1.8 volts (max)

Two 1N4148?

The critical sensor in question, an SGP30 gas sensor, is normally off, but draws 50mA running its heaters, so needs a proper 1.8-volt regulator. Adafruit uses an MIC5225, but that's too big, so I'm eyeing NCP170. And an BSS138PS, for level shifting two I2C lines.

Some FPGA and uP cores run at 1 volt, or even less.

A gentleman I know who works at Intel on portions of their most recent designs would call 1 volt "high voltage"

I'm confused. Is the chip rated for 1.8 volts max, or 1.8 volts nominal?

For level shifting I like to use switch parts. I'm not sure if you can fin d parts that are intended to shift 3.3 down to 1.8, but the parts I used fo r interfacing 5.0 to 3.3 volts might work with the lower voltages. But I'm using 24 pin package parts, I expect you would want smaller devices.

This says it's 1.8 - 5v

NT

yeah, but there's 7 extra parts on that board.

Now I'm eying the 150mA TPS71718 in SC-70-5.

OK, right. It's 1.98 volts absolute maximum.

A simple MOSFET is the ultimate switch part. :-)

. 3.3V ,-----+--- 1.8V . | | | . 10k G 10k . 3.3V I2C logic --+-- D S --+-- 1.8V I2C logic . "S" "D"

The bidirectional I2C standard uses pulldown, with pullup resistors to the local supply voltage. So a simple MOSFET can handle the level shifting in both directions. For low-to-high voltage, it's a cascode. For high-to-low voltage, you can see the substrate diode doing some work, but the drain of the MOSFET becomes its "source", with the gate at 1.8 volts, the "drain" pulls down with Rds(on) = 2 ohms for BSS138. The '138PS is a dual version in SOT-363.

Yes, that's what the switch parts are, pass FETs. But the gate needs a hig her voltage than the low supply to turn on adequately over the voltage rang e. Otherwise you get weak passive pullups and slower performance.

The parts I use include a diode to shift the upper supply down enough to al low low resistance across the FET close to the lower V+ rail yeilding fast transitions by the drivers on either side. The drivers have to drive both load resistors in parallel so each should be twice the value. In your desi gn once the 3 volt side releases the pull down and the voltage starts to ri se, the FET will quickly become high resistance because of the 1.8 volt gat e drive. So the low voltage side will only be pulled up by the single pull up which will be half as fast.

I haven't used I2C to know if this is an issue or not, but it should be con sidered. Likely you can use full strength pull ups on both sides and the d rivers will drive them in parallel ok.

Oh, you might want to check the specs on your FET. The ON data sheet has a typical graph that shows the resistance rising *rapidly* below 2 volts Vgs . That might not be the best part to use.

The core voltage of the latest 5 and 3nm feature-size processors under development is probably significantly below 0.5 volts but specifics isn't information they give out to anybodies

It'd be interesting to learn what level-translator chip you used. We have up to five feet of ugly cable. Worried about possible pickup, we've added 100pF caps on both lines, and will run the bus slowly, 100kHz. So maybe we're OK.

Yes, using a 60-volt part at 3.3 volts seems silly, even if many others are using it. I eyeballed NXP's curves to estimate the 2-ohm number, but there must be a better part. I do like the BSS138PS small SOT-363 package.

How can they distribute tens, or maybe hundreds, of amps on a chip?

very carefully I guess :)

no wonder some motherboards have voltage regulators with 8 or even more phases

Intel was using non-planar "3D" transistors in mass production as of the

The 22nm process is pushing a decade ago, now, they're on to "Trigate" and other topologies that again, my buddy cannot talk to me about and possibly isn't even familiar with - he probably was barely old enough to drive when that article was written! He does on-die cache memory controller design/layout AFAIK that's all he does, Intel can afford to super-specialize like that

Even my dumpy msi AMD Phenom II desktop mobo from 7 years ago has a four-phase main regulator for the CPU with a pleasant set of blue LED indicators to indicate activity

they cheeped out on the design tho and uses a single power MOSFET as a phase-splitter in the row of synchronous bucks instead of a proper high-side driver, and it blows them repeatedly under high load :-(

The design and mask set costs for a bleeding-edge chip is approaching a billion dollars.

And when I plunk down $600 for a 2066-pin X299 Intel CPU for my mobo, and know it's going to be blisteringly fast, I'm making a contribution.