Synch buck DC/DC controller ICs capable of constant current

[Siwastaja]

Hi,

This has been a pet peeve of mine.

I frequently need to design a DC/DC converter as a part of some project, where I need the controller to be able to supply somewhat regulated constant current to the load:
1) without blowing up
2) without entering latched fault mode
3) without entering hickup mode
4) without self-destroying or refusing to work if the load is pre-biased (i.e., a capacitor, or a battery).

Usually I end up needing a synchronous buck.

It's a bit irritating, because I know that literally hundreds of ICs could do this, if it wasn't intentionally or unintentionally prevented in design (e.g., by adding a non-disableable "protection" logic) since this is not "the most typical" use case.

Haven't found any distributor with parametric search allowing for this option. Searching for synch buck always results in about 3773 relevant products. I have spent days just opening each and every datasheet and trying to find out if it can do it or cannot, sometimes you need to read quite far before being sure. Not going to find them this way. I suspect the relevant products mostly don't claim this miraculous CC feature on the front page. Those that do claim it, are too-specialized super-expensive and often totally broken-by-design battery management ICs or similar.

Ages old ICs, non-synch or non-bootstrap (P-FET), or voltage mode ("oscillate and blow up on capacitive load") controllers do seem to have CC mode more widely available, but I don't want to use them for obvious reasons except as a last resort. They could be used in more creative ways, circumventing their issues, but I'm still looking for modern level of integration.

So far, I have found one single product to suit this need:
LM25117
It's fairly good! Synch buck for good efficiency, capable of sane fsw (not some ancient 50kHz crap), current mode control for stability expected today, and, a way to disable both latched and hickup failure modes, so it can run against the current mode limit for unlimited time without issues -> a crude, but often accurate enough CC mode as is, plus an option to make it more accurate using an additional feedback with a few external components. So, a true CC-CV part.

This part also proves how difficult it is to find these parts: the front page of the datasheet only tells you about the hickup current limit, which would make you disregard the part on the first sight. You always need to read the whole freaking datasheet to find the needle in the haystack!

I have used this as a combined power supply and battery charger without issues (a battery-backed up DC link, not a rare application IMHO).

But it was a tedious task to find: Digikey parametric search, for example, doesn't even list this part as having a "current limit" at all! Found it using TI WEBENCH, which does have a relevant search option for true (unlimited time, no failure modes) current limit.

Although TI is fairly reliable, I'm looking for alternatives and options to give me a wider margin of design choice in the future, so if you know of any, please help me list them here, as I guess someone else might have the same need.

[ovnr]

You can usually bodge a CC mode onto it with just an opamp and some assorted passives, if you don't have very tight tolerance requirements.

Otherwise, consider looking for a CC driver that'll let you set the maximum voltage; there are a few LED drivers, for instance, that easily let you set the open-circuit voltage, and thus will operate in a CV-CC mode; some examples being the LT3571 and MP2483. I know, they're not synchronous; I didn't bother looking too hard.

[David Hess]

The universal solution is to drive the regulator's feedback pin or network with an operational amplifier and current shunt.  If you have access to the compensation pin which is the output of the regulator's transconductance amplifier, then you can drive that instead and bypass the internal error amplifier or combine them to produce CC/CV operation.

Older but still ubiquitous switching regulator controllers often have two independent transconductance error amplifiers allowing easy CC/CV operation but offhand I do not know of any which directly support synchronous operation.

[T3sl4co1l]

Indeed, the amount of time spent looking for this, may prove better spent doing it yourself...

Is discrete a non-option, due to size/cost?

Tim

[Siwastaja]

Indeed, the amount of time spent looking for this, may prove better spent doing it yourself...

Is discrete a non-option, due to size/cost?

Tim

Thanks, sometimes I do semi-discrete; fully discrete would be quite large and expensive. This time, I have an MCU available which would provide me with fast comparator and DAC for setpoint, would only need a current sense amp and a gate driver to create a hysteretic or peak current mode controller. But I'd need a bootstrap gate driver capable of 36V input, and it simply doesn't seem to exist (anymore: I sure remember using some cheap&obtainable half-bridge bootstrap drivers with 48V input just a few years back; now everything is hugely specialized and max 100kHz only), and designing a bootstrap high side driver with the necessary UVLO and/or desat detection would get close to the area of "non-option due to size and cost". This is frustrating, since there are about 3000 ICs that contain this gate driver I need (and everything else, including the unwanted fault/shutdown logic).

Right now, it seems the best idea would be to get the cheapest and smallest integrated part possible and simply hack the current measurement signal into the VFB pin.

[T3sl4co1l]

Yeah, "discrete" at the RTL level, or with a programmable device; like this phase-shift PWM generator,



I don't know of an IC that provides this function in a usable form (free of fault logic, with user defined controls), so, made one myself.  (Also, I wanted continuous-time control, excluding most MCUs.)

What's wrong with gate drivers?  HIP2101 is 2A and 100V, good old IR2110 is 2A 500V, UCC27714 is 4A 600V, FAN7390 is similar, etc.  Haven't checked any of these (except HIP I know) if they have overlap prevention logic, but usually there's a different part in the family with, without, and with different logic inputs (inverting, noninverting, polarity/enable, etc.).

There's also some very nice parts, with integrated magnetic, electric or opto isolators, and monitoring and protection circuitry and feedback, for driving bigger things (IGBTs).

Not a lot for high side GaN applications, yet, or that I've noticed (high dV/dt, high freq).  Plenty beyond 100kHz though?

Tim