|
Hello, :
May I ask, is there any limitation when using the behavioral source (bv, bi) ? I found it seems there are some differences in my simulations, when I tried to add the BV behavioral voltage source to do some math manipulation of a feedback signal in a application schematic (eg: flyback, etc), before adding the BV into the feddback, the simulation ran quite normal as expected, but after adding the BV to the feedback, the "FB" signal seems having larger swing amplitude which makes the output to switch separated in chunks (many time durations), the only thing I doubt caused those differences is the timestep, the simulation without adding the BV model, ran into smaller timestep while the one with BV didn't, the worse resolution of the simulation made the errors.
So , may I confirm, what's the difference(/limitations) when using BV in those kinds of applications ?
Thank you very much
Have a nice day!
|
Hi, :
I'm sorry, I have reduced the transfer gain to almost 1+/- 2u, which I thought the +/- 2u (1e-6) shouldn't take any effect to the feedback, at least not result in a +/- 0.03 V variation. The example isn't flyback, it includes some properties , I forgot to simplify the transfer gain to single "1", maybe tomorrow. But if this means, there aren't any known issues about the BV,BI behaviors.
Is it true, not any limitation found yet ?
Thank you very much. ---In LTspice@..., <ericsson.sunshine@...> wrote : Hello, :
May I ask, is there any limitation when using the behavioral source (bv, bi) ? I found it seems there are some differences in my simulations, when I tried to add the BV behavioral voltage source to do some math manipulation of a feedback signal in a application schematic (eg: flyback, etc), before adding the BV into the feddback, the simulation ran quite normal as expected, but after adding the BV to the feedback, the "FB" signal seems having larger swing amplitude which makes the output to switch separated in chunks (many time durations), the only thing I doubt caused those differences is the timestep, the simulation without adding the BV model, ran into smaller timestep while the one with BV didn't, the worse resolution of the simulation made the errors.
So , may I confirm, what's the difference(/limitations) when using BV in those kinds of applications ?
Thank you very much
Have a nice day!
|
of course, the input isn't constant "1 + +/- 2u", it's varying calculated result,? the result was modified to a very tiny limited range which is approx to 2u.
I just tried a simple schematic with constant gain = 1+0.1m, the result is same as without the bv device. I am thinking how to repeat the same phenenmeon in other circuit topology.
Thank you very much. ---In LTspice@..., <ericsson.sunshine@...> wrote : Hi, :
I'm sorry, I have reduced the transfer gain to almost 1+/- 2u, which I thought the +/- 2u (1e-6) shouldn't take any effect to the feedback, at least not result in a +/- 0.03 V variation. The example isn't flyback, it includes some properties , I forgot to simplify the transfer gain to single "1", maybe tomorrow. But if this means, there aren't any known issues about the BV,BI behaviors.
Is it true, not any limitation found yet ?
Thank you very much. ---In LTspice@..., <ericsson.sunshine@...> wrote : Hello, :
May I ask, is there any limitation when using the behavioral source (bv, bi) ? I found it seems there are some differences in my simulations, when I tried to add the BV behavioral voltage source to do some math manipulation of a feedback signal in a application schematic (eg: flyback, etc), before adding the BV into the feddback, the simulation ran quite normal as expected, but after adding the BV to the feedback, the "FB" signal seems having larger swing amplitude which makes the output to switch separated in chunks (many time durations), the only thing I doubt caused those differences is the timestep, the simulation without adding the BV model, ran into smaller timestep while the one with BV didn't, the worse resolution of the simulation made the errors.
So , may I confirm, what's the difference(/limitations) when using BV in those kinds of applications ?
Thank you very much
Have a nice day!
|
ericsson.sunshine asked about (I think) timestep effects, and Behavioral sources, and feedback.
I can't figure out what you are asking.??However, I guess the answer is to say yes, there can be limitations when adding anything to a simulation.? Everything interacts.? There might be more limitations when considering B-sources.
Regarding timestep and Bv/Bi sources, take a careful look at the tripdv and tripdt parameters of those functions.??The Help file description says little more than "Tripdv and tripdt control (time)step rejection."??There is a more complete explanation (although one that is almost as confusing) on the LTwiki webpage, here:
? ??
I have read that description a dozen times, but I still don't understand it.
Regarding feedback signals, you should be aware that the inputs to a B-source should not depend on instantaneous feedback from its own output.? This is mentioned on the Help page, this way:
? ? "However, it is assumed that the circuit element current is varying quasi-statically, that is, there is no instantaneous feedback between the current through the referenced device and the behavioral source output. Similarly, any ac component of such a device current is assumed to be zero in a small signal linear .AC analysis."
It is not clear exactly what that means.? It SEEMS to say that you shouldn't use a B-source anywhere near a feedback loop -- which seems like an extreme restriction.? On the other hand, the restriction stated above was given in the context of using pin currents as the inputs to the B-source, which would imply that maybe the restriction only applies to B-source formulas that use pin currents of the form Ib(Q1) or Id(M5), and not in general to all B-sources.? Regardless, I have come across mention of this restriction before, about instantaneous feedback and signals that vary faster than "quasi-statically" (meaning: they can only be DC or changing very slowly).
So I can't answer your question directly (because I don't understand the question).? But I can say this:
?- B-sources have some restrictions when used in feedback circuits;
?- B-sources have some restrictions even when not used in feedback circuits, regarding what signals can be used as their "inputs";
?- B-sources can have issues with timestep control;
?- B-sources have extra parameters to help remedy the problems with timestep control.
Regards, Andy
|
Hi, Andy:
Thank you for your replies, I have read the documentation in help or on-line, neither could I understand it very well , though it cost me much looooong time trying to read it for the first time, made me step back thinking if I should read it for the 2nd time.
I have uploaded an example file, which contain 2 simulation files, one is from the original jigs, the other is added the BV=V(FB) simply doing the signal unity transferring, but the result is much different, I don't know if this could completely represent the original question I ask (due to the confidential reason), but it shows similar phenomemon, -> logically nothing changed, but severe result happens.
The file named "20181115_BV,BI behavioral device FB pin test failure desc.zip" uploaded at the temp folder. If you could kindly take a look at it and try it.
? ? "However, it is assumed that the circuit element current is varying quasi-statically, that is, there is no instantaneous feedback between the current through the referenced device and the behavioral source output. Similarly, any ac component of such a device current is assumed to be zero in a small signal linear .AC analysis."I guess this means, the .AC analysis is based on the DC biased point, then doing small signal analysis, and the small signal injection doesn't include the BV,BI which is not typical SPICE devices. But my question was ran in .TRAN analysis.
Thank you very much.
Have a nice day! ---In LTspice@..., <ai.egrps@...> wrote : ericsson.sunshine asked about (I think) timestep effects, and Behavioral sources, and feedback.
I can't figure out what you are asking.??However, I guess the answer is to say yes, there can be limitations when adding anything to a simulation.? Everything interacts.? There might be more limitations when considering B-sources.
Regarding timestep and Bv/Bi sources, take a careful look at the tripdv and tripdt parameters of those functions.??The Help file description says little more than "Tripdv and tripdt control (time)step rejection."??There is a more complete explanation (although one that is almost as confusing) on the LTwiki webpage, here:
? ??
I have read that description a dozen times, but I still don't understand it.
Regarding feedback signals, you should be aware that the inputs to a B-source should not depend on instantaneous feedback from its own output.? This is mentioned on the Help page, this way:
? ? "However, it is assumed that the circuit element current is varying quasi-statically, that is, there is no instantaneous feedback between the current through the referenced device and the behavioral source output. Similarly, any ac component of such a device current is assumed to be zero in a small signal linear .AC analysis."
It is not clear exactly what that means.? It SEEMS to say that you shouldn't use a B-source anywhere near a feedback loop -- which seems like an extreme restriction.? On the other hand, the restriction stated above was given in the context of using pin currents as the inputs to the B-source, which would imply that maybe the restriction only applies to B-source formulas that use pin currents of the form Ib(Q1) or Id(M5), and not in general to all B-sources.? Regardless, I have come across mention of this restriction before, about instantaneous feedback and signals that vary faster than "quasi-statically" (meaning: they can only be DC or changing very slowly).
So I can't answer your question directly (because I don't understand the question).? But I can say this:
?- B-sources have some restrictions when used in feedback circuits;
?- B-sources have some restrictions even when not used in feedback circuits, regarding what signals can be used as their "inputs";
?- B-sources can have issues with timestep control;
?- B-sources have extra parameters to help remedy the problems with timestep control.
Regards, Andy
|
Ericsson.sunshine, that paragraph (from LTspice's Help) about "instantaneous feedback", applies to .TRAN analysis too.
The last sentence mentions .AC analysis, to say that it is not only .TRAN analysis, it also applies to .AC analysis.
Regards, Andy
|
Ericsson.sunshine,
Regarding your circuit in 20181115_BV,BI behavioral device FB pin test failure desc.zip:
I think you forgot about the current through R2 and the voltage drop across it.
The original circuit had about 83uA current through R2.? So there was 11V voltage drop across R2.? The altered circuit has no current through R2.? Therefore, the voltage on node FB would be 11V greater than it was in the original circuit, if everything else was the same.? That's going to cause a big difference.
Looking quickly at the LT3580 datasheet, I think they depend on a significant voltage drop across the feedback resistor.? The current into the part's "FB" pin is tightly controlled, within 2%-3%, and the normal voltage on that pin is small, only +1.215V.? In your simulation it has?+4.7V (+7.1V at first) on the pin which is too much and could kill the LT3580.
Regards, Andy
|
Hi, Andy:
Thank you for the replies, I'm sorry, it's too rush, so made this mistake. I have uploaded another example which uses the LT1249, typically the FB is resistor divided, the current sinked into the IC's normally treated very tiny, the voltage reference almost viewed as the voltage divided by resistors.
The new example is more approaching to what problem I described. Is that caused by the load effect ? I mean the current sinked into the IC, or the errop's peripherals ,those affect the difference of the result ?
What do you think of those ? Could it be solved by other methods ? Bypass with other device , models or ... ?
Thank you for the opinions.
Have a nice day!
---In LTspice@..., <ai.egrps@...> wrote : Ericsson.sunshine,
Regarding your circuit in 20181115_BV,BI behavioral device FB pin test failure desc.zip:
I think you forgot about the current through R2 and the voltage drop across it.
The original circuit had about 83uA current through R2.? So there was 11V voltage drop across R2.? The altered circuit has no current through R2.? Therefore, the voltage on node FB would be 11V greater than it was in the original circuit, if everything else was the same.? That's going to cause a big difference.
Looking quickly at the LT3580 datasheet, I think they depend on a significant voltage drop across the feedback resistor.? The current into the part's "FB" pin is tightly controlled, within 2%-3%, and the normal voltage on that pin is small, only +1.215V.? In your simulation it has?+4.7V (+7.1V at first) on the pin which is too much and could kill the LT3580.
Regards, Andy
|
Hi, Andy:
Sorry, I saw the blindpoint. The BV didn't the connection of two nodes, it's a new voltage source. So the errOP's peripheral won't work as expect, after I added a 10K resistor between the BV and IC's 'Vsense', the simulation woks as expect.
I think I know what's going on.
Sorry for the disturbance,
Best regards. ---In LTspice@..., <ericsson.sunshine@...> wrote : Hi, Andy:
Thank you for the replies, I'm sorry, it's too rush, so made this mistake. I have uploaded another example which uses the LT1249, typically the FB is resistor divided, the current sinked into the IC's normally treated very tiny, the voltage reference almost viewed as the voltage divided by resistors.
The new example is more approaching to what problem I described. Is that caused by the load effect ? I mean the current sinked into the IC, or the errop's peripherals ,those affect the difference of the result ?
What do you think of those ? Could it be solved by other methods ? Bypass with other device , models or ... ?
Thank you for the opinions.
Have a nice day!
---In LTspice@..., <ai.egrps@...> wrote : Ericsson.sunshine,
Regarding your circuit in 20181115_BV,BI behavioral device FB pin test failure desc.zip:
I think you forgot about the current through R2 and the voltage drop across it.
The original circuit had about 83uA current through R2.? So there was 11V voltage drop across R2.? The altered circuit has no current through R2.? Therefore, the voltage on node FB would be 11V greater than it was in the original circuit, if everything else was the same.? That's going to cause a big difference.
Looking quickly at the LT3580 datasheet, I think they depend on a significant voltage drop across the feedback resistor.? The current into the part's "FB" pin is tightly controlled, within 2%-3%, and the normal voltage on that pin is small, only +1.215V.? In your simulation it has?+4.7V (+7.1V at first) on the pin which is too much and could kill the LT3580.
Regards, Andy
|
Ericsson.sunshine,
Without running any simulation yet, it looks like your second uploaded example (LT1249) could be a problem too.
The Vsense and VAout pins are the input and output pins of the error amplifier.? The original circuit has an RC network (C1, C2, R1) in that feedback loop, where it provides AC feedback around that amp.? Your added Bv source is applied directly to the Vsense pin, so it shorts out that feedback.? That might be a big change.
I don't think there is very much loading difference on the resistors (R5 & R6) with B1 added.? But I'm sure there is a little.
Running the simulation, it looks like the output voltage is still about right, but it is more "ragged" with a sawtooth shape.? This is only a first guess, but I'm thinking the error amplifier is not stable without its AC feedback path, and it's hunting back and forth.? Indeed, in the original circuit VAout stabilizes at around?+4V with a little bit of variation; but in the modified circuit, it swings wildly between 5.5V and almost 12V, and doesn't settle down.
Every change has a consequence.
Let me ask this:? What are you trying to do?? Why add the Bv sources?
I can't answer your last questions (e.g., "Could it be solved by other methods?") without knowing what you are trying to do.
Regards, Andy
|
An interesting thing......is ...
Those phenomemon in other software eg: PSIM, the math parsing won't take any effect with those loading effect. This might confused me to do those kind of contradiction, if I ran both simulators in a same while. I can't believe , it took me many minutes to recall why I do this kind of mistake (neglected the loading effect), then I done some hard effort to figure out the way why I connect wires like this, to the final, because the different fashion in different simulators.
Some strange experience. ---In LTspice@..., <ericsson.sunshine@...> wrote : Hi, Andy:
Sorry, I saw the blindpoint. The BV didn't the connection of two nodes, it's a new voltage source. So the errOP's peripheral won't work as expect, after I added a 10K resistor between the BV and IC's 'Vsense', the simulation woks as expect.
I think I know what's going on.
Sorry for the disturbance,
Best regards. ---In LTspice@..., <ericsson.sunshine@...> wrote : Hi, Andy:
Thank you for the replies, I'm sorry, it's too rush, so made this mistake. I have uploaded another example which uses the LT1249, typically the FB is resistor divided, the current sinked into the IC's normally treated very tiny, the voltage reference almost viewed as the voltage divided by resistors.
The new example is more approaching to what problem I described. Is that caused by the load effect ? I mean the current sinked into the IC, or the errop's peripherals ,those affect the difference of the result ?
What do you think of those ? Could it be solved by other methods ? Bypass with other device , models or ... ?
Thank you for the opinions.
Have a nice day!
---In LTspice@..., <ai.egrps@...> wrote : Ericsson.sunshine,
Regarding your circuit in 20181115_BV,BI behavioral device FB pin test failure desc.zip:
I think you forgot about the current through R2 and the voltage drop across it.
The original circuit had about 83uA current through R2.? So there was 11V voltage drop across R2.? The altered circuit has no current through R2.? Therefore, the voltage on node FB would be 11V greater than it was in the original circuit, if everything else was the same.? That's going to cause a big difference.
Looking quickly at the LT3580 datasheet, I think they depend on a significant voltage drop across the feedback resistor.? The current into the part's "FB" pin is tightly controlled, within 2%-3%, and the normal voltage on that pin is small, only +1.215V.? In your simulation it has?+4.7V (+7.1V at first) on the pin which is too much and could kill the LT3580.
Regards, Andy
|
Hello, Andy: Let me ask this:? What are you trying to do?? Why add the Bv sources?
Just some quick transfer trying for compensation to a newer shape as the secondary reference, concurrently with the first priority feedback still connected. The description expressed in the fashion of could apply to any topology.
Thank you for the suggestions.
Best regards.
|
For example, (not adapted by me), the typically active PF controller makes the PF upto very high 99%, but after EMI filter circuit added, maybe some phase difference may be caused, those phase difference causes cos(theta) decreases some, so if maybe improve it, but doing some compensation, such kinds of things. Especially when using in heavy loading. (Heavy EMI filter.)
Thank you very much. ---In LTspice@..., <ericsson.sunshine@...> wrote : Hello, Andy: Let me ask this:? What are you trying to do?? Why add the Bv sources?
Just some quick transfer trying for compensation to a newer shape as the secondary reference, concurrently with the first priority feedback still connected. The description expressed in the fashion of could apply to any topology.
Thank you for the suggestions.
Best regards.
|
I still don't know what the Bv is for.
Since it represents an unreal component, are you thinking about using it to experiment with an altered circuit, before designing that "alteration" using real components?? Or do you want to add an extra signal into your circuit?? Otherwise, I can't imagine why a Bv would be a useful thing, here.
Regardless of the reason, you need to be more careful to consider the effect of adding the Bv to the existing circuit.? In both examples, it has a profound effect even though it is only a unity-gain buffer.? Remember that Bv unity-gain buffers have infinite input impedance and zero output impedance and are uni-directional, so breaking a connection and inserting a Bv source in its place, is a significant change.
I don't know whether the following might be useful.? But a simple Bv unity-gain buffer could be replaced by a unity-gain E-element.? As far as I know, E-elements do not have the restrictions that a Bv source might have.? Therefore, if you were concerned about those Bv-source restrictions, you could replace it by an E-element.? While it is true that LTspice sometimes elevates an E-element back to a B-element, I think that it does not for a simple linear buffer.? So, you could use an E-element to check whether changes to your simulations were due to Bv-source's peculiar limitations, or something else.
Regards, Andy
|
Hello, Andy:
The mentioned ideas was not from me, and not related to what I said about 'confidential'. Just a brief, sudden thoughts, but I certainly remembered that someone who had written some thing about that (I surveyed that years ago) , though not very detailed about implementation. I guess its phase is not from detection, I would presume it's from pre-calculation, since many parameters could be obtained, Irms, Vrms, jwC, jwL, etc. Then generate whatever signal to the input of the compensation. The last thing is about the math & equations of math in the nature/circuits.
I know some of the rule of the nature, and some culture of some countries, but the world is huge, I don't understand everything, including rules everywhere and nowhere. But if possible, I would like keep going further, not for any matters, maybe for fun, I think.
Thank you very much. ---In LTspice@..., <ai.egrps@...> wrote : I still don't know what the Bv is for.
Since it represents an unreal component, are you thinking about using it to experiment with an altered circuit, before designing that "alteration" using real components?? Or do you want to add an extra signal into your circuit?? Otherwise, I can't imagine why a Bv would be a useful thing, here.
Regardless of the reason, you need to be more careful to consider the effect of adding the Bv to the existing circuit.? In both examples, it has a profound effect even though it is only a unity-gain buffer.? Remember that Bv unity-gain buffers have infinite input impedance and zero output impedance and are uni-directional, so breaking a connection and inserting a Bv source in its place, is a significant change.
I don't know whether the following might be useful.? But a simple Bv unity-gain buffer could be replaced by a unity-gain E-element.? As far as I know, E-elements do not have the restrictions that a Bv source might have.? Therefore, if you were concerned about those Bv-source restrictions, you could replace it by an E-element.? While it is true that LTspice sometimes elevates an E-element back to a B-element, I think that it does not for a simple linear buffer.? So, you could use an E-element to check whether changes to your simulations were due to Bv-source's peculiar limitations, or something else.
Regards, Andy
|
There are contradiction in real world ,too. If I just only study the textbook which contain lots math equations to describe the phenomemon of physics/electrics, etc. I tend to be confused, but if I just do the works ,practices of physical effort, I couldn't understand more deep theoretical mechanism behind, and unfortunately I, for an only one instance existing, I couldn't be two or more instances existing instantaneously to learn theory in the meanwhile do the practical things, and then connect them, so the fundamental materials is important, which is somewhat similar to , sometimes, we need to do FFT to transfer the time domain things to frequency domain, then revert the frequency domain things to time domain, make things simpler/clear to see, though the process is more complicated.
By the way, I really had suffered severe math equations confusion very badly, before . But, I still believe the learning curve of based on the basis/fundamental things ,finally would be likely the exp effect. Though, I don't think it's very impressive (I don't feel that / think so) , about the time step, but if you doubt or misunderstand.
Best regards. ---In LTspice@..., <ericsson.sunshine@...> wrote : Hello, Andy:
The mentioned ideas was not from me, and not related to what I said about 'confidential'. Just a brief, sudden thoughts, but I certainly remembered that someone who had written some thing about that (I surveyed that years ago) , though not very detailed about implementation. I guess its phase is not from detection, I would presume it's from pre-calculation, since many parameters could be obtained, Irms, Vrms, jwC, jwL, etc. Then generate whatever signal to the input of the compensation. The last thing is about the math & equations of math in the nature/circuits.
I know some of the rule of the nature, and some culture of some countries, but the world is huge, I don't understand everything, including rules everywhere and nowhere. But if possible, I would like keep going further, not for any matters, maybe for fun, I think.
Thank you very much. ---In LTspice@..., <ai.egrps@...> wrote : I still don't know what the Bv is for.
Since it represents an unreal component, are you thinking about using it to experiment with an altered circuit, before designing that "alteration" using real components?? Or do you want to add an extra signal into your circuit?? Otherwise, I can't imagine why a Bv would be a useful thing, here.
Regardless of the reason, you need to be more careful to consider the effect of adding the Bv to the existing circuit.? In both examples, it has a profound effect even though it is only a unity-gain buffer.? Remember that Bv unity-gain buffers have infinite input impedance and zero output impedance and are uni-directional, so breaking a connection and inserting a Bv source in its place, is a significant change.
I don't know whether the following might be useful.? But a simple Bv unity-gain buffer could be replaced by a unity-gain E-element.? As far as I know, E-elements do not have the restrictions that a Bv source might have.? Therefore, if you were concerned about those Bv-source restrictions, you could replace it by an E-element.? While it is true that LTspice sometimes elevates an E-element back to a B-element, I think that it does not for a simple linear buffer.? So, you could use an E-element to check whether changes to your simulations were due to Bv-source's peculiar limitations, or something else.
Regards, Andy
|
