Vbe stuff

[Robert Baer]

Following what Mr Pease had to say about Vbe, ie read, at 27C that it
measures 799mV at 100uA, 640mV at 10uA, and 580mV at 1uA for 60mV per
decade.
But.
Measuring a 1N4006 at 21C i get 474mV at 100uA and 387mV at 10uA for
87mV per decade.
Measuring a MPSA42 in DCT mode at 21C i get 564mV at 100uA and 505mV
at 10uA for 59mV per decade.
Seems to make sense; the larger die area will have a lower current
density than a small die area, given the same current.
Comments?

 

[[email protected]]

Following what Mr Pease had to say about Vbe, ie read, at 27C that it
measures 799mV at 100uA, 640mV at 10uA, and 580mV at 1uA for 60mV per
decade.
But.
Measuring a 1N4006 at 21C i get 474mV at 100uA and 387mV at 10uA for
87mV per decade.
Measuring a MPSA42 in DCT mode at 21C i get 564mV at 100uA and 505mV
at 10uA for 59mV per decade.
Seems to make sense; the larger die area will have a lower current
density than a small die area, given the same current.
Comments?

Ebers-Moll, Gummel-Poon - have you looked at what the classical
transistor models predict?

As far as I know, the Gummel-Poon 64-parameter model of the bipolar
transistor is the one normally realised in Spice simulators, while the
simpler (and less accurate) Ebers-Moll model is the one taught by lazy
academics to undemanding students.

The evil behaviour of the 1N4006 is well known - I haven't a clue why
rectifier diodes behave oddly, but I am well aware that they do.

 

[Clifford Heath]

The evil behaviour of the 1N4006 is well known - I haven't a clue why
rectifier diodes behave oddly, but I am well aware that they do.

Something to do with being PIN diodes, perhaps?

 

[Jim Thompson]

Something to do with being PIN diodes, perhaps?

Probably

...Jim Thompson

 

[Jim Thompson]

Following what Mr Pease had to say about Vbe, ie read, at 27C that it
measures 799mV at 100uA, 640mV at 10uA, and 580mV at 1uA for 60mV per
decade.
But.
Measuring a 1N4006 at 21C i get 474mV at 100uA and 387mV at 10uA for
87mV per decade.
Measuring a MPSA42 in DCT mode at 21C i get 564mV at 100uA and 505mV
at 10uA for 59mV per decade.
Seems to make sense; the larger die area will have a lower current
density than a small die area, given the same current.
Comments?

Density has nothing to do with vbe change per decade _provided_ you're
not into where resistance is affecting the measurement.

...Jim Thompson

 

[Eeyore]

Following what Mr Pease had to say about Vbe, ie read, at 27C that it
measures 799mV at 100uA, 640mV at 10uA, and 580mV at 1uA for 60mV per
decade.
But.
Measuring a 1N4006 at 21C i get 474mV at 100uA and 387mV at 10uA for
87mV per decade.
Measuring a MPSA42 in DCT mode at 21C i get 564mV at 100uA and 505mV
at 10uA for 59mV per decade.
Seems to make sense; the larger die area will have a lower current
density than a small die area, given the same current.
Comments?

Bulk resistance ?

Graham

 

[Fred Bloggs]

Density has nothing to do with vbe change per decade _provided_ you're
not into where resistance is affecting the measurement.

...Jim Thompson

I suppose it has more to do with a design goal minimizing conduction
losses in the high current diode thereby prohibiting the use of very
large ratios of majority doping material density as with the transistor.
This will cause a departure from the simplified minority carrier density
at the transition region boundaries as a function of forward bias
because the minority carrier density on both sides of the transition
region change significantly with a coupled dependence.

 

[Jim Thompson]

I suppose it has more to do with a design goal minimizing conduction
losses in the high current diode thereby prohibiting the use of very
large ratios of majority doping material density as with the transistor.
This will cause a departure from the simplified minority carrier density
at the transition region boundaries as a function of forward bias
because the minority carrier density on both sides of the transition
region change significantly with a coupled dependence.

As someone else pointed out, the 1N4xxx devices are also very lightly
doped (almost PIN-diode-like) to get the "high-voltage" performance.

But, Fred, your dissertation sounds just like that... dissertation out
of the mouth of some _twisted_ PhD ;-)

...Jim Thompson

 

[Jonathan Kirwan]

As someone else pointed out, the 1N4xxx devices are also very lightly
doped (almost PIN-diode-like) to get the "high-voltage" performance.

But, Fred, your dissertation sounds just like that... dissertation out
of the mouth of some _twisted_ PhD ;-)

All the same, it makes some sense to me. Shockley's equation develops
from an assumption that the forward current is entirely due to
minority carrier diffusion in neutral regions, I think. I need to
think more about it, but on first cut it sounds like the right
direction.

Jon

 

[[email protected]]

Following what Mr Pease had to say about Vbe, ... 60mV per
decade.
But.
Measuring a 1N4006 at 21C i get 474mV at 100uA and 387mV at 10uA for
87mV per decade.
Measuring a MPSA42 in DCT mode at 21C i get 564mV at 100uA and 505mV
at 10uA for 59mV per decade.

The diode equation (Shockley, "Theory of p-n Junctions", Bell System
Technical Journal
#28, p 435-489 , 1949)
applies to planar semiconductor junctions.

I = Is * (exp( (Qe*V) / (k * T) ) - 1)

Is is dependent on the current-conducting area, T is the junction
temperature

The base-emitter junction of a modern transistor (MPS-A42) is a good
fit.

A 1N4006 is intended for high current, and there's a fault condition
called
thermal runaway that could funnel ALL the current in the diode through
one small
hot spot. To keep that fault from occurring, this kind of diode has
some
distributed electrical resistance added to the diode; the diode
equation might
still fit well, but you have to consider that you're measuring both the
diode and
the series resistor that's built-in.

At the high end of diode-equation compliance, low noise transistors can
follow
the equation for about eight decades in current (the deviations from
the equation
behavior are one part in a hundred million). At the low end, items
like point-contact
diodes have (literally) different diode area at different biases. Your
crystal radio
diode might not follow the formula well at all.

In addition to 1N4006 diodes not being good logarithmic-response
elements, high
current transistors are similarly treated. The 1N4006 has a surge
rating of 10A,
and the highest current rating I could find for a near-ideal transistor
was about
the 2A current rating (TIP 31).

 

[Jim Thompson]

All the same, it makes some sense to me. Shockley's equation develops
from an assumption that the forward current is entirely due to
minority carrier diffusion in neutral regions, I think. I need to
think more about it, but on first cut it sounds like the right
direction.

Jon

Large ratios of doping level are precluded by (primarily) breakdown
voltage and forward resistance considerations.

...Jim Thompson

 

[Ban]

Ebers-Moll, Gummel-Poon - have you looked at what the classical
transistor models predict?

As far as I know, the Gummel-Poon 64-parameter model of the bipolar
transistor is the one normally realised in Spice simulators, while the
simpler (and less accurate) Ebers-Moll model is the one taught by lazy
academics to undemanding students.

The evil behaviour of the 1N4006 is well known - I haven't a clue why
rectifier diodes behave oddly, but I am well aware that they do.

Well, in the theory there is a factor I=Is e^(Vf/mVt) with m being 1...2 for
diodes, which predicts this result. On the upper current range there comes a
point where the density of the injected minority carriers equals the
impurity concentration, so only majority carriers can flow more. Higher
blocking voltage has lower doping and this effect starts early.
For very low currents the recombination current(long charge carrier
lifetime) becomes higher than the diffusion current and forces m again to 2.

 

[Jonathan Kirwan]

Large ratios of doping level are precluded by (primarily) breakdown
voltage and forward resistance considerations.

But isn't forward resistance a valid part of this measurement being
discussed?

Jon

 

[Jim Thompson]

On Sun, 15 Oct 2006 09:51:18 -0700, Jim Thompson
[snip]

All the same, it makes some sense to me. Shockley's equation develops
from an assumption that the forward current is entirely due to
minority carrier diffusion in neutral regions, I think. I need to
think more about it, but on first cut it sounds like the right
direction.

Jon

Large ratios of doping level are precluded by (primarily) breakdown
voltage and forward resistance considerations.

But isn't forward resistance a valid part of this measurement being
discussed?

Jon

Certainly. I believe someone (was it you, Jon?) mentioned PIN diodes.
I believe that the 1N4xxx series is very PIN-like in doping. I'll ask
someone at ON Semi.

...Jim Thompson

 

[Eeyore]

But isn't forward resistance a valid part of this measurement being
discussed?

I'm confident of it. I've never seen rbb specified for devices like the MPSAs
but I have a vague recollection of seeing a figure of ~ 1k somewhere once for
small signal parts. If so, the 100uA would account for 100mV in the resisitve
part of Vbe.

Graham

 

[Winfield Hill]

Eeyore wrote...

I'm confident of it. I've never seen rbb specified for devices like
the MPSAs but I have a vague recollection of seeing a figure of ~
1k somewhere once for small signal parts. If so, the 100uA would
account for 100mV in the resisitve part of Vbe.

Whoa, 100mV is a serious error! Let's talk 10mV; that'd be 10uA
of base current, which would correspond to about 2mA of collector
current. Since the measured characteristics don't deviate from
the ideal by that much by 2mA, I'd imagine rbb' must be under 1k.
But we're probably in the right ballpark. A 1k rbb' corresponds
to 4nV of lowest en noise at high collector currents. I don't
know the lowest en value for mppa parts, but I imagine they're
in the 2 to 5nV region; certainly not below 2nV (250 ohms).

 

[Clifford Heath]

Certainly. I believe someone (was it you, Jon?) mentioned PIN diodes.
I believe that the 1N4xxx series is very PIN-like in doping.

No, that was me. I learned it here, during a discussion
I started about Tom McEwan avalanching them (at LLNL) to
get 100ps pulses peaking around 100KW. He used a 1000V
step of a couple of ns into a pulse-forming network, and
the MOSFET step generator was also extensively discussed.

 

[Eeyore]

Eeyore wrote...

Whoa, 100mV is a serious error!

Look at the numbers. The OP did mention 100uA btw.

Let's talk 10mV; that'd be 10uA
of base current, which would correspond to about 2mA of collector
current. Since the measured characteristics don't deviate from
the ideal by that much by 2mA, I'd imagine rbb' must be under 1k.
But we're probably in the right ballpark. A 1k rbb' corresponds
to 4nV of lowest en noise at high collector currents. I don't
know the lowest en value for mppa parts, but I imagine they're
in the 2 to 5nV region; certainly not below 2nV (250 ohms).

One of the few parts I've seen En for was a Rohm part ( 2SB737) which was ~
600pV. That had an rbb' of ~ 10 ohms IIRC.

Graham

 

[Robert Baer]

Ebers-Moll, Gummel-Poon - have you looked at what the classical
transistor models predict?

As far as I know, the Gummel-Poon 64-parameter model of the bipolar
transistor is the one normally realised in Spice simulators, while the
simpler (and less accurate) Ebers-Moll model is the one taught by lazy
academics to undemanding students.

The evil behaviour of the 1N4006 is well known - I haven't a clue why
rectifier diodes behave oddly, but I am well aware that they do.

OK; scratch the 1N4006; the MPSA42 did not conformeither; it just was
"closer".

 

[Robert Baer]

Density has nothing to do with vbe change per decade _provided_ you're
not into where resistance is affecting the measurement.

...Jim Thompson

The MPSA42 has a multi-milliamp rating,and should follow the log
relationship to at least 10mA or more; i measured at 100uA max - so
internal resistance is not significant.