breadboarding fast, tiny stuff

message

works

shelf so

isolated pad

Right: you gave me this chip/tip, the best use for old 486's.

I've been quoted BeO in custom-cut slabs, and it's not terribly expensive, close to AlN and a couple times more than alumina. The sales guy said it's fine to handle, even break, but don't grind it up.

Of course arsenic is forbidden RoHS stuff, so all the europeans will gave to stop using mesfets and phemts and mmics. Serves'em right.

John

The Eclips ecl gates (the MC10EL and EP parts) swing almost 0.9 volts in about 220 ps, which is perfect for slamming the gates of modern phemts. Analog Devices makes some comparators that are a lot faster, ballpark 40 ps with ecl swings. I'll have to try driving a phemt from one of them, just to see.

Some of the lvds line receivers will swing 3.3 or 5 volts in roughly

John

I can't name any off the top of my head, but perhaps xemacs qualifies.

I'm surprised about the price. I was relying on various descriptions; I've not quoted it.

Anyway, I don't believe our units have it, or that they'd be a danger even if they did.

I been using the self-same x-acto blade ... for 32 years. It's still sharp and good as new. I slice copper traces with it (but avoid the epoxy-glass), carve, cut, etc. I keep it shaving-sharp. (Good: sharp knives and sharp women.)

My great friend Ray gave it to me, so it has sentimental value as well.

Hey, as for carving proto traces, a broken tungsten carbide PCB drill bit, mounted in a Dremel. You can carve copper almost as if writing with a pencil. Dremel makes some cutters in high speed steel (HSS), but fiberglass eats them.

With warm regards, James Arthur

John,

Have you ever experimented with CML logic gates instead of ECL? I mean something like the NBSG16M. The swing is somewhat reduced (about

Thanks, Allan

We use an NBSG16VS as a variable-swing pulse generator. It gens a fast, programmable-level pulse that we feed into a couple of Hittite distributed amps, to finally get a 0-6 volt pulse that drives an optical modulator. It all works nicely, very clean and fast. All three chips are power-pad mounted with heatsinks on the back side of the board. The NBSG costs $32, and the Hittites are about $200 each.

PCB:

ftp://66.117.156.8/AmpTop.jpg

and output into 50 ohms, 1 volt/cm:

ftp://66.117.156.8/NewSq2.jpg

The NBSG would be a nice TDR step generator.

John

...

Superb waveform! Are the hittite amps DC coupled or do you clamp the output somehow?

Today I have been playing with a 74LVC00A which was available in some drawer to generate "narrow" pulses from a square signal in the quick and dirty way:

----

---·-----| \\ | |NAND---NOT-- |-NOT-| / ----

In -------- ------- -----| |-------| _ _ Out------| |------------| |----

Output pulse width is rougly 2.5ns with a leading edge rise time of 1ns. It fills the spectrum quite nicely up to 1 GHz. (Currently targeting the

The NBSG16VS would probably allow to take this some steps further. Have also looked at the MC10EP05. However, from the ON-Semi datasheets I can't figure out how fast the input edges should be. Is there some schmitt-trigger built into the inputs? How fast will the output edges be for say 2ns input edges? Do I only have to care on the time it takes through the V_IL-V_IH zone?

Pere

All ac coupled. This thing gates the NIF laser beamlines, which run at an absurdly low duty cycle. We generate a 0-30 ns wide pulse at 960 Hz. The laser itself fires maybe a few times a day.

Take a look at the NL37WZ16 triple buffer. With all three sections in parallel, it will put 5 volts into 50 ohms in about 600 ps. Some of the LVDS-to-ttl receivers are about that fast, but not quite so fierce a drive.

The spec sheet typ EP05 risetime is 130 ps. These things aren't schmitts but they do have very high voltage gains, so they don't need a super-fast input edge. Something pokey like 1 volt per ns should do fine.

John

and then some!

Cheers Terry (who doesnt wear a watch, as one nearly got him killed)

that

Never been a Gamer, but Today's XKCD was sureal, one of the best

martin

that

He's the best techie cartoonist. Damn that emacs! I await the next Richard Stallman cartoon :)

Cheers Terry

that

Futurama has some goodies as well

martin

that

Coincidentally, I just choked on my coffee over this older one:

He's the new Scott Adams, just as surely as Python is the new BASIC.

Best regards, Spehro Pefhany

Thanks again for the info, John. I have been looking at the NL37WZ16 datasheet, but can't infer the 600ps figure from anywhere. Either I am blind or the only mention to rise and fall times is in Fig.3 where it says 2.5ns. It seems that OnSemi does not have any faster triple buffers in that series either. Perhaps this is an "undocumented" behavior when you connect the outputs together? There is no mention to such low output impedances either (500R is the lowest that appears) although the 24mA output sink and source capability per output suggest somewhat lower values.

I will try as soon as I get some!

Pere

that

That's brilliant, I like this one

martin

The risetime isn't spec'd. Fig 3 refers to propagation delay; I suppose Tr and Tf refer to the input, not the output. The typical Voh and Vol values hint at the drive capability of three sections in parallel.

John