBENCH.pdf

Another screen shot of the OS with a 34V input.

The design is completely unstable at this point.

Somebody from Webench team needs to look at this to figure out  why the ripple circuit is not shown.

Continuing the discussion otherwise.

Is RADJ connected?

What is the load in all these cases?

What is the frequency of operation? The webench schematic is not very clear. Is it possible share a clearer version?

The load at the time was pure resistive @ 1Amp.

Here's the webench report.  I had to re-run the design tool. It gives me the same circuit but slightly different values.  2502.1602.WebenchReportsLM5085.pdf

Yes.  Per the webench recommendation.

Radj is connected and the frequency of operation is right around 360kHz.

Hi,

I used the quickstart calculator (excel based) on the LM5085 website and it predicts ~450 kHz operation with RT=59k ohm.

The previous Webench analysis showed the frequency at 330kHz with Rt at 64.9k Ohm.  This is what I used for the initial build and the actual frequency came out to be more like 460kHz.  The hand calculations based on the datasheet equations for Rt showed that it needed to be between 70k Ohm and 90k Ohm.  Since the frequency didn't match the Webench prediction, I raised Rt to 82k Ohm which brought the frequency down to around 360kHz.

The problem depicted in the scope plots was the same and this has had no effect in changing the charger circuit behavior.

The behavior can be related to:

1. insufficient ripple at the output. But this tends to get worse at lower VIN. Not the case here as you mentioned. But still worth checking with higher resistor or smaller inductor to increase the ripple current in the inductor.

2. Current limit mistriggering. This is possible. But difficult diagnose as this could be related to layout.

I have a board that I'm experimenting with.  Should I disconnect the current limiter?

You could try to increase the RADJ to 2k or something close to that in the user guide.

Vijay,

I've increased Radj to 2.1k and this did not fix the problem.  It made it a little better, but did not fix the problem.   I also tried to input the t-filter depicted in the data sheet and it made it really unstable.  For a device that is advertised as not requiring any compensation, this thing sure is unstable.

All of these components were recommended by Webench and I followed the layout recommendations in the datasheet.  But, to no avail.

Any further ideas? 

Thanks,

Seth

Vijay,

I have another clue.  In reviewing the datasheet, the design requires a Bypass Capacitor on the Vin.  The Webench output didn't include this capacitor nor did my design.

I'll be back in my lab tomorrow night and will experiment by adding a bypass cap to the protoboards.

Seth

Input bypass cap close to the IC is a must. Good find.