0
\$\begingroup\$

I'm working on a portable project that takes 4 (appropriately-rated) Li-Ion cells in a 4S configuration (14.46-16.8V) and boosts it to 27V for a 10-20 Ohm resistive load (soldering iron). The majority of the purpose of the project is because I want the learning experience; I know there are commercial offerings.

Running my numbers, I'll be pumping ~70-80W around, which is higher power than anything I've done before. I spent a good amount of time trying to find a DC/DC converter that could handle the amperage I needed, but ultimately came up short. I DID, however, find an interesting example of a "2-phase converter" in the LT3579-1 documentation (side note: if I drop the out voltage a bit, that converter could handle the load at just about it's absolutely maximum rating... so I have a fallback).

This 2-phase converter had the FB pins and the Error Amplifier Output (VComp) pins of 2 ICs tied together, and the ICs were also synchronized (well, actually un-synchronized, 180 degrees out-of-phase) by way of some included circuitry. That chip is more than I want to spend, but it got me wondering if the approach would work generally for a step-up converter:

  1. Build N (or maybe N+1 to be safe) identical DC/DC Step-Up circuits on the board
  2. Tie all the FB lines together, as well as all the VComp lines
  3. Since the outputs on the DC/DC regulators are already going through a diode, I don't need to worry about the current from one flowing back into another
  4. Generate a frequency/clock that is (N or N+1) times as high as the IC's spec, then put it through an (N or N+1) divider to clock each IC as out-of-phase with the others as possible.

Is this a reasonable approach? I figure I'll get better efficiency at lower (not TOO low) current, too - so less total heat, plus I'll be dividing the heat that is generated across multiple chips, so it should be easier to manage all around. Am I missing something? Is this a good approach? Are there things I should watch out for - specific regulators that won't work in this configuration, etc? Thank you!

\$\endgroup\$
6
  • 2
    \$\begingroup\$ 4 would be ideal. If you have constant current limitation on each boost converter, they should parallel nicely. But is this a one-off or mass production? There are controllers with balancing features for the very purpose. \$\endgroup\$ – winny Dec 18 '20 at 8:19
  • \$\begingroup\$ @winny "4 would be ideal" as in "You should use 4 of them in parallel" or "The bullet item #4 is a good idea"? Just trying to make sure I'm understanding everything, this is past the limit of my experience so I'm really trying to learn. \$\endgroup\$ – Helpful Dec 18 '20 at 16:59
  • 1
    \$\begingroup\$ Sorry about the confusion. Four parallel boost converters separated by 90 degree is a technically very pleasing solution. \$\endgroup\$ – winny Dec 18 '20 at 19:48
  • \$\begingroup\$ @winny Thank you for clarifying! Is there anything I could read to learn more about why (aside from "more converters/redundancy") that's better than, say, 3 that are 120 degrees out of phase? \$\endgroup\$ – Helpful Dec 18 '20 at 19:50
  • 1
    \$\begingroup\$ It’s quickly diminishing results of reduced ripple by 1/N by paralleling. There are however special cases if you have fixed input and output voltage where you can make very favorable choices of N. Look at PC motherboards with 12 V input and say 1 V output with 12 phases. Simulate it and you’ll see something quite magical. Boost works in a similar fashion. \$\endgroup\$ – winny Dec 18 '20 at 20:08
2
\$\begingroup\$

Is this a reasonable approach?

Yes and personally, I'd just make several boost circuits like this: -

enter image description here

And have one controller that cycles through each MOSFET producing the right level of duty cycle to transfer (using DCM) the correct energy to deliver 27 volts for the 10 Ω load: -

enter image description here

It looks like all three are in parallel but it's down to how you alternate each one in terms of the control inputs CA, CB and CC. All three are sharing the power losses when driving the load.

So, where you and I probably differ is in the implementation; you are trying to use multiple controllers whereas I'm favouring one controller and some sequencing logic to drive individual boost circuits.

\$\endgroup\$
8
  • \$\begingroup\$ I'll see if I can dig into this better to understand it, but I appreciate you responding! IMO the advantage of going with the separate ICs is that the MOSFET is built-in, so there's no chance of me screwing up the selection for it, haha. Your approach definitely seems more economical, though. \$\endgroup\$ – Helpful Dec 18 '20 at 16:58
  • 1
    \$\begingroup\$ @Helpful the LT3579-1 uses an external transistor so I'm not sure what you mean here. Appreciation is usually given in this normal way on stack exchange. \$\endgroup\$ – Andy aka Dec 18 '20 at 17:44
  • 1
    \$\begingroup\$ That appreciation has been given in the normal way - apologies, been a while since I was around here. The LT3579-1 was an example where I found the polyphasic converter, but I'm probably going to use a different (more efficient and cheaper) chip with an external clock \$\endgroup\$ – Helpful Dec 18 '20 at 17:53
  • 1
    \$\begingroup\$ If your heating element needs to be isolated (soldering iron tip and ESD stuff) then maybe a 100 watt flyback design might be preferable. \$\endgroup\$ – Andy aka Dec 18 '20 at 17:55
  • 1
    \$\begingroup\$ No, are there no advantages except theoretically slightly easier on the math when deriving. The math is still very similar between flyback and boost. \$\endgroup\$ – Andy aka Dec 18 '20 at 18:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.