really nice little switcher

Looks similar to their TPS6xxx inverting charge pumps. They work pretty good too. Not much noise.

Cheers

The data sheet doesn't say much about thermals, and it looks scary to me to make a 15 watt switcher in a SOT23-6 without a power pad, so I got the eval board and tried it.

It's unusual in that it accepts 28 volts input and has internal soft-start and spread-spectrum. And it's cheap.

I think 3 amps is pushing this little guy pretty hard, but 2 amps, maybe even 2.5, looks reasonable. I don't know where the heat comes out... most likely pin 1, ground. I guess I'll put a lot of copper on all the pins. ===============================================================================

Can't you just take a pic with your IR camera and see how hot each pin is where it enters the body? If not on the eval board, do a proto with the same pad area at each pin and drive it just hard enough to make it warm up a little and see which pin(s) get hottest.

----- Regards, Carl Ijames

The pins themselves are shiny in the thermal IR, so their temperature display is very wrong. I didn't quite have enough resolution to image the epoxy around each pin. Maybe I'll paint the pins with something with high emissivity to try to resolve pin temps. I wish I could image the heat flow!

I probably don't need over 2 amps on any supply, so I'll just lay out the real board. These are cheap enough that I can use a lot of them for the various supplies.

This one will work in the inverting configuration, but it can't quite do +24 to -5... that's one volt over its abs max rating.

I can switch down to +5 with one of these, and flip it to -5 with one more. I need some +5 anyhow.

The NXP RF PA's (see recent discussion) show some thermal photos after the entire reference design was sprayed with a uniform black. The actual paint used was unlike any black I've seen - I suspect because it was extremely thin and high-conductivity, to avoid changing the heat flow. Does anyone know what it is?

You know the approximate thermal conductivity, and the heat flow is just the temperature derivative multiplied by the conductivity. Even just convolving the raw image to produce the derivative would be enlightening. I think most image manipulation tools have such a thing.

Clifford Heath.

Stick some kapton tape on the leads. It is black to LWIR.

Black whiteboard marker seems to have high emissivity, and it wipes or washes off easily without making a mess. I have a Testor model paint set, and I suspect that's high e, so I could paint the pins and just leave it that way. Most paints are high e.

I'd need pretty high imaging resolution to see the temp gradient along a pin. The FLIR could probably manage that, but not handheld! Maybe I could hack up a tripod sort of thing to hold it steady. It would be interesting to compara the temp gradients on all six pins.

But the heat flow will still depend on the board layout. TI should better discuss heat flow on a SOT23 part that can switch 30 watts. Oh well, playing with parts like this is fun, a break from being a grownup.

I do that sometimes. These pins are really tiny, but it could be done.

Here's a piece of Kapton on copperclad FR4:

Nice, thanks for the data. Did you try stressing it to the thermal limit shutdown? I always like to do that... let it thermal cycle on and off for a few hours and make sure it still runs after that.

George H.

Black whiteboard marker seems to have high emissivity, and it wipes or washes off easily without making a mess. I have a Testor model paint set, and I suspect that's high e, so I could paint the pins and just leave it that way. Most paints are high e.

I'd need pretty high imaging resolution to see the temp gradient along a pin. The FLIR could probably manage that, but not handheld! Maybe I could hack up a tripod sort of thing to hold it steady. It would be interesting to compara the temp gradients on all six pins.

But the heat flow will still depend on the board layout. TI should better discuss heat flow on a SOT23 part that can switch 30 watts. Oh well, playing with parts like this is fun, a break from being a grownup. =========================================================================

I would be afraid of the variable thicknesses you will get with any kind of hand painting. I'd just give it a light spritz with some aerosol flat black paint, from a foot or more away. Should give a nice even coat, and you can hit it a few times if you think it's not dense enough. Doesn't matter so much exactly what the emissivity is, so long as it is the same everywhere and much less shiny than the metal pins, and you are getting the same insulation value everywhere. You should also be able to look at the package where the pins exit, instead of the pins, for an accurate temp (given sufficient resolution in the flir). Anyway, just some thoughts since I knew you had that nice ir camera, but I should have known that you would already have tried it :-).

----- Regards, Carl Ijames

The problem with the eval board is that the pins have various amounts of heat sinking. A better thermal breadboard would have identical traces on all pins, to estimate the heat flow from each. Too much work. I'll just put lots of copper on the ground pin and the output node, probably the main heat dumps, which I would have dome for electrical reasons anyhow.

This chip should come in a power-pad version.

Den fredag den 2. september 2016 kl. 04.16.21 UTC+2 skrev John Larkin:

the datasheet does mention that copper on GND and Vcc helps, which makes sense since they are the pins that doesn't "move" so you can add lots of copper, you wouldn't want lots of area on sensitive or switching nodes

-Lasse

One thing I didn't find in the datasheet was any limit on the 'dropout' voltage. For example, can I develop 5V out for a 6V input? Sometimes this is expressed as a maximum on-time ratio.

Good looking efficiency curve, though, from those nice large-area internal switches!

it says it can do 100% duty cycle as long as Vboot-Vsw > 2.1V (wonder how they intend you'd do that at 100%)

the schematic for a 5V supply says 8-28V input, that's probably a hint

-Lasse

It thermal cycles at 4 amps out, and the top of the chip is 107C. The thermal image looks like a heartbeat. It will run continuously at 5.2 volts out, 3.5 amps, 18 watts out. Not bad for a SOT23.

Like many buck switchers, voltage goes UP as load current increases.

My fave is Krylon #1602 Ultra Flat Black. It's mildly conductive due to its high density of carbon black vs. binder, but it's very black over a huge wavelength range--at least vacuum UV to thermal IR.

You have to get it from industrial supply places these days, but it's worth a Grainger or Fastenall order.

(For us optics guys, it's also a spectacularly good refractive index match to fused silica, so a dab or two on a lens or prism will get rid of annoying ghosts and fringes to the tune of 40 dB optical (80 dB electrical).

Cheers

Phil Hobbs

The datasheet, page 21, shows large copper areas for Vin and Gnd, and four vias on the SW pin to "internal SW node copper". Those are the 3 pins connected to the MOSFETs. Isn't that the SW pin that's overheating? I see their PCB has a long SW trace before reaching 12 vias to a bottom trace, which it fails to make as big as it could. It should be easy to do better than that.

They don't mention inverting mode. I always worry about using parts in the inverting mode if they do 100% duty cycle, when the output is lower than programmed, "the TPS54302 is designed to operate at 100% duty cycle as long as the BOOT to SW pin voltage is greater than 2.1 V." You have to rely on a maximum-input-current shutoff. They specify 5A for that, which is pretty ugly.