Shut down of LM3481 not working

Hello Akshay,

You are completely right. I overlooked this fact.

I will have a look at the SEPIC design. There a capacitor blocks the DC path.

I had build a SEPIC before and posted a question about the FET getting so hot. never understood why. maybe it needs a small gate resistor? I started with an FET in TO-252, but that one desoldered itself from the board. so lots of power dissipation. 

I replaced it with an TO-220 package and it needed an heatsink. quite large. never had an answer what I could do to lower the dissipation. All lines in the design were as short as possible. also, the design worked wel irt the functionality. but the dissipation in the FET was too high. 

hope you can help.

kind rgards,

rob van vreden

Hello Rob,

We can help with your SEPIC design. The main factors to large power dissipation are:

1) not enough heat sink on the board under the FET. 

2) very high switching frequency. 

3) chosen FET has larger RdsOn than allowable. 

You could attach your schematic with the BOM and the layout images and hopefully we could give some answers. 

Regards,

Akshay

Hello Akshay,

The design parameters in webench were:

input Voltage min8Volt/ max 35V

Vout -=24 Volt at Iout 2Amp.

I used L-in 10uH and Lout 56uH.. C sep I made with two capacitors of 2x2.2uF

The working frequency is at 550kHz

The Fets I tried: AUIRF1010EZ MOSFET,W DIODE,N CH,60V,75A,TO220AB on a large heatsink

and

AUIRLR3636 on a moderate copper area. surely not enough to handle the dissipation  but the fet desoldered itself in a few seconds.

all the values as you derive from webench. the setup works ok, except for the too hot fets.

Probably the board layout is not to good.

one other thing I had/suspected is that the current limiter was not working as is should. I expect it to limit arround 2 Amps, but one could draw much more than that from the design. I decoupled the I sense at best, but the current was still too high, to my opinion. 

I had to raise Rsense to 4x the value as assumed in Webench to keep the current somewhat neat the 2 Amps. For me an indication that the setup was not working ok in some area's.

Hope this clarifies some bits. on the schematics you see also a current cut out. it was not connected at the time of testing.

I use the LM2587 at this moment as an substitute of the design above. It works starting from 4 Volts and delivers easily 2,5Amps at 24 Volts. It works flawless. but it is not the route I want to follow. I want a more modern design.

kind regards,

rob van vreden

Hello Rob,

Your board layout needs to be revisited. There are quite a few items that you could change and would benefit from those. I will outline some basic stuff here and by tomorrow, attach a presentation that our group has made that can help you understand the basics of SMPS board layout. Firstly, are the FETs that you used from what Webench actually suggested? It seems to me that the FETs have fairly high gate charge and transient times. At higher frequency, these large values will cause more dissipation in form of the switching power loss.

1) The switch node should not be long and thin. It should be short, fat and just about enough in area. Your switch node is very long and thin and this would make your design very noisy.

2) You should not have relief connects in your layout. They add small amount of parasitic inductance and also reduce your effective thermal performance.

3) Your ground plane is not quite optimized. The distance between the FET ground and the output cap ground is very large. They should be as close as possible.

The current limit in the LM3481 is set by an internal comparator. The trip voltage is about 160mV typical. The sense resistor is calculated assuming 2A is the allowable load and about 20% over it to trip the current limit. This is why you were able to draw more current out of the circuit. Webench would not know if you want 2A as your current limit. If your typical load is lower, you could try to recreate your design using that value and surely you should get a higher sense resistor value.

I hope this helps. 

Regards,
Akshay 

Hello Rob,

I apologize for the delay. Here are the files.

Regards,
Akshay

2570.PCB layout and EMI mitigation.pptx