AM
Alan Melia
Sat, Aug 14, 2010 11:47 PM
Surely all this is a case of an engineer being able to
read......specifications :-))
Manufactures specify to sell parts, most of the important parameter are
minimum values, and the range is as wide as it can be. If you want a
parameter "bracketed" then you must specify that when buying and pay the
premium or take the expense of testing incomming batches for sensitive
parameters. I used to give lectures to new engineers on reading
specifications .....and what "Absolute maximum" actually means. Clever
engineers also love using operations that rely on unspecified parameters.
For example using a stock switching diode as a spike quencher in a relay
driver. The usually work (but not always!) but are generally not specified
for that service.
Spice parameters are probably a few "typical" samples of even "the one I
happened to choose to measure"!!
I had a case of some +/-12v transistor logic which made it was into a big
telecoms project. the prototypes worked fine. By the time of bulk
manufacture the supplier of the the main transistor had had several cost
improvement stages. The result was the kit was unreliable. The problem was
that first the base junction were swung to -12v by the "0" logic state and
second from a transistor with a fairly simple round dot geometry emitter,
the device had been replaced by an interdigital high ft chip (it still met
the basic "greater than" specs.)
The interdigital device suffered severe loss of gain after a period of
avalanching at several millamps (I think it also suffered electromigration)
much more than the circular geometry transistors. I think this was due to
the increased emitter periphery length. It was solved by placing a simple
diode in the emitter leg, or a clamp on the base. The engineer had not read,
or not understood the meaning of the Vebmax = 5v parameter, and his
prototype had worked. The production side were not very happy but eventually
were forced to junk thousands of 8inch square pcbs and do a re-layout. This
was in the days before computer simulation !! It might have helped!
Alan G3NYK
----- Original Message -----
From: "jimlux" jimlux@earthlink.net
To: "Discussion of precise time and frequency measurement"
time-nuts@febo.com
Sent: Saturday, August 14, 2010 10:42 PM
Subject: Re: [time-nuts] Simulation
Line got moved to other side of big ocean. Process got " tweaked" beta
I'm pretty sure that military grade parts have paperwork and processes
cover that case. I expect it costs a lot.
I think the same sort of service is available to high volume customers
less than military price.
just so..
In the space business, we call it "traceability to sand"... you haven't
lived til someone has a failed 2n2222, somwhere on some piece of
critical hardware, and they issue a GIDEP alert, and then the mission
assurance folks call you up and ask, "you don't by any chance have
2N2222's in your flight hardware do you?".. then there's the whole
manufacturer and date code hunt.. Looking through the build
documentation to find out. (or worse yet, if you had decided for some
reason to use the prototype, which you didn't keep such good records on,
but which you have photos of, and trying to read the date codes off the
assembled item with a magnifying glass)
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to
and follow the instructions there.
Surely all this is a case of an engineer being able to
read......specifications :-))
Manufactures specify to sell parts, most of the important parameter are
minimum values, and the range is as wide as it can be. If you want a
parameter "bracketed" then you must specify that when buying and pay the
premium or take the expense of testing incomming batches for sensitive
parameters. I used to give lectures to new engineers on reading
specifications .....and what "Absolute maximum" actually means. Clever
engineers also love using operations that rely on unspecified parameters.
For example using a stock switching diode as a spike quencher in a relay
driver. The usually work (but not always!) but are generally not specified
for that service.
Spice parameters are probably a few "typical" samples of even "the one I
happened to choose to measure"!!
I had a case of some +/-12v transistor logic which made it was into a big
telecoms project. the prototypes worked fine. By the time of bulk
manufacture the supplier of the the main transistor had had several cost
improvement stages. The result was the kit was unreliable. The problem was
that first the base junction were swung to -12v by the "0" logic state and
second from a transistor with a fairly simple round dot geometry emitter,
the device had been replaced by an interdigital high ft chip (it still met
the basic "greater than" specs.)
The interdigital device suffered severe loss of gain after a period of
avalanching at several millamps (I think it also suffered electromigration)
much more than the circular geometry transistors. I think this was due to
the increased emitter periphery length. It was solved by placing a simple
diode in the emitter leg, or a clamp on the base. The engineer had not read,
or not understood the meaning of the Vebmax = 5v parameter, and his
prototype had worked. The production side were not very happy but eventually
were forced to junk thousands of 8inch square pcbs and do a re-layout. This
was in the days before computer simulation !! It might have helped!
Alan G3NYK
----- Original Message -----
From: "jimlux" <jimlux@earthlink.net>
To: "Discussion of precise time and frequency measurement"
<time-nuts@febo.com>
Sent: Saturday, August 14, 2010 10:42 PM
Subject: Re: [time-nuts] Simulation
> Hal Murray wrote:
> >> Line got moved to other side of big ocean. Process got " tweaked" beta
is
> >> now 4x what it was.
> >
> > I'm pretty sure that military grade parts have paperwork and processes
to
> > cover that case. I expect it costs a lot.
> >
> > I think the same sort of service is available to high volume customers
at a
> > less than military price.
>
> just so..
>
> In the space business, we call it "traceability to sand"... you haven't
> lived til someone has a failed 2n2222, somwhere on some piece of
> critical hardware, and they issue a GIDEP alert, and then the mission
> assurance folks call you up and ask, "you don't by any chance have
> 2N2222's in your flight hardware do you?".. then there's the whole
> manufacturer and date code hunt.. Looking through the build
> documentation to find out. (or worse yet, if you had decided for some
> reason to use the prototype, which you didn't keep such good records on,
> but which you have photos of, and trying to read the date codes off the
> assembled item with a magnifying glass)
>
>
> _______________________________________________
> time-nuts mailing list -- time-nuts@febo.com
> To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
BC
Bob Camp
Sun, Aug 15, 2010 12:33 AM
Hi
Simulation might or might not have helped.
- Was Vbe breakdown even included in the Spice model
- If so did it ring bells (rare) or did it just clip without error ( common )
- Would the same designer who didn't understand it in the first place have seen it clipping at -5 and concluded "looks to be in spec at -5"
- Would any of it be reviewed in light of the new transistor or was the guy on another project by then
My favorite in the absolute max Vbe category is the typical class C RF amp. Look at the spec, check a few thousand working boards. Scratch head and move on. Lots of reverse bias on the base and they run forever.
Bob
On Aug 14, 2010, at 7:47 PM, "Alan Melia" alan.melia@btinternet.com wrote:
Surely all this is a case of an engineer being able to
read......specifications :-))
Manufactures specify to sell parts, most of the important parameter are
minimum values, and the range is as wide as it can be. If you want a
parameter "bracketed" then you must specify that when buying and pay the
premium or take the expense of testing incomming batches for sensitive
parameters. I used to give lectures to new engineers on reading
specifications .....and what "Absolute maximum" actually means. Clever
engineers also love using operations that rely on unspecified parameters.
For example using a stock switching diode as a spike quencher in a relay
driver. The usually work (but not always!) but are generally not specified
for that service.
Spice parameters are probably a few "typical" samples of even "the one I
happened to choose to measure"!!
I had a case of some +/-12v transistor logic which made it was into a big
telecoms project. the prototypes worked fine. By the time of bulk
manufacture the supplier of the the main transistor had had several cost
improvement stages. The result was the kit was unreliable. The problem was
that first the base junction were swung to -12v by the "0" logic state and
second from a transistor with a fairly simple round dot geometry emitter,
the device had been replaced by an interdigital high ft chip (it still met
the basic "greater than" specs.)
The interdigital device suffered severe loss of gain after a period of
avalanching at several millamps (I think it also suffered electromigration)
much more than the circular geometry transistors. I think this was due to
the increased emitter periphery length. It was solved by placing a simple
diode in the emitter leg, or a clamp on the base. The engineer had not read,
or not understood the meaning of the Vebmax = 5v parameter, and his
prototype had worked. The production side were not very happy but eventually
were forced to junk thousands of 8inch square pcbs and do a re-layout. This
was in the days before computer simulation !! It might have helped!
Alan G3NYK
----- Original Message -----
From: "jimlux" jimlux@earthlink.net
To: "Discussion of precise time and frequency measurement"
time-nuts@febo.com
Sent: Saturday, August 14, 2010 10:42 PM
Subject: Re: [time-nuts] Simulation
Line got moved to other side of big ocean. Process got " tweaked" beta
I'm pretty sure that military grade parts have paperwork and processes
cover that case. I expect it costs a lot.
I think the same sort of service is available to high volume customers
less than military price.
just so..
In the space business, we call it "traceability to sand"... you haven't
lived til someone has a failed 2n2222, somwhere on some piece of
critical hardware, and they issue a GIDEP alert, and then the mission
assurance folks call you up and ask, "you don't by any chance have
2N2222's in your flight hardware do you?".. then there's the whole
manufacturer and date code hunt.. Looking through the build
documentation to find out. (or worse yet, if you had decided for some
reason to use the prototype, which you didn't keep such good records on,
but which you have photos of, and trying to read the date codes off the
assembled item with a magnifying glass)
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to
and follow the instructions there.
Hi
Simulation might or might not have helped.
1) Was Vbe breakdown even included in the Spice model
2) If so did it ring bells (rare) or did it just clip without error ( common )
3) Would the same designer who didn't understand it in the first place have seen it clipping at -5 and concluded "looks to be in spec at -5"
4) Would any of it be reviewed in light of the new transistor or was the guy on another project by then
My favorite in the absolute max Vbe category is the typical class C RF amp. Look at the spec, check a few thousand working boards. Scratch head and move on. Lots of reverse bias on the base and they run forever.
Bob
On Aug 14, 2010, at 7:47 PM, "Alan Melia" <alan.melia@btinternet.com> wrote:
> Surely all this is a case of an engineer being able to
> read......specifications :-))
> Manufactures specify to sell parts, most of the important parameter are
> minimum values, and the range is as wide as it can be. If you want a
> parameter "bracketed" then you must specify that when buying and pay the
> premium or take the expense of testing incomming batches for sensitive
> parameters. I used to give lectures to new engineers on reading
> specifications .....and what "Absolute maximum" actually means. Clever
> engineers also love using operations that rely on unspecified parameters.
> For example using a stock switching diode as a spike quencher in a relay
> driver. The usually work (but not always!) but are generally not specified
> for that service.
>
> Spice parameters are probably a few "typical" samples of even "the one I
> happened to choose to measure"!!
>
> I had a case of some +/-12v transistor logic which made it was into a big
> telecoms project. the prototypes worked fine. By the time of bulk
> manufacture the supplier of the the main transistor had had several cost
> improvement stages. The result was the kit was unreliable. The problem was
> that first the base junction were swung to -12v by the "0" logic state and
> second from a transistor with a fairly simple round dot geometry emitter,
> the device had been replaced by an interdigital high ft chip (it still met
> the basic "greater than" specs.)
>
> The interdigital device suffered severe loss of gain after a period of
> avalanching at several millamps (I think it also suffered electromigration)
> much more than the circular geometry transistors. I think this was due to
> the increased emitter periphery length. It was solved by placing a simple
> diode in the emitter leg, or a clamp on the base. The engineer had not read,
> or not understood the meaning of the Vebmax = 5v parameter, and his
> prototype had worked. The production side were not very happy but eventually
> were forced to junk thousands of 8inch square pcbs and do a re-layout. This
> was in the days before computer simulation !! It might have helped!
>
> Alan G3NYK
>
>
> ----- Original Message -----
> From: "jimlux" <jimlux@earthlink.net>
> To: "Discussion of precise time and frequency measurement"
> <time-nuts@febo.com>
> Sent: Saturday, August 14, 2010 10:42 PM
> Subject: Re: [time-nuts] Simulation
>
>
>> Hal Murray wrote:
>>>> Line got moved to other side of big ocean. Process got " tweaked" beta
> is
>>>> now 4x what it was.
>>>
>>> I'm pretty sure that military grade parts have paperwork and processes
> to
>>> cover that case. I expect it costs a lot.
>>>
>>> I think the same sort of service is available to high volume customers
> at a
>>> less than military price.
>>
>> just so..
>>
>> In the space business, we call it "traceability to sand"... you haven't
>> lived til someone has a failed 2n2222, somwhere on some piece of
>> critical hardware, and they issue a GIDEP alert, and then the mission
>> assurance folks call you up and ask, "you don't by any chance have
>> 2N2222's in your flight hardware do you?".. then there's the whole
>> manufacturer and date code hunt.. Looking through the build
>> documentation to find out. (or worse yet, if you had decided for some
>> reason to use the prototype, which you didn't keep such good records on,
>> but which you have photos of, and trying to read the date codes off the
>> assembled item with a magnifying glass)
>>
>>
>> _______________________________________________
>> time-nuts mailing list -- time-nuts@febo.com
>> To unsubscribe, go to
> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>> and follow the instructions there.
>
>
> _______________________________________________
> time-nuts mailing list -- time-nuts@febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
>
AM
Alan Melia
Sun, Aug 15, 2010 10:36 AM
Hi Bob yes that was a point raised by Prof Nat Sokal after I published some
data, or rather he pointed out it happened in RF amps. I guess if you take
an used PA transistor out of service and measure it you might find the base
emitter junction very leaky, but does few mA of leakage matter so much in a
low impedance high drive power circuit.
Reliability asks "does it do the job it was designed for" not "is it as good
as new now"
Alan
G3NYK
----- Original Message -----
From: "Bob Camp" lists@rtty.us
To: "Discussion of precise time and frequency measurement"
time-nuts@febo.com
Sent: Sunday, August 15, 2010 1:33 AM
Subject: Re: [time-nuts] Simulation
Hi
Simulation might or might not have helped.
- Was Vbe breakdown even included in the Spice model
- If so did it ring bells (rare) or did it just clip without error (
- Would the same designer who didn't understand it in the first place
have seen it clipping at -5 and concluded "looks to be in spec at -5"
- Would any of it be reviewed in light of the new transistor or was the
guy on another project by then
My favorite in the absolute max Vbe category is the typical class C RF
amp. Look at the spec, check a few thousand working boards. Scratch head and
move on. Lots of reverse bias on the base and they run forever.
Hi Bob yes that was a point raised by Prof Nat Sokal after I published some
data, or rather he pointed out it happened in RF amps. I guess if you take
an used PA transistor out of service and measure it you might find the base
emitter junction very leaky, but does few mA of leakage matter so much in a
low impedance high drive power circuit.
Reliability asks "does it do the job it was designed for" not "is it as good
as new now"
Alan
G3NYK
----- Original Message -----
From: "Bob Camp" <lists@rtty.us>
To: "Discussion of precise time and frequency measurement"
<time-nuts@febo.com>
Sent: Sunday, August 15, 2010 1:33 AM
Subject: Re: [time-nuts] Simulation
> Hi
>
> Simulation might or might not have helped.
>
> 1) Was Vbe breakdown even included in the Spice model
> 2) If so did it ring bells (rare) or did it just clip without error (
common )
> 3) Would the same designer who didn't understand it in the first place
have seen it clipping at -5 and concluded "looks to be in spec at -5"
> 4) Would any of it be reviewed in light of the new transistor or was the
guy on another project by then
>
> My favorite in the absolute max Vbe category is the typical class C RF
amp. Look at the spec, check a few thousand working boards. Scratch head and
move on. Lots of reverse bias on the base and they run forever.
>
> Bob
JF
J. Forster
Sun, Aug 15, 2010 10:17 PM
It depends on whether the leakage continues to rise or stabilizes after a
"burn in" period.
-John
============
Hi Bob yes that was a point raised by Prof Nat Sokal after I published
some
data, or rather he pointed out it happened in RF amps. I guess if you take
an used PA transistor out of service and measure it you might find the
base
emitter junction very leaky, but does few mA of leakage matter so much in
a
low impedance high drive power circuit.
Reliability asks "does it do the job it was designed for" not "is it as
good
as new now"
Alan
G3NYK
----- Original Message -----
From: "Bob Camp" lists@rtty.us
To: "Discussion of precise time and frequency measurement"
time-nuts@febo.com
Sent: Sunday, August 15, 2010 1:33 AM
Subject: Re: [time-nuts] Simulation
Hi
Simulation might or might not have helped.
- Was Vbe breakdown even included in the Spice model
- If so did it ring bells (rare) or did it just clip without error (
- Would the same designer who didn't understand it in the first place
have seen it clipping at -5 and concluded "looks to be in spec at -5"
- Would any of it be reviewed in light of the new transistor or was the
guy on another project by then
My favorite in the absolute max Vbe category is the typical class C RF
amp. Look at the spec, check a few thousand working boards. Scratch head
and
move on. Lots of reverse bias on the base and they run forever.
It depends on whether the leakage continues to rise or stabilizes after a
"burn in" period.
-John
============
> Hi Bob yes that was a point raised by Prof Nat Sokal after I published
> some
> data, or rather he pointed out it happened in RF amps. I guess if you take
> an used PA transistor out of service and measure it you might find the
> base
> emitter junction very leaky, but does few mA of leakage matter so much in
> a
> low impedance high drive power circuit.
> Reliability asks "does it do the job it was designed for" not "is it as
> good
> as new now"
>
> Alan
> G3NYK
> ----- Original Message -----
> From: "Bob Camp" <lists@rtty.us>
> To: "Discussion of precise time and frequency measurement"
> <time-nuts@febo.com>
> Sent: Sunday, August 15, 2010 1:33 AM
> Subject: Re: [time-nuts] Simulation
>
>
>> Hi
>>
>> Simulation might or might not have helped.
>>
>> 1) Was Vbe breakdown even included in the Spice model
>> 2) If so did it ring bells (rare) or did it just clip without error (
> common )
>> 3) Would the same designer who didn't understand it in the first place
> have seen it clipping at -5 and concluded "looks to be in spec at -5"
>> 4) Would any of it be reviewed in light of the new transistor or was the
> guy on another project by then
>>
>> My favorite in the absolute max Vbe category is the typical class C RF
> amp. Look at the spec, check a few thousand working boards. Scratch head
> and
> move on. Lots of reverse bias on the base and they run forever.
>>
>> Bob
>
>
> _______________________________________________
> time-nuts mailing list -- time-nuts@febo.com
> To unsubscribe, go to
> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
>
>
J
jimlux
Mon, Aug 16, 2010 12:16 AM
It depends on whether the leakage continues to rise or stabilizes after a
"burn in" period.
-John
============
Hi Bob yes that was a point raised by Prof Nat Sokal after I published
some
data, or rather he pointed out it happened in RF amps. I guess if you take
an used PA transistor out of service and measure it you might find the
base
emitter junction very leaky, but does few mA of leakage matter so much in
a
low impedance high drive power circuit.
Reliability asks "does it do the job it was designed for" not "is it as
good
as new now"
Alan
We get into the argument about "still works ok in the circuit" vs
"doesn't meet databook specs" all the time at work. To some folks, "not
meet datasheet" = "failure", while if you have a circuit that needs a
gain of 10, and the part has a gain of 1000, and degrades to 500, it's
hardly failed.
And rarely, do you have the budget or time to do a real "life test" to
prove it experimentally.
In radiation environments, it's more of a continuous aging effect.. more
dose, more leakage. And what drives the conflict between designer and
reliability guys is that the effect isn't particularly predictable,
particularly between lots (because it's not something controlled in the
process)
(which makes life testing all that much more fun)
J. Forster wrote:
> It depends on whether the leakage continues to rise or stabilizes after a
> "burn in" period.
>
> -John
>
> ============
>
>
>> Hi Bob yes that was a point raised by Prof Nat Sokal after I published
>> some
>> data, or rather he pointed out it happened in RF amps. I guess if you take
>> an used PA transistor out of service and measure it you might find the
>> base
>> emitter junction very leaky, but does few mA of leakage matter so much in
>> a
>> low impedance high drive power circuit.
>> Reliability asks "does it do the job it was designed for" not "is it as
>> good
>> as new now"
>>
>> Alan
>
We get into the argument about "still works ok in the circuit" vs
"doesn't meet databook specs" all the time at work. To some folks, "not
meet datasheet" = "failure", while if you have a circuit that needs a
gain of 10, and the part has a gain of 1000, and degrades to 500, it's
hardly failed.
And rarely, do you have the budget or time to do a real "life test" to
prove it experimentally.
In radiation environments, it's more of a continuous aging effect.. more
dose, more leakage. And what drives the conflict between designer and
reliability guys is that the effect isn't particularly predictable,
particularly between lots (because it's not something controlled in the
process)
(which makes life testing all that much more fun)
MD
Magnus Danielson
Mon, Aug 16, 2010 12:43 AM
On 08/16/2010 02:16 AM, jimlux wrote:
We get into the argument about "still works ok in the circuit" vs
"doesn't meet databook specs" all the time at work. To some folks, "not
meet datasheet" = "failure", while if you have a circuit that needs a
gain of 10, and the part has a gain of 1000, and degrades to 500, it's
hardly failed.
And rarely, do you have the budget or time to do a real "life test" to
prove it experimentally.
In radiation environments, it's more of a continuous aging effect.. more
dose, more leakage. And what drives the conflict between designer and
reliability guys is that the effect isn't particularly predictable,
particularly between lots (because it's not something controlled in the
process)
(which makes life testing all that much more fun)
But armed with the knowledge of what parameters do degrade, you may do
simulations of various grades of degradation and show that with the
current knowledge of degradation and assumptions on its effects and
degrees, it is reasonable to expect functionality.
Right?
Simulation can be off assistance, if used wisely...
(or a tool of self deception if not used wisely)
Cheers,
Magnus
On 08/16/2010 02:16 AM, jimlux wrote:
> We get into the argument about "still works ok in the circuit" vs
> "doesn't meet databook specs" all the time at work. To some folks, "not
> meet datasheet" = "failure", while if you have a circuit that needs a
> gain of 10, and the part has a gain of 1000, and degrades to 500, it's
> hardly failed.
>
> And rarely, do you have the budget or time to do a real "life test" to
> prove it experimentally.
>
>
> In radiation environments, it's more of a continuous aging effect.. more
> dose, more leakage. And what drives the conflict between designer and
> reliability guys is that the effect isn't particularly predictable,
> particularly between lots (because it's not something controlled in the
> process)
>
> (which makes life testing all that much more fun)
But armed with the knowledge of what parameters do degrade, you may do
simulations of various grades of degradation and show that with the
current knowledge of degradation and assumptions on its effects and
degrees, it is reasonable to expect functionality.
Right?
Simulation can be off assistance, if used wisely...
(or a tool of self deception if not used wisely)
Cheers,
Magnus
