Interesting experiment

[Robert]

Interesting demonstration of the principle that the return current for a
high
speed signal line over a ground plane follows the top trace even if it's not
the shortest path, to minimize inductance.

A gent took an experiment by Benjamin Franklin and turned it into a visual
demonstration of that principle.

They bend the trace on the top layer and you can see the return current
flow change to follow the top layer high speed signal.

And they do it with a Marx Generator.

http://www.sigcon.com/Pubs/news/8_08.htm

Robert

 

[Deefoo]

Interesting demonstration of the principle that the return current for a
high
speed signal line over a ground plane follows the top trace even if it's not
the shortest path, to minimize inductance.

A gent took an experiment by Benjamin Franklin and turned it into a visual
demonstration of that principle.

They bend the trace on the top layer and you can see the return current
flow change to follow the top layer high speed signal.

And they do it with a Marx Generator.

http://www.sigcon.com/Pubs/news/8_08.htm

Robert

Much like those (british?) homing pigeons that follow highways and
roundabouts to find their way home instead of going in a straight line.
Hmm... so maybe homing pigeons are in fact homing electrons? That would
explain why there are so many of them. But how do they put them in
batteries? Honey, I shrunk the pigeons?

--DF

 

[PeteS]

Interesting demonstration of the principle that the return current for a
high
speed signal line over a ground plane follows the top trace even if it's not
the shortest path, to minimize inductance.

A gent took an experiment by Benjamin Franklin and turned it into a visual
demonstration of that principle.

They bend the trace on the top layer and you can see the return current
flow change to follow the top layer high speed signal.

And they do it with a Marx Generator.

http://www.sigcon.com/Pubs/news/8_08.htm

Robert

I had an interesting problem related to return current paths a little
while ago. We had an LM87 triple temperature monitor, and the ground
pin was on a nice solid plane, but in the return path zone for high
currents. During operation, we would get wild readings. I asked about
it, and the only answer I got was that the device was sensitive to
ground noise. Note I had a bypass cap from ground pin to Vdd right next
to the device, but the cure was to *capacitively decouple* the ground
pin to a nearby ground plane that was not in the return current zone.
Worked fine after that.

Cheers

PeteS

 

[John Larkin]

Interesting demonstration of the principle that the return current for a
high
speed signal line over a ground plane follows the top trace even if it's not
the shortest path, to minimize inductance.

A gent took an experiment by Benjamin Franklin and turned it into a visual
demonstration of that principle.

They bend the trace on the top layer and you can see the return current
flow change to follow the top layer high speed signal.

And they do it with a Marx Generator.

http://www.sigcon.com/Pubs/news/8_08.htm

Robert

Sounds like HoJo is saying that it takes current to charge a
capacitor, and that if I show up at some trade show, and pay him a lot
of money to attend one of his seminars, he's willing to argue with me
about it.

Hey, I already paid for his silly book! In the Xilinx ads, this guy is
referred to as "the world's foremost authority on signal integrity."

John

 

[Joel Kolstad]

Hey, I already paid for his silly book! In the Xilinx ads, this guy is
referred to as "the world's foremost authority on signal integrity."

I'm sure he's perfectly humble in person. He's got plenty of company
too... Eric Bogatin, Doug Smith, and even Henry Ott, etc. all tend to overlap
his area of expertise to some degree. My experience has been that what those
guy teachs at seminars really is useful information to a lot of people --
there are a lot of, e.g., BSEE programs that are completely lacking in some
course along the lines of "practical high speed design" or somesuch. I've
never tried it myself, but I'm willing to bet you that a significant fraction
of new graduates couldn't correctly sketch the return current path of say, a
100MHz toggling output.

(And to demonstrate just how little I knew out of college, I failed an
interview question along the lines of "In a typical inverting or non-inverting
op-amp setup, how does the amount of noise at the output relate to that at the
input?" Oops...)

---Joel

 

[Phil Hobbs]

Sounds like HoJo is saying that it takes current to charge a
capacitor, and that if I show up at some trade show, and pay him a lot
of money to attend one of his seminars, he's willing to argue with me
about it.

Hey, I already paid for his silly book! In the Xilinx ads, this guy is
referred to as "the world's foremost authority on signal integrity."

John

Yeah, and it's a totally bogus demo anyway--a complete snake-oil job.
The inductance you'd need to make any significant difference to a 150 kV
spark isn't in the 1-microhenry range of that tiny loop--1 uH at 1 GHz
is only 6k ohms, and I doubt that Marx generator gets anywhere near
that, even in the high harmonics.

What's really going on is that the E field near the corners of the
squares is huge when the other conductor is underneath, which causes a
corona discharge on those squares, which causes the return path to
follow the signal.

Notice that the return wants to go diagonally when the trace turns a
corner. This is pure electrostatics, not inductance. Honestly, Howie,
can't you do any better than that?

(Anyone want to buy a slightly used copy of the "black magic" book? Cheap?)

Cheers,

Phil Hobbs

 

[John Larkin]

I'm sure he's perfectly humble in person. He's got plenty of company
too... Eric Bogatin, Doug Smith, and even Henry Ott, etc. all tend to overlap
his area of expertise to some degree. My experience has been that what those
guy teachs at seminars really is useful information to a lot of people --
there are a lot of, e.g., BSEE programs that are completely lacking in some
course along the lines of "practical high speed design" or somesuch. I've
never tried it myself, but I'm willing to bet you that a significant fraction
of new graduates couldn't correctly sketch the return current path of say, a
100MHz toggling output.

I design gigahertz/picosecond/microvolt/kilovolt stuff, and I probably
couldn't either. Never tried, actually.

John

 

[Joel Kolstad]

I design gigahertz/picosecond/microvolt/kilovolt stuff, and I probably
couldn't either. Never tried, actually.

Some people have naturally good instincts? I'm sure you'd admit there's
plenty of crappy designs out there!

 

[Joel Kolstad]

"Noise? I was never very good with noise - my first guess is that it's
amplified, either by AVol or by (R1 + R2)/R2."

I believe they were looking for an answer along the lines of, "Oh, it's about
1/beta," where beta -- the feedback factor -- is R2/(R1+R2) for both the
non-inverting AND inverting case...

I imagine if I'd done better, I suspect that next up would have been something
along similiar lines such as, "For an otherwise ideal op-amp with a 100MHz GBW
product, with a gain of -1, what's the resulting system GBW?" -- It's 50MHz,
although right out of college I'd have flubbed that too.

I did get a few questions regarding the return loss going through some
mismatched terminations correct...

So, what was the right answer, or do they just tell you "You got this one
wrong, you don't get the job"?

They were in "polite" mode where there wasn't any direct feedback of whether
or not you were giving them the correct answers. In the end, I wasn't offered
a job, and I had meanwhile conceded I wasn't qualified for it anyway, so I
wasn't particularly unhappy. (They were looking for someone to do high speed
op-amp IC design, and that really wasn't me...)

---Joel

 

[John Larkin]

Yeah, and it's a totally bogus demo anyway--a complete snake-oil job.
The inductance you'd need to make any significant difference to a 150 kV
spark isn't in the 1-microhenry range of that tiny loop--1 uH at 1 GHz
is only 6k ohms, and I doubt that Marx generator gets anywhere near
that, even in the high harmonics.

What's really going on is that the E field near the corners of the
squares is huge when the other conductor is underneath, which causes a
corona discharge on those squares, which causes the return path to
follow the signal.

Notice that the return wants to go diagonally when the trace turns a
corner. This is pure electrostatics, not inductance. Honestly, Howie,
can't you do any better than that?

(Anyone want to buy a slightly used copy of the "black magic" book? Cheap?)

Cheers,

Phil Hobbs

But the book is valuable for its humor. Reread section 7.4, "Return
Current And Its Relationship To Vias" for a good chuckle.

John

 

[Steve Sousa]

But the book is valuable for its humor. Reread section 7.4, "Return
Current And Its Relationship To Vias" for a good chuckle.

Hello:

So, "High-speed digital design : a handbook of black magic / Howard W.
Johnson, Martin Graham " is not a good reference??

Best Regards

Steve Sousa

 

[[email protected]]

Tim Williams skrev:

Well that's quite impressive, especially considering it takes a minimum of
five lines just to print "hello fucking world" in C ;-)

Tim

nah, only two lines. one line to include stdio.h and one line for the
code

-Lass

 

[John Larkin]

Hello:

So, "High-speed digital design : a handbook of black magic / Howard W.
Johnson, Martin Graham " is not a good reference??

Half of it is good stuff and half is nonsense. If you know enough to
tell which is which, you don't need the book.

John

 

[John Larkin]

Tim Williams skrev:


nah, only two lines. one line to include stdio.h and one line for the
code

-Lass

In PowerBasic, it's


PRINT "Hello, cruel world!"


or maybe


TCP OPEN PORT 2000 AT "192.345.678.001" AS # 1 TIMEOUT 500

PRINT #1, "Hello, internet!"



Either compiles to a single, few-kilobyte .EXE file.

John