dc dc converter 12v to 48v. High power.

Hello Gennadiy,

I would need more information to refine the recommendation.  Not less than 100 W is vague.  Exactly what ripple voltage do you need?  What operating temperature range do you need - consumer, industrial, or automotive?

The output ripple of any switcher can always be reduced with a second-stage LC type filter.

Based on the limited information you provided today I would take a look at the UCCx808A-x, and the UCCx808x push-pull families.

Regards,

Eric

OK.  Avoiding a transformer is an important requirement.  I think a BOOST topology makes more sense for your application.

I used TI's online WEBENCH designer with the following requirements;

VIN = 8V to 18V (more realistic for automotive), Vout = 48V, Iout = 2A (96W).

WEBENCH produced a design based on the LM51561-Q1 controller.  The PDF report is attached below.

48Vo, 2A Boost.pdf

The report has quite a bit of detailed information:  Schematic, BOM, operating conditions, efficiency, etc.

I hope this provides a good starting point.

Regards,

Eric

Hi Gennadiy,

The UC3843 or UC3843A can work for you.  They have Vin UVLO Start/Stop thresholds of 8.4V/7.6V.

The output voltage depends on the feedback resistor divider from Vout to the FB pin:

The first page of most data sheets show the maximum switching frequency:

TI has a nice power design tool, POWER STAGE DESIGNER.  This tool analyzes the power component requirements and can be used with many different topologies.  You should watch the training video.  Choose BOOST and enter your Vin, Vout, and Iout specifications.

Regards,

Eric

Hi Gennadiy,

The UC2843AN is the same controller, just higher temperature rating.

Usually, customers use a medium frequency, like 200 kHz (not MHz).

For any regulator, the ratio of the feedback resistors (RFB1 and RFB2) depends on Vout and VFB:

Regards,

Eric

Hi Gennadiy,

For the boost diode a schottky would be the best because it will have the lowest forward voltage drop and dissipate the least amount of heat.

However, the boost diode must be rated for at least Vout_max (55V in your case).  It may be difficult to find a schottky rated this high.

Regards,

Eric

Hi again,

I used Power Stage Designer to calculate the RMS current in the boost diode, the FET, and the inductor with your typical operating conditions.

The following data will help you choose components.

DIODE:

MOSFET:

INDUCTOR:

You should download and use Power Stage Designer.

Regards,

Eric

The secondary inductor (L2) can be rated about 3A.  No need for 15A in that position.

Connect both diodes in parallel.

Your input voltage is relatively low.  For the R/C values to VCC I would start with 1 or 2 ohms and 1uF to 2.2uF.

Calculating compensation components (FB to COMP) is one of the challenges of a boost regulator.  Here is one application note from TI that might help:  https://www.ti.com/lit/an/slva636/slva636.pdf?ts=1654802983383&ref_url=https%253A%252F%252Fwww.google.com%252F

There are many other articles on the web about boost compensation.

I'm sure the diode needs to be on a heatsink.  The MOSFET probably should have a heatsink too.

Regards,

Eric

So, if you could (eventually) find some LM5122 controllers you could replicate this reference design from TI.  The following link shows test results, schematics, BOM, layout files, gerbers, etc.

https://www.ti.com/tool/PMP21274

This is a synchronous boost (no diode).  It significantly reduces losses and achieves much higher efficiency.

Eric

Hello Gennadiy,

R5 will conduct the same current as the MOSFET.  So the power can be calculated with ~(13Arms)**2 x 0.1ohms x MaxDutyCycle.  I think 3W should be adequate.  If you want to be conservative, have the option for two 3W resistors in parallel, but only populate one.

It is good to have options for a secondary output filter to reduce high frequency noise, especially for your application.  You may not need the filter but best to include it in the PCB layout.  I think 4x 470uF + 1x 470uF is more than enough.  Maybe place 2x 470uF on each side of L1 (4 total).

The specifications for the 470uF electrolytic capacitors look relatively good.  However, to address very high frequency noise (MHz region) it is best to include some ceramic capacitors too.  You should have some ceramic capacitance in parallel with both C5 (input) and C9 (output) and C7/C8 (output #2).  It's important to have the ceramic capacitor(s) located as close as possible to the controller.  See PCB layout recommendations below.

Is 0.1 ohm for R5 low enough?  The current limit will occur at 1V/0.1ohms = 10A, which is too low.  The peak currents can be at least 13A or more.  The value for R5 should be reduced to less than 75 mohms.  Maybe start with ~50 mohms.

You need to include an RC low pass filter between R5 and the Isense pin to remove current (switching) spikes.  I would start with 1k and 100pF.

Also, for peak current mode control with duty cycles above 50%, you must have a "slope compensation" circuit.  Without slope compensation current mode control is unstable above 50% duty cycle.  We typically use an NPN transistor as an emitter follower to buffer the RT/CT voltage and AC couple it into the Isense pin.  Here is what slope compensation and current sense filtering should look like:

Lastly, the error amplifier compensation circuit is not entirely correct.  R13 should be changed to a capacitor (about 100pF) to form a high frequency roll-off of the gain.

One final note.  PCB layout is an important part of constructing any switch mode regulator.  The following application note discusses the important aspects of layout for a boost regulator.  Pay attention to minimizing the loops around the input and output capacitors.

https://www.ti.com/lit/an/slva773/slva773.pdf?ts=1654869304360&ref_url=https%253A%252F%252Fwww.google.com%252F

Regards,

Eric

Hello again Gennadiy,

I entered your specifications into Power Stage Designer and used the Loop Calculator to obtain better compensation values.

Note that I used a 50mohm sense resistor (Rs), and 2x 470uF electrolytic (940uF) along with 5x 10uF ceramic (50uF) output capacitance.

Based on this analysis you should start with R6=250k (presently 6.81k), C3=15nF (presently 1nF), and R13=100pF (presently showing 100k).

At 10Vin and 3A load your crossover frequency will be 0.47 kHz with a phase margin of 85 degrees.  The gain margin will be 15.4 dB.  This is a very stable operating point.

Regards,

Eric

Hi Gennadiy,

It's OK to use 249k for R12.

I think you can remove C10 and C11.  C7 and C8 (2x 470uF) should be adequate.

Also, C5 is quite large (470uF).  To save some cost and PCB area I think 22uF or 47uF would be more than enough.

The output voltage ripple after the second filter should be very low;  10's of mV.

Regards,

Eric

You should remove L3 and C5.

The PWM switcher will create some noise.  Physically separating it from other sensitive components is a good idea.  Make sure you have adequate bypass capacitors at each downstream device (amplifier and microphone).

If you are not sensitive to cost, then having two power supply units seems to make sense.  Maybe you can have provisions for two independent power supply units but start by populating only one.

Two more things.  Whenever a switcher has noise issues we do two things.  1.  Add a resistor between the OUPUT pin and the gate of the MOSFET (include provisions for 1206 size resistor, but start with 0 ohms), and 2) Add an R/C snubber from the SW node to ground.  The following paper discusses boost noise.  See section 2.4 for placement and calculations for the RC snubber.

https://www.ti.com/lit/an/slva790a/slva790a.pdf?ts=1655142562228&ref_url=https%253A%252F%252Fwww.google.com%252F

Eric

Even if the snubber were impossible to calculate, you should include options for RC snubber, in case you need them.

I see one via in many places.  If you're connecting to ground or Vin there should be 3 or 4 vias at each connection point.

Make sure the Vin and Vout loops are small, like the layout guideline shows.

Regards,

Eric

Yes, multiple vias looks good.

You can start with 220 pF (100V) and 220 ohms.  Make sure the resistor is a 1206 size, or larger because it will dissipate a low/medium amount of power.

Eric

Using the ground plane is easiest layout.  If you place the ceramic input and output capacitors to minimize the loops it should be good.

Eric

Hello Gennadiy,

For a ceramic capacitor (which filter high frequency), it is best to have vias located at the capacitor.  Locating the vias at the capacitor minimizes the trace inductance and allows the ceramic capacitor to have optimal performance at high frequencies.

As for grounding, I found a very good reference on PCB routing of a boost regulator.  In the following article, pay attention to how they minimize the input and output power loops when placing the power components, and use a dedicated signal ground for the control components.  A dedicated signal ground prevents the power ground from coupling into the control.  This is good standard procedure for any switching regulator.

https://www.powerelectronicsnews.com/the-dc-dc-boost-converter-part-3-power-supply-design-tutorial-section-5-3/

Regards,

Eric

When you do the place and route do your best to make the loop small.  Don't place the diode or MOSFET far from each other.  If possible, connect the source of the MOSFET to the low side of Cout on 1 layer, then make the connection to ground.  Don't use vias in that loop.

You can start with R6=0 and increase the value when the PCBs arrive.

This thread is 2 weeks old and hasn't closed.  Please consider clicking RESOLVED and open a new thread if you need further assistance.

Thanks,

Eric