Desiging a dc to dc boost converter

[ctiger]

Hi all,

I am trying to design a dc to dc 12V to 24V converter for a A/C unit and the converter need to be able to output of 40A rated under full load, the Vin is 12V and Vout is 24V. The power is 960W, I'm looking at different chips at LT demo circuits http://www.linear.com/designtools/software/demo_circuits.php  and https://www.maximintegrated.com/en/products/power/switching-regulators/applications/industrial.html

I could not seem to find a chip to able to handle 40A, do I need a current limit switch or a current amplifier to have that 40A rated. I think I need a heat sink for the circuit but I do not know what kind of heat sink for the heat dissipation on the circuit.

I have the duty cycle of 0.5, P=960W, Vin=12V, Vout=24V. Please do let me know if any information is needed and I will provide as much to my knowledge

I am new to designing a converter any help or advice would be much appreciated.

[David Hess]

Pulse width modulating 80 amps on the input is not going to be much fun never mind the ripple current rating of the capacitors.  Using multiple phases with separate switching transistors, inductors, and capacitors is one option and they make switching regulator controllers which support this.  I might just use a high frequency inverter instead so the output voltage is fixed by the transformer ratio and the output tracks the input; this has the advantage of much lower ripple current.

[rx8pilot]

This project sounds like a very poor candidate to learn the delicate nature of power electronics.

Personally, I would go with a polyphase topology best deal with higher currents.

Short and misplld from my mobile......

[jbb]

The power is 960W, I'm looking at different chips at LT demo circuits.

I am new to designing a converter any help or advice would be much appreciated.

960W is quite a lot of power.  For example, that's going to be like 100A at the input (considering efficiency and a 10V input).  I think that you should warm up on some smaller stuff first.

Regarding topology
I totally agree with David Hess and rx8pilot that a multi-phase solution would be required. Multi-phase is very much like 'strapping some smaller ones  together,' only you need a little extra circuitry so that they all synchronise with each other.

In addition, you will need external MOSFETs (perhaps several in parallel) and synchronous rectification.  Something like the LT3784 might be a possibility. Check out page 34 for a 24V 20A circuit.  It's a 4 phase interleaved synchronous buck converter.

Some numbers
As a quick ballpark for the 24V 20A circuit, you would aim for 95% or higher efficiency, so the total losses would be somewhere around (480/95% - 480)=(505 - 480)=25W.  Assuming that around 1/3 goes to inductor losses and 2/3 to the MOSFETs, we see that (50*2/3)=16W is spread over 8 MOSFETs, yielding (very roughly) 2W per MOSFET.

The HAT2169H has a thermal resistance to the PCB of 4.2 C/W.  With a big exposed pad on the PCB we might be able to hit ~ 30 C/W from PCB to air.  So thats ~ 34C/W * 2W = 68C temperature rise.  With a 50C ambient temperature, we get silicone junction temperature Tj = 50 + 68 = 118 C, which looks OK.  However, the PCB could be around 50C + 30C/W*2W = 110C, which is perhaps too hot for long term reliability.

So if you double the circuit from the LTC3785 data sheet to have 4 phases, you might get your 960W output.  Total losses would be 55W or so, which is manageable with care, a fan, and a somewhat special PCB.

[David Hess]

With a 24 volt output, synchronous rectification is not strictly necessary; diode rectification at that voltage only contributes 4% loss and that is on another heat sink mounted part anyway.

[jbb]

With a 24 volt output, synchronous rectification is not strictly necessary; diode rectification at that voltage only contributes 4% loss and that is on another heat sink mounted part anyway.

I guess so.  It's just making me twitch a bit...

[rx8pilot]

For that power level.....it's just too easy to use a synchronous design and ease the thermal challenges. If size and noise is no concern, maybe. Just use a big heatsink and a fan.

Short and misplld from my mobile......

[ctiger]

I am trying to retrofit the new design with the current unit, I have take the parts apart and shown in the pictures. I going to make a new design to have better control and in the long run to optimize.

In addition, you will need external MOSFETs (perhaps several in parallel) and synchronous rectification.  Something like the LT3784 might be a possibility. Check out page 34 for a 24V 20A circuit.  It's a 4 phase interleaved synchronous buck converter.

I could not find the page 34 on the datasheet for LTC3785, may I ask where is the page?

This project sounds like a very poor candidate to learn the delicate nature of power electronics.

Yes I know I am a very poor candidate, I been doing readings on power electronics and what I have learn from undergrad is totally different than the real application. And I hope the community would guide me this greenhorn.

[rx8pilot]

Yes I know I am a very poor candidate, I been doing readings on power electronics and what I have learn from undergrad is totally different than the real application. And I hope the community would guide me this greenhorn.

If you don't know, you are an excellent candidate to learn. What I was suggesting earlier is that this particular project is not the best place to start. It was only a few years ago that I got started in power electronics, and I am still considered a beginner in the big scheme of things. Early on, I was surprised at the complexity of switching power supplies - some of the complexity is obvious and some of it seems to come out of nowhere.

Some classic examples of surprise challenges are:

• Probing various sections of the circuits
• PCB layout is just as critical as the schematic itself
• Transient response - capacitive and other dynamic loads can freak out your design
• Ground bounce (related to PCB layout and probing)

Of course, there are other challenges as well, but these are the ones that stick out in my mind at the moment. I personally had a some good learning experiences from buying reference designs from top manufacturers like TI and Linear. Those allowed me to learn how to measure the circuit since you have properly captured signals as screen grabs in the datasheet of the reference PCB. I read a lot about probe design, probing technique, etc until I was able to see a similar signal on my scope as the reference from the manufacturer.

Once I was comfortable measuring things, I laid out a PCB based on a reference schematic. On the first few attempts - the circuit ran and appeared to be just fine. After closer looks - I found stability problems, noise, poor efficiency, etc. All of these problems were related to PCB layout problems. An absolutely flawless schematic is useless if the PCB layout is poor - as I discovered. Again.....lots of reading, watching videos, studying reference designs, and doing experiments to learn the critical and not-obvious details of PCB layout of SMPS.

The project that you stated at the beginning of this thread is no trivial challenge. At the power levels you are talking about - mistakes end with flames and fireworks that destroy the PCB and burn holes in your bench. If you learn with smaller projects, mistakes end with a fizzle, smoke, and some solder re-work to try again.

[jbb]

I could not find the page 34 on the datasheet for LTC3785, may I ask where is the page?

Yes I know I am a very poor candidate..

Bad link, sorry. You want the LTC3874[\b].

Like rx8pilit says, there is nothing wrong with being a beginner.  This is just a very big first project. There will be flames, and bangs, and times when one mistake destroys everything (see also: flames).

 Maybe a better introduction would be to investigate the controls of the existing device. It looks like a single phase ####-you design. I’m interested in the copper bus at design: you can see how it passes past DC cap banks to the switching stage. And that large film cap might be a ‘whoopsie’ addition (ie added late in the design when a problem became apparent). (Bus at leakage inductance is a killer of efficiency and transistors.)

What resources do you have access to? Large 12V supplies and electronic loads? Oscilloscope with >= 200MHz bandwidth? Differential probes? Current probes? Mechanical enjnerknf support? The original designer?

[ctiger]

 Maybe a better introduction would be to investigate the controls of the existing device. It looks like a single phase ####-you design. I’m interested in the copper bus at design: you can see how it passes past DC cap banks to the switching stage. And that large film cap might be a ‘whoopsie’ addition (ie added late in the design when a problem became apparent). (Bus at leakage inductance is a killer of efficiency and transistors.)

What resources do you have access to? Large 12V supplies and electronic loads? Oscilloscope with >= 200MHz bandwidth? Differential probes? Current probes? Mechanical enjnerknf support? The original designer?

Yes I have access to large 12V supplies, currently no oscilloscope, Fluke 369 FC True-rms Leakage Current Clamp Meter, no mechanical engineer support, I have tried contacting the vendor for the schematics of the converter but no luck. I bought this unit from https://www.made-in-china.com/ and its like an ebay or amazon where there are different vendors selling their product.

Like rx8pilit says, there is nothing wrong with being a beginner.  This is just a very big first project. There will be flames, and bangs, and times when one mistake destroys everything (see also: flames).

Yes I do know I will expect some sort of flames but I am prepared for it. Its like how a person do not know how to swim and push the person to the pool and learn how to swim.

I am trying the LTC3874 4-Phase 480W Single Output Boost Converter design and work from there.

[rx8pilot]

I will confidently say that successfully designing something like this without a decent scope and probes is extremely unlikely. You may get voltage at the output, but that is all you would know.

Perhaps like successfully designing and building a car with only a pencil and screwdriver.

[jbb]

First up, sorry for weird formatting. I’m using my phone...

Once again I agree with rx8pilot. Without a scope this moves from ‘hard’ to ‘hell no.’

I have the impression that you’re doing this for work. If that’s the case, tell your boss a) it’s crazy without equipment and b) it’ll take ages to make something that will beat what you’re currently using.

[David Hess]

Someone with experience could do it without an oscilloscope.  A beginner will likely not be able to do it even with an oscilloscope.

Someone with experience might elect for a lower performance inverter which is easier to get right.  Hmm, that gives me an idea; build a multiphase inverter.

[RES]

http://www.feec.ece.vt.edu/NewProjects/FuelCellConverter.php

https://powerelectronicsworld.net/article/78745/1000W_Step-Up_DC-DC_Converter_Modules

[oldway]

You did not said if the output must be isolated from the 12V battery or not.

Nor if the output must be electronically protected against short circuits or not.

If not, you only have to add 12V to the battery , so you need only a 12V x 40A = 480W additive converter.

And what about output regulation ? How much may it vary ?

I wonder how could you design a dc to dc boost converter without these specifications !