LM2576ADJ

Hi Bento,

We've received your question and will get back to you shortly. 

Thanks,

Anston

Hi Bento,

LM2576 is an acceptable device for implementing the 5V @1A rail. Keep in mind though that the voltage drop across the switch is comparatively high by today's standards. Newer devices have a boost capacitor to provide a voltage source to the top side switch during the switch ON-time, thereby reducing the switch voltage drop. 

There are two issues with using LM2576 for + or - 18V. Since, your 24V input rail is from a battery we assume you would like to extend battery run time as the battery discharges while maintaining regulation. The LM2576 will hit drop-out fairly quickly in the 24V to +18V application and as you pointed out in the -18V application the abs max rating of the device is exceeded.

May we suggest that for the + & - 18V rails you use LM3481 configured as a SEPIC for +18V, and a Cuk for -18V. The Cuk converter will require an additional Op-amp to invert the output feedback signal. Both of these converter stages require two winding transformers but they can use the same part for the two applications. At this power level, you will have much better success with coupled inductors rather than standalone non coupled inductor pairs.

Thanks,

Anston

Hi, I looked on datasheet and didn´t have application with CUK converter for lm3481, the site sugest LM5576 for inverter converter, what kind it´s better, do you have application with CUK converter for lm3481?

Airton.

Hi Airton,

We are working on a schematic for the Cuk converter. We will get back to you with a design very soon.

Thanks,
Anston

Good morning Anston!!

In the first case I use lm2576 for 5V  1A what new component it´s better, you talked there are new component for this application.

Thanks,

Airton.

Good night Anston!!

I received the samples and I did the board with lm3100 --> 5V@1A - ok, LM3481 -> 18V@1A - ok, you sent me an schematic with cuk converter with lm3481, I did and without load the out is -->  -18V, but when I put @500mA the voltage decrease to -14V, the schematic it is atached.

in this schematic the VCC and GND of the LM321 Was wrong, the Gnd it is 2pad and VCC it is 5 pad, but the rest it is ok.

Hello,

I haven't spent much time on this schematic, but before I do, I wanted to suggest you to reduce the value of R5. You are using a 33mohms resistor. Have you tried to calculate its value and arrive at 33mohms? Would you try to see what response you obtain at 20mohms? 

Also please send us your board layout.

Regards,
Akshay

Good morning!!

I reduced the R5, but didn´t improve, I didn´t calculate the R5, I use the same that simulator did to +18V. I will send you my board layout and the complete schematic for you analize.2134.ESQ.FTMPCI2014.pdf5468.LAY.FTMPCI2014.pdf

Could you see my schematic and layout board ?

Thanks.

Hello Airton,

A couple of reasons when a particular DC-DC circuit works without load and doesn't when loaded are that the input supply is current limiting and is not able to provide the power for the load or that the layout of the circuit is not optimized and the resulting noise is interfering with the operation. If there is noise on the sensitive nodes such as the current sense, FB or COMP then the IC could be false triggering protection. Are you able to drive currents lower than 500mA at all? Is any load possible at all? You should probe the CS pin, SW node and the inductor currents for more information.

I just got a chance to look at the layout. Have you hidden the copper pours from the layout or have you not used them? For a SMPS layout, you should have flooded all the unused area on the board with copper and also have enough copper pour as heat sinking for power dissipative components such as the MOSFET. If enough copper area is not provided, the FET's RdsON will not be the same as mentioned by its datasheet but higher.

If the layout is not done properly, the resulting EMI can interfere with proper functioning and cause the design to not behave properly. The high frequency switching current is carried in the loop created by R5, U2, C5, D1, and C6. If this loop is too large, then the noise can interfere with other nodes. Noisy loops as that should also be run on the same plane. If they are dropped through vias to different planes, then the additional inductance of the vias can add to the noise. I am attaching a presentation on good layout techniques. I'd suggest that you look at the eval board layout for the LM3481 SEPIC EVM and take examples from there. 

I hope this helps.

Regards,
Akshay

Hello!

I did tests and the result it is atached. I put too the layout board per layer. 7635.SINAIS CAPTADOS NOS TESTES.zip5810.BOT.FTMPCI2014.pdf6622.TOP.FTMPCI2014.pdf7776.TOPSILK.FTMPCI2014.pdf

The best result was when we change the R3 to 200kohms, the frequency 100KHz, with this values the out = -18V, but with noise. Can you see the signals atached, I Am looking the file that you sent me.

Hello,

If you change the switching frequency, then you have to change the BOM along with it. Otherwise the design is not going to be stable. Are the signal plots after you change the switching frequency? They look like the system is not stable at all. Does lowering the frequency allow you to apply a load? You haven't answered any questions from by previous reply.

Regards,
Akshay

I change the frequency and  changed of the inductors l1=150uh and l2=82uh, this signal its after the change and with 500mA resistive load, but when i put on with the real load, doesn´t work, it seem the protection act.

If you need more information I send for you.

Hello,

Did you also change the compensation components along with the change in L and Cout? It looks as if the design is unstable. You could use Webench again as you did before to obtain the new values for compensation. Also, please send me your updated schematic.

Regards,
Akshay