LM5085 misses some pulses

Just another info: I replaced the mosfet to a smaller one (19 mOhm 3300pF). The only result is the higher mosfet-temperature, nothing else.

Hi Bela, did you solve this problem?  I have a similar problem with a much lower current design with 8 dc/dcs on a single pcb.  I believe the issue is noise causing the LM5085 to go into current limit shutdown.  This causes large dropout spikes on the output voltage cap.

Hello Jesse,

unfortunately, the problem still exists despite the several new layouts I have drawn.

My experience is that the noise-feedback from the switching point to the reference input can be critical and the filtering of overcurrent setting resistor is also critical. One should have to be very careful, where these filter capacitors are grounded. I use a ground plane on an intermittent layer, but a few millimeter here or over there still matters.

But, again: I couldn't find a stable, producebar solution. I'm geting to hate this IC.

Hi,

I am trying to understand your circuit. What is the driver IC for? Also what is the GND of the driver IC connected to.

Are the inputs common to all 4 circuits? What kind of load do you have?

Dear Mr Choudhary,

in the last half year I had to change my schematic a bit. I attach a newer version, please consider it as the actual circuit.

U22, Q14 and the components around serves as a security system. This circuit should only work, when a heartbeat pulse comes, otherwise it should always be off.

U25 driver is necessary, because Q12's input capacitance is too large, U20 alone cannot drive it. At the beginning, I didn't have a driver, and U20 got extreme warm in a fraction of a second.

R143 is wrong on this schematic, it is acually only 1 Ohm.

The GND of the driver IC is already on the schematic, it is always 8V below the input supply. This driver IC is just a complementer emitter-follower. Other, dedicated mosfet drivers doesn't work with LM8085 because they all have a delay the controller cannot tolerate. This driver is quick enough.

The input is common for all the DC/DC converters on the PCB. We had to separate the GND's and we had to employ input filters (L6-C50-C62-C49) due to EMC reasons.

The load is purely resistive, actually we're heating up some iron parts in out instrument. The heating is controlled through the heartbeat pulses (on-off control).

I also modified the feedback circuit. This way it reflects a bit higher stability (at least it works somehow, the oricinal circuit from the datasheet is so unstable, it is unuseful), but isn't stable enough to be released.

C66 goes to GND instead of the input voltage, it also improves the stability of the circuit.

C53 is phisically close to U20 (few millimeters), C60 is phisically close to U25.

The layout is made on a 4-layer PCB, each of the 4 circuits has their GND plane, also, the input voltage is a plane. Both are 150 um dick.

I'm trying to make some scope-pictures, if I can, I will post them here.

A comment: C47 and C55 should connect to VOUT and FB. Please double check the ripple circuit with the application schematic in the datasheet.

Well, it's been the original version.

Unfortunately, I couldn't reach a stable operation. Since C55 goes to GND and it is 100nF instead of 22nF, and C47 is reduced to 22nF, sometimes it is stable (at least 2 of my 4 converters are working stable).

The other 2 circuit got also stable yesterday, by changing R66 from 10k to 100k.

I don't understand th esituation, these 4 circuits are exactly the same, even the layout is the same. The only difference is the load, it is between 6A and 13A.

Also, changin C66 from 1nF to 100nF and tieing to GND instead of the input rail improved the stability.

I still have constraints if this circuit can be released to production (but it should be, deadline is over).