UCC33420 dc/dc converter eval

Yikes, what a piece of junk. Small maybe, but lots of parts needed for cleanup.

What does it do to the input rail?

Cheers

Phil Hobbs

I didn't scope the input rail. I expect it has lots of noise and DC load in bursts.

The architecture seems to be a simple forward converter with lots of step-up, and regulates bang-bang in inefficient bursts. I expect that if I reduced the input voltage, it woud get more efficient as the bursts approach 100% duty cycle and it starts to lose regulation.

I took the rig down. It won't work in my application, as a high side gate drive supply in a GaN half-bridge. Way too noisy.

I love the idea of a tiny cheap dc/dc converter on a chip, but not this one.

$100 for this eval board is silly too. It comes with a generic application brochure that says "go to the web site"

I now have, in theory, threeTI support engineers, whose universal support is "go to the forums."

I remember in the olden days when our TI guys would show up with a briefcase full of data books and unexpected new samples.

Enjoy it while it lasts. In six months they’ll be shoveling AI-generated summaries of the forums. :(

As I’ve said before, as a young teenager I had a subscription to the Motorola Update. For a *very* small fee, four times a year they sent me a big box of databooks and app notes.

It was magical stuff, most of which I understood dimly at best—bit slice processors, TTL register files, MNOS nonvolatile memory, all sorts of things.

It made a big impression on my subsequent career.

Cheers

Phil Hobbs

I think this high frequency noise should be easy to filter by some CLC Filter.

Not really. I need it to be close to the load, and that load is super noise sensitive. I didn't check, but there is no doubt some considerable and nasty emitted mag fields too.

In addition to all the high frequency crud, there's the basic burst and its side effects in the hundreds of KHz range, would would take giant parts to filter.

There are interesting parts around now that include their own power isolation. Some are capacitive, not inductive, so may be better.

Of course, the vincinity of any isolated DCDC is super noisy. If nothing else helps, think about battery power.

Look at LDOs with good PSRR at high frequency. From the datasheet, the TPS7A20XX is still in the 40 db range at 1 MHZ. The LT1963 may be better, but is much more expensive, the TPS7A20 is rather cheap and available.

What part do you expect to have capacitive power transport? Capacitive signaling is however common.

B.t.w, looking at a ISOW7841, I also see frequency in the 50 Mhz range.

Wuerth WPME-CDIP Capacitive Digital Isolator Powered talks about 13 MHz

I'm considering a bootstrap circuit to handle the switching power assisted by a photovoltaic source for steady-state. I can bootstrap off my own output pulse.

I suspect it uses similar power isolation as the UCCpart.

ADE9113 has power isolation too, and it also sounds magnetic. I thought I saw a capacitive power isolator somewhere.

That sounds like fun--if you do that, please let us know how it works!

Jim Williams used to write about piezoelectric transformers, back in his CC-fluorescent backlight days. (May they never return--fluorescent lights suck eggs.)

Cheers

Phil Hobbs

The PVs are affordable and of course marvelously quiet, but they max out typically below 100 uA. That gets tricky.

I could consider a tiny supercap, assuming that my users won't try to make weeks-long pulses.

I always thought the piezo transformer was cute. The real HV isolator is a motor and a generator and a fiberglass shaft, all surface-mount of course.

I threw out a bunch of incandescent bulbs and wish I'd kept a few. I don't miss CFs at all.

You can get more powerful ones for "power over fibre" with a laser at the other end. I vaguely remember there being an example in AOE3.

I would love such a power optocoupler, if it were a reasonable size and price.

There must be a market for a really quiet isolated dc/dc converter. Maybe a sine wave thing.

People go spread-spectrum to pass FCC and CE EMI tests, but that doesn't change the actual voltage noise at all.

I have a couple of old Luxo adjustable desk lamps that I love--one at my drafting table and one at my workbench. They have a medium-base light socket surrounded by a 22-W circline fluorescent tube. I mostly use LED bulbs, but for testing things like photoreceivers, where the 40-kHz flicker is obnoxious, I have a medium-sized stock of 100-W, 1690-lumen tungsten bulbs that I collected back in 2012.

I'll happily send you a couple of packages if that would be useful.

Cheers

Phil Hobbs

Yes I have been thinking about that, a really quiet DC-DC, e.g. with a magnetically-shielded transformer having both windings thoroughly electrostatically screened, or an optical isolator (but there is a worse limit to the efficiency for optical). The DC-DC inside Keithley Sourcemeters is interesting in its construction though I have no idea how well it performs.

Solar cells are cheap, and high-power IR LEDs are fairly cheap too, but the combination won't be super-efficient nor small. If you want to avoid putting multiple solar cells in series, you could connect one solar cell to a step-up transformer, and modulate the LED current so that there is some AC for the transformer to step-up. To avoid DC in the windings you could even put two solar cells in anti-parallel across the low voltage winding of the transformer, and illuminate the pair of solar cells separately with two IR LEDs driven with opposite phase AC. Big solar cells have a lot of capacitance though, so the frequency would have to be lowish. If you want more isolation voltage, the light could be guided through a acrylic rods like a fat optical fibres. It'd be large, and not as efficient as a transformer.

Hard-drive motors can run for a long time. Maybe you could make a motor-generator that lasts long enough. A little line-shaft across your PCB for power distribution, with tiny flat belts to generators where you need extra outputs. It'd look nice all made from ebonite and polished brass.

If you hermetically seal the entire assembly in a metal box with feedthroughs, you could use perovskite solar cells for a nice bump in efficiency to lower the total area of cells you would need, and have great EMI shielding. Just include a little pkg of silica gel in the box to soak up moisture and the perovskites should last longer than you need them to.

If you can choose the wavelength of illumination, there is no need to use perovskites, as their ability to be tuned to match the spectrum of sunlight has little benefit. I work with perovskite cells that stay in a glove box full of very pure nitrogen (not me, I stay outside in the air). Lots of things damage them, though they are getting better.

Interestingly, white LEDs don't like being inside the glove box in pure nitrogen. They rapidly lose efficiency if they are operated in there, but they recover if a little bit of oxygen is added (which we can't do, because it harms the perovskites). So the LED solar simulator has to stay outside. It has dozens of different LED wavelengths, some of them are not bothered by being in nitrogen.

With an optical power isolator, the area of solar cells can be kept small if the illumination can be prevented from spreading out much. So, if a laser diode puts a few watts down a fibre, the receiving cell could be tiny, perhaps only limited by not wanting it to melt.

It would be possible to couple a single high-powered fibre laser to several photodiodes - perhaps connected in series to get a more useful voltage - using a fibre beam splitter. Such splitters are very cheap if you choose an infra-red wavelength compatible with passive optical networking (GPON or XGPON) . John

Solar cells are silicon, which don't work at all with wavelengths exceeding 900 nanometers, or shorter than about 500 nm.

Joe

I tried to order a few of that chip, to xray one just for fun. My purchasing lady says none are available, not even from TI.

Mouser shows the eval board as obsolete. Curious.

The camera in my phone just about responds to 850nm. However, there are plenty of 1310 and 1550nm fibre-coupled photodiodes around. Whether they would survive more than a few tens of mW is another matter. John