Vbe stuff

[Jim Thompson]

On Fri, 20 Oct 2006 08:10:41 GMT, Robert Baer



Are you referring to these Pease articles:

What's All This Logarithmic Stuff, Anyhow?
Bob Pease | ED Online ID #6068 | June 14, 1999
http://www.elecdesign.com/Articles/ArticleID/6068/6068.html

What's All This Logarithmic Stuff, Anyhow? (Part 2)
Bob Pease | ED Online ID #11301 | November 7, 2005
http://www.elecdesign.com/Articles/ArticleID/11301/11301.html

Just curious if we are discussing the same articles here.

Glen

Methinks Robert is soldering on resistors to set current ;-)

...Jim Thompson

 

[Jonathan Kirwan]

Assuming T is fixed is ignoring the elephant in the room. T is the
most significant variable by far, that is why diode connected
transistors or Vbe multipliers make such good temperature sensors,
especially if you cancel out the poorly controlled Is with the two
current method which has been discussed here extensively.

Several things come to mind in your comment, Glen. (1) is the fact
that the T in the equation provided above actually isn't the only
effect that leads to using these junctions for temperature -- Is
itself is a function of temperature, enough so that it actually
reverses the sign of the rate of change. So the above equation
doesn't tell you the picture there. I can post the Is equation, if
you'd like (already done it before, here.) (2) at Vbe of a few tenths
of a volt and currents in the tens of microamps, self heating of the
die shouldn't be a problem, yes? And we are starting at around 300K,
after all, which means even for dT of 20K, we are talking about 6-7%.
(3) Is has a huge effect when comparing Vd of diodes and transistors.
(4) 'n' varies widely in diodes -- very widely.

So, would you please elaborate why you say that T is the most
significant variable?

Jon

 

[Jim Thompson]

Several things come to mind in your comment, Glen. (1) is the fact
that the T in the equation provided above actually isn't the only
effect that leads to using these junctions for temperature -- Is
itself is a function of temperature, enough so that it actually
reverses the sign of the rate of change. So the above equation
doesn't tell you the picture there. I can post the Is equation, if
you'd like (already done it before, here.) (2) at Vbe of a few tenths
of a volt and currents in the tens of microamps, self heating of the
die shouldn't be a problem, yes? And we are starting at around 300K,
after all, which means even for dT of 20K, we are talking about 6-7%.
(3) Is has a huge effect when comparing Vd of diodes and transistors.
(4) 'n' varies widely in diodes -- very widely.

So, would you please elaborate why you say that T is the most
significant variable?

Jon

Because it's in an exponential function that defines IS ?

...Jim Thompson

 

[Jonathan Kirwan]

Because it's in an exponential function that defines IS ?

Might as well post it:

Is(T) = Is(Tnom)*(T/Tnom)^3*e^[-(Eg/k)*(1/T-1/Tnom)]

The derivative of this with respect to T, combined with the kT/q
effect results in the -2mV/K net slope circa T=300K, yes?

So yes, there is a 3rd power there and a power of the inverse kelvins.
Good catch!

Still, we are talking about such little power for self-heating and the
likelihood of ambient temp changes that are beneath the (87-59)/2 or
14K difference needed. On the other hand, it's possible that handling
the parts by hand could do that. 37C internal body temps might,
through the fingers, create a useful difference between measurements I
suppose.

More to the point, though, this still doesn't deal with the basic Vd
differences in absolute magnitide, only the rate of change. It's
almost dead certain, from my limited experience, that this is due to
Is differences. Diodes very often have much larger Is than BJTs, from
my meager experiences.

So Jim, do you really imagine this is the most important explanation
in this case? My own bet is on the huge differences in Is and the
likely difference in 'n', possibly combined to a lesser degree with T
(granted.)

No matter what, I'll learn something from this discussion.

Jon

 

[Glen Walpert]

Several things come to mind in your comment, Glen. (1) is the fact
that the T in the equation provided above actually isn't the only
effect that leads to using these junctions for temperature -- Is
itself is a function of temperature, enough so that it actually
reverses the sign of the rate of change. So the above equation
doesn't tell you the picture there. I can post the Is equation, if
you'd like (already done it before, here.) (2) at Vbe of a few tenths
of a volt and currents in the tens of microamps, self heating of the
die shouldn't be a problem, yes? And we are starting at around 300K,
after all, which means even for dT of 20K, we are talking about 6-7%.
(3) Is has a huge effect when comparing Vd of diodes and transistors.
(4) 'n' varies widely in diodes -- very widely.

So, would you please elaborate why you say that T is the most
significant variable?

Jon

Just to make sure we are talking about the same thing, the question as
I understood it is, for a given (forward) I through a diode at "room
temperature", what is the resulting voltage, Vbe? I will take "room
temperature" to be between about 65 and 85 F, (my winter/summer
thermostat settings), or 291 to 302 K. Since we are talking about Vbe
as a hopefully useful and repeatable function of I we should restrict
the discussion to diode connected small signal transistors having
decent gain over the range of currents considered.

Considering the approximation

I = Is(e^qVbe/kT - 1)

I = is the collector current.
Is= Emitter saturation current, typically well below 10^-13 amps
in a well made silicon transistor. Very temperature dependent.
q = The unit-charge constant, 1.602?*10^-19 coulombs.
Vbe = Base-emitter voltage.
k = Boltzmann's Constant, 1.38062*10~-23 joules/deg K.
T = T-absolute, K (273.15 at 0C).

Then (neglecting the -1):

Vbe = kT/q ln(I/Is)

Vbe varies directly with T, at approx 2 mV per degree K, or about 22
mV over "room temperature"

Leaving the question, how much does Is cause Vbe to vary. Is is not
included in many transistor data sheets, but from my notes I have a
few data points at 25 C for some transistors commonly used for logging
(source unremembered):

Tr Vbe @ 1 mA Is
MAT01 .63 2.2 * 10^-14
MAT03 (PNP) .585 1.3 * 10^-13
LM194/394 .58 1.6 * 10^-13
MAT02/04 .57 2.3 * 10^-13

So, with around 60 mV variation due to Is it looks like you were
right! But IME with the transistors I have used for temp measurement
such as 2n2222a or 2n3906 the part to part variation due to Is has
been much less than the variation over the range of "room temperature"
- I think the MAT01 comes closest to your typical high gain small
signal transistor; the others listed have relatively large dies for
low bulk resistance resulting in better log conformance at higher
currents.

But any way you look at it, T cannot be ignored even over the range of
"room temperature".

Glen

 

[Robert Baer]

Well, I did explain it and as simply as I can. I must bow out, now.
Best of luck.

Jon

I do not think you said why the discrepancy.

 

[Robert Baer]

Assuming T is fixed is ignoring the elephant in the room. T is the
most significant variable by far, that is why diode connected
transistors or Vbe multipliers make such good temperature sensors,
especially if you cancel out the poorly controlled Is with the two
current method which has been discussed here extensively.




Pease just threw about some typical values for near room temp,
entirely correct as far as he went but not even close to a complete
analysis. I see no reason to expect to duplicate his approximate
typical values with any accuracy, and I am sure he did not say you
could. If you want the complete analysis, read the book, Barrie
Gilbert knows the subject fairly well. Vbe is in fact very well
behaved and predictable within well understood limits, I have used the
effect many times in 6-decade log amplifiers, multipliers and
temperature sensors with no difficulty. But then I read the book .
IMO a few SED posts are not a good subsitute for reading a
comprehensive reference book written by experts if you really want to
understand the subject.

All well and good, but he stated a value and used the word "fact".
Maybe i should re-do the measurement with the collector open...

 

[Robert Baer]

Assuming T is fixed is ignoring the elephant in the room. T is the
most significant variable by far, that is why diode connected
transistors or Vbe multipliers make such good temperature sensors,
especially if you cancel out the poorly controlled Is with the two
current method which has been discussed here extensively.




Pease just threw about some typical values for near room temp,
entirely correct as far as he went but not even close to a complete
analysis. I see no reason to expect to duplicate his approximate
typical values with any accuracy, and I am sure he did not say you
could. If you want the complete analysis, read the book, Barrie
Gilbert knows the subject fairly well. Vbe is in fact very well
behaved and predictable within well understood limits, I have used the
effect many times in 6-decade log amplifiers, multipliers and
temperature sensors with no difficulty. But then I read the book .
IMO a few SED posts are not a good subsitute for reading a
comprehensive reference book written by experts if you really want to
understand the subject.

As far as using a transistor with Vcb=0, with care and selection, one
can approach 11 decades of useage.

 

[Robert Baer]

Jonathan Kirwan wrote:
[snip]

You can do a one-point calculation of Is, just as I did earlier. That
works and is simple, but it may not be enough at times. You can do
better than this by doing curve-tracing of Ic versus Vce with a range
of fixed Vbe's set up. Then take each fixed Vbe that you used and
find it on the Vce (horizontal) axis and mark them. Take a vertical
line up from there and mark its intercept with the associated Vbe
curve you traced out. Do that for each Vbe you used. That will give
you points where Vbc=0. (Not done like you did with B and C shorted.)
This allows for a more accurate estimation of Is, drawing the line
back to its intercept with the Ic axis.

Jon

Forget all of that. KISS.
I made a measurement and it did not even come close to what Mr Pease
said.
I made another measurement at a different current with the same result.
I am puzzled about that discrepancy.

Robert, Post your lash-up. You must be doing something terribly
wrong.

...Jim Thompson

Very crude; the base and collector leads were twisted together and
poked into the "sideways: hole in the banana connector which then was
cinched up; the emitter lead was treated in a similar manner to the
other banana connector of an HP5326B in DVM mode.
An adhustable HP6228B supply in series with a fairly large resistor
and current meter supplied the current to the lash-up at the banana
connector plugin holes.
When i have time, i will try with the collector open.

 

[Robert Baer]

On Fri, 20 Oct 2006 08:10:41 GMT, Robert Baer




Are you referring to these Pease articles:

What's All This Logarithmic Stuff, Anyhow?
Bob Pease | ED Online ID #6068 | June 14, 1999
http://www.elecdesign.com/Articles/ArticleID/6068/6068.html

What's All This Logarithmic Stuff, Anyhow? (Part 2)
Bob Pease | ED Online ID #11301 | November 7, 2005
http://www.elecdesign.com/Articles/ArticleID/11301/11301.html

Just curious if we are discussing the same articles here.

Glen

Look at:
http://www.national.com/rap/Story/vbe.html

 

[Fred Bartoli]

Robert Baer a écrit :

As far as using a transistor with Vcb=0, with care and selection, one
can approach 11 decades of useage.

Where did you get that figure?
11 decades is 1pA to 100mA or 10pA to 1A.

1pA is dangerously close to Is for good linearity.
100mA, let alone 1A, will get you into trouble with self heating and
current crowding.

 

[Glen Walpert]

All well and good, but he stated a value and used the word "fact".
Maybe i should re-do the measurement with the collector open...

Thanks for the proper article reference. Two things come to mind as
possibilities: Pease did say he was discussing the characteristics of
a particular transistor ("This transistor ..."), and in the 5th from
last para discussing "what other factors" he mentions a Vbe difference
of "with monolithic npn .. only 5 or 10 mV", where variations in Is
would be the primary cause of differences as I now see it, and the
differences would be larger between devices not on the same die - up
to 60 mV just between a few transistors specifically designed for
logging, see data in another post. Secondly, your leads are fairly
well thermally connected to the power dissipating and presumably
somewhat warm HP5326B; I think you need to thermally isolate and
preferably closely control or at least accurately measure the temp of
your DUT.

 

[Jim Thompson]

Jonathan Kirwan wrote:
[snip]

You can do a one-point calculation of Is, just as I did earlier. That
works and is simple, but it may not be enough at times. You can do
better than this by doing curve-tracing of Ic versus Vce with a range
of fixed Vbe's set up. Then take each fixed Vbe that you used and
find it on the Vce (horizontal) axis and mark them. Take a vertical
line up from there and mark its intercept with the associated Vbe
curve you traced out. Do that for each Vbe you used. That will give
you points where Vbc=0. (Not done like you did with B and C shorted.)
This allows for a more accurate estimation of Is, drawing the line
back to its intercept with the Ic axis.

Jon

Forget all of that. KISS.
I made a measurement and it did not even come close to what Mr Pease
said.
I made another measurement at a different current with the same result.
I am puzzled about that discrepancy.

Robert, Post your lash-up. You must be doing something terribly
wrong.

...Jim Thompson

Very crude; the base and collector leads were twisted together and
poked into the "sideways: hole in the banana connector which then was
cinched up; the emitter lead was treated in a similar manner to the
other banana connector of an HP5326B in DVM mode.
An adhustable HP6228B supply in series with a fairly large resistor
and current meter supplied the current to the lash-up at the banana
connector plugin holes.
When i have time, i will try with the collector open.

Don't do that, you will further confuse yourself ;-)

What is it that is still troubling you?

...Jim Thompson

 

[Jim Thompson]

Look at:
http://www.national.com/rap/Story/vbe.html

Aha! The most important part, for you at least, is in the first few
paragraphs... "IS" is NOT a constant, but is a rather hairy function
of temperature.

The rest of it is typical Pease-eisms... off the wall bull-crap. He
should stick to talking about mountain climbing ;-)

The only important thing to remember is that DELTA VBE is
well-defined.

...Jim Thompson

 

[Fred Bloggs]

The rest of it is typical Pease-eisms... off the wall bull-crap.

Pease knows that but he has to simplify things for the masses of
bull-crap people...

 

[Jim Thompson]

Pease knows that but he has to simplify things for the masses of
bull-crap people...

No. It's just to fill the page ;-)

...Jim Thompson

 

[John Larkin]

Pease knows that but he has to simplify things for the masses of
bull-crap people...

Pease knows that he has to come up with a page of blather once a
month, and he does. Check the latest Electronics Design to see his
vacation pictures.

But Electronics Design is mainly blather, so he fits right in.

John

 

[Robert Baer]

Robert Baer a écrit :



Where did you get that figure?
11 decades is 1pA to 100mA or 10pA to 1A.

1pA is dangerously close to Is for good linearity.
100mA, let alone 1A, will get you into trouble with self heating and
current crowding.

I am only passing on what i read over 10 years ago..

 

[Robert Baer]

Thanks for the proper article reference. Two things come to mind as
possibilities: Pease did say he was discussing the characteristics of
a particular transistor ("This transistor ..."), and in the 5th from
last para discussing "what other factors" he mentions a Vbe difference
of "with monolithic npn .. only 5 or 10 mV", where variations in Is
would be the primary cause of differences as I now see it, and the
differences would be larger between devices not on the same die - up
to 60 mV just between a few transistors specifically designed for
logging, see data in another post. Secondly, your leads are fairly
well thermally connected to the power dissipating and presumably
somewhat warm HP5326B; I think you need to thermally isolate and
preferably closely control or at least accurately measure the temp of
your DUT.

When i have some spare time, i will try just the E-B junction and use
a transistor socket wired up for the purpose.
Other than measuring the environment (21C) and letting the transistor
"soak" in it for over 1/2 hour, how do you suggest i "measure" the
transistor temperature?
A Tektronix 7D13 that measures temperature with an external
transistor agrees with other measurements of the environment
temperature; and i used the same MPSA42.

 

[Robert Baer]

Jim Thompson wrote:

On Fri, 20 Oct 2006 08:10:41 GMT, Robert Baer



Jonathan Kirwan wrote:

[snip]


You can do a one-point calculation of Is, just as I did earlier. That
works and is simple, but it may not be enough at times. You can do
better than this by doing curve-tracing of Ic versus Vce with a range
of fixed Vbe's set up. Then take each fixed Vbe that you used and
find it on the Vce (horizontal) axis and mark them. Take a vertical
line up from there and mark its intercept with the associated Vbe
curve you traced out. Do that for each Vbe you used. That will give
you points where Vbc=0. (Not done like you did with B and C shorted.)
This allows for a more accurate estimation of Is, drawing the line
back to its intercept with the Ic axis.

Jon

Forget all of that. KISS.
I made a measurement and it did not even come close to what Mr Pease
said.
I made another measurement at a different current with the same result.
I am puzzled about that discrepancy.


Robert, Post your lash-up. You must be doing something terribly
wrong.

...Jim Thompson

Very crude; the base and collector leads were twisted together and
poked into the "sideways: hole in the banana connector which then was
cinched up; the emitter lead was treated in a similar manner to the
other banana connector of an HP5326B in DVM mode.
An adhustable HP6228B supply in series with a fairly large resistor
and current meter supplied the current to the lash-up at the banana
connector plugin holes.
When i have time, i will try with the collector open.

Don't do that, you will further confuse yourself ;-)

What is it that is still troubling you?

...Jim Thompson

The measurement i made of the MPSA452 in DCT mode at 21C was
significantly lower than what Mr Pease stated (implied as fact for
virtually all transistors).

Why do i have to repeat this so many times?