breadboarding fast, tiny stuff

[John Larkin]

We got some samples of an NEC hj fet and were wondering what its
time-domain response might be like. The part is only 2x2 mm and the
leads are 1.2 mm pitch, and I hadn't previously had a lot of luck
breadboarding stuff like this.

We found two tricks:

Get a piece of copperclad, epoxy-glass or preferably teflon; the
teflon is easier to cut. Cut out "pads" with a very sharp xacto knife,
under a Mantis magnifier. This will make horrible burrs and shorts, so
the first trick is to scrub it really hard with a Scotchbrite pad
between cuts. This cleans it up beautifully.

The second trick is to use small patches of kapton tape as insulators.
like where parts join or whatever. Soldering doesn't bother it at all.

ftp://66.117.156.8/FetTest.zip

Here, the fet is in a first-pass test circuit, just to see how fast we
can turn it on and off. The TDR pulse from the sampling head is the
gate drive, 0 (Idss) to -0.5 (pretty much off) at 50 ohms source z.

The drain is pulled up through a 47 ohm resistor, and the 150 ohm
resistor off to the side is an "attenuator" into the other scope
channel. The turnon fall is very clean, no nasty ringing or whatever,
with a 190 ps fall time. Turnoff is similar; these things don't store
charge! The TDR of the gate (lower trace) indicates that the gate
capacitance is loading the drive, so we need a bigger gate swing, from
a lower source impedance, to make this thing switch really fast. That
will be next.

John

 

[Joerg]

We got some samples of an NEC hj fet and were wondering what its
time-domain response might be like. The part is only 2x2 mm and the
leads are 1.2 mm pitch, and I hadn't previously had a lot of luck
breadboarding stuff like this.

We found two tricks:

Get a piece of copperclad, epoxy-glass or preferably teflon; the
teflon is easier to cut. Cut out "pads" with a very sharp xacto knife,
under a Mantis magnifier. This will make horrible burrs and shorts, so
the first trick is to scrub it really hard with a Scotchbrite pad
between cuts. This cleans it up beautifully.

The second trick is to use small patches of kapton tape as insulators.
like where parts join or whatever. Soldering doesn't bother it at all.

ftp://66.117.156.8/FetTest.zip

You need to clean that solder iron sponge ;-)

I am going to have to prototype with DFN packages soon. 0.5mm pitch and
no leads. Not looking forward to that one. Maybe I'll attach litz wire
strands to it.

Here, the fet is in a first-pass test circuit, just to see how fast we
can turn it on and off. The TDR pulse from the sampling head is the
gate drive, 0 (Idss) to -0.5 (pretty much off) at 50 ohms source z.

The drain is pulled up through a 47 ohm resistor, and the 150 ohm
resistor off to the side is an "attenuator" into the other scope
channel. The turnon fall is very clean, no nasty ringing or whatever,
with a 190 ps fall time. Turnoff is similar; these things don't store
charge! The TDR of the gate (lower trace) indicates that the gate
capacitance is loading the drive, so we need a bigger gate swing, from
a lower source impedance, to make this thing switch really fast. That
will be next.

Nice ramps. Which part number of their FETs is that?

 

[John Larkin]

You need to clean that solder iron sponge ;-)

I am going to have to prototype with DFN packages soon. 0.5mm pitch and
no leads. Not looking forward to that one. Maybe I'll attach litz wire
strands to it.



Nice ramps. Which part number of their FETs is that?

That's the NE3508, a slightly bigger version of the 3509. The 3509 is
amazing; pinchoff is about -0.4 volts, RDSon is 6 ohms, drain C is
0.35 pF. I've got to do the corresponding measurements on the 3508.

This fet would make a nice footstool for an ant.

John

 

[Joerg]

That's the NE3508, a slightly bigger version of the 3509. The 3509 is
amazing; pinchoff is about -0.4 volts, RDSon is 6 ohms, drain C is
0.35 pF. I've got to do the corresponding measurements on the 3508.

Thanks! The NE3509 seems even more hot than the BF862. Nice. The NE3508
datasheet states a similar pinch-off and Rdson looks like 4-ohmish.

http://www.cel.com/pdf/datasheets/ne3508m04.pdf

This fet would make a nice footstool for an ant.

Better than what I just found in a review of a unit from the field: A
weird kind of spider in there.

 

[John Larkin]

Thanks! The NE3509 seems even more hot than the BF862. Nice. The NE3508
datasheet states a similar pinch-off and Rdson looks like 4-ohmish.

http://www.cel.com/pdf/datasheets/ne3508m04.pdf

Measures 3.6, down to under 2 ohms if you enhance it a few tenths of a
volt.

I could do a whole product line around this part. I think I will.

John

 

[JosephKK]

We got some samples of an NEC hj fet and were wondering what its
time-domain response might be like. The part is only 2x2 mm and the
leads are 1.2 mm pitch, and I hadn't previously had a lot of luck
breadboarding stuff like this.

We found two tricks:

Get a piece of copperclad, epoxy-glass or preferably teflon; the
teflon is easier to cut. Cut out "pads" with a very sharp xacto knife,
under a Mantis magnifier. This will make horrible burrs and shorts, so
the first trick is to scrub it really hard with a Scotchbrite pad
between cuts. This cleans it up beautifully.

The second trick is to use small patches of kapton tape as insulators.
like where parts join or whatever. Soldering doesn't bother it at all.

ftp://66.117.156.8/FetTest.zip

Here, the fet is in a first-pass test circuit, just to see how fast we
can turn it on and off. The TDR pulse from the sampling head is the
gate drive, 0 (Idss) to -0.5 (pretty much off) at 50 ohms source z.

The drain is pulled up through a 47 ohm resistor, and the 150 ohm
resistor off to the side is an "attenuator" into the other scope
channel. The turnon fall is very clean, no nasty ringing or whatever,
with a 190 ps fall time. Turnoff is similar; these things don't store
charge! The TDR of the gate (lower trace) indicates that the gate
capacitance is loading the drive, so we need a bigger gate swing, from
a lower source impedance, to make this thing switch really fast. That
will be next.

John

Hell, you have a webpage to work with post gif's not zip's.

 

[John Larkin]

Hell, you have a webpage to work with post gif's not zip's.

I'm offering free data and advice, and you're whining about the price.

And it's not a web page, it's an FTP site.

And my camera makes jpeg's, not gif's.


Did I leave anything out?

John

 

[Joerg]

I'm offering free data and advice, and you're whining about the price.

And it's not a web page, it's an FTP site.

And my camera makes jpeg's, not gif's.


Did I leave anything out?

Yes, one item: It works

(Both the web site and the circuit)

 

[John Larkin]

Yes, one item: It works

(Both the web site and the circuit)

I'm right now doing the next version, a new board, with (I hope)
improved copper-hacking technique [1]. It will include an MC10EL gate
driver, SO-8, to really wail the thing. I'd post more pics, if I
didn't think I was boring people.

John


[1] Score two parallel lines in the copper with an x-acto, and cut the
ends, to make, say, a long, skinny rectangle, like a 50 ohm CPW gap
for example. Tin the strip, or dab it with liquid flux. Now place a
soldering iron near one end, and lift the trace; this is the tricky
part, getting started. Once the end is free, pull it up gently with
tweezers and run the iron along the trace, peeling up behind the tip.
The heat softens the epoxy and the copper comes off like a zipper.
Very clean cutouts of, say, 30 mil width or bigger can be done.

Do they make really tiny Dremel router bits? That could be
interesting.

 

[Ecnerwal]

Do they make really tiny Dremel router bits? That could be
interesting.

(annoying website - hit the mute button unless you like random noises)

http://microcutusa.com/

Is 0.002 inches small enough for you? ;-)

 

[Nico Coesel]

You need to clean that solder iron sponge ;-)

I am going to have to prototype with DFN packages soon. 0.5mm pitch and
no leads. Not looking forward to that one. Maybe I'll attach litz wire
strands to it.

I did that a long time ago. Using very thin wire is the key to succes.
The surface tension of the solder will pull the wire automatically to
te center of the pad. It is in fact much easier than -for instance-
soldering ribbon cable wires to a header connector.

 

[John Larkin]

You need to clean that solder iron sponge ;-)

I am going to have to prototype with DFN packages soon. 0.5mm pitch and
no leads. Not looking forward to that one. Maybe I'll attach litz wire
strands to it.

Why not buy (or fab) a little PC board adapter thing, like the Bellin
stuff?

John

 

[John Larkin]

(annoying website - hit the mute button unless you like random noises)

http://microcutusa.com/

Is 0.002 inches small enough for you? ;-)

Nice. Something like this...

http://www.microcutusa.com/stub.php

could be chucked into a Dremel and slid along a steel straightedge as
a cutting guide.

I'll get some and try.

John

 

[Joel Koltner]

Do they make really tiny Dremel router bits? That could be
interesting.

I have a diamond-coated bit that's probably no bigger than 1/16" that works
well for slicing through copper. Old dental burrs are also frequently
available (cheaply -- they're new, but I guess they can only sit on a shelf so
long before they're no longer certified to be used on people?) and pretty
tiny.

Do you have one of those diamond "donut cutter" bits? (Makes an isolated pad
of copper about 1/8" in diameter.)

 

[John Larkin]

I have a diamond-coated bit that's probably no bigger than 1/16" that works
well for slicing through copper. Old dental burrs are also frequently
available (cheaply -- they're new, but I guess they can only sit on a shelf so
long before they're no longer certified to be used on people?) and pretty
tiny.

Do you have one of those diamond "donut cutter" bits? (Makes an isolated pad
of copper about 1/8" in diameter.)

The ideal cutter would be 10-20 mills in diameter, square edged like
an end mill, but with a smooth fat shank to chuck into the dremel and
slide along a guide of some sort, maybe just a steel ruler. The
Microcut stuff looks good.

FR-4 sure eats x-acto blades. Three today!

John

 

[Joerg]

Yes, one item: It works

(Both the web site and the circuit)

I'm right now doing the next version, a new board, with (I hope)
improved copper-hacking technique [1]. It will include an MC10EL gate
driver, SO-8, to really wail the thing. I'd post more pics, if I
didn't think I was boring people.

Please do. This stuff is never boring. Engineers who'd get bored by it
should either send back their degrees or go into sales.

John


[1] Score two parallel lines in the copper with an x-acto, and cut the
ends, to make, say, a long, skinny rectangle, like a 50 ohm CPW gap
for example. Tin the strip, or dab it with liquid flux. Now place a
soldering iron near one end, and lift the trace; this is the tricky
part, getting started. Once the end is free, pull it up gently with
tweezers and run the iron along the trace, peeling up behind the tip.
The heat softens the epoxy and the copper comes off like a zipper.
Very clean cutouts of, say, 30 mil width or bigger can be done.

Do they make really tiny Dremel router bits? That could be
interesting.

Try a shop that caters to glass cutters and people who build those fancy
lead-glass lamps.

 

[Joerg]

Why not buy (or fab) a little PC board adapter thing, like the Bellin
stuff?

John

I don't think they come in DFN package sizes. It's just for a one-off
prototype, see if the concept works. Support by semi mfgs has become so
dismal that you have to test everything they didn't write in their
datasheets :-(

 

[John Larkin]

I don't think they come in DFN package sizes. It's just for a one-off
prototype, see if the concept works. Support by semi mfgs has become so
dismal that you have to test everything they didn't write in their
datasheets :-(

The RF parts are especially bad. We are getting some Nitronex GaN fet
samples, but the datasheet has no transfer curves, no DC data, no
drain curves, so we'll have to measure it all. Apparently you're
supposed to pump RF into the gate and tweak the gate bias until
something works. 18 gazillion Smith and load-pull charts, though.

Ditto lots of "DC-XX GHz" mmics and RF switches that apparently DC
bias themselves to some undocumented voltages, and work for some
unspecified range of "DC."

But these little NEC fets work, as the man says, "like a trained pig."

John

 

[Joerg]

The ideal cutter would be 10-20 mills in diameter, square edged like
an end mill, but with a smooth fat shank to chuck into the dremel and
slide along a guide of some sort, maybe just a steel ruler. The
Microcut stuff looks good.

FR-4 sure eats x-acto blades. Three today!

That's because we are spoiled members of a throw-away society. Grandpa
had a sharpening stone right next to him and kept that wet all the time.
He wouldn't throw any knife or blade away until it was worn down to less
that half its regular size. I still have (and use!) his pocket knife
that had made it through two world wars.

 

[John Larkin]

That's because we are spoiled members of a throw-away society. Grandpa
had a sharpening stone right next to him and kept that wet all the time.
He wouldn't throw any knife or blade away until it was worn down to less
that half its regular size. I still have (and use!) his pocket knife
that had made it through two world wars.

One of those purple ceramic Intel 486 CPU's makes a damned fine x-acto
sharpening stone.

John