CSP light sensitivity / google groups

[sea moss]

A friend of mine recently root caused some strange circuit behavior down to LED light affecting a "chip scale package" (CSP) analog IC.

I did some internet browsing and only found a couple mentions of this phenomenon... I can't help but wonder if the app note was written as a band-aid after customers first discovered the behavior.

Has anyone here experienced this? And for bonus points, which pn junctions are most sensitive: do they necessarily need to be connected to a pin (e.g. ESD diodes), or could light get into the internal circuits as well?

And I couldn't resist posting something on google groups on 2/21/24... Been a pleasure lurking in this forum, hope to see all the seasoned (and sometimes salty) posters continue on Usenet elsewhere.


https://www.silabs.com/documents/public/application-notes/an0878_methods_of_reducing_light_sensitivity_in_csp_packages.pdf

[john larkin]

On Wed, 21 Feb 2024 13:45:29 -0800 (PST), sea moss

Post by sea moss
A friend of mine recently root caused some strange circuit behavior down to LED light affecting a "chip scale package" (CSP) analog IC.
I did some internet browsing and only found a couple mentions of this phenomenon... I can't help but wonder if the app note was written as a band-aid after customers first discovered the behavior.
Has anyone here experienced this? And for bonus points, which pn junctions are most sensitive: do they necessarily need to be connected to a pin (e.g. ESD diodes), or could light get into the internal circuits as well?
And I couldn't resist posting something on google groups on 2/21/24... Been a pleasure lurking in this forum, hope to see all the seasoned (and sometimes salty) posters continue on Usenet elsewhere.
https://www.silabs.com/documents/public/application-notes/an0878_methods_of_reducing_light_sensitivity_in_csp_packages.pdf

The first plastic-packaged GE NPN transistors were potted in a
translucent plastic and were photosensitive. They picked up hum from
flourescent lights and your DC offets would go crazy if your boss
leaned over your bench and blocked the light.

I wonder if the EPC GaN fets are photosensitive. They are BGA bare
die. I might try that.

[Lasse Langwadt]

Post by john larkin
On Wed, 21 Feb 2024 13:45:29 -0800 (PST), sea moss

Post by sea moss
A friend of mine recently root caused some strange circuit behavior down to LED light affecting a "chip scale package" (CSP) analog IC.
I did some internet browsing and only found a couple mentions of this phenomenon... I can't help but wonder if the app note was written as a band-aid after customers first discovered the behavior.
Has anyone here experienced this? And for bonus points, which pn junctions are most sensitive: do they necessarily need to be connected to a pin (e.g. ESD diodes), or could light get into the internal circuits as well?
And I couldn't resist posting something on google groups on 2/21/24... Been a pleasure lurking in this forum, hope to see all the seasoned (and sometimes salty) posters continue on Usenet elsewhere.
https://www.silabs.com/documents/public/application-notes/an0878_methods_of_reducing_light_sensitivity_in_csp_packages.pdf

The first plastic-packaged GE NPN transistors were potted in a
translucent plastic and were photosensitive. They picked up hum from
flourescent lights and your DC offets would go crazy if your boss
leaned over your bench and blocked the light.
I wonder if the EPC GaN fets are photosensitive. They are BGA bare
die. I might try that.

the first Raspberry pi 2 would reset if photographed with flash light,
turned out to be the power supply chip package that was transparent to
the light from a photo flash, and the light upset it

[a***@spenarnc.xs4all.nl]

Post by john larkin
On Wed, 21 Feb 2024 13:45:29 -0800 (PST), sea moss

Post by sea moss
A friend of mine recently root caused some strange circuit behavior down to LED light affecting a "chip scale package" (CSP) analog IC.
I did some internet browsing and only found a couple mentions of this phenomenon... I can't help but wonder if the app note was written as a band-aid after

customers first discovered the behavior.

Post by sea moss
Has anyone here experienced this? And for bonus points, which pn junctions are most sensitive: do they necessarily need to be connected to a pin (e.g. ESD

diodes), or could light get into the internal circuits as well?

Post by sea moss
And I couldn't resist posting something on google groups on 2/21/24... Been a pleasure lurking in this forum, hope to see all the seasoned (and sometimes

salty) posters continue on Usenet elsewhere.

Post by sea moss
https://www.silabs.com/documents/public/application-notes/an0878_methods_of_reducing_light_sensitivity_in_csp_packages.pdf

The first plastic-packaged GE NPN transistors were potted in a
translucent plastic and were photosensitive. They picked up hum from
flourescent lights and your DC offets would go crazy if your boss
leaned over your bench and blocked the light.

I have an OC3 somewhere. (Early Philips germanium transistor.
Before germanium were called AC### ).
I thought the enclosure was glass, painted black
and the paint was removable. If the enclosure was plastic, it could
easily be made black plastic. Need it check it though.

Post by john larkin
I wonder if the EPC GaN fets are photosensitive. They are BGA bare
die. I might try that.

All electronic devices are sensitive for EM radiation as long as it
enters the crystal.

[Clive Arthur]

On 22/02/2024 10:04, Martin Brown wrote:

<snip>

Some early germanium transistors were potted in hazy epoxy inside a
bullet shaped glass envelope with matt black paint on the outside. OC71
being a notable Mullard part that is still curiously popular in guitar
fuzz circuits even today. You could convert the cheaper part into a fair
approximation of the more expensive phototransistor just by scraping off
the black paint. This no longer worked when transistors came in metal
cans. The latter went with a much bigger bang when they blew.

A lesson learnt as a technician when fixing an old outside broadcast
mixer at the BBC. Terrible mains hum, but try to see it on a 'scope and
it disappeared. Took a while to realise that leaning over with a 'scope
probe was shielding the Germanium transistors with worn paint from the
overhead fluorescent striplights.

Separately, some of the RF Germanium transistors, eg OC44, used some
sort of blue putty to protect the junctions. I believe, on very little
evidence, that it was what came to be known as 'Silly Putty', or at
least was very similar.

[piglet]

Post by Clive Arthur
<snip>

Some early germanium transistors were potted in hazy epoxy inside a
bullet shaped glass envelope with matt black paint on the outside. OC71
being a notable Mullard part that is still curiously popular in guitar
fuzz circuits even today. You could convert the cheaper part into a fair
approximation of the more expensive phototransistor just by scraping off
the black paint. This no longer worked when transistors came in metal
cans. The latter went with a much bigger bang when they blew.

A lesson learnt as a technician when fixing an old outside broadcast
mixer at the BBC. Terrible mains hum, but try to see it on a 'scope and
it disappeared. Took a while to realise that leaning over with a 'scope
probe was shielding the Germanium transistors with worn paint from the
overhead fluorescent striplights.
Separately, some of the RF Germanium transistors, eg OC44, used some
sort of blue putty to protect the junctions. I believe, on very little
evidence, that it was what came to be known as 'Silly Putty', or at
least was very similar.

That blue stuff was silicone gel.

[Jan Panteltje]

On a sunny day (Wed, 28 Feb 2024 11:58:42 -0000 (UTC)) it happened piglet

Post by piglet

Post by Clive Arthur
<snip>

Some early germanium transistors were potted in hazy epoxy inside a
bullet shaped glass envelope with matt black paint on the outside. OC71
being a notable Mullard part that is still curiously popular in guitar
fuzz circuits even today. You could convert the cheaper part into a fair
approximation of the more expensive phototransistor just by scraping off
the black paint. This no longer worked when transistors came in metal
cans. The latter went with a much bigger bang when they blew.

A lesson learnt as a technician when fixing an old outside broadcast
mixer at the BBC. Terrible mains hum, but try to see it on a 'scope and
it disappeared. Took a while to realise that leaning over with a 'scope
probe was shielding the Germanium transistors with worn paint from the
overhead fluorescent striplights.
Separately, some of the RF Germanium transistors, eg OC44, used some
sort of blue putty to protect the junctions. I believe, on very little
evidence, that it was what came to be known as 'Silly Putty', or at
least was very similar.

That blue stuff was silicone gel.

I still have a Valvo OC140 somewhere, the black paint is still intact though
https://www.radiomuseum.org/tubes/tube_oc140.html

In the late fifties or early sixties I build an audio preamp with OC71
just an open box, one day there was a noise.. then it was gone
turned out to be caused by the sun shining on the circuit.
Sun is dynamic (eruptions) but that much modulated?
Maybe alien signal ;-)
I have used all those OC13, OC16 (power), OC44 (RF to 15 Mhz) and killed a few over time.
The OC16 had a metal case
OC13 was the first transistor I ever used I think.
https://www.radiomuseum.org/tubes/tube_oc13.html

[Martin Brown]

Post by Jan Panteltje
On a sunny day (Wed, 28 Feb 2024 11:58:42 -0000 (UTC)) it happened piglet

Post by piglet

Post by Clive Arthur
<snip>

Some early germanium transistors were potted in hazy epoxy inside a
bullet shaped glass envelope with matt black paint on the outside. OC71
being a notable Mullard part that is still curiously popular in guitar
fuzz circuits even today. You could convert the cheaper part into a fair
approximation of the more expensive phototransistor just by scraping off
the black paint. This no longer worked when transistors came in metal
cans. The latter went with a much bigger bang when they blew.

A lesson learnt as a technician when fixing an old outside broadcast
mixer at the BBC. Terrible mains hum, but try to see it on a 'scope and
it disappeared. Took a while to realise that leaning over with a 'scope
probe was shielding the Germanium transistors with worn paint from the
overhead fluorescent striplights.
Separately, some of the RF Germanium transistors, eg OC44, used some
sort of blue putty to protect the junctions. I believe, on very little
evidence, that it was what came to be known as 'Silly Putty', or at
least was very similar.

That blue stuff was silicone gel.

I still have a Valvo OC140 somewhere, the black paint is still intact though
https://www.radiomuseum.org/tubes/tube_oc140.html
In the late fifties or early sixties I build an audio preamp with OC71
just an open box, one day there was a noise.. then it was gone
turned out to be caused by the sun shining on the circuit.
Sun is dynamic (eruptions) but that much modulated?

No but the light after travelling through our atmosphere and any air
currents in the room can have a fair bit of variation. The amount of
variation twinkling in stars (less for planets) can be used to infer the
diameter of unresolved objects by finding the separation between two
scopes at which they cease to be correlated. Hanbury-Brown and Twiss
first made the intensity interferometer work in the optical.

Michaelson & Peas beat them too it with a steel girder add-on in front
of the Mt Wilson 100", but it required an experimentalist of
Michaelson's calibre to make it work properly. They measured the
diameters of several of the brighter nearby stars with it.

Post by Jan Panteltje
Maybe alien signal ;-)
I have used all those OC13, OC16 (power), OC44 (RF to 15 Mhz) and killed a few over time.
The OC16 had a metal case
OC13 was the first transistor I ever used I think.
https://www.radiomuseum.org/tubes/tube_oc13.html

AF116 and AC128 were the ones in my first electronics kit.

Followed closely by BC107 and Ferranti e-line tinning failures that
merely required patient application of a soldering iron to fix them.

[Jan Panteltje]

On a sunny day (Thu, 29 Feb 2024 10:40:48 +0000) it happened Martin Brown

Post by Martin Brown

Post by Jan Panteltje
On a sunny day (Wed, 28 Feb 2024 11:58:42 -0000 (UTC)) it happened piglet

Post by piglet

Post by Clive Arthur
<snip>

Some early germanium transistors were potted in hazy epoxy inside a
bullet shaped glass envelope with matt black paint on the outside. OC71
being a notable Mullard part that is still curiously popular in guitar
fuzz circuits even today. You could convert the cheaper part into a fair
approximation of the more expensive phototransistor just by scraping off
the black paint. This no longer worked when transistors came in metal
cans. The latter went with a much bigger bang when they blew.

A lesson learnt as a technician when fixing an old outside broadcast
mixer at the BBC. Terrible mains hum, but try to see it on a 'scope and
it disappeared. Took a while to realise that leaning over with a 'scope
probe was shielding the Germanium transistors with worn paint from the
overhead fluorescent striplights.
Separately, some of the RF Germanium transistors, eg OC44, used some
sort of blue putty to protect the junctions. I believe, on very little
evidence, that it was what came to be known as 'Silly Putty', or at
least was very similar.

That blue stuff was silicone gel.

I still have a Valvo OC140 somewhere, the black paint is still intact though
https://www.radiomuseum.org/tubes/tube_oc140.html
In the late fifties or early sixties I build an audio preamp with OC71
just an open box, one day there was a noise.. then it was gone
turned out to be caused by the sun shining on the circuit.
Sun is dynamic (eruptions) but that much modulated?

No but the light after travelling through our atmosphere and any air
currents in the room can have a fair bit of variation.

Yes

Post by Martin Brown
The amount of
variation twinkling in stars (less for planets) can be used to infer the
diameter of unresolved objects by finding the separation between two
scopes at which they cease to be correlated. Hanbury-Brown and Twiss
first made the intensity interferometer work in the optical.

That was a bit more difficult, I took to wikipdia
https://en.m.wikipedia.org/wiki/Hanbury_Brown_and_Twiss_effect
https://en.m.wikipedia.org/wiki/Intensity_interferometer
l photo*

Post by Martin Brown
Michaelson & Peas beat them too it with a steel girder add-on in front
of the Mt Wilson 100", but it required an experimentalist of
Michaelson's calibre to make it work properly. They measured the
diameters of several of the brighter nearby stars with it.

Post by Jan Panteltje
Maybe alien signal ;-)
I have used all those OC13, OC16 (power), OC44 (RF to 15 Mhz) and killed a few over time.
The OC16 had a metal case
OC13 was the first transistor I ever used I think.
https://www.radiomuseum.org/tubes/tube_oc13.html

AF116 and AC128 were the ones in my first electronics kit.

Yes I remember those

Post by Martin Brown
Followed closely by BC107 and Ferranti e-line tinning failures that
merely required patient application of a soldering iron to fix them.

Back in those Ge days there were some Amroh kits too,
I once had one
http://www.hansotten.com/other-kits/amroh-step-by-step/
the one with the blue housing and the white knob.
Used it to test the range of my transistor medium wave transmiter..
Also some Philips one with tubes before that

Photo multipliers are fun
https://panteltje.nl/panteltje/pic/sc_pic/
link is from 2010 I think
This uses a Russian PMT (in the cardboard tube) and is still in use to see what radiates when the bomb falls:
Loading Image...
Loading Image...
Loading Image...
The whole thing runs on 2 AA batteries...

These days they use some photo diodes as detector I think, should get and try some...

[Jasen Betts]

Post by john larkin
On Wed, 21 Feb 2024 13:45:29 -0800 (PST), sea moss

Post by sea moss
A friend of mine recently root caused some strange circuit behavior down to LED light affecting a "chip scale package" (CSP) analog IC.
I did some internet browsing and only found a couple mentions of this phenomenon... I can't help but wonder if the app note was written as a band-aid after customers first discovered the behavior.
Has anyone here experienced this? And for bonus points, which pn junctions are most sensitive: do they necessarily need to be connected to a pin (e.g. ESD diodes), or could light get into the internal circuits as well?
And I couldn't resist posting something on google groups on 2/21/24... Been a pleasure lurking in this forum, hope to see all the seasoned (and sometimes salty) posters continue on Usenet elsewhere.
https://www.silabs.com/documents/public/application-notes/an0878_methods_of_reducing_light_sensitivity_in_csp_packages.pdf

The first plastic-packaged GE NPN transistors were potted in a
translucent plastic and were photosensitive. They picked up hum from
flourescent lights and your DC offets would go crazy if your boss
leaned over your bench and blocked the light.

Some early germanium transistors were potted in hazy epoxy inside a
bullet shaped glass envelope with matt black paint on the outside. OC71
being a notable Mullard part that is still curiously popular in guitar
fuzz circuits even today. You could convert the cheaper part into a fair
approximation of the more expensive phototransistor just by scraping off
the black paint. This no longer worked when transistors came in metal
cans. The latter went with a much bigger bang when they blew.

Post by john larkin
I wonder if the EPC GaN fets are photosensitive. They are BGA bare
die. I might try that.

A fair number of consumer LCD digital watch circuits were photosensitive
in that they could not survive a photoshoot flash gun at close range.

I've seen watches that could not withstand direct sunlight. It was 1978 or
something like that, I think they were being given away at "Pauls
camera House" or similar shop as part of some promotion.

[Chris Jones]

Post by sea moss
A friend of mine recently root caused some strange circuit behavior down to LED light affecting a "chip scale package" (CSP) analog IC.
I did some internet browsing and only found a couple mentions of this phenomenon... I can't help but wonder if the app note was written as a band-aid after customers first discovered the behavior.
Has anyone here experienced this? And for bonus points, which pn junctions are most sensitive: do they necessarily need to be connected to a pin (e.g. ESD diodes), or could light get into the internal circuits as well?
And I couldn't resist posting something on google groups on 2/21/24... Been a pleasure lurking in this forum, hope to see all the seasoned (and sometimes salty) posters continue on Usenet elsewhere.
https://www.silabs.com/documents/public/application-notes/an0878_methods_of_reducing_light_sensitivity_in_csp_packages.pdf

If you have a really bright light, you might even get a chip to latch
up. If there is a pnpn structure e.g. formed by some grounded n-well in
a p-substrate process, that can be basically a SCR across the power
rails. If there are enough substrate contacts grounding the p substrate
near the grounded n-well then that can be ok (it's like putting a low
value resistor from the gate to cathode on the SCR making it hard to
trigger), but if there aren't enough substrate contacts and the light is
bright enough, you can turn it on, and possibly destroy the chip, or at
least make it need power-cycling.

This used to happen to me quite a lot when I was debugging chips on a
probe station, and I was cutting off tracks with a pulsed laser (to
determine which ones were coupling RF from one part of the chip to
another). To be safe one would turn off the power before cutting, and
turn it back on afterwards to see the effect, but sometimes I didn't
turn it off, because I was cutting a track that was necessary to start
up the chip and after the track was cut it would no longer be possible
to configure the chip registers to the desired state, and so sometimes I
triggered latchup.

Once I noticed this, out of curiousity I defocused the laser and scanned
around the chip to find all of the other places where I could make it
latch up easily, and then on the next mask revision I improved the
substrate contacts and/or removed the need to have grounded n-wells in
those places, and made the chip much less prone to latch-up.

[Bill Sloman]

Post by Chris Jones

Post by sea moss
A friend of mine recently root caused some strange circuit behavior
down to LED light affecting a "chip scale package" (CSP) analog IC.
I did some internet browsing and only found a couple mentions of this
phenomenon...  I can't help but wonder if the app note was written as
a band-aid after customers first discovered the behavior.
Has anyone here experienced this?  And for bonus points, which pn
junctions are most sensitive: do they necessarily need to be connected
to a pin (e.g. ESD diodes), or could light get into the internal
circuits as well?
And I couldn't resist posting something on google groups on
2/21/24...  Been a pleasure lurking in this forum, hope to see all the
seasoned (and sometimes salty) posters continue on Usenet elsewhere.
https://www.silabs.com/documents/public/application-notes/an0878_methods_of_reducing_light_sensitivity_in_csp_packages.pdf

If you have a really bright light, you might even get a chip to latch
up. If there is a pnpn structure e.g. formed by some grounded n-well in
a p-substrate process, that can be basically a SCR across the power
rails. If there are enough substrate contacts grounding the p substrate
near the grounded n-well then that can be ok (it's like putting a low
value resistor from the gate to cathode on the SCR making it hard to
trigger), but if there aren't enough substrate contacts and the light is
bright enough, you can turn it on, and possibly destroy the chip, or at
least make it need power-cycling.
This used to happen to me quite a lot when I was debugging chips on a
probe station, and I was cutting off tracks with a pulsed laser (to
determine which ones were coupling RF from one part of the chip to
another). To be safe one would turn off the power before cutting, and
turn it back on afterwards to see the effect, but sometimes I didn't
turn it off, because I was cutting a track that was necessary to start
up the chip and after the track was cut it would no longer be possible
to configure the chip registers to the desired state, and so sometimes I
triggered latchup.
Once I noticed this, out of curiosity I defocused the laser and scanned
around the chip to find all of the other places where I could make it
latch up easily, and then on the next mask revision I improved the
substrate contacts and/or removed the need to have grounded n-wells in
those places, and made the chip much less prone to latch-up.

In theory, when we were doing electron beam probing on unpackaged chips,
we could have run into this problem, but the chip was running under
vacuum in an evacuated chamber to let the electron beam get at it.

The chamber didn't have to be light-tight all the time, but we detected
the secondary electrons by letting them hit a scintillator
(Everhart–Thornley detector) so it did have to be light-tight when we
were using the electron beam to look at the chip.