Happy New Year!!

[Jim Thompson]

Just got a New Years E-mail announcement from a customer that
*another* of my mixed-signal integrated circuit designs came out of
the foundry WORKING ON THE FIRST PASS!

Whoooopeeeee!

...Jim Thompson

 

[Tom Del Rosso]

In Jim Thompson typed:

Just got a New Years E-mail announcement from a customer that
*another* of my mixed-signal integrated circuit designs came out of
the foundry WORKING ON THE FIRST PASS!

Whoooopeeeee!

Congrats!

But isn't this a fairly routine thing for you? How often are revisions
necessary after prototyping? Does the layout often need revising when
the circuit is okay?

 

[Tim Auton]

Just got a New Years E-mail announcement from a customer that
*another* of my mixed-signal integrated circuit designs came out of
the foundry WORKING ON THE FIRST PASS!

Well done Jim.

Out of curiosity, how does an IC designer fault-find when an IC
doesn't work? I guess you can't stick your scope probes in different
parts of the circuit to see what's happening like you can with a PCB.


Tim

 

[Jim Thompson]

In Jim Thompson typed:

Congrats!

But isn't this a fairly routine thing for you? How often are revisions
necessary after prototyping? Does the layout often need revising when
the circuit is okay?

Usually. But I've had a recent spate of two-turners (*). I thought I
was losing my touch.

(*) Mostly due to customers not really specifying what they wanted...
but *I* should have more firmly enforced specification reviews. I've
now made that a company policy with specifications more thoroughly
spelled out and rigorous sign-off procedures, with penalties for
specification changes after first silicon.

...Jim Thompson

 

[Mike Engelhardt]

Tim,

...Out of curiosity, how does an IC designer fault-find when an IC
doesn't work? I guess you can't stick your scope probes in different
parts of the circuit to see what's happening like you can with a PCB.

IC's are scoped out much like PCB's. One method is with a microprober.
There the die is placed on a 3D micro-positioning stage under an
optical microscope and oscilloscope probes as tiny wires are touched
to the traces. This method losses steam with about a 1 micron feature
size. The next method is to use a scanning electron microscope to image
the die and then capture the voltage waveforms with a phenomenon called
voltage contrasting and boxcar integration. This works well for
repetitive waveforms and has the advantage that it doesn't load the
circuit like a mechanical probe capacitance, so rise times of <100ps
can be measured if you have a bright enough electron source. Additional,
propriety techniques exist that are developed by the lead microprocessor
companies, or by companies in contract with them, to probe their IC's.

--Mike

 

[Jim Thompson]

Well done Jim.

Out of curiosity, how does an IC designer fault-find when an IC
doesn't work? I guess you can't stick your scope probes in different
parts of the circuit to see what's happening like you can with a PCB.


Tim

Actually you can (sometimes, if you provide test pads that are on the
top metal layer). They're called pico-probes. Also I tend to make up
a test metal pattern that allows checking cells out individually for
complex chips.

...Jim Thompson

 

[Mike]

Tim,


IC's are scoped out much like PCB's. One method is with a microprober.
There the die is placed on a 3D micro-positioning stage under an
optical microscope and oscilloscope probes as tiny wires are touched
to the traces. This method losses steam with about a 1 micron feature
size.

I spent much of December probing a 0.18u 4-level-metal GHz-frequency chip.
It wasn't as painful as I thought it might be: probe points were added
using FIB (focused ion beam), and picoprobes can handle GHz signals with no
problem these days. FIB is great, but requires ~0.5u clearance around the
point on a wire where you want to make contact.

-- Mike --

 

[Mike Engelhardt]

Mike,

I spent much of December probing a 0.18u 4-level-metal GHz-frequency chip.
It wasn't as painful as I thought it might be: probe points were added
using FIB (focused ion beam), and picoprobes can handle GHz signals with no
problem these days. FIB is great, but requires ~0.5u clearance around the
point on a wire where you want to make contact.

Yes! I'm familiar with the techniques, I've been involved in design
both e-beam probes and FIB machines. For the original poster, putting
test points on IC's either as part of initial design for testability or
forensically by FIB is analogous to putting test points on PCB's.
Like I said, you need something of the order of 1um to mechanically
probe. Also, e-beam probing not only has response way over 10GHz, it
does it with essentially no loading of the circuit. Capacitive loading
is *so* low, you have some capability to probe traces buried under Si02
passivation or even a Nitride cap by measuring the voltage fluctuations
capacitively coupled to the top surface of passivation. The probe
capacitive loading of an e-beam robe is even a tiny fraction of that
capacitance.

--Mike

 

[Bill Sloman]

Tim,


IC's are scoped out much like PCB's. One method is with a microprober.
There the die is placed on a 3D micro-positioning stage under an
optical microscope and oscilloscope probes as tiny wires are touched
to the traces. This method losses steam with about a 1 micron feature
size. The next method is to use a scanning electron microscope to image
the die and then capture the voltage waveforms with a phenomenon called
voltage contrasting and boxcar integration. This works well for
repetitive waveforms and has the advantage that it doesn't load the
circuit like a mechanical probe capacitance, so rise times of <100ps
can be measured if you have a bright enough electron source. Additional,
propriety techniques exist that are developed by the lead microprocessor
companies, or by companies in contract with them, to probe their IC's.

I worked on a couple of E-beam testers at Cambridge Instruments from 1982 to
1991.

We never got the stroboscopic pulse width below 500psec, and even getting
there was sort of interesting - the electrons are only travel at about 10%
of the speed of light in the sort of low voltage column you use for voltage
contrast, so the beam-blanking electrodes have to be shorter than 15mm (or
some kind of travelling wave structure) which can be tricky. I got one of my
(two) patents out of that.

With that sort of narrow pulse, you are often only letting through one
electron per pulse or less (not even one electron on some pulses), which
makes it rather slow to get a decent image/waveform - a high-brightness
source would have helped. We used lanthenum boride sources - a hot field
emission source would have been brighter, and probably sufficiently stable
for the job, but Cambridge Instruments didn't have access to a suitable
source at the time.

Fun stuff.

 

[Mike Engelhardt]

Bill,

I worked on a couple of E-beam testers at Cambridge Instruments from 1982
to 1991.

My own association was with the Fairchild e-beam group has had been
purchased by Schlumberger. I wrote the charged-particle optic simulator,
designed some of the microscope objective/spectrometers and designed some
beam deflection electronics.

We never got the stroboscopic pulse width below 500psec, and even getting
there was sort of interesting - the electrons are only travel at about 10%
of the speed of light in the sort of low voltage column you use for voltage
contrast, so the beam-blanking electrodes have to be shorter than 15mm (or
some kind of travelling wave structure) which can be tricky. I got one of my
(two) patents out of that.

With that sort of narrow pulse, you are often only letting through one
electron per pulse or less (not even one electron on some pulses), which
makes it rather slow to get a decent image/waveform - a high-brightness
source would have helped. We used lanthenum boride sources - a hot field
emission source would have been brighter, and probably sufficiently stable
for the job, but Cambridge Instruments didn't have access to a suitable
source at the time.

Yes., a LAB6 source isn't bright enough for this. We used a TEM in the
released product.

--Mike

 

[Aubrey McIntosh]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

With that sort of narrow pulse, you are often only letting through
one electron per pulse or less (not even one electron on some
pulses), which makes it rather slow to get a decent image/waveform
- a high-brightness source would have helped. We used lanthenum
boride sources - a hot field emission source would have been
brighter, and probably sufficiently stable for the job, but
Cambridge Instruments didn't have access to a suitable source at
the time.

Fun stuff.

Could the lanthinum boride source be adapted to be a cold mass
spectrometer (electron) source?
Do you have some entries to the literature? I'm off campus this
semester and can't get into the indexes.




-----BEGIN PGP SIGNATURE-----
Version: PGPfreeware 7.0.3 for non-commercial use <http://www.pgp.com>

iQA/AwUBP/xixgKlSw8yssF7EQJgQACdHQ+zRJaa6xtVmmomQqoQpCZROoIAoLEx
PPbNnr+EqPkNUT/eFTWSgRrI
=yefo
-----END PGP SIGNATURE-----

 

[John Woodgate]

I read in sci.electronics.design that Aubrey McIntosh

Could the lanthinum boride source be adapted to be a cold mass
spectrometer (electron) source?

Lanthenum, lanthinum!! It's LANTHANUM. I don't normally pick up spelling
unless it causes confusion, but...


It's not even lanthanium in UK! (;-)

 

[Tim Auton]

Mike,

[snip info]

Thanks for the info all. Most interesting.


Tim

 

[Bret Cannon]

Lanthanum hexaboride is used for surface ionization in mass spectrometry to
produce negative ions. As I recall, there are problems with some types of
samples of contamination of the surface increasing the work function and so
decreasing the efficiency for producing negative ions.

Bret Cannon

 

[Bill Sloman]

My own association was with the Fairchild e-beam group has had been
purchased by Schlumberger. I wrote the charged-particle optic simulator,
designed some of the microscope objective/spectrometers and designed some
beam deflection electronics.

Then we probably have a common acquaintance in Neil Richardson.

Graham Plows' firm Lintech sold the first commercial electron beam
tester as an add-on unit, back around 1982, and my first job at
Cambridge Instrunments was getting it working on a couple of Cambridge
Instruments electron microscopes - not any easy task, because Graham
Plows had a salesman's approach to development, which was to maximise
the number of features on his system, rather than spending more time
on a smaller number of features that would work reliably.

Neil Richardson was his in-house electronic engineer at the time, and
I had some contact with him before he took one of the early units off
to Fairchild in California, to get it working and train the people who
were going to use it.
Effectively, he never came back, and Graham had to hire himself
another electronic engineer.

With Graham's example in front of him, Neil did a much better job on
the Fairchild-Schlumberger system, and once their unit was on the
market, Graham never sold another machine, and ended up closing down
Lintech in 1988, and coming to work at Cambridge Instruments as
Technical Director (and my boss), where we built a rather better
machine than Schlumberger's, which got canned after we'd got it to the
stage of a fully working prototype, in part because Graham had
resigned.

Interesting story, containing many libellous details only availalbe by
e-mail.

Yes., a LAB6 source isn't bright enough for this. We used a TEM in the
released product.

We were looking at a hot field emission source, and second column for
ion-milling and tungsten deposition, when the EBT2000 project got
canned.

The EBT2000 had a digital data acquisition system which allowd us to
sample stroboscopicly at close to 12MHz (it should have been 25MHz)
independent of the repetition rate of the waveform we were looking at
- which gave the LaB6 source a rather higher effective brightness than
it had in the then competing machines.

 

[Mike Engelhardt]

Bill,

...Then we probably have a common acquaintance in Neil Richardson...

Yep. Neil's vision of an e-beam probe did extremely well. As I
recall, it had 98% percent market share. Our group was featured on
the front cover the Schlumberger annual report one year. I always
had a very high regard for him. He moved on from the e-beam group
to go to New York to a position something like Grand Chief Technical
Executive Adviser or some other wizard-type title of Schlumberger.
Then he left that to head up a critical dimension group at KLA, or
something else even more executive. There was legal action to
prevent him from hiring too many people away from Schlumberger.
Myself, as the person who wrote the simulator and did the later
optic design Schlumberger make a stink. In the end, I left to
write bigger and better simulators and never looked back.

--Mike