EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: 741 on January 06, 2025, 09:36:07 pm
-
A "microprocessor RESET" circuit that actually oscillated in practice. Shown in "Troubleshooting Analog Circuits" by Bob Pease, p109
"The little transistor would run at over 10mA and, with a bypass capacitor at its base, the transistor would oscillate" [at a "couple hundred megahertz"]
(The simulation does not oscillate).
What is the mechanism of the oscillation (maybe akin to a phase-shift oscillator?) - do I need to add any real-world parasitics to the simulation? Maybe just a current kick / impulse?
-
A "microprocessor RESET" circuit that actually oscillated in practice. Shown in "Troubleshooting Analog Circuits" by Bob Pease, p109
"The little transistor would run at over 10mA and, with a bypass capacitor at its base, the transistor would oscillate" [at a "couple hundred megahertz"]
(The simulation does not oscillate).
What is the mechanism of the oscillation (maybe akin to a phase-shift oscillator?) - do I need to add any real-world parasitics to the simulation? Maybe just a current kick / impulse?
Schematic in form viewable in a browser?
Photo of construction technique?
Without knowing those, one guess is shown at https://entertaininghacks.wordpress.com/2024/03/16/practical-traps-with-a-one-transistor-audio-amplifier-solderless-breadboards-and-oscilloscopes/ which shows
-a one transistor audio amplifier using jellybean components, an audio function generator, a voltmeter and an oscilloscope
-how a gross problem is easily missed, not only with simulation but also experimentally
-what’s necessary to reveal the problem
-a subtle variation in the way the problem manifests itself
-common tools showing differing measurement results
-the problem’s cause, and how it can be verified with adequate simulation
-professional products which rely on such “problems” for their basic operation
-
Heres the circuit the op was referring to.
[attachimg=1]
-
I managed to get it to oscillate at about 350MHz with some minor tweaks, one of the important things for SPICE is to reduce the timestep, otherwise it just steps over any oscillation.
-
Heres the circuit the op was referring to.
(Attachment Link)
I suspected as much :)
We don't know the physical implementation, but I'll bet the cause and resolution is given in the article I referenced.
-
Yeah, emitter-followers aren't as harmless as they look. I've experienced this enough when making constant-current sources for LEDs.
The trick is to include a base resistor, that normally stops the behaviour.
-
Yeah, emitter-followers aren't as harmless as they look. I've experienced this enough when making constant-current sources for LEDs.
The trick is to include a base resistor, that normally stops the behaviour.
So in general, what value of "stopper" resistor ought to be used? (Assuming a small non-power transistor.)
Something in the ballpark of 100Ω or less?
-
Yeah, emitter-followers aren't as harmless as they look. I've experienced this enough when making constant-current sources for LEDs.
The trick is to include a base resistor, that normally stops the behaviour.
So in general, what value of "stopper" resistor ought to be used? (Assuming a small non-power transistor.)
Something in the ballpark of 100Ω or less?
It's like the gate resistor applied to MOSFETS, it should be as near to the base as possible and 100R would be fine in this case maybe even as low as 10R could work but it is a practical decision based on the circuit parasitics, in the case of the sim I posted 10R works.
-
Yeah, emitter-followers aren't as harmless as they look. I've experienced this enough when making constant-current sources for LEDs.
The trick is to include a base resistor, that normally stops the behaviour.
Or a ferrite.
Or have decent layout/construction techniques, based around an understanding of what is happening and why.
IMHO understanding is preferable to suck-it-and-see coupled with "it worked with last year's batch of transistors". Replacing original 1980s 2N3055s with modern 2N3055s anecdotally enables them to oscillate in some cases. Presumed cause: a much higher fT than the specified 2.5HMz min.
-
Yeah, emitter-followers aren't as harmless as they look. I've experienced this enough when making constant-current sources for LEDs.
The trick is to include a base resistor, that normally stops the behaviour.
A fundamental issue with emitter followers is their operation into a capacitive load.
(This is also true with cathode followers and source followers.)
The base (grid, gate) stopper spoils the negative resistance component that appears at its input causing oscillation: the full diagram including parasitic capacitances resembles a Colpitts oscillator.
-
Yeah, emitter-followers aren't as harmless as they look. I've experienced this enough when making constant-current sources for LEDs.
The trick is to include a base resistor, that normally stops the behaviour.
A fundamental issue with emitter followers is their operation into a capacitive load.
(This is also true with cathode followers and source followers.)
The base (grid, gate) stopper spoils the negative resistance component that appears at its input causing oscillation: the full diagram including parasitic capacitances resembles a Colpitts oscillator.
I.e. operating as a common-base amplifier with lots of voltage gain (cf the presumed common emitter amplifier with voltage gain <1)
-
A "microprocessor RESET" circuit that actually oscillated in practice. Shown in "Troubleshooting Analog Circuits" by Bob Pease, p109
"The little transistor would run at over 10mA and, with a bypass capacitor at its base, the transistor would oscillate" [at a "couple hundred megahertz"]
(The simulation does not oscillate).
What is the mechanism of the oscillation (maybe akin to a phase-shift oscillator?) - do I need to add any real-world parasitics to the simulation? Maybe just a current kick / impulse?
Simulators "lie in their teeth".
Unless they are provided with typical values of stray inductance & capacitance, they assume both Ls & Cs are zero. The real world is very different, & common values of both may well cause VHF oscillation.
-
Simulators "lie in their teeth".
I disagree. A model is an approximation of reality, if it doesn't match reality then your approximation is too coarse and you need to do as you say (consider what else in your circuit might matter, eg parasitics). See also: GIGO theory (garbage in, garbage out). A simulator is just a tool, it's up to the user to use it appropriately.
-
Simulators "lie in their teeth".
Unless they are provided with typical values of stray inductance & capacitance, they assume both Ls & Cs are zero. The real world is very different, & common values of both may well cause VHF oscillation.
You underestimate simulators.
Fed with the right models (including parasitics), they provide excellent results, but an engineering brain has to accompany them.
Ngspice was actually the reason I found out why a constant-current source was misbehaving, and led me to the dangers of the common-emitter circuit.
-
Simulators "lie in their teeth".
Bob Pease certainly agreed with your sentiment as he threw his computer from the roof of the building, but they have very practical uses if you accept the limitations.
-
If you add in the parasitics, there can be enough feedback to cause oscillation.
Also reminds me of this thread: https://www.eevblog.com/forum/projects/challenge-thread-the-fastest-breadboard-oscillator-on-the-mudball/ (https://www.eevblog.com/forum/projects/challenge-thread-the-fastest-breadboard-oscillator-on-the-mudball/)
-
A class of transistors also known as Low VCE-SAT transistors meant for switching circuits are excellent oscillators when used in emitter followers.
An example here by Dave. At around 6 minutes into this video the output stage transistor is replaced by a FZT849 and Dave wrongly assumes that the compensation around the op amp needs adjustment.
https://www.youtube.com/watch?v=b7UQVZaqxg0 (https://www.youtube.com/watch?v=b7UQVZaqxg0)
Most of the comments state the same and trow wild values around for the compensation circuit while it is not the op amp which is oscillating but the unfortunate choice of transistor which makes the output stage a good oscillator and the op amp is trying to "compensate" the oscillating output stage.
What is needed is a resistor in series with the FZT849 base to compensate for the negative input resistance caused by a more often than not invisible capacitive load such that input resistance stays positive. Finding out how large this resistor must be is the difficult part.
-
Here is an image of the LTSpice circuit as requested (I thought maybe it would be disallowed to post a photo of the page in the book).
Also shown is the circuit of Reply #3, which does oscillate.
-
A "microprocessor RESET" circuit that actually oscillated in practice. Shown in "Troubleshooting Analog Circuits" by Bob Pease, p109
"The little transistor would run at over 10mA and, with a bypass capacitor at its base, the transistor would oscillate" [at a "couple hundred megahertz"]
(The simulation does not oscillate).
What is the mechanism of the oscillation (maybe akin to a phase-shift oscillator?) - do I need to add any real-world parasitics to the simulation? Maybe just a current kick / impulse?
Simulators "lie in their teeth".
Unless they are provided with typical values of stray inductance & capacitance, they assume both Ls & Cs are zero. The real world is very different, & common values of both may well cause VHF oscillation.
Everybody should read learn and inwardly digest the aphorism "all models are wrong, but some are useful".
Too many think "X true because the computer says so". (I use the word "think" loosely, of course).
-
Quote from: vk6zgo on Today at 01:26:43 (https://www.eevblog.com/forum/index.php?topic=449609.msg5773699#msg5773699)...
Too many think "X true because the computer says so". (I use the word "think" loosely, of course).
I'd suggest: Too many believe "X true because the computer says so".
A professional needs the ability to make rough estimates and checking the plausibility of a calculator result. The extend of a "rough estimate" varies. When I design a circuit, this may well be equations and transforms that span several pages. For stuff like that I combine a pencil, sheets of paper and the computer (LTspice and Excel work surprisingly well for me).
Incidentally: Bob Pease suggests a remedy (series base resitor) on the very same page, above the diagram. However, he leaves the detailed analysis to the reader.
-
Quote from: vk6zgo on Today at 01:26:43 (https://www.eevblog.com/forum/index.php?topic=449609.msg5773699#msg5773699)...
Too many think "X true because the computer says so". (I use the word "think" loosely, of course).
I'd suggest: Too many believe "X true because the computer says so".
A professional needs the ability to make rough estimates and checking the plausibility of a calculator result. The extend of a "rough estimate" varies. When I design a circuit, this may well be equations and transforms that span several pages. For stuff like that I combine a pencil, sheets of paper and the computer (LTspice and Excel work surprisingly well for me).
Incidentally: Bob Pease suggests a remedy (series base resitor) on the very same page, above the diagram. However, he leaves the detailed analysis to the reader.
Except that this case is not a matter of simple calculations and rough estimates. It requires good insight into analog circuit design to identify such unintentional invisible oscillators as I stated before in my earlier post with a nice example.
-
Quote from: vk6zgo on Today at 01:26:43 (https://www.eevblog.com/forum/index.php?topic=449609.msg5773699#msg5773699)...
Too many think "X true because the computer says so". (I use the word "think" loosely, of course).
I'd suggest: Too many believe "X true because the computer says so".
A professional needs the ability to make rough estimates and checking the plausibility of a calculator result. The extend of a "rough estimate" varies. When I design a circuit, this may well be equations and transforms that span several pages. For stuff like that I combine a pencil, sheets of paper and the computer (LTspice and Excel work surprisingly well for me).
Incidentally: Bob Pease suggests a remedy (series base resitor) on the very same page, above the diagram. However, he leaves the detailed analysis to the reader.
Except that this case is not a matter of simple calculations and rough estimates. It requires good insight into analog circuit design to identify such unintentional invisible oscillators as I stated before in my earlier post with a nice example.
As I stated (but was snipped before your response) "Everybody should read learn and inwardly digest the aphorism "all models are wrong, but some are useful"."
Design requires
- theoretical understanding. Without that it is blind fumbling and/or cargo-cult engineering
- models of the system being considered
- understanding what the models don't address. Models cannot address everything, because by definition they are a simplification of reality
- practical and theoretical understanding of the extent to which the system being considered is not valid
- experiment to validate all the above
- repeat all the above when a discrepancy is found
In the case of a circuit that is presumed to be common collector amplifiers (i.e. emitter followers) with voltage gain <1, parasitic components can cause it to act as a common-base amplifier with voltage gain >1.
-
Still missing perhaps a dozen of parasitics there..
159MHz..
PS: 100mm long straight wire is aprox 100nH..
-
Still missing perhaps a dozen of parasitics there..
159MHz..
:-DD
-
Still missing perhaps a dozen of parasitics there..
159MHz..
PS: 100mm long straight wire is aprox 100nH..
Nice. :)
-
Still missing perhaps a dozen of parasitics there..
159MHz..
PS: 100mm long straight wire is aprox 100nH..
Nice. :)
Particularly relevant w.r.t. easy to (mis)use dupont flying leads on solderless breadboards.
The 1nH/mm rule of thumb has interesting consequences w.r.t. SMD components at RF, and 6inch scope probe grounds with a tip capacitance of 15pF.
-
The 1nH/mm rule of thumb has interesting consequences w.r.t. SMD components at RF, and 6inch scope probe grounds with a tip capacitance of 15pF.
That rule-of-thumb can be thrown in the garbage can.
It applies to inductive loops that have been sectioned into to 1 mm pieces. It has zero significance to this thread.
-
The 1nH/mm rule of thumb has interesting consequences w.r.t. SMD components at RF, and 6inch scope probe grounds with a tip capacitance of 15pF.
That rule-of-thumb can be thrown in the garbage can.
It applies to inductive loops that have been sectioned into to 1 mm pieces. It has zero significance to this thread.
Single straight wire in free space. The formulae are easily found, as at calculators.
Loops have a higher inductance.
-
The 1nH/mm rule of thumb has interesting consequences w.r.t. SMD components at RF, and 6inch scope probe grounds with a tip capacitance of 15pF.
That rule-of-thumb can be thrown in the garbage can.
It applies to inductive loops that have been sectioned into to 1 mm pieces. It has zero significance to this thread.
Single straight wire in free space. The formulae are easily found, as at calculators.
Loops have a higher inductance.
I think you've forgotten wire radius. I don't believe a 1 mm thick, 100 mm diameter disc will confirm to your formula, although it will be a 1 mm long piece of wire.
-
The 1nH/mm rule of thumb has interesting consequences w.r.t. SMD components at RF, and 6inch scope probe grounds with a tip capacitance of 15pF.
That rule-of-thumb can be thrown in the garbage can.
It applies to inductive loops that have been sectioned into to 1 mm pieces. It has zero significance to this thread.
Single straight wire in free space. The formulae are easily found, as at calculators.
Loops have a higher inductance.
I think you've forgotten wire radius. I don't believe a 1 mm thick, 100 mm diameter disc will confirm to your formula, although it will be a 1 mm long piece of wire.
That's changing the subject away from loop vs straight wire.
1nH/mm of wire is a rule of thumb, i.e. a simple encapsulation of common circumstances useful for indicating whether effects might need to be considered in more detail.
Does anybody other than you think that 10cm disk 1mm thick constitutes a wire?
It would be difficult to find a 10cm diameter conductor for any high power distribution system, let alone for low power electronics. And especially SMD components and scope probes!
-
Here is the paper Jim Williams cited which discusses emitter follower oscillation.
-
The 1nH/mm rule of thumb has interesting consequences w.r.t. SMD components at RF, and 6inch scope probe grounds with a tip capacitance of 15pF.
That rule-of-thumb can be thrown in the garbage can.
It applies to inductive loops that have been sectioned into to 1 mm pieces. It has zero significance to this thread.
Single straight wire in free space. The formulae are easily found, as at calculators.
Loops have a higher inductance.
I think you've forgotten wire radius. I don't believe a 1 mm thick, 100 mm diameter disc will confirm to your formula, although it will be a 1 mm long piece of wire.
That's changing the subject away from loop vs straight wire.
1nH/mm of wire is a rule of thumb, i.e. a simple encapsulation of common circumstances useful for indicating whether effects might need to be considered in more detail.
Does anybody other than you think that 10cm disk 1mm thick constitutes a wire?
It would be difficult to find a 10cm diameter conductor for any high power distribution system, let alone for low power electronics. And especially SMD components and scope probes!
Could be modeled, but gradients of resistance and inductance across the plane of the disc would dominate, similar to dead bugging a uhf circuit on a ground plane, the return through that has to be taken into account.
Common sense should be applied that the wire would have to be Length >> diameter in any "the rule of thumb".
-
Here is the paper Jim Williams cited which discusses emitter follower oscillation.
@David Hess, thanks for digging out this paper. :-+ :-+