|
Hi,
I am trying to model emitter follower oscillation to learn how to use LT Spice to determine when to use a base resistor and when it is not necessary.
I have tried several designs that I know oscillate, but could not get them to do so in the simulator. I tried the Cshunt parameter, and I tried adding capacitance to the base and emitter.
In particular, a current mirror made with KSC3503, KSA1220, and other devices oscillate without base resistors (I learned this the hard way).
Has anyone had success modeling emitter follower oscillation?
Thanks,
Morty
|
Simulations often won't start oscillating
unless given a 'kick' out of a stable 'zero' state. There are
different methods of kicking, depending on the configuration.
You can apply a short current pulse to the base, or start the
simulation with a charge on a capacitor. You need to experiment
to find the way that works.
On 2023-11-11 21:03, mtarr00 via
groups.io wrote:
переключити цитоване повідомлення
Показати цитований текст
Hi,
I am trying to model emitter follower oscillation to learn how to
use LT Spice to determine when to use a base resistor and when it
is not necessary.
I have tried several designs that I know oscillate, but could not
get them to do so in the simulator. I tried the Cshunt parameter,
and I tried adding capacitance to the base and emitter.
In particular, a current mirror made with KSC3503, KSA1220, and
other devices oscillate without base resistors (I learned this the
hard way).
Has anyone had success modeling emitter follower oscillation?
Thanks,
Morty
|
If you consider a Colpitts Oscillator as an emitter follower oscillator (I do), then, yes, many times. There are several tricks you may need to apply.
1) Oscillators with high-Q resonators (e.g crystals) can be VERY slow to start. Wait at least the time of Q cycles.
2) Transistor models are noiseless, so some oscillators will not start without help. A narrow current pulse (pulse width less than 1/4 oscillation period) will often do it. Amplitude of a few mA is often sufficient.
Hi,
I am trying to model emitter follower oscillation to learn how to use LT Spice to determine when to use a base resistor and when it is not necessary.
I have tried several designs that I know oscillate, but could not get them to do so in the simulator. I tried the Cshunt parameter, and I tried adding capacitance to the base and emitter.
In particular, a current mirror made with KSC3503, KSA1220, and other devices oscillate without base resistors (I learned this the hard way).
Has anyone had success modeling emitter follower oscillation?
Thanks,
Morty
|
Morty asked if others have simulated oscillations with simple emitter followers.
This might not help you -- but be aware that simulating oscillations can be slow and difficult. If a circuit will oscillate in a simulation, you might need to run the simulation for a rather long time (many seconds of simulated time), and you might need to stimulate the oscillation with a good "kick in the pants" -- a transient added in the right place.
I imagine that oscillation may depend on circuit parasitics, including lead inductance and perhaps inter-lead capacitance. Simple transistor models lack any lead inductance, since they model just the die portion of the transistor.
For me, the answer to your question is no, I never tried to model emitter follower oscillation. It's an interesting question.
Andy
|
Yes, emitter-follower oscillation can be due
to inductance between emitter and ground. This can create a
negative input impedance at the base. Depending on the frequency
range and the transistor packaging, the inductance may be
external or just the emitter lead (if any) and bonding wire
inductance.
On 2023-11-11 21:18, Andy I wrote:
переключити цитоване повідомлення
Показати цитований текст
Morty asked if others have simulated oscillations with simple
emitter followers.
This might not help you -- but be aware that simulating
oscillations can be slow and difficult. If a circuit will
oscillate in a simulation, you might need to run the simulation
for a rather long time (many seconds of simulated time), and you
might need to stimulate the oscillation with a good "kick in the
pants" -- a transient added in the right place.
I imagine that oscillation may depend on circuit parasitics,
including lead inductance and perhaps inter-lead capacitance.
Simple transistor models lack any lead inductance, since they
model just the die portion of the transistor.
For me, the answer to your question is no, I never tried to model
emitter follower oscillation. It's an interesting question.
Andy
|
What about uploading your .asc file(s) so we could try to help
you?
Le 11/11/2023 à 22:03, mtarr00 via
groups.io a écrit :
переключити цитоване повідомлення
Показати цитований текст
Hi,
I am trying to model emitter follower oscillation to learn how to
use LT Spice to determine when to use a base resistor and when it
is not necessary.
I have tried several designs that I know oscillate, but could not
get them to do so in the simulator. I tried the Cshunt parameter,
and I tried adding capacitance to the base and emitter.
In particular, a current mirror made with KSC3503, KSA1220, and
other devices oscillate without base resistors (I learned this the
hard way).
Has anyone had success modeling emitter follower oscillation?
Thanks,
Morty
|
I have spent some time simulating emitter follower (common collector) circuits to understand the stability problem. Like other things in using SPICE, exploring oscillating circuits becomes rather easy with practice.
The mechanism that causes the real part of the input impedance (looking into the transistors base) to be negative, does not need package parasitics to function. But when your model includes the package inductance, it becomes more difficult to prevent oscillation by tailoring the impedance connected to the base.
The MCH4009 ( https://www.onsemi.com/download/data-sheet/pdf/ena0389-d.pdf) is a Ft=20GHz NPN bipolar made by onsemi. The data sheet supplies the SPICE model and an 11 element package model. The package parasitics really change the transistor's behavior. This is the device I used when I was fooling around with this topic.
What about uploading your .asc file(s) so we could try to help
you?
Le 11/11/2023 à 22:03, mtarr00 via
groups.io a écrit :
Hi,
I am trying to model emitter follower oscillation to learn how to
use LT Spice to determine when to use a base resistor and when it
is not necessary.
I have tried several designs that I know oscillate, but could not
get them to do so in the simulator. I tried the Cshunt parameter,
and I tried adding capacitance to the base and emitter.
In particular, a current mirror made with KSC3503, KSA1220, and
other devices oscillate without base resistors (I learned this the
hard way).
Has anyone had success modeling emitter follower oscillation?
Thanks,
Morty
|
I remember an Elektor circuit idea for a very low THD RC oscillator using only an emitter follower.
The striking feature was that it didn't need inductors and/or a complicated automatic amplitude control with
a thermistor or incandescent lamp (because, of course, the gain was only slightly larger than 1).
Analog art, 40 years ago.
-marcel
|
In general, the condition that causes instability with emitter
followers is capacitive loading on the emitter, not inductive. In
combination with Cbe, the resulting feedback sets up negative
impedance on the base. And thence with an inductive source on the
base forms the classic Colpitts configuration. So, adding the series
resistor on the base, kills the Q of any resonant condition.
It is very easy to show this in a non-oscillating
schematic that shows the range over which return gain and
negative impedance is displayed on the base with modest capacitance
loading on the emitter. Potentially, this circuit can oscillate over
this entire frequency range if the appropriate inductive loading is
placed on the base.
--
Regards,
Tony
On 11/11/2023 22:30, John Woodgate
wrote:
Yes,
emitter-follower oscillation can be due to inductance between
emitter and ground. This can create a negative input impedance at
the base. Depending on the frequency range and the transistor
packaging, the inductance may be external or just the emitter lead
(if any) and bonding wire inductance.
|
That is very interesting. Do you have any
idea where I could find details, because I need such a circuit?
I see that there is an archive of Elektor issues, but of course
it would be a long search without any input data. I don't see
how an emitter follower can have a gain of 1, let alone more
than 1.
On 2023-11-12 17:56, mhx@... wrote:
переключити цитоване повідомлення
Показати цитований текст
I remember an Elektor circuit idea for a very low THD RC oscillator using only an emitter follower.
The striking feature was that it didn't need inductors and/or a complicated automatic amplitude control with
a thermistor or incandescent lamp (because, of course, the gain was only slightly larger than 1).
Analog art, 40 years ago.
-marcel
|
John wrote, "I don't see how an emitter follower can have a gain of 1, let alone more than 1."
It has power gain, whether or not a voltage gain.
Andy
|
|
|
Have you never used an emitter follower in a 2nd order filter where
the Q is greater than 0.707?
--
Regards,
Tony
On 12/11/2023 19:29, John Woodgate
wrote:
I
don't see how an emitter follower can have a gain of 1, let alone
more than 1.
|
A Colpitts oscillator exploits it. The “dual” of the Colpitts (the Hartley) is easy to see how it works. Colpitts has always been rather mysterious to me. I know that the two capacitors essentially transform the negative emitter resistance to appear across the tank circuit, but I’ve never dug past that point.
On Nov 12, 2023, at 3:22 PM, John Woodgate < jmw@...> wrote:
Good point. Now, how to exploit it?
On 2023-11-12 23:18, Andy I wrote:
John wrote, "I don't see how an
emitter follower can have a gain of 1, let alone more than 1."
It has power gain, whether or not a voltage gain.
Andy
|
I've never felt the need to do that.
On 2023-11-12 23:29, Tony Casey wrote:
переключити цитоване повідомлення
Показати цитований текст
Have you never used an emitter follower in a 2nd order filter
where the Q is greater than 0.707?
--
Regards,
Tony
On 12/11/2023 19:29, John Woodgate
wrote:
I
don't see how an emitter follower can have a gain of 1, let
alone more than 1.
|
I think that the Colpitts can be explained
without mentioning the negative resistance, even though it's
there.
On 2023-11-12 23:36, Jim Wagner wrote:
переключити цитоване повідомлення
Показати цитований текст
A Colpitts oscillator exploits it. The “dual” of the Colpitts (the
Hartley) is easy to see how it works. Colpitts has always been
rather mysterious to me. I know that the two capacitors
essentially transform the negative emitter resistance to appear
across the tank circuit, but I’ve never dug past that point.
Jim Wagner
On Nov 12, 2023, at 3:22 PM, John Woodgate
< jmw@...>
wrote:
Good point. Now,
how to exploit it?
On 2023-11-12 23:18, Andy I
wrote:
John wrote, "I don't see how an emitter follower can
have a gain of 1, let alone more than 1."
It has power gain, whether or not a voltage gain.
Andy
|
Colpitts is a bit unusual because you can ALSO analyze it on the basis of loop gain (loop gain greater than one with appropriate phase shift at some frequency),
This shows how some circuit behaviors can have more than one analysis view (e.g. ascribed cause and effect) and all of those analyses can be “correct”.
Interestingly, the oscillation amplitude of a Colpitts oscillator will grow until it starts limiting on positive or negative peaks. It will stabilize at the amplitude for which the AVERAGE gain (averaged over one signal cycle) drops to the critical gain required for oscillation OR the net averaged negative resistance shifts to be close to zero (I think that is the correct criterion).
Jim
A Colpitts oscillator exploits it. The “dual” of the Colpitts (the Hartley) is easy to see how it works. Colpitts has always been rather mysterious to me. I know that the two capacitors essentially transform the negative emitter resistance to appear across the tank circuit, but I’ve never dug past that point.
Jim Wagner On Nov 12, 2023, at 3:22 PM, John Woodgate < jmw@...> wrote:
Good point. Now, how to exploit it?
On 2023-11-12 23:18, Andy I wrote:
John wrote, "I don't see how an
emitter follower can have a gain of 1, let alone more than 1."
It has power gain, whether or not a voltage gain.
Andy
|
The reason oscillation is possible with an active device with less than unity voltage gain, is that the network (2 caps, one inductor) that provides the feedback has a voltage gain greater than one at the frequency of oscillation.
I agree that thinking about this as a grounded collector colpitts oscillator makes sense.
Imagine a LC tank circuit. where the lower terminals of L and C are grounded and the top terminals are connected together. Further imagine that the capacitor is actually two capacitors in series with equal values.
inductor: L capacitor connected to inductor: C1 capacitor connected to ground: C2
Let's say we make the junction of L and C1 the input, and the junction of C1 and C2 the output. If you drive the input at the circuit's resonant frequency, the gain (output / input) is 1/2. If you reverse things, and instead drive the 'output' the gain (input / output) is 2. This only works this simply if the circuit Q is high, but it illustrates the principle.
If we have an oscillating voltage follower with its input terminal grounded. L is the bonding wire and package inductance in series with the transistor base. C1 is the transistors base-emitter capacitance which is approximately Cbe=40*Ie/(2pi*Ft). And C1 is the capacitance loading the emitter.
Colpitts is a bit unusual because you can ALSO analyze it on the basis of loop gain (loop gain greater than one with appropriate phase shift at some frequency),
This shows how some circuit behaviors can have more than one analysis view (e.g. ascribed cause and effect) and all of those analyses can be “correct”.
Interestingly, the oscillation amplitude of a Colpitts oscillator will grow until it starts limiting on positive or negative peaks. It will stabilize at the amplitude for which the AVERAGE gain (averaged over one signal cycle) drops to the critical gain required for oscillation OR the net averaged negative resistance shifts to be close to zero (I think that is the correct criterion).
Jim
A Colpitts oscillator exploits it. The “dual” of the Colpitts (the Hartley) is easy to see how it works. Colpitts has always been rather mysterious to me. I know that the two capacitors essentially transform the negative emitter resistance to appear across the tank circuit, but I’ve never dug past that point.
Jim Wagner On Nov 12, 2023, at 3:22 PM, John Woodgate < jmw@...> wrote:
Good point. Now, how to exploit it?
On 2023-11-12 23:18, Andy I wrote:
John wrote, "I don't see how an
emitter follower can have a gain of 1, let alone more than 1."
It has power gain, whether or not a voltage gain.
Andy
|
You mean you have never implemented a 3rd order (or higher)
Butterworth filter with emitter followers? That requires the 2nd
order section to have a Q of 1, which results in peaking of 1.25dB.
A 4th order requires one of the sections to have a Q of 1.848, with
a peaking of 5.66dB.
--
Regards,
Tony
On 13/11/2023 01:17, John Woodgate
wrote:
I've never felt the need to do that.
On 2023-11-12 23:29, Tony Casey
wrote:
Have you never used an emitter follower in a 2nd order filter
where the Q is greater than 0.707?
--
Regards,
Tony
On 12/11/2023 19:29, John Woodgate
wrote:
I
don't see how an emitter follower can have a gain of 1, let
alone more than 1.
|
Why would I choose emitter followers rather
than op-amps? The design, including the DC biasing, would be
much more complicated, I think, and the board area would
probably be larger. For me, it is an interesting theoretical
exercise, but not so good for implementation. Perhaps you would
post an example .ASC?
On 2023-11-13 10:00, Tony Casey wrote:
переключити цитоване повідомлення
Показати цитований текст
You mean you have never implemented a 3rd order (or higher)
Butterworth filter with emitter followers? That requires the 2nd
order section to have a Q of 1, which results in peaking of
1.25dB. A 4th order requires one of the sections to have a Q of
1.848, with a peaking of 5.66dB.
--
Regards,
Tony
On 13/11/2023 01:17, John Woodgate
wrote:
I've never felt the need to do that.
On 2023-11-12 23:29, Tony Casey
wrote:
Have you never used an emitter follower in a 2nd order filter
where the Q is greater than 0.707?
--
Regards,
Tony
On 12/11/2023 19:29, John
Woodgate wrote:
I
don't see how an emitter follower can have a gain of 1, let
alone more than 1.
|
John Woodgate
Tony Casey