some breadboards

Are these your first pass board designs that you plan to put into production?

I usually breadboard to evaluate parts that are underspecified (or whose specs I don't trust) or to try tricky circuits. This is usually nanosecond/picosecond stuff and occasionally things like the power inverter. I never breadboard or prototype entire products.

Testing parts is not the same thing as prototyping products. If we really understand the parts and design carefully, a product *will* work first etch, without prototype spins. IC manufacturers do their best to hid defects and weirdnesses, making us discover them over and over all around the world [1].

I especially like to test parts to destruction, to see what sorts of margins I'm going to have in real life. Earlier this week I blew up a bunch of 1206 resistors and SSRs to see how a product might behave if arbitrarily over-voltaged, to estimate which parts would fail, and when, so I could put some limits on a datasheet. I now have, essentially, SOAR curves for these parts.

And one of my guys just finished testing a bunch of high-ohm resistors for shot and excess noise. That is a surprisingly difficult thing to do, but crucial to a product we're working on. If we'd just dropped some standard 100M surface-mount cermet resistors into the gadget, it wouldn't have worked.

John

[1] somebody with time should start an ICs Bug web site.

Nice! ;-)

If I test the parts I don't fully understand, the products will work first pass. *That* is nice.

A breadboard is not a prototype. It's a way to better understand a part. The "circuit" of such a breadboard is unlikely to appear in any deliverable product.

John

First pass design means first pass design. This should all be modeled and you should not be wasting someones time to build these boards when a handy dandy simulator program is available.

OK - I'm just ribbing you. You obviously know your business and how to most effectively get to production, but on size does not fit all in electrical design.

Just bustin' your chops, John. I'm on your side in that kerfuffle (though don't often come up to your standards), but funny is funny.

You can't usefully simulate a system without good part models, and lots of parts have no or bad models. Especially RF parts used in nonlinear or time-domain apps.

I like soldering and playing with parts once in a while, and blowing things up. When we make measurements like this, we write it up (or just photograph the breadboards/notes/whiteboards) and add it to the datasheets and stuff we keep in our parts database. So we don't lose it.

What's tricky is to decide how far you might go beyond a part's published specs, or how to make you own limits from measurements when there aren't suitable specs.

1206 resistor at 26 watts for 1 millisecond!

John

Where we got burned years ago was using 1206 resistors in high peak power very low duty cycle applications. The average power dissipation was well within spec, but the peak voltages were ruining the resistors. (Domain Knowledge learned the hard way)

My case is just for an overload rating for the 4-20 mA i/o channel thing we're doing. We're specifying that the customer can apply up to

50 volts across our pins without damaging it. We of course hope nobody ever will. The resistors will dissipate about 26 watts, and the supervisory CPU should shut off a series SSR in a couple of hundred microseconds. The resistors explode at 65 watts in 1 millisecond, pulsed at 1 Hz, so we have a pretty decent margin.

The 50 volts is arbitrary, just based on what we think is safe. We could spec it at 40 if we wanted.

Time to cook some salmon cakes.

John

And besides, I need to solder and play with parts now and then. If I ever quit soldering, just have me recycled.

John

John, I also like the "Reality Approach"!

And sometimes the 'input signal' to a device isn't what you really thought it was. Even if (especially if?) it is coming from another of my designs :-)

Then what the device DOES with that signal and therefore what the output signal really is, can be a nasty surprise.

As my favorite tech at IBM used to say, "If what the scope shows us is Wrong or Impossible that is it's way of telling us that there are things we don't understand!" (Later they made him a Manager. What a waste! )

Regards, Terry King ..On the Red Sea at KAUST snipped-for-privacy@terryking.us

[...]

Some guys have all the fun!

Frank McKenney

--
  The preservation of truth values when translating scientific prose
  is very nearly as delicate a task as the preservation of resonance
  and emotional tone in the translation of literature. Neither can
  be fully achieved; even responsible approximation requires the
  greatest tact and taste.
                    -- Thomas S. Kuhn / The Road Since 'Structure'

Models are also bad in the following cases:

Basically no op-amp model correctly shows the power supply current.

Op-amps that are crashed into the rails or slew rate limited don't recover like the model suggests.

Linears switcher chips are about the only ones that have anything like models that work.

The recovery of the body diode in power MOSFETs is almost never fully specified in datasheets. Many have a long tail of noisy current.

The noise performance of DACs are seldom fully characterized.

Gate leakage in JFETs is usually not modeled.

Those parts seem relatively easy. In some cases I have a PCB made to test a particular piece of circuitry. That often leads to interesting discoveries!

--
Failure does not prove something is impossible, failure simply
indicates you are not using the right tools...
nico@nctdevpuntnl (punt=.)
--------------------------------------------------------------

A simulator doesn't reveal all pitfalls like latch-up effects. And in many circuits (even a switching regulator) the PCB is a part itself.

--
Failure does not prove something is impossible, failure simply
indicates you are not using the right tools...
nico@nctdevpuntnl (punt=.)
--------------------------------------------------------------

Yup. This one had several versions of an output driver, plus some other handy parts testers, and a couple of filters and attenuators.

ftp://jjlarkin.lmi.net/Z250A.jpg

When we do a multilayer like this, we pile on anything we can think of.

Somebody could make a home business out of consolidating peoples' needs for test circuits, doing the layout, ordering boards a couple of times a week, and shearing them up/mailing them out.

John

Except for the actual layout part, I think this is what places like PCBCart already do?

For the layout itself I'd think many people would want to do it themselves anyway, at least for switchers or anything high-speed or RF-related.

Their turn around time is in weeks. See the FAQ.

hamilton

[shows a 74ACT04 with connection points]

But, what's the point of breadboarding a fast gate without the required bypass capacitors on the power pins? It's not gonna behave like the final application does.