[John Larkin]

300-ball BGA. ;-)

There are some BGA mosfets, tiny packages with absurdly low Rds-on
values.

John

 

[[email protected]]

Well, that approximation certainly works well at low currents.

Maybe it holds true generally including at large currents, I honestly don't know
about that (that's why I'm asking), but of course as the value of gm gets higher,
real device bulk resistances will raise their ugly heads, enter the equation and
change that relationship.





It reduces it proportionally by the numerical value of the feedback factor.

It never eliminates it. Indeed NFB can never *eliminate* distortion, that's why
for purists, designing circuits that are initially as linear as possible *before*
applying feedback is so important.

Graham

I believe there is a very old paper on negative feedback where a
mathematical proof is done to show that if you take an amplifier with
just 2nd harmonic distortion and apply feedback, the harmonics are now
infinite. Smaller of course, but now an infinite series rather than
just 2nd.

 

[Jim Thompson]

On Sat, 28 Jul 2007 13:51:32 -0700, [email protected] wrote:

[snip]

I believe there is a very old paper on negative feedback where a
mathematical proof is done to show that if you take an amplifier with
just 2nd harmonic distortion and apply feedback, the harmonics are now
infinite. Smaller of course, but now an infinite series rather than
just 2nd.

Probably written by some audiophool trying to support the "no
feedback" concept.

...Jim Thompson

 

[Michael A. Terrell]

Probably written by some audiophool trying to support the "no
feedback" concept.

Come on Jim! Audiophools don't write papers, they write ad copy.


--
Service to my country? Been there, Done that, and I've got my DD214 to
prove it.
Member of DAV #85.

Michael A. Terrell
Central Florida

 

[John Larkin]

Well, lessee if I can wrap my headbone around this: base current is
related to base voltage by that exponential curve with temperature in it,

Base input impedance, for a small ac signal, is about 26 * Beta / Ic,
where Ic is the dc collector current in mA. So for a garden-variety
NPN with beta=100, running at 10 mA collector current, an ac signal
pumped into the base sees a load of 260 ohms. Transconductance will be
(1/260) * 100, about 0.4 mhos.

Which is why I like fets.

and beta is either the collector current divided by the base current, or
the emitter current divided by the base current; so multiplying those
together would give you collector current/base volts, which is
transconductance, right?

Yup. But if you finagle the numbers, it turns out that
transconductance is independent of beta.

Transconductance is delta_Ic / delta-Vbe, which is a small-signal,
incremental slope thing that's only meaningful in the context of some
existing DC bias current.

Assume some dc bias currents. If you apply a small additional ac
voltage to the base, the base current is inverse with the base input
impedance. If beta were to, say, double, the base impedance would
double, the base ac input current would drop in half, but since the
beta doubled, that smaller base current will be amplified twice as
hard, so the ac collector current is unchanged. Ta-daah, Gm is
independent of beta!

For a fairly ideal transistor, Gm = 40 * Ic.

So it's a pretty big number, and temperature-dependent?

Fraction of a mho typically, and it varies with temperature only if
the dc collector current depends on temperature.

Certain parties will predictably nit-pick.

John

 

[Jim Thompson]

On Sat, 28 Jul 2007 16:29:06 -0700, John Larkin

[snip]

For a fairly ideal transistor, Gm = 40 * Ic.


Fraction of a mho typically, and it varies with temperature only if
the dc collector current depends on temperature.

Certain parties will predictably nit-pick.

John

John, Are you going to call me a nit-picker if I point out that...

"40" = q/(kT)

So your temperature-independence statement is dead wrong. You get
flat gm if IC _varies_linearly_ with temperature.

...Jim Thompson

 

[John Larkin]

On Sat, 28 Jul 2007 16:29:06 -0700, John Larkin

[snip]

For a fairly ideal transistor, Gm = 40 * Ic.


Fraction of a mho typically, and it varies with temperature only if
the dc collector current depends on temperature.

Certain parties will predictably nit-pick.

John

John, Are you going to call me a nit-picker if I point out that...

"40" = q/(kT)

So your temperature-independence statement is dead wrong. You get
flat gm if IC _varies_linearly_ with temperature.

...Jim Thompson

You mean with absolute tempearture? -273 is a long distance away, and
transistor betas vary by 5:1 or more on the datasheets, so it hardly
matters. Calcs like this don't have to be right by even 2:1, unless
you are designing ICs, which I don't think we are.

If I'd filled up a screen with equations, I'd be "right" but wouldn't
convey a lot of understanding. Having designed electronics for 40
years or so, transistor math at this level is plenty good enough.

OK, how about "hardly varies with temperature"?

John

 

[Jim Thompson]

On Sat, 28 Jul 2007 16:29:06 -0700, John Larkin

[snip]

For a fairly ideal transistor, Gm = 40 * Ic.

So it's a pretty big number, and temperature-dependent?

Fraction of a mho typically, and it varies with temperature only if
the dc collector current depends on temperature.

Certain parties will predictably nit-pick.

John

John, Are you going to call me a nit-picker if I point out that...

"40" = q/(kT)

So your temperature-independence statement is dead wrong. You get
flat gm if IC _varies_linearly_ with temperature.

...Jim Thompson

You mean with absolute tempearture? -273 is a long distance away, and
transistor betas vary by 5:1 or more on the datasheets, so it hardly
matters. Calcs like this don't have to be right by even 2:1, unless
you are designing ICs, which I don't think we are.

If I'd filled up a screen with equations, I'd be "right" but wouldn't
convey a lot of understanding. Having designed electronics for 40
years or so, transistor math at this level is plenty good enough.

OK, how about "hardly varies with temperature"?

John

BS, John! You're wrong! Beta doesn't have anything to do with gm
except in a very miniscule way...

gm = q/(kT)*ß/(ß+1)

For some reason all of my TVG's in sonar's and ultra-sounds have bias
current varying with temperature... and have beta correction as well.
I posted one for Fred once when he was bragging about his analysis
skills. I've yet to see his analysis.

The ratio between -40°C and +140°C is (273+140)/(273-40) = 1.773, not
inconsequential for REAL designers ;-)

...Jim Thompson

 

[John Larkin]

On Sat, 28 Jul 2007 16:29:06 -0700, John Larkin

[snip]

For a fairly ideal transistor, Gm = 40 * Ic.

So it's a pretty big number, and temperature-dependent?

Fraction of a mho typically, and it varies with temperature only if
the dc collector current depends on temperature.

Certain parties will predictably nit-pick.

John


John, Are you going to call me a nit-picker if I point out that...

"40" = q/(kT)

So your temperature-independence statement is dead wrong. You get
flat gm if IC _varies_linearly_ with temperature.

...Jim Thompson

You mean with absolute tempearture? -273 is a long distance away, and
transistor betas vary by 5:1 or more on the datasheets, so it hardly
matters. Calcs like this don't have to be right by even 2:1, unless
you are designing ICs, which I don't think we are.

If I'd filled up a screen with equations, I'd be "right" but wouldn't
convey a lot of understanding. Having designed electronics for 40
years or so, transistor math at this level is plenty good enough.

OK, how about "hardly varies with temperature"?

John

BS, John! You're wrong! Beta doesn't have anything to do with gm
except in a very miniscule way...

gm = q/(kT)*ß/(ß+1)

I just posted an explanation of why beta doesn't change Gm. b/(b+1) =
1 for all practical purposes. Nit picking.

Who the hell is going to design something that needs precise beta *or*
Gm? I do most transistor calculations in my head, to 10% or so, and
everything just works.

For some reason all of my TVG's in sonar's and ultra-sounds have bias
current varying with temperature... and have beta correction as well.
I posted one for Fred once when he was bragging about his analysis
skills. I've yet to see his analysis.

Doesn't matter. Even if he can do analysis, he can't design, so he has
nothing to analyze.

The ratio between -40°C and +140°C is (273+140)/(273-40) = 1.773, not
inconsequential for REAL designers ;-)

Inconsequential when beta and temperature-tolerant design is done.

Again, we're not designing ICs. The level of approximation that I use
is plenty adequate for discrete circuit design, and I have many other
fish to fry. I suppose the level you use is adequate for IC design.
Neither is correct.

John

 

[Jim Thompson]

On Sat, 28 Jul 2007 16:52:12 -0700, Jim Thompson

On Sat, 28 Jul 2007 16:29:06 -0700, John Larkin

[snip]

For a fairly ideal transistor, Gm = 40 * Ic.

So it's a pretty big number, and temperature-dependent?

Fraction of a mho typically, and it varies with temperature only if
the dc collector current depends on temperature.

Certain parties will predictably nit-pick.

John


John, Are you going to call me a nit-picker if I point out that...

"40" = q/(kT)

So your temperature-independence statement is dead wrong. You get
flat gm if IC _varies_linearly_ with temperature.

...Jim Thompson

You mean with absolute tempearture? -273 is a long distance away, and
transistor betas vary by 5:1 or more on the datasheets, so it hardly
matters. Calcs like this don't have to be right by even 2:1, unless
you are designing ICs, which I don't think we are.

If I'd filled up a screen with equations, I'd be "right" but wouldn't
convey a lot of understanding. Having designed electronics for 40
years or so, transistor math at this level is plenty good enough.

OK, how about "hardly varies with temperature"?

John

BS, John! You're wrong! Beta doesn't have anything to do with gm
except in a very miniscule way...

gm = q/(kT)*ß/(ß+1)

I just posted an explanation of why beta doesn't change Gm. b/(b+1) =
1 for all practical purposes. Nit picking.

Who the hell is going to design something that needs precise beta *or*
Gm? I do most transistor calculations in my head, to 10% or so, and
everything just works.

For some reason all of my TVG's in sonar's and ultra-sounds have bias
current varying with temperature... and have beta correction as well.
I posted one for Fred once when he was bragging about his analysis
skills. I've yet to see his analysis.

Doesn't matter. Even if he can do analysis, he can't design, so he has
nothing to analyze.

The ratio between -40°C and +140°C is (273+140)/(273-40) = 1.773, not
inconsequential for REAL designers ;-)

Inconsequential when beta and temperature-tolerant design is done.

Again, we're not designing ICs. The level of approximation that I use
is plenty adequate for discrete circuit design, and I have many other
fish to fry. I suppose the level you use is adequate for IC design.
Neither is correct.

John

I design those sensitivities out because I can... you have to live
with it, and do ugly things like have to use uP's to do
auto-calibrations ;-)

...Jim Thompson

 

[John Larkin]

On Sat, 28 Jul 2007 17:09:11 -0700, John Larkin

On Sat, 28 Jul 2007 16:52:12 -0700, Jim Thompson

On Sat, 28 Jul 2007 16:29:06 -0700, John Larkin

[snip]

For a fairly ideal transistor, Gm = 40 * Ic.

So it's a pretty big number, and temperature-dependent?

Fraction of a mho typically, and it varies with temperature only if
the dc collector current depends on temperature.

Certain parties will predictably nit-pick.

John


John, Are you going to call me a nit-picker if I point out that...

"40" = q/(kT)

So your temperature-independence statement is dead wrong. You get
flat gm if IC _varies_linearly_ with temperature.

...Jim Thompson

You mean with absolute tempearture? -273 is a long distance away, and
transistor betas vary by 5:1 or more on the datasheets, so it hardly
matters. Calcs like this don't have to be right by even 2:1, unless
you are designing ICs, which I don't think we are.

If I'd filled up a screen with equations, I'd be "right" but wouldn't
convey a lot of understanding. Having designed electronics for 40
years or so, transistor math at this level is plenty good enough.

OK, how about "hardly varies with temperature"?

John

BS, John! You're wrong! Beta doesn't have anything to do with gm
except in a very miniscule way...

gm = q/(kT)*ß/(ß+1)

I just posted an explanation of why beta doesn't change Gm. b/(b+1) =
1 for all practical purposes. Nit picking.

Who the hell is going to design something that needs precise beta *or*
Gm? I do most transistor calculations in my head, to 10% or so, and
everything just works.

For some reason all of my TVG's in sonar's and ultra-sounds have bias
current varying with temperature... and have beta correction as well.
I posted one for Fred once when he was bragging about his analysis
skills. I've yet to see his analysis.

Doesn't matter. Even if he can do analysis, he can't design, so he has
nothing to analyze.

The ratio between -40°C and +140°C is (273+140)/(273-40) = 1.773, not
inconsequential for REAL designers ;-)

Inconsequential when beta and temperature-tolerant design is done.

Again, we're not designing ICs. The level of approximation that I use
is plenty adequate for discrete circuit design, and I have many other
fish to fry. I suppose the level you use is adequate for IC design.
Neither is correct.

John

I design those sensitivities out because I can... you have to live
with it, and do ugly things like have to use uP's to do
auto-calibrations ;-)

...Jim Thompson

Last delay generator I did, I measured the pcb temperature with an
LM71 SPI temp sensor. The uP does three different temperature
compensations, two linear and one polynomial, and drives 10 different
dacs. Among other things, the through-the-box delay tc was reduced
from 40 ps/K, mediocre, to something like 3, hard to measure.

CMOS has a rotten delay tempco. I suppose one could program Vcc versus
temp and help a lot. I'll have to try that some day.

Really, all this uP stuff is liberating. I can poke all these
compensation coefficients into a calibration table that's saved in
eeprom, and change them any time it might be needed. That sure
pipelines the design process, sort of like zener-zapping on steroids.

The downside, of course, is that eventually you have to write a bunch
of damned code, which is what I'm supposed to be doing this particular
instant.

John

 

[Jim Thompson]

On Sat, 28 Jul 2007 17:19:25 -0700, Jim Thompson

On Sat, 28 Jul 2007 17:09:11 -0700, John Larkin

On Sat, 28 Jul 2007 16:52:12 -0700, Jim Thompson

On Sat, 28 Jul 2007 16:29:06 -0700, John Larkin

[snip]

For a fairly ideal transistor, Gm = 40 * Ic.

So it's a pretty big number, and temperature-dependent?

Fraction of a mho typically, and it varies with temperature only if
the dc collector current depends on temperature.

Certain parties will predictably nit-pick.

John


John, Are you going to call me a nit-picker if I point out that...

"40" = q/(kT)

So your temperature-independence statement is dead wrong. You get
flat gm if IC _varies_linearly_ with temperature.

...Jim Thompson

You mean with absolute tempearture? -273 is a long distance away, and
transistor betas vary by 5:1 or more on the datasheets, so it hardly
matters. Calcs like this don't have to be right by even 2:1, unless
you are designing ICs, which I don't think we are.

If I'd filled up a screen with equations, I'd be "right" but wouldn't
convey a lot of understanding. Having designed electronics for 40
years or so, transistor math at this level is plenty good enough.

OK, how about "hardly varies with temperature"?

John

BS, John! You're wrong! Beta doesn't have anything to do with gm
except in a very miniscule way...

gm = q/(kT)*ß/(ß+1)

I just posted an explanation of why beta doesn't change Gm. b/(b+1) =
1 for all practical purposes. Nit picking.

Who the hell is going to design something that needs precise beta *or*
Gm? I do most transistor calculations in my head, to 10% or so, and
everything just works.


For some reason all of my TVG's in sonar's and ultra-sounds have bias
current varying with temperature... and have beta correction as well.
I posted one for Fred once when he was bragging about his analysis
skills. I've yet to see his analysis.

Doesn't matter. Even if he can do analysis, he can't design, so he has
nothing to analyze.


The ratio between -40°C and +140°C is (273+140)/(273-40) = 1.773, not
inconsequential for REAL designers ;-)

Inconsequential when beta and temperature-tolerant design is done.

Again, we're not designing ICs. The level of approximation that I use
is plenty adequate for discrete circuit design, and I have many other
fish to fry. I suppose the level you use is adequate for IC design.
Neither is correct.

John

I design those sensitivities out because I can... you have to live
with it, and do ugly things like have to use uP's to do
auto-calibrations ;-)

...Jim Thompson

Last delay generator I did, I measured the pcb temperature with an
LM71 SPI temp sensor. The uP does three different temperature
compensations, two linear and one polynomial, and drives 10 different
dacs. Among other things, the through-the-box delay tc was reduced
from 40 ps/K, mediocre, to something like 3, hard to measure.

CMOS has a rotten delay tempco. I suppose one could program Vcc versus
temp and help a lot. I'll have to try that some day.

Really, all this uP stuff is liberating. I can poke all these
compensation coefficients into a calibration table that's saved in
eeprom, and change them any time it might be needed. That sure
pipelines the design process, sort of like zener-zapping on steroids.

The downside, of course, is that eventually you have to write a bunch
of damned code, which is what I'm supposed to be doing this particular
instant.

John

I know, I be yanking your chain (as if you haven't been yanking mine
;-).

I just finished a chip that features auto-cal/auto-zero before _every_
measurement. It doesn't use a µP, just a timing chain to sequence
analog events.

I even find myself offering clients digital solutions to their analog
problems ;-)

Auto-cal, auto-zero, and auto-DC-restore can be really nice features
;-)

...Jim Thompson

 

[John Larkin]

I know, I be yanking your chain (as if you haven't been yanking mine
;-).

You can't open a clause with a left parenthesis and close it with a
smiley! You'll get a fatal compile error every time.

John

 

[[email protected]]

On Sat, 28 Jul 2007 13:51:32 -0700, [email protected] wrote:

[snip]

I believe there is a very old paper on negative feedback where a
mathematical proof is done to show that if you take an amplifier with
just 2nd harmonic distortion and apply feedback, the harmonics are now
infinite. Smaller of course, but now an infinite series rather than
just 2nd.

Probably written by some audiophool trying to support the "no
feedback" concept.

No, it was a very early paper on negative feedback. The infinite
harmonics were proven using a Taylor series. Remember, there was a
time when engineers were good at math.

 

[Tim Williams]

(Bah to you I say! I've been doing that for years! ;-)

Tim

 

[Eeyore]

No, it was a very early paper on negative feedback. The infinite
harmonics were proven using a Taylor series. Remember, there was a
time when engineers were good at math.

A good simulation ought to prove this.

Graham

 

[Fred_Bartoli]

Eeyore a écrit :

A good simulation ought to prove this.

A good simulation might put you on track but it *proves* nothing.

 

[Jonathan Kirwan]

A good simulation ought to prove this.

Proof is a nice, closed form mathematical solution. A simulation may
only be sufficiently consistent with or essentially inconsistent with
the math (or else uninterpretable.)

Jon

 

[MooseFET]

[snip]

I believe there is a very old paper on negative feedback where a
mathematical proof is done to show that if you take an amplifier with
just 2nd harmonic distortion and apply feedback, the harmonics are now
infinite. Smaller of course, but now an infinite series rather than
just 2nd.

Probably written by some audiophool trying to support the "no
feedback" concept.

No, it was a very early paper on negative feedback. The infinite
harmonics were proven using a Taylor series. Remember, there was a
time when engineers were good at math.

You don't need a lot of math to argue the point. You have a non-
linear system that when fed a sinewave, produces a harmonic
component. When you enclose that non-linear system in a feedback path
you are now applying harmonics of the signal to the non-linear device
and thus end up with harmonics of the harmonics and all posible IM
products between harmonics.

This is part of why you always see the harmonics near the gain
crossover point point.

 

[Eeyore]

Eeyore a écrit :

A good simulation might put you on track but it *proves* nothing.

I like the idea of the rigid mathematical anyalsyis too actually. A simulation
using a SPICE based tool is just so much quicker.

Graham