Our mechanical engineer decided that a fan in a large (~200W) battery charger
we designed wasn't blowing enough air, so he found himself a beefier fan to
use instead (due to volume & weight constraints using multiple fans or a
physically larger fan wasn't readily doable). Only problem: There's just 5V
in the box, and the fan wants ~12V. This is a test board that consists of an
LT3580 boost converter (top of PCB) producing 13.1V @ 500mA from 4.5-5.5V, a
simple FET switch (IRF7455, middle of PCB) used for load step response
testing, and a Max6643 fan controller (bottom of PCB, varies fan speed based
on temperature).

It all works like it should. The boost converter ends up being ~85%
efficient, running at ~638kHz. The LT3580 has an internal power switch, so
other than changing the switching frequency about the only way to improve
efficiency would be with a lower-loss inductor... and I decided the current
inductor is already large enough.

---Joel

Here's The Brat's first PCB layout. U1 is an LT3580 in the leadless
DFN package. It's working as a Cuk inverter, making -2.5 from +12.

This is a very cool part. One resistor sets the frequency and one
resistor sets the output voltage, and it's smart enough to get the
feedback right whether it's an inverting or non-inverting converter.

The whole thing works!

John

"John Larkin" wrote in message
...
Here's The Brat's first PCB layout. U1 is an LT3580 in the leadless
DFN package. It's working as a Cuk inverter, making -2.5 from +12.

Nice. Did you guys hand solder that board or use paste and an oven? I've
done leadless DFNs by hand, but I choose the QSOP package for mine since it's
easier to hand solder.

This is a very cool part.

Agree -- "it just works," which is more than I can say for some switcher IC
I've used.

That's your ~DC to 1GHz amplifier board, isn't it, John? What uses the -2.5V?

---Joel

On Mon, 2 Feb 2009 09:34:01 -0800, "Joel Koltner"
wrote:

"John Larkin" wrote in message
.. .
Here's The Brat's first PCB layout. U1 is an LT3580 in the leadless
DFN package. It's working as a Cuk inverter, making -2.5 from +12.

Nice. Did you guys hand solder that board or use paste and an oven? I've
done leadless DFNs by hand, but I choose the QSOP package for mine since it's
easier to hand solder.

That was stencil pasted and reflowed. My production people found an
outfit that makes cheap laser-cut stencils out of mylar. They don't
last as long as stainless, but are as lot faster and cheaper. And they
know we're probably going to change something on the board sooner or
later.

I'm not a great fan of DFNs, but somebody picked this one for some
reason.


This is a very cool part.

Agree -- "it just works," which is more than I can say for some switcher IC
I've used.

That's your ~DC to 1GHz amplifier board, isn't it, John?

Yup. Really DC, not ~DC!

What uses the -2.5V?

The opamps. It's a gross overkill, as we need maybe 50 mA.

John

"John Larkin" wrote in message
...
Yup. Really DC, not ~DC!

Ah, neat!

Hmm... and it's not immediately obvious to me how you get DC and AC gains to
track perfectly. Clearly you have a few tricks up your sleeve...

They're thermal testing my little boost converter today... -12C (easy) to +54C
(hmm...).

---Joel

On Mon, 2 Feb 2009 15:40:52 -0800, "Joel Koltner"
wrote:

"John Larkin" wrote in message
.. .
Yup. Really DC, not ~DC!

Ah, neat!

Hmm... and it's not immediately obvious to me how you get DC and AC gains to
track perfectly. Clearly you have a few tricks up your sleeve...

No trick here. Just use fast opamps!

John

"John Larkin" wrote in message
news
No trick here. Just use fast opamps!

Ah, OK, I was thinking that you were using one of the cheap MMIC amplifiers
from the likes of Hittite/MiniCircuits/etc. and letting them ride on a DC
offset or somesuch.

What's the gain and flatness of your design?

On Tue, 3 Feb 2009 09:31:29 -0800, "Joel Koltner"
wrote:

"John Larkin" wrote in message
news
No trick here. Just use fast opamps!

Ah, OK, I was thinking that you were using one of the cheap MMIC amplifiers
from the likes of Hittite/MiniCircuits/etc. and letting them ride on a DC
offset or somesuch.

Too much work, for a mere GHz or so.


What's the gain and flatness of your design?


Gain = 100, or 50 loaded, using two THS4303's. It could be less, using
THS4302's and/or some optional resistors here and there. There are
some hooks for peaking capacitors, to equalize pcb losses and such,
but I haven't played with them yet. These things typically need a
little tuning to get clean step response, which is what we're after
here.

John