Hi all,
The gain of an inverting op-amp amplifier is given as -Rf/R1. However, there can be many possible configurations depending on these resistor values, all providing the same gain. For eg: a) Rf=100M, R1=10M b) Rf=100K, R1=10K c) Rf=100, R1=10. All gives gain=10. So which one is the best? And what are all the reasons for choosing it?
Regards,
Raghu
Hi Raghuram Tr,
Op-amp output current depends on the selected values of Rf and R1. The current at the inverting terminal is Vin-0/R1. if the R1 increases current reduces, if current reduces the total power across the load reduces. minimum resistor values are chosen, where the current is a major parameter at the output. reason might be at higher resistance values thermal noise is very high due to high power dissipation.
Even i have the same doubt.. i am expecting reasonable answers to this question....
Large resistors is limited for practicable design, they will be more complicated to design, and consume more power with a large chip area. However, the important thing in implementing fully integrated analogue large resistors is to find a proper form for resistors that are available in a standard CMOS technology. Their value has tolerance of about 20% to 50%, in integrated form, which is not very accurate, resulting from fabrication process and temperature variations when polysilicon or diffused lines are used as resistors.
Another drawback of using fully integrated large resistors is the self-heating of the resistor due to non-uniformity of the composition inside the component which causes change in the current distribution, leading to non-uniform power dissipation.
As mentioned by some others, one concern is the output current limits of the opamp.
[Due to assymetry of an opamp's output stage, these current limits may be different for sourced and sinked cases, i.e. for Vo>0 and Vo Vo,max / Io,lim
Besides this lower limit, an upper limit can be defined for Rf.
For a very large Rf value, the input currents (Iin) of the opamp (which are usually negligible) may be comparable with Vo/Rf (the current of Rf). To avoid any gain errors due to this, one should choose Rf such that
Rf
For a discrete op-amp the quality of the resistors can be chosen, but get more expensive the better they are. One can also get matched pairs of gain resistors e.g. "Vishay Thin Film MPM Matched Pair Resistor Network provides ± 2 ppm/°C tracking and a ratio tolerance as tight as 0.01 %".
The actual values to use depend on the application. For precision and lowest resistor noise the smaller value is often in the hundreds of ohms range. For lowest power use it may well be 100K or higher. For RF however it can be 50 ohms, or lower.
In an inverting op-amp the first input resistor is summing into a virtual ground, so this sets the input impedance. It is common practise to have a same value resistor on the non-inverting input, to balance any small voltage/current biases.
There seems almost as many op-amp types as grains of sand on a beach, so it is always good to look closely at the datasheet information. Most datasheets will have specific guidelines and recommendations based on their part’s characteristics (low noise, low pow, high speed, etc.).
Finally watch the gain product bandwidth, that f3 corner can sneak up on one unexpectedly and since an op-amp is a complex circuit the signal at the f3 point can be quite distorted before it significantly reduces in amplitude.
Ibtisam, I disagree with your statement that "large resistors... consume more power" in this arrangement. The powers are P1 = VIN2/R1, P2 = VOUT2/R2 and the total power is P = (VIN + VOUT)2/(R1 + R2). So, at the same voltage, larger resistors should consume less power...
Another disadvantage of using high resistances (especially true for Rf) is the influence of the resistor's stray capacitance which, in the case of Rf, makes the amplifier behave as an integrator. The classic solution of this and other problems mentioned above, is the T-network...
... but now, under the influence of an interesting discussion about the negative impedance converter (NIC), a new idea occurred to me - to reduce the ratio Rf/R1 by introducing a positive feedback in addition to the negative one. For this purpose, we should add another voltage divider between the op-amp output and its non-inverting input. Thus we can strengthen the negative feedback by decreasing the ratio Rf/R1 but, at the same time, we will weaken it by the additional positive feedback...
Get a Spice simulator, and "play" with the circuits and values.
LTspice is a good start and has accurate models for a wide range of parts:
http://www.linear.com/designtools/software/
It is free and although it has a slighty eccentric user interface once on has it figured out it is really useful to be able to sketch out ideas.
Lots of examples too, to get going with.
... but do not blindly rely on simulators. You must know what such a simulator can do - and when it give unrealistic (false) results.
Example: Opamp with positive and negative feedback. If - by error - the pos. feedback dominates, the simulator (dc and ac simulatioins) will produce false (stable) results. With other words: "Playing" without thinking is dangerous and does not help at all.
As Lutz says, do not use simulators with your eyes closed.
This problem with calculated results is the same when using a digital calculator instead of a slide rule. The slide rule may not be so precise, but one has to keep track of orders of magnitude during the calculation steps. Hands up who still have their slide rules (mine is raised), and who actually uses it on a regular bases for solving problems?
As to opamps, perhaps the best place to start is with a discrete transistor implementation; then one can see what is going on inside the black box at the top level (I am assuming here that the user isn't a Spice expert). Plenty of stuff to look at if one does a web search.
And as I previously mentioned, watch carefully the f3 knee on the gain-bandwidth curve. Don’t assume that it will roll off gracefully; some op-amps will, others will turn a sine into a triangle wave before appreciably dropping the level.
Also anything to do with inputs or outputs getting close or to the power rails needs testing. Rail to rail specifications often have a number of conditions in the small print that won’t necessarily show in generic op-amp simulations.
It depends on the transistor level configuration inside op-amp. Any configuration can be used in experiments. For integrated circuit design, the low value of resistor should be preferred to save silicon area. But I have some experience that the real voltage gain from oscilloscope is not the same as ideal value you ever design. Because of many types of non ideality such as output offset voltaqe and input offset voltage. Current inject from signal source is another problem which need iterative modification of the conventional inverting amplifier schematics. It might not practical in many value of resistors!
Generally speaking, the choice of absolute resistors values for the feedback network of an opamp should be traded off considering different aspects: low resistance values give rise to improved performance in terms of DC accuracy (expecially when input bias currents are not negligible), lower noise, reduced impact of parasitics (e.g. the impact of 1pF parasitic capacitance in parallel to a 100Ohm resistor is negligible up to more than 1GHz, while the impact of the same 1pF parasitic in parallel to a 1MOhm resistor could be relevant at less than 200kHz), and hence better high-frequency behavior. Moreover in integrated designs, low resistance values also mean reduced area occupancy.
On the other hand, a reduced resistance values mean an increased power consumption and additional loading effect on the output (as a matter of fact, the feedback network is connected in parallel of the actual "load" and wastes part of the opamp output stage driving capability). Unless there are valid motivations to do otherwise, it is recommended that the current flowing through the feedback network is negligible (say less than 5%) with respect to the current flowing through the intended load.
The above considerations apply to any opamp-based circuit in any resistive feedback configuration.
Considering the specific case of the "inverting opamp" circuit, a more critical point needs to be observed. The "inverting amplifier", unlike the "non-inverting" opamp-based amplifier is not a "true" voltage amplifier, since its input impedance is not (ideally) infinite, but is given by R1. As a consequence, the value that you choose for R1 is also the input impedance of the amplifier! If you expect that the cicuit behaves like an "inverting voltage amplifier", this input impedance needs to be significantly larger than the internal resistance of the source providing the input signal. Otherwise, input loading effect could be relevant.
I think Paulo 's answer has comprehensively covered the considerations for the choice of absolute resistor values in the feedback network.
It's interesting to note that op-amps intended for low noise applications are often designed with strong output drive capability. This permits the op-amp to work with relatively low-valued resistors in the feedback network, in order to allow the designer to minimize the noise contribution from the resistors in the operational circuit.
.
The problem in using low resistances such as 100 ohms and 1k is in the loading that the ik provides at output. At higher end leakage resistances of boards come in. So a good choice is as 100k and 1 meg, subject to bias current errors. if bias current offset has to be reduced, us a 91k in Positive lead. The effect of bias current is negligible in FET op amps at less than 45 degC. but at 120 degC this can become a significant error. yes one would expect that this is o at low frequencies such as less than 100kHz. At higher frequencies, it may be better to use 1k (Ri) and 10k (Rf). Of course the input resistance is reduced... to 1k in this case. Lower resistances such as 1k is preferred as bias / offset current errors reduce with resistance. So the first thing I will do is to have a look at bias current and its variation with temperature. Then the max operating temperature... Based on this one can choose the resistors. For low current applications, the non availability of resistors beyond about 10 Meg may be a problem. So choose 1meg and 10 meg.. when using FET op amps below 45 degC. See also Gray and Meyer: Analysis of analog integrated circuits. A good, tough book.
Raghuram:
The help given in all of the above replies,
particularly that provided by Paolo, should
answer your original question. But I am a
bit concerned that you needed to ask it in
the first place. It is quite evident there are
now very many types of op amp available,
and no one answer covers all possibilities.
There is no substitute for acquiring your
own body of fundamental principles; then
there will be no need for relying on simple
answers. As a skilled designer, you must
defend every aspect of your solutions and
never have to rely on such quick answers.
If this sounds like a "hard" answer, forgive
me. But it is the best advice I can offer in
circumstances like these.
Barrie
As Barrie says, the best path is to build up one's own experience.
In my experience, testing out parts experimentally is an essential part of understanding a particular amplifier's function. The data sheet may well be very extensive, but without the practical experience to anchor it into real operation the very comprehensiveness of the information can be overwhelming to any design decision process.
Simulation is not reality, and whilst incredibly useful can only be as good (or bad) as the models used and the tests themselves. When simulating a circuit for example, does one put in power supply noise or does one always simulate with ideal voltage/current sources?
Additionally how is the "grounding" being done in reality. Physically one should have in mind the various current paths as poor grounding in the real world can lead to anomalous results.
Finally amplifiers tend to be grouped into categories… low noise, high power, &c. Beware of missing out on something because it’s labelled in this manner and therefore gets discounted because of the way search tables may be organised.
Best, Susan.
One of my longtime technician / engineers, R.I.P., had a saying for a quick pick for a resistor / cap value? "Use a 10k and a point 01"
It looks like 10 k is a favorite value... maybe because we use a decimal number system? If it was hexadecimal/octal, probably 16/8 k would be the preferable value:)
Many valuable answers have already been given. I would like to add that resistors around 1 to 10 K ohms is the right choice. If you consider the input impedance of an op amp it is around a M Ohms where the output impedance is few ohms. Therefore we have to choose the passive components such that their values are too low compared to M ohms and too high compared to Ohms such that ideal op amp approximation remains valid. Possibly that suggests 1~10 K resistors.
Dr. Bandyopadhyay
Agree.
Everything to reduce analog interactions must be considered.
My OpAmps have Z(in) aprox. 1 M Ohm.
My OpAmps have Z(out) aprox. 50 Ohm.
Using the 10x ratio , to effectively minimize interaction ,
the R(in) and R(feedback) would be between 500 Ohm and 100K Ohm.
That is just a safe place to begin.
As Dr. Crovetti points out, the original Source Impedance
and the target Load Impedance must be considered.
As Dr. Parker points out, the F3 knee and system roll-off
must be considered.
Also, we must include the project specifications for accuracy.
Student projects may tolerate 1% error,
but real-world projects by Lutz, Puncochar, Lindberg
may require 0.001% error accuracy.
At some point, we must select
Resistors and Capacitors and OpAmps
which fit the application/project and error tolerance.
Analog design is a wonderful experience in fluid effects
wherein every specification is affected by every other spec !
Hi Glen,
Not a "Dr.", although I do have hopes of doing a PhD here in Physics at Imperial, just a "Mrs". :)
"Mrs." Parker,
Thank you for inserting
your very many perceptive comments at RGN.
You have obviously read widely and astutely.
https://tse1.mm.bing.net/th?id=Ac8ebd85a64c641d8f01d07ae232b0a44&w=171&h=183&c=8&rs=1&qlt=90&pid=3.1&rm=2
I came very late to this question. I would like to thank the colleagues for covering different sides of this interesting question.
If you do not have any other requirement on your amplifiers except gain you are free to choose the absolute value of one resistor and the other will be determined by the gain.
However, there are more functional considerations and practical consicerations
In fact you have to characterize your amplifier to satisfy specific system parameters and requirements where the amplifier is built in.
From this point of view, any amplifier has a set of performance parameters that must be defined not only gain such as input impedance , output impedance. bandwidth, linearity ,power consumption, ..etc. As Barrie said you have to be of wide and deep insight to be successful circuit designer.
I give an example:
You say that the amplifier gain Av= - Rf/R1,
This is only valid for an input voltage source to the amplifier. In real world the sources are not ideal and the signal voltage source has a specific resistance Rs. So, for the validation of this gain one has to make R1 much greater than RS. As this inequality must be satisfied, then R1 value is determined and accordingly Rf.
Then comes the practical considerations that our elements and components are not ideal. The resistors them self has fabrication tolerances. If the gain tolerance is defined then one has to choose resistors with the proper tolerances. Practically smaller value resistors have better tolerances than large value resistors.
Also the sensitivity of the resistors to the temperature variations is to taken into consideration.
Mathematically, if your design problem is formulated such that in your mathematical model , the no. of algebraic equations is equal to the number of the unknowns then you get absolute value for each unknown. If your unknowns are larger than your equations, then you can free choose some unknowns and fix the remaining from the equations.
What i want to conduct is that the design problem may give specific values for the independent parameters or ranges of values for them depending on performance parameters required.
.
Abdelhalim:
.
Unless I missed it in reading your piece,
you did not mention the most important
consideration in selecting an IC op amp
for a practical task; and that is the plain
fact that the OPEN-LOOP gain of these
components is very modest - even low -
at practical frequencies. Let me take as
an example that old war-horse, the 741.
.
The last time I checked (many years ago)
its unity-gain point occurs at roughly one
megahertz (it's never very accurate due to
the monolithic R and C components that
determine this frequency, and allowance
should generally be made for as much as
a +/-30% variation). We can call this f1.
.
For all internally-compensated amplifiers,
using the so-called "dominant-pole" HF
stabilization practice, the open-loop gain
increases in inverse proportion to fs, the
signal frequency (assuming just one, for
the moment); that is, the OL gain is just
f1 / fs. This is often not a very big number.
It is certainly not "infinite!" ! For example,
at the upper end of the audio range it is
a paltry 50.
.
Furthermore, the terribly low slew-rate of
this IC op amp will cause severe transient
intermodulation distortion at all moderate
output levels.
.
Choose your op amp carefully - with its f1
very much in mind!
.
The Lone Arranger
Barrie,
Thank you for complementing my answer by explaining the limitations of the op amp parameters one has to consider in choosing the right op amp. Really it is the first task to do for designing an amplifier. But you did not tell how it could be applied to determine the values of RF and R1 as is required from the question.
Best wishes
.
Abdelhalim:
.
This forum is not meant to be the best place
to teach Analog 1.01. But a quick answer to
your question is that, in addition to the poles
used for HF compensation, there will be (at
least) one pole generated at the feedback
node by the parallel sum of the resistors and
the particular op amp's input capacitance -
which each user has to look up in the Data
Sheet. This extra pole may be large enough
(especially if the chosen feedback network
should use large-value resistors) to cause
anything from mild ringing in the transient
response to outright oscillation.
.
There is no simple answer that applies to
all situations. As for any trade, the design
of analog circuits has to be learned.
.
Barrie
I have never supposed that the calculation of this simple 2-resistor circuit can be such a big design problem... Just choose the minimum possible resistance values (and a little above:) with the desired ratio between them... and of course, make them multiples of 10:)
I rather had problems, many years ago, with the understanding of the very need for these two resistors:) I have told this fun story in the attached Wikibooks story below...
https://en.wikibooks.org/wiki/Talk:Circuit_Idea/Voltage_Compensation#History
Indeed!
The designing of analog circuits is a learning process
that must be experienced through many small failures.
This process is very fluid,
and some interactions may not become evident
until the formula are applied and measured.
Thank you Barrie for pointing out the simple
"one pole generated at the feedback node"
which escaped my observation.
It is affected by the input poles and has an effect on the input poles.
Cyril, what do you think about the idea that an OpAmp ,
being a system of feedback and re-generation,
virtually "talking with itself" ,
may have a "life of its own" ?
just an apprentice with questions...
Glen, you turn the circuitry into poetry with your metaphors:) I love them!
Yes, I also think that when applying a feedback to an op-amp, it begins "living its own life". In the case of a negative feedback, it strives for the point of equilibrium, and in the case of a "self-reinforcing" positive feedback - for the one of the supply rails...
I discovered this discussion just minutes ago.
To add my little bit: if the input source exhibits a rather constant source resistance and power consumption is not an issue: use RS for R1. RF comes 'automatically' then. OTOH if RS varies wildly with the input voltage, it is wise to go high-Z to alleviate the influence of these variations on the gain. Either case it is wise to consider the source resistance when calculating the feedback resistance value. And the same considerations apply when calculating capacitor values for all kinds of filters.
.
Gentlemen and at least One Lady:
.
As for the autonomous behaviour
of analog circuits surely you know
that they are living organisms just
as much as are flora and fauna?
.
In a book I am writing, my teacher
character says much about this -
but I will only send a bit of the text.
In the attachment to this note I've
yellow highlighted the relevant bit.
.
The Lone Arranger
Cyril, the opamp should live the way we wish - not as he wishes. That is the essence of the good design of analog circuits :-))
Josef
.
Josef:
.
The "good design of analog circuits"
is known as "Taming the Wild Beast".
An analog circuit never knows about
your personal desires for it; it simply
obeys all of the physical laws, and it
will sing whenever it wishes, if you're
inclined to neglect even the slightest
detail of those primitive laws.
.
The Lone Arranger
U. Dreher,
I would further develop your "economizing" idea above if we assign the op-amp output resistance to perform the role of RF:) Thus we will get an inverting amplifier without any resistors... consisting only of an op-amp:) The only problem here is where to get the output signal from:) We have to find some internal point before the output resistance:)
Another "clever" trick can be to use the wire resistance between the op-amp output and the inverting input as RF... and the wire resistance between the perfect input voltage source and the inverting input.. if only the input source and the op-amp output are powerful enough:)))
Now a bit more seriosly... In your configuration, where RS acts as R1, we can think of the combination of VIN+RS as an input current source driving a transimpedance amplifier (the op-amp+RF). This current source seems to be perfect (almost ideal)... but actually it is an imperfect "resistor-type" current source that works at ideal load conditions (shorted output)... it is an imperfect current source acting as a perfect one...
@ Cyril
A unique idea. Might be a bit difficult to get the opamp with the output resistance for the individual circuit's needs.
The consideration of the input signal's source resistance is not so far-fetched: there are a lot of applications where you first have a buffer because the source resistance is too high or too variable for any reasonable opamp gain stage.
And even more seriously... The variations (tolerances) of the sum RS+R1 (the entire input resistance) determines the gain accuracy. So, changing the proportion between these resistances, we change the role of each of them on the gain accuracy. In your extreme case (R1 = 0), RS entirely determines the gain accuracy while in the opposite extreme case (RS = 0), R1 entirely determines the accuracy.
From this perspective, we can think of R1 as of an element reducing the impact of RS on the gain accuracy... depreciating the role of RS on the gain...
Barrie
In 1998 I wrote:
...Therefore, we need to examine everything we want to use. First through basic considerations, theories, simulations. The final judgment, however, give a practical experiment only. The one, in contrast to pencil and paper (or computer), automatically and always respects all existing laws, because it is directly "in them" seated.
Josef
Below is the original in Czech :-)
Barry,
Very nicely written extracted text
on the Life and Times of an OpAmp !
You made a clear distinction between Analog and Digital.
These things are not taught in academic school,
only in the school-of-hot-chips.
Looking at the participants' comments above, a timid belief is growing in my mind that in some moments, the "human" (intuitive, imaginative, emotional) thinking still prevails over the "machine" (formal, analytical, sterile) thinking of the participants in this forum...
This gives me the hope that I can make an emotional appeal to reveal basic circuit ideas with the same passion, with which we discuss even the smallest constructive details... or, as they figuratively say, "to see the forest for the trees"...
This is my next attempt to win followers of this noble idea after I tried to do this in 2002 on the Circuit Fantasia site... and later, in 2007, in the Circuit Idea wikibook. The latter is a wiki project, in which everyone (including you) can participate. I started this venture with great enthusiasm with the idea for it to be my "life work"... but my hopes did not come true... and I finally stopped developing it... Perhaps your participation would bring this project to life?
But let's go back to our topic and apply all this there... What will you say if, apart from discussing how big the two resistances are supposed to be, we decide to find out why the two resistors are needed at all? What is their role in this circuit? What are their functions?
The answer to this question would be no less important than the previous one...
http://www.circuit-fantasia.com
https://en.wikibooks.org/wiki/Circuit_Idea
.
Lutz:
.
That old chestnut about the op amp
was probably started by Bob Pease,
who had all kinds of unwarranted and
naive things to say about SPICE.
.
That particular "problem" has a very
simple explanation, once you know a
little about matrix arithmetic, and the
"mathematical perfection" that often
prevails with overly simple circuits.
.
The wrong solution is a "metastate".
.
This "problem" arises whenever the
user of a simulator doesn't have very
clear ideas about its limitations, and
can easily be avoided by the routine
use of transient analysis and minute
excitations that upset the matrix just
enough to make the solution fall into
the correct state. Often, a 1mV pulse
of 1ns width is quite enough to bring
the matrix-solver to its senses.
.
Also, you must be very familiar with the
importance of getting all the tolerance
variables into the right conditions for a
particular kind of work you're pursuing.
.
I am attaching a PPT file that shows
why such care is essential. There is
a great deal more in here that can be
skipped if it's old news to you, maybe
even boring, especially Section. 3 on
the fundaments of the Electronic Life.
Be sure to use the PPT "Slide Mode"
and have your volume turned all the
way up for extra effect!
.
ADDENDUM:
It seems this gateway doesn't allow
large files to be attached. I will try
again later.
.
The Lone Arranger
Barry,
Yep! Everything has its real world limits.
Some of my files have been split chapter-by-chapter, etc.
Barry,
am I right in assuming that your last contribution (adressed to me) concerns my problem as described in
https://www.researchgate.net/post/Where_are_the_limitations_of_idealized_opamp_models?_tpcectx=qa_overview_following&_trid=nlwA1iFErk3LW1yPZkDJCuQx_
Quote: What will you say if, apart from discussing how big the two resistances are supposed to be, we decide to find out why the two resistors are needed at all? What is their role in this circuit? What are their functions?
Hello Cyril - may I ask you:
What kind of new insights do you expect from a discussion about the role of these two resistors? Advanteges/disadvantages of negative feedback?
Hello, Cyril,
Thanks a lot for your references:
http://www.circuit-fantasia.com
https://en.wikibooks.org/wiki/Circuit_Idea
Best regards
ERIK
Dear Barrie,
Thank you so much for the Extract from TEXTBOOK.docx
Please provide a proper reference: author, book, publisher, year.
Your yellow marking: - - - - - - What was that? Living organs? We are here groping down to the deepest roots of a rarely expressed philosophy. Just as a snake differs from a spider, analogue circuits are highly specialized. The living organisms are bounded by nature as determined by their genes. A similar kind of ‘bindingto-nature’ can be discerned in the analogue world. Each type of analogue cell is endowed with specific but limited behaviour, along with a closetful of inherited weaknesses, which must also be well understood. And each individual instantiation of the nominally identical cell is not exactly the same. Analog cells have a mind of their own; they are mischievous, cantankerous and just a bit untrustworthy. - - - - - -
created a philosophical resonance in me: Electrical circuits are manmade subsystems of nature. Because Human beings are natural subsystems of nature electrical circuits are also natural subsystems of nature. With reference to the oscillator principle of nature, electrical circuits are fractal patterns of coupled oscillators when we take the parasitic components into account. We use electrical circuits as models for subsystems in nature, so we can make experiments in order to understand the mechanisms of these subsystems. FROM the galaxies VIA solar systems, planets and smaller systems in our environment TO the superstrings in theoretical physics, we see the same patterns of coupled oscillators. We observe steady state limit cycle or chaotic oscillations. We observe noise in our circuits. Question: Are noise just chaotic oscillations? We distinguish between analogue and digital circuits and sometimes we forget, that the kernels of the digital circuits are analogue cells. Many more questions pop up now. Questions are very important. Question: Do you have an idea about the answer, when you set up your question? - - - asofh - - - (and so on for hours)
We live in a wonderful world with singing analogue circuits around us
ERIK, a lonely traveller on the fractal border between science and religion
Eric,
Very poetically put, by our "resonant" engineer.
You are far away from the 'lonely' group.
I moved from Quantum Mechanics
to Quantum Wave Theory decades ago,
yet,
I cannot escape the "singing" that occurs when I "observe".
Erik:
The extract I recently sent out was from my own
textbook-in-the-making. The yellow highlighting
was simply to draw your attention to the idea of
analog circuits being much closer to nature than
are digital circuits. This is as much as anything,
I believe, due to the high degree of connectivity
and thus parallelism in an almost real-time way.
.
I don't subscribe to the idea that "noise is chaos".
Each of these has a very exact meaning, with no
overlap, except that any formally chaotic system
may generate time-sequences that appear to be
pure noise. Even the complex modulation found
in a modern today looks "just like noise".
.
The Lone Arranger
Barrie,
You have provided us with some excerpts of your „book in preparation“. Here, I can find the terms „digital part“ and „digital cell“.
I understand that your book primarily will not be application-oriented. Instead, it will have a - more or less - general philosophical „touch“. In this context, I have the following question:
Are there really „digital parts“ or shouldn`t we better speak about parts which are operated outside their quasi-linear (analog) characteristics? As a typical example, we have the CMOS-inverter which also can be used as a small-signal analog amplifier.
Thank you
Lutz
.
Lutz:
,
The fact that a circuit cell optimized for its
for its value in digital logic might also be of
use as a (crude) analog amplifier is hardly
the way to teach the latter subject with the
depth and professionalism it deserves.
.
The Lone Arranger
Barrie,
does this really apply - in your opinion - also to the CMOS-Inverter?
Is it really a "digital part"?
Why shouldn`t we define something "in between"?
For example: We also have something between analog and digital signal processing: Continuous amplitude and discrete time (sampled data systems like S/C circuits).
And - what about the comparator-IC? Is it a digital or an analog device?
All phenomena in circuits are analog. Only "use is different" :-)
Josef
.
Lutz:
.
Of course a logic inverter is an amplifier;
frequently that's its main objective. Is it a
"good" solution for general use in analog
signal paths? Probably not. But arguing
along these lines, one might just as well
ask whether a transistor is a logic device
or an analog device; or a piece of wire, if
it should come to that. What are electrons
"good for"?
.
You choose your method to "understand"
analog circuits, and I have chosen mine.
The question, surely, is "What skills are
you trying to inculcate in your students?"
.
Yes, many signal processing techniques
in use today are hard to classify. A text
should make it clear at the outset: what
is going to be emphasized and what will
be set aside for another day? There are
too many things that might be discussed;
surely, the best text-books focus on what
must always be limited in scope. Focus
is always a good thing for a teacher.
.
The Lone Arranger
,
Dear Barrie,
Your book project is great. I am looking forward to reading it. Your English is of course much richer than mine. I have learned new words. Thank you.
With all respect, I have a few comments.
Please provide a list of acronyms used.
Construction of shorter sentences may increase the readability.
e.g.
The electronics field is undergoing a radical transition. It is hard to say for how long this has been happening. It is a historical fact that electronics got its jump-start over a century ago in 1906 when Lee de Forest invented the triode vacuum tube.
Concerning fundamental concepts. You are dealing with electromagnetic fields, so energy represented as the charge Q and the flux F should be used as variables in your models. It is difficult to measure Q and F so we use the time derivatives i.e. the current I = dQ/dt and the voltage V = dF/dt instead. A discussion of the consequences and implications of this approach should be given.
Concerning simulation, the SCHEMATIC-CAPTURE scheme is nice for creating a picture of your circuit, but you should always take a look at the NET-LIST. It is faster and more secure to make changes in the netlist than in the schematics.
Good luck with your book project
ERIK
amateur pocket philosopher
Susan Parker, Esq,
Certainly you mean that
a thoughtful researcher should "experiment'" with their Spice.
I use "ngSpice" via internet browser at PartSIM.com .
However, I know of several others. LT-spice, EDA-spice, P-spice.
Some have FFT, but PartSIM does not.
One leads directly into PCB manufacture, easyeda.com,
but PartSIM will not export gerber files.
Some have a library of device models, PartSIM is just developing this.
Some will handle only a few circuit stages,, strictly for students,
but PartSim will handle as many as 34 OPA filter stages .
Lutz,
Agree !
"Garbage in, garbage out" !
The researcher must "understand" what his Spice is doing.
You have astutely advised me that Spice calculates via Linear Equations, but the real world is Non-Linear by nature. Cuidado!
just an old apprentice
Cyril,
You make me think !
"Thus we can strengthen the negative feedback by decreasing the ratio Rf/R1 but, at the same time, we will weaken it by the additional positive feedback..." Indeed, I have juggled this Watts-IN = Watts-OUT concept. I applied this Positive-FB idea in a Band-Pass filter circuit, only as an experiment, but it worked within measurable bounds. As I recall, too much Positive FB produced Oscillations, and too little Positive FB brought about a Notch. But, you are right about making more power available at the Vout node.
Glen,
please, can you give some short explanations to your "positive feedback" statement above? (In such a simple amplifier configuration I cannot see any pos. feedback.)
Thank you
Dear Lutz,
Glen probably means my extravagant idea to connect in the inverting amplifier a voltage divider (additional two resistors) between the op-amp output and its non-inverting input. This will introduce a (smaller) positive feedback... and, as a result, the ratio RF/R1 can be smaller.
Actually, the inverting amplifier will become a VNIC.
Dear Cyril - yes, perhaps. The background of my question was only to avoid misunderstandings.
Later it came into my mind that Glen eventually was referring to high-frequency effects where the neg. feedback turns into a positive one.
It is interesting that here the virtual ground point is located somewhere inside RIN... and a part of this resistance is "eaten" by the circuit:)
Cyril, and Lutz, Always appreciate your informed comments.
I was referring to a thread that Ponchochar was in, with us all.
I think we started by discussing a oscillator, which Barrie has advised should be driven by a clean ping.
I wrote ""As I recall, too much Positive FB produced Oscillations, and too little Positive FB brought about a Notch
The discussion turned to a Filter, ( which I commonly drive with a noise signal) to allow the filter effect to show, measurably. ""
The following are notes from my recorded project and from the long-ago thread.
GC_ET_RB_Josef_NonInvert-BandPass-Fx_Positive-FB_Schem is attached showing a tuned-negative-feedback changed to positive-feedback for an oscillator.
PFB-rg-v1-S-3-161214-1332-3N.png is my revision.
PFB-rg-v1-B-1-161117-1729.jpg are the variations of +/-25% of f(0) which was my control target / goal.
In effect, we started with a Positive-Feedback Oscillator, added a tuned feedback network, and this allowed adjusting the f(0) as a Filter.
In tinkering, I applied aprox. Q=3 positive-feedback and generated an aprox. Q=7 BandPass V(out).
Results:
(1) more complex than needed for a Q=7 BandPass Filter,
(2) feedback signal is delayed, filtered, and eventually applied as a positive control bias,
(3) background noise (low level) indicates a tendency to oscillate,
(4) Vout always contained this background noise as shown in PFB-rg-v1-T-3-161213-1411.jpg.
(5) Obtaining a Null at under-bias and a Filter at middle-bias and a Oscillator at over-bias indicates Gross Sensitivity to the Driving Signal Amplitude and makes this unsuitable for my radio filtering projects.
The thread was interesting in that it showed some proof of concept for a positive-feedback filter, but it is not a well engineered design when driven by a noise signal ( typical of a filtering project ). Obviously, I was having 'fun' and more serious work would need to be done. However, I already have far superior circuits working in my radio circuits projects already, so this diversion was put to 'rest' .
(link http://www.geocities.ws/glene77is/ )
just an old apprentice
I will develop even more my idea above by revealing the role of the op-amp in both inverting configurations:
- In the "pure" op-amp inverting amplifier with negative feedback, the op-amp "eats" only the resistance RF thus acting as a negative resistor with resistance -RF. As a result, R1 only remains... and the virtual ground is between R1 and RF.
- In the "mixed" op-amp inverting amplifier with additional positive feedback (VNIC), the op-amp "eats" the sum of resistances RF + R1'' thus acting as a negative resistor with resistance -(RF + R1''). As a result, R1' only remains... and the virtual ground enters R1.
Here R1' (from the left) and R1'' (from the right) are the two parts of R1 separated by the internal virtual ground (R1 = R1' + R1'').
One can also incorporate a divider into the feedback loop... just to keep life interesting and feedback resistor values more reasonable.
http://www.radio-electronics.com/info/circuits/opamp_inverting_high_impedance/op_amp_inverting_high_imp.php
May well need a bit of external compensation...
Susan
I do not think this is completely right Av = - R2 (R3 + R4) / (R1 x R4 )
It is valid only if R2 is much larger than R4
In general Av = - (R2 R3 +R3 R4+R2 R4) / (R1 x R4 )
Josef
Susan and Josef,
I have known this circuit from many years... but then I could not understand it because it was presented to me as an "inverting amplifier with a T-bridge in the feedback". I only understood it when I managed to see this voltage divider in the negative feedback...
Now, after I glanced at it, a new idea struck me - can we remove the resistor R2 as its function will be performed by the equivalent R3||R4 resistance (according to the Thevenin's theorem)?
Cyril
If we remove R2, the gain will be just - R3/R1 - because R4 as attached to the virtual ground has no influence (ideally).
Josef
.
Susan is of course quite correct.
.
It was not stated (but it should be
obvious) that the "mathematics"
changes, and the resistor values
need to be altered for the gain to
be preserved at its original value.
.
I am truly amazed - perhaps the
word is stunned - by the amount
of time spent on trivial matters.
It's a total mystery to me how the
various and diverse 'advisors' in
this thread can hope to address
real challenges in analog design.
.
This is certainly not a deserving
topic for a "research" thread. I've
called it "Analog Design 1.01", in
the past, and stand by that.
.
The Lone Arranger
Josef,
Thanks for the detailed answer.
It now remains only to explain what is wrong in my "naive" suggestion. Maybe the Thevenin's theorem is not valid here:)?
Cyril
Maybe I was wrong to express. If we substitute R2 for a shortcut, what I said earlier is true.
Josef
Guys,
We are so fortunate to be reminded that we are tinkering with "dits and dahs "
when the new worlds of Analog #1001, #2001, #3001 are ours to create.
.
For this current question :
I first think of values that cause the OPA to be balanced, with V(out) = 0 .
Remembering that I must stay withing the bounds of current handling capacity,
and Gain Curve capacity, internal cross-talk, heat-dissapation , etc.
That is Analog 000 Introductory .
...
After that, with my math models sitting nearby , I define a "Analog Target 101" :
To produce that end, I have used Thought and Proto-Type circuits containing
(1) linear passive circuits,
(2) non-linear dynamic circuits,
(3) resonant OPA filter circuits, in the feedback loop,
(4) off-board control signals pumped into the feed-back loop
(5) and some-times little night-dreams
.
all to move my "101" OPA circuit towards my target V(out) state, in the real-world.
I only adjust or balance numbers after my conceptual design is heading in the right direction.
...
And
... I am still aware that these design tools are "Analog 101",
[ still inside the academic "box" that we learned in school ]
...
Only when I can build a New Box to contain a New Idea
will I be thinking outside of the academic box.
...
Normally, our research leads us to improving the past idea, still inside the Academic Box.
Real Research should lead us to Seek for the New Idea that is OutSide of the Academic Box.
.
That is the old apprentice' "Soap Box" speech for today .
...
Raghuram,
You ONLY asked "So which one is the best? And what are all the reasons for choosing it?"
My standard is 10K R(in) and R(FB) between 1K and 100K , because that fits my usual criteria.
But, you gave no criteria ,
so now, you should give the real-world requirements or environment of your circuit.
Then , we may adjust these passive components to better fit the new criteria.
.
Raghu:
.
There is one aspect of this op amp
drama that does not seem to have
been raised in all of the preceding
advice, and I have to confess that
also includes my own. Working in
analog design for over 7 decades
(since before teen years) I probably
make too many assumptions about
what is, or what is not, “obvious” in
this field.
.
In the inverting op amp configuration
you raised in your original query, the
resistor R1 loads the primary source
of the signal, whatever that might be.
This loading can affect the waveform
and/or peak voltage of the source, so
you need to make R1 high enough to
avoid that outcome.
.
But there's another consideration: the
the output resistance of the source is
perhaps comparable with the R1; and
in this case of course, the overall gain
calculation must include this aspect
of the source, if the gain reduction is
significant. The effect is usually quite
small, and may be no worse than the
error due to the resistor tolerances.
.
Now, while this could suggest that it
is advisable to use the highest value
for R1, this view is counter-indicated
by the need to minimize the Johnson
noise of the resistors. To get a fairly
good estimate of this noise, you can
invoke the following relationship:
.
The noise spectral density (NSD) of a
resistor evaluates to 129 picovolts per
Hertz (of channel bandwidth) times the
square-root of its ohmic value.
.
So, a 1k-ohm resistor will generate an
NSD of 4.08 nV/Root-Hz, amounting
to a noise voltage of about 4 uV in a
1 MHz channel bandwidth – whatever
the actual range of signal frequencies.
It is important to realize that the latter
part of this calculation assumes that
both upper (HF) and lower (LF) band
edges are "brick-wall" transitions; that
is, the channel gain drops suddenly to
magnitude zero. This is never quite the
case in any real analog circuit (though
it can be in digital channels). Plenty of
guidance is to be found in
.
http://analog.intgckts.com/equivalent-
noise-bandwidth/
.
A common case is when the bandwidth
falls on a single pole to a -3 dB corner
at frequency fc. In this case the effective
(“brick-wall”) noise bandwidth is simply
(pi/2)fc. For example, assuming a value
of 1 MHz for the -3 dB point, the noise
bandwidth will be about 1.5 MHz.
.
Finally, do not forget that the input bias
current of many earlier BJT op amps is
not very low, and so it is customary to
include an offset-compensating resistor
in the unused input node (in your case,
the non-inverting input) whose value is
equal to the parallel sum of R1 and RF.
When the additional noise contributed
by the extra resistor is significant it can
be shunted by a high-quality capacitor
large enough to eliminate the mid-band
noise and above.
.
Probably the best book ever written on
the topic of voltage feedback amplifiers
is the one by Jim Roberge, back when
he was at MIT. It’s dated 1975 but that
should not deter you from acquiring a
second-hand copy of this outstanding
work: “Operational Amplifiers: Theory
and Practice”.
.
The Lone Arranger
Barrie,
Interesting comment on using a tuned circuit on the non-inverting input.
Your example reads as a high-pass ( RC parallel ).
@ Josef Punčochář
Your suggestion on "High input impedance inverting operational amplifier circuit" , I quote " If we remove R2, the gain will be just - R3/R1 - because R4 as attached to the virtual ground has no influence (ideally)" is not acceptable if the word "remove is used to mean "delete". Because when you remove R2 the R3 R4 series combination acts as a . load for the operational amplifier. One should remember that "virtual ground" and the real ground points are two separate nodes and they are not connected. Therefore removing R2 will remove feedback to the inverting input of the operational amplifier. If, however the word "remove" is used to make r2 =0 then the situation may be handled as follows. One can definitely use Thevenin's theorem to calculate the gain of the original circuit which is explained below:
Let VTH = Thevenin's equivalent voltage and RTH = thevenin's impedance.
-VTH/(R2+RTH) = Vi/R1
Then
(VTH/Vi) = - (R2 +RTH)/R1
On can calculate
RTH = R3R4/(R3+R4)
and
VTH = V0 R4/(R3+R4)
where V0 is the output voltage of the operational amplifier.
Then (V0/Vi) =-( (R3+R4)/R4)( (R2+RTH)/R1)
Dear Anup Kumar Bandyopadhyay,
I recently wrote to Cyril:
Cyril
Maybe I was wrong to express. If we substitute R2 for a shortcut, what I said earlier is true.
Josef
Talking about the role of the offset-compensating resistor connected in the unused op-amp input, I make an interesting parallel with TTL circuits...
One of the difficult things to understand from my students there is that these circuits can be driven not only by input voltage but also by "input resistance". I demonstrate this trick by connecting a variable resistor (rheostat, photoresistor, etc.) between the input and ground. When I increase its resistance, at some point (over 1 k) the logical element switches. This is of course due to the fact that TTL circuits source current from their inputs.
Similarly, for curiosity (and fun:), here we can connect a high-resistive sensor in the op-amp input... and varying its resistance to observe how the output voltage of the op-amps changes... Then we can ask the inquisitive students, "Where is the input source here?"
If they are observant enough, they could find that the input source is inside the op-amp... and it is not a voltage source, but a current source creating an input voltage across the external resistor...
Any source resistance is going to provide a drop depending on the current flowing. It is a difficult concept for many students to understand that for a Lo the input should be less than the max allowed voltage of .8 including the drop due to current flowing, and a current of 1.6mA flows in TTL towards the source! The fact that the input is returned to zero volt alone does not determine the state, it also depends on the resistance through which it is returned. The difficult part to follow is that the source should be able to sink a current of about 1.6 mA for a TTL circuit. So if you return the input to zero voltthrough a resistance the max value hte resistance will have to have is 500 ohms (=.8V/1.6mA). If the voltage returned to is .2V then the max resistance allowed is (.8-.2)/1.6mA= 400 ohms. Often I have found them to be unaware of the importance of source resistance and that causes eroneous results! Of course in amy other situations the current is lesser, as it is .4mA with LSTTL and much lesser with CMOS devices.... That does not mean zero current! hence one cannot have infinit resistance as source resistance.
Note also that manufacturers do not often indicate the direction of flow of bias current in op amps. The LM 358 has current flowing into the op amp while 741 has current flowing out of the op amp. These are current sources....
Exactly, Mani... I will even further delve into the philosophy of these odd TTL circuits...
In fact, the input source of a TTL element is inside it... and it is a current source. It is simply implemented by the internal pull-up resistor RB connected between VCC and the base of the input multiple-emitter transistor. The previous stage (the input source) only diverts (or not) this current through itself (what they call current steering).
So, when a TTL gate drives another TTL gate, its output does not act as a voltage source. It is simply a switch that commutates the else's current source...
Let's continue with the op-amp "input" philosophy... which is also difficult to understand for students...
The op-amp has two input sources - a "true" input voltage source and a bias current source... The former is external (the previous stage), the latter is internal... and, for some reason, they are connected in series - the bias source passes its current through the input voltage source. This concept is also difficult to understand for students... It is interesting to see why...
Thanks Cyril.
Ability to use a device properly means a fair understanding of how it works. So instead of jumping to op amps and its uses. it will be good to speak about internal construction of a simple oBJT op amp... may be a simple op amp haing a differential stage and a class B output. It will perhaps help the students to know what bias current and offset current are and their various effects. Another difficulty is to undertsand the difference between offset voltage V offset as specified by the manufacturer (which is for low resistance at inputs) and the offset referred to input (Voffset rti)that comprises in addition the effects of bias currents.. [fundamentals are the most useful but dificult to undertstand ] I feel that in many modern situations people tend to work from top to bottom, which leaves the fundas at last priority to understand....Perhaps I am of OLD school....
Exactly, Mani... I am also of this old school:)
Regarding the difference between Voffset and Voffset rti, maybe it would be clear to say that the former is a property of the bare op-amp while the latter is a property of the op-amp circuit?
IMO there is something misleading in the Voffset definition - "the voltage that should be applied at the op-amp input to zero its output voltage"... maybe because it rather expresses the way it is measured, not its nature... which prevents its understanding... Maybe it is better to imagine the real Voffset is somewhere inside the chip... and to measure it, we compensate it with equal external voltage... which is a copy of Voffset...
.
Friends of the Electron:
.
You might be interested in this Competition,
organised by the Solid-State Circuits Society
of the IEEE:
.
https://sscs.ieee.org/education/2017-2018-circuit-analysis-design-contest
.
The Lone Arranger
.
Thank you, Victor:
.
I find the definition of the SSCS "problem"
is incomplete in important ways, related
to LF accuracy. The function has already
been implemented, many years ago, in a
circuit I designed at Analog Devices (the
AD639; see attachment). In fact, the key
translinear cell was implemented much
early, during my Tektronix years. . So the solution, or a least, a very efficient one, using the excellent 'translinearity' of
BJTs, is already in the public domain. It's
a simple matter to translate the circuit to
sub-threshold CMOS form -- even easier
in view of the fact that the accuracy was
not specified (nor other important things,
like behaviour over temperature).
.
There was also a JSSC paper about this
circuit; it is attached. I see it was written
35 years ago. How the years fly by......
.
The Lone Arranger
Hi folks,
Just watching the show, it's probably time for the curtains!
After three years of "solid-state" discussion and lots of maths and circuit simulations and models let's conclude - and for that I would like to go with the Lone Arranger" to arrange that the dot has not changed its position - it remains at Analog 1.01 - so it is not 10.1
We cannot be spending three years just to decide what combination of resistances we would use for a simple opamp circuit. After all we have to design chips not just one stage circuit.
Have a great time!
We can continue (a few years more:) this so exciting discussion about the 2-resistor circuit connected to the inverting op-amp input if we include a similar 2-resistor circuit to the non-inverting input...
Dear Cyril and Mani
Concerning offset and switches
DC offset.
The mean amplitude of a waveform (originally, a direct-current ("DC") waveform)
Frequency offset.
The difference between a source and a reference frequency
https://en.wikipedia.org/wiki/Frequency_offset
Phase offset.
In sinusoidal functions or in waves "phase" has two different, but closely related meanings. One is the initial angle of a sinusoidal function at its origin and is sometimes called phase offset or phase difference. Another usage is the fraction of the wave cycle that has elapsed relative to the origin.
https://en.wikipedia.org/wiki/Phase_(waves)
TTT, things take time
A switch is a resistor (current controlled voltage source) which in a time interval change its value from very large (OFF) to very small (ON).
All the best to all of you
ERIK
.
Prasanna:
.
There are none so deaf as those who
do not wish to listen to sound advice.
.
The Lonely Arranger
.
.
For anyone with a taste for self-flagellation
attached is a copy of this entire conversation.
.
It seems that MOS for SSCS design needs to work in weak inversion region (sub-threshold region.)
- Badges
- Science topic
- Similar topics
- Books