4
\$\begingroup\$

This is more of a meta question, so it may get canned.

I'm wondering if anyone has any techniques/workflow advice for how to handle the IC shortages particular to switching regulator circuits.

For 10 years or so I always used some derivative of the same local power supply design. That chip now has 52 week lead time, so for every new board I have to go out and find something in stock and redesign a power supply around it.

If I'm lucky, when I finish the design it will still be in stock. 0% of the time will it be in stock the next time I need to build a board.

My question is, "Is this really the way?" Is everyone on this pain train with me, or am I missing some vendor/design path that can make it easier?

\$\endgroup\$
4
  • 2
    \$\begingroup\$ We're all in the same boat. 52 week lead time is good these days as we're getting 2 to 3 year lead times on some parts. Having a good relationship with vendors and factory reps doesn't help right now. You can go through parts scalpers, but you might pay 10x to 100x the list price. \$\endgroup\$
    – qrk
    Apr 15, 2022 at 17:54
  • \$\begingroup\$ Matt - Hi, As you highlighted at the beginning, I'm in two minds about the validity of this question here. Although this one is specific to switching regulators, if we're not careful, we'll get others e.g. specific to transistors ,then PNP transistors, then PNP transistors in SOT-70 package etc. and it gets out of hand. So there are no promises about the longevity of such questions. FYI "our" question which tries to be canonical for tips about the parts shortage is here. Please follow the 3 points mentioned in that question, which aim to make any answers widely applicable. TY \$\endgroup\$
    – SamGibson
    Apr 15, 2022 at 18:20
  • 2
    \$\begingroup\$ RE: "If I'm lucky, when I finish the design it will still be in stock." Don't wait. Buy now, complete the design after you have secured inventory. \$\endgroup\$
    – The Photon
    Apr 15, 2022 at 23:16
  • \$\begingroup\$ could buy a lot to at least have enough stock to cover your product for a while, and become a "scalper" to cover the extra cost. \$\endgroup\$
    – Abel
    Apr 15, 2022 at 23:16

2 Answers 2

3
\$\begingroup\$

I've embraced more discrete approaches that use generic parts with lots of equivalents. If I can, I trade off some performance for an implementation with a control loop using an op-amp or two - for which I make sure there are ample substitutes - and the switch driver and miscellaneous circuitry are discrete transistors in tiny packages. The stuff I work on isn't usually area- nor power-constrained. Thus I can afford a bit more area and a bit higher dissipation for a design that'll be infinitely easier to get parts for, since none of the parts are a "switcher". In a few cases I could get away with using BJTs as switches as well, since those tend to have much better availability than mosfets.

But when you need the best performance, or have no time to roll your own, you'll need to stick with integrated solutions, and select switchers that have drop-in alternates available from multiple sources. There are a few of those.

For a few cost-sensitive applications I've made the MCU boostrap itself: it starts up in a low power mode using a resistive divider straight from USB Vbus, and then carefully spins up a software control loop that closes around a discrete mosfet driven from a few paralleled GPIO pins. It starts generating 1.8V for the core, steps up the clock speed and in unison also steps up the switching speed and control loop bandwidth. Eventually it performs like any off-the-shelf switcher, except uses mostly built-in peripherals - a fast ADC, a comparator, high drive strength GPIO pads. The switcher control loop runs on a dedicated realtime core along with other hard-realtime functions, and this works well. That core runs a small amount of robust code, and survives all kinds of upsets to the general-purpose application cores, including a full reset. Of course not all MCUs have that, but often the complete application is simple enough that realtime performance good enough for self-power-regulation is there to do the job.

\$\endgroup\$
1
  • \$\begingroup\$ Really good point about just going lower level in the design. \$\endgroup\$
    – Matt
    Apr 16, 2022 at 14:24
1
\$\begingroup\$

I see the generic link, but here’s my take on switcher sourcing.

First, check distribution for ‘in stock’ parts. The ones that are there are popular and will tend to be re-ordered, helping ensure a steady supply.

Second, choose older regulators that have as many second sources as possible, and try to make your design tolerant of differences.

This means perhaps eschewing specialty multi-output types or latest-and-greatest technology in favor of older, single-output parts.

For example, for 1-2A stuff I long ago settled on a common SOT23-6 synchronous type that has footprint compatible devices from US, Taiwan and China sources. Sometimes these have small pinout variations, so strap options can help.

Bigger regulators? Might need to consider external switch types using mature controller chips and FETs. Or perhaps using standardized POL modules.

Third, everyone has their favorite vendors. I personally try sourcing MPS, Diodes Inc, and TI first, followed by Maxim and Analog Devices. These companies are in it to win it.

I used to use Intersil, but now that they’re part of Renesas, be careful: Renesas tends to aggressively purge SKUs from its product lines in my experience.

On the other hand, International Rectifier is part of Siemens now. They’re a good choice for automotive/high rel and have a commitment to long term sourcing.

At the risk of self-contradiction, emerging vendors looking for growth, like AMS for example, might be easier to work with.

Finally, look for BOM leverage. If there are several different parts that you can source from the same vendor sometimes they will give you priority for your low-value devices. TI was big on this in my experience. Never hurts to ask.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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