Feedback Loop: Calling a spade a spade

[Faringdon]

Thanks...

Do you agree, page 6 of the document “Practical Feedback Loop Design considerations for switch mode power supplies” is  somewhat general and non-commital in its attire?

Practical Feedback Loop....
https://www.pdffiller.com/jsfiller-desk10/?requestHash=c955daa6377db4d605d6ae3210244f1cf38b021219fe93e1df94e97d016d993b&projectId=1362206773&loader=tips&MEDIUM_PDFJS=true&PAGE_REARRANGE_V2_MVP=true&isPageRearrangeV2MVP=true&jsf-page-rearrange-v2=true&jsf-new-header=false&LATEST_PDFJS=false&routeId=2bc220699173ef97f6b3d8196f006eac&mode=force_choice#846a5be8c7734ec4a3736531a2e1f463


…..I am referring to equation (18) of page 6, ie, the value “Km = delta IDSpk/delta VEA”.
This “Km” value occurs in the Power stage transfer function of the BuckBoost in equation (21) of the same page.

However, the equation (18) is incomplete in its attire. So too is the diagram below it in figure (10).
The “Modulation Gain” that’s being referred to is often  actually called as  “R/Ri”

Where:
R = Load resistance
Ri = Current sense resistor in source of FET.

There is usually another gain associated with the modulator (actually an attenuation), which is the factor by which the “VEA” error amplifier output voltage is divided down before it is taken into the PWM comparator. Usually a factor of 0.25 or so. However, the page does not actually mention this.

The document TN-203...(at the bottom of page 9, for buckboost...)
“Voltage mode, current mode (and hysteretic control)”, as follows
https://www.microsemi.com/document-portal/doc_view/124786-voltage-mode-current-mode-and-hysteretic-control

..Gets the Modulation gain correct, they however, for whatever reason, call the Sense resistor, the “RMAP” …ie, the “PWM ramp voltage divided by the ramp current”…….this is Ohms Law, and is in fact the sense resistance , Ri.

However, they unfortunately don’t make any reference to the gain (attenuation) involving the dividing down of the error amplifier output voltage.

Do you know why none of these papers can “call a spade a spade”?

[mtwieg]

…..I am referring to equation (18) of page 6, ie, the value “Km = delta IDSpk/delta VEA”.
This “Km” value occurs in the Power stage transfer function of the BuckBoost in equation (21) of the same page.

However, the equation (18) is nonsense. So too is the diagram below it in figure (10).

It's not nonsense. Km is simply defined as the change in peak current per change in error amplifier output voltage.

The “Modulation Gain” that’s being referred to should actually be “R/Ri”

Where:
R = Load resistance
Ri = Current sense resistor in source of FET.

Substituting this for Km would give wrong results (unless you change something else in the equations).

There is usually another gain associated with the modulator (actually an attenuation), which is the factor by which the “VEA” error amplifier output voltage is divided down before it is taken into the PWM comparator. Usually a factor of 0.25 or so. However, the page does not actually mention this.

The document TN-203...(at the bottom of page 9, for buckboost...)
“Voltage mode, current mode (and hysteretic control)”, as follows
https://www.microsemi.com/document-portal/doc_view/124786-voltage-mode-current-mode-and-hysteretic-control

..Gets the Modulation gain correct, they however, for whatever reason, call the Sense resistor, the “RMAP” …ie, the “PWM ramp voltage divided by the ramp current”…….this is Ohms Law, and is in fact the sense resistance , Ri.

However, they unfortunately don’t make any reference to the gain (attenuation) involving the dividing down of the error amplifier output voltage.

Do you know why none of these papers can “call a spade a spade”?

What your describing is indeed how a lot of peak current mode control chips operate, but certainly not the only way. How exactly these chips operate internally is beyond the scope of this review paper. It's treating the PWM modulator as a black box with some effective transconductance, which it calls Km. Just because it doesn't express this transconductance in the exact terms you're accustomed to doesn't make it nonesense.

[jonpaul]

FTTS: Your posts are an infinite and unstable feedback loop.
  get some psyc counceling.  Will solve all your problems!

And save our time and bandwidth,

j

[Faringdon]

Just because it doesn't express this transconductance in the exact terms you're accustomed to doesn't make it nonesense.

Thanks.
Thanks, ayk, they have omitted the inevtable scaling factor that is used to scale the VEA, the error amplifier output voltage...this is an important feedback loop parameter, i believe you would agree.
As you well know, The VEA gets scaled before going to pwm comparator...then there is another "effective scaling factor" due to the fact the user may use a low value sense res, and at max load it may not peak at the threshold, but some way below it...this imprtant scaling, is omitted.....i am sure Choi knows it...but he doesnt say it.....the other paper fails to spell this out...but gives Ri, which accounts for it anyway

[mtwieg]

Just because it doesn't express this transconductance in the exact terms you're accustomed to doesn't make it nonesense.

Thanks, ayk, they have omitted the inevtable scaling factor that is used to scale the VEA, the error amplifier output voltage...this is an important feedback loop parameter, i believe you would agree.

No.

As you well know, The VEA gets scaled before going to pwm comparator...

And in some cases the current sense voltage is amplified instead (or a combination of amplifying one and attenuating the other). And usually there is some offset thrown in somewhere before the PWM comparator. Regardless of how the internal circuitry works, you ultimately get some roughly linear relationship between error amplifier voltage and peak current. For the purpose of this review paper, that's what matters. There's no reason for them to look inside the black box.

[Faringdon]

And in some cases the current sense voltage is amplified instead (or a combination of amplifying one and attenuating the other). And usually there is some offset thrown in somewhere before the PWM comparator. Regardless of how the internal circuitry works, you ultimately get some roughly linear relationship between error amplifier voltage and peak current.

Thanks, i wish you'd written the paper..that was the bit that was missing.

[f4eru]

FTTS: Your posts are an infinite and unstable feedback loop.

We need to move some zeroes around here.

[Faringdon]

you ultimately get some roughly linear relationship between error amplifier voltage and peak current.

Thanks, yes indeed, and its interesting, because as we say, in virtually all  current mode SMPS control chips, the error amplifier output voltage is divided down before entry into the PWM comparator....this scaling factor needs to be accounted for when making out the feedback loop Bode plots. Of course, this then  means that the voltage across the sense resistor is less...and this in turn then ends up increasing the gain in the feedback loop....since as Choi well states....the Modulator gain involves "Peak IDS/ Peak voltage into PWM comparator". (OK , he doesnt *exactly* state that, but this is to what he somewhat loosely refers).

So the scaling down of the error amplifier output voltage....gives rise to  an attenuation, and also, a gain, in the feedback loop.

[mtwieg]

Right, when estimating the value of Km (or whatever we call it), you would need to analyze the internals of the PWM controller chip or its datasheet. Choi chooses not to go into detail on how to do so in this review paper. IMO this is reasonable because it depends heavily on the specific control chip you're using.

Do you have a specific circuit/controller you want help analyzing?

[Faringdon]

Thanks, its just general. but most typically, i need to do it with Flybacks with NCP12XX type current mode controllers, which feature an opto in common emitter, pulling down a primary side feedback pin.

The process i follow, is,
1...First convert the flyback to its equivalent BuckBoost...refer stuff to secondary side, so can do the Choi "Post LC filter"  thing more readily.
2...Then , if in CCM at low mains, which it often is, then use euqation in Fig 8, page 9 of the following... TN-203 App Note..
https://www.microsemi.com/document-portal/doc_view/124786-voltage-mode-current-mode-and-hysteretic-control

..ayk, ,this gives the power stage transfer function. (Sometimes instead use the more fuller , attached Basso version, which includes the slope compensation)

3...Then the opto/TL431 error amp transfer function is gotten with equation on page 19 of TND381 app note...
https://www.onsemi.com/pub/Collateral/TND381-D.PDF

4....Then there is the final bit, involving the scaling of the error amplifier output voltage.....the equation in (2) above has the "R/RMAP" factor......which kind of gets to me a bit, so i use the equivalent Choi method of R(load) x (IDS peak/Peak PWM comparator voltage).

[Faringdon]

Hi, So anyway, On the subject of this post, it is still considered that feedback loop docs on the web are very poor.....

As such, Why is it that the Feedback loop books/docs do not properly explain how to do feedback loop
calculation for eg an offline flyback?
The first job is to convert the Flyback to its "equivalent buckboost".
Best to "refer everything to secondary" in this process. (will not explain why here)

So basically you find that buckboost which has the ......
a) same duty cycle
b) same vout
c) same iout (max)

The inductor of this "equivalent buckboost" will be equal (in inductance) to the secondary inductance of the flyback.

You simply find the "effective input voltage" of the buckboost which allows you these requirements.
Then you have your "equivalent buckboost". You can then shovel its values into the Power Stage transfer function
of the feedback loop.

So, Supposing its in CCM, You then subject it to the power stage transfer function on page 9, fig 8 of TN203...

https://www.microsemi.com/document-portal/doc_view/124786-voltage-mode-current-mode-and-hysteretic-control

....But in fig 8, we are told that "RMAP" is "PWM ramp voltage divided by the corresponding sensed current".
-But This just isnt a good explanation, its not just the amplitude of the "ramp" bit...its the actual peak current
value that goes into the calculation of RMAP.

Also, we have to remember that we converted our flyback to a (equivalent) buckboost, and in many cases the amplitide of the sensed current will now be much bigger than it
was when we had the original flyback. However (and this isnt explained anywhere on the web)...you must still calculate this bit as if you were still
working with the original flyback, and not the "buckboost equivalent". So in other words, the RMAP will be actually be the value of the
primary sense resistance
used in the flyback's FET source connection. But the docs dont tell you this....they leave students for dead by telling them to convert to the
"equivalent buckboost", and then leaving them lost from there on.

The docs also dont tell you that you must put in the attenuation factor that is exhibited by the attenuation that the output voltage of the error
amplifier gets exposed to in order to give the peak voltage into the PWM comparator. (usually about 0.3 or so) And of course, this is from the original flyback situation, since
this situation didnt really exist at all with the "equivalent buckboost" situation, because there never really was an "equivalent buckboost" on the
bench.

Obviously your optocoupled error amp calculation will take the form of page 19 of the TND381 App note...

https://www.onsemi.com/pub/Collateral/TND381-D.PDF

So we see students getting well lost with all this, and we have no texts or docs to refer them to...do you know of any?

[Andy Chee]

So we see students getting well lost with all this, and we have no texts or docs to refer them to...do you know of any?

I found Christophe Basso's books gripping reading:

https://www.amazon.com/Switch-Mode-Power-Supplies-Second-Simulations-ebook/dp/B00IZX3XZ8

https://www.amazon.com/Designing-Control-Linear-Switching-Supplies-ebook/dp/B00BEZNHI2

[jonpaul]

usual FTTS garbage. only FTSS has this issues.

Problem is psyc not engineering.



Farringdon Read  a book recommend "Control System Theory" DelToro, etc. 1971

j

[Faringdon]

Thanks, i cant find that book  to order online.
I have the two Basso books  that Andy Chee kindly recomends....i am of the impression that these two are the best out there on this kind of subject?
One of them has fallen to bits from such high useage.

[jonpaul]

 Principles of Control Systems Engineering
 Del Toro, Vincent
 Parker, Sydney R.
 New York et al.: McGraw-Hill Book Co., 1960

Any university engineering library should have a copy.

Happy reading !
 

Del Toro taught my CCNY course about 1971

Jon