LTSpice

[mikebg]

HOW TO DESIGN P-I-D CONTROLLER

HOW TO DESIGN P-I-D CONTROLLER

There is in the WEB an excellent textbook on P-I-D control:
http://www.controlguru.com/pages/table.html
I use for design a paragraph in page 1152 of German textbook:
Ulrich Tietze, Christoph Schenk "Halbleiter-Schaltungstechnik: [neuer Teil: Nachrichtentechnische Schaltungen]"
http://books.google.bg/books?id=h6A0...page&q&f=false
The paragraph is shown below.

There are English editions:
Ulrich Tietze, Ch. Schenk, Eberhard Schmid "Advanced electronic circuits"
Ulrich Tietze, Ch. Schenk, Christoph Schenk, Eberhard Schmid "Electronic circuits: design and applications" Springer-Verlag, 1991
Ulrich Tietze, Christoph Schenk, Ch. Schenk, Eberhard Gamm "Electronic circuits: handbook for design and application" Springer, 2008 (see page 1112 for PID-controller).

The algorithm for practical design of PID-controller is:
1. Disable derivative action (minimize D-gain).
2. Disable integrative action (minimize I-gain)
3 Decrease slowly the proportional action (P-gain) until ringing becomes decaying sinusoid.
4. Measure period of ringing Ts.
5. Calculatre integrating timeconstant Ti ten times more than period of ringing, ie Ti=10Ts.
6. Calculate derivating timeconstant Td equal to period of ringing, ie Td=Ts.
7. Enable integrative action and decrease proportional gain until overshoot disappears.
8. Enable derivative action, the system will operate with minimal settling time.
9. Check for noise. To suppress noise, increase HF filtering for derivative signal.

I will illustrate how to make this with SPICE for above circuit diagram.

[Davor]

Sure, but guess what happens when you come across a conductive patch and your coil's Q changes - your overblown PID regulator will show it's ugly face once more. Just leave the integrating component away and everything else is yofi tofi.

[mikebg]

Russian colleagues are angry that in the Russian translation of the book the mentioned paragraph is missing, and it is difficult to translate it from German. For palliative, I will say that this is the old edition. There's a new Russian edition of the textbook:

Ульрих Титце, Кристоф Шенк "Полупроводниковая схемотехника" (комплект из 2 книг) Лет Ме Принт, 2012

I will translate the paragraph, but my translation will probably be different from that in the new edition.

Look what's at stake. Increase the P-gain until oscillation occurs. Measure the vibration period Ts. From now follow the above described algorithm.

[mikebg]

GE-> Aus dem Schwingversuch - also dem Auftreten einer schwach gedämpften Schwingung - Iassen sich alle für die
Dimensionierug des PID-Reglers erforderlichen Daten auch berechnen: Die Schwingfrequenz ist die kritische
Frequenz: f_S = 1/T_S = f_k . Die Schwingverstärkung ergibt die P-Verstärkung A_RS = A_P . Darin ist A_RS die
P-Verstärkung des P-Reglers an der Stabilitätsgrenze. Die Differentiationsgrenzfrequenz wählt man gIeich
der Schwingfrequenz f_D = f_S und die Integrationsgrenzfrequenz gleich einem Zehntel der Schwingfrequenz f_I = f_S/10. Damit ergibt sich zusammenfassend die folgende Dimensionierungsanleilung für einen PID-Regler:
A_P~A_S T_D=T_S T_I = 10T_S

EN-> The vibration test, as well the appearance of a low damped vibration, enables to calculate all data of PID-controller necessary for the design. The vibration frequency is the critical frequency: f_S = 1/T_S = f_k .
The gain at vibration is the P-gain A_RS = A_P , where A_RS is the proportional gain of the P-controller at the stability limit.
The derivation cutoff frequency is selected to be equal to the vibration frequency f_D = f_S and the integration cutoff
frequency - equal to one tenth the vibration frequency f_I = f_S/10. This results in summary, the following directions for
design of a PID-controller:
A_P~A_S T_D=T_S T_I = 10T_S

RU-> Вибрационные испытания, а также появление медленно затухающего колебания, позволяет рассчитать
все данные ПИД-регулятором, необходимые для проектирования. Частота колебания является критической
частоты: f_S = 1/T_S = f_k .
Усиление при колебанию является П-усиление A_RS = A_P , где A_RS - пропорциональное усиление П-регулятора
на пределе устойчивости. Частота среза дифференциатора выбирается равна частоте колебаний f_D = f_S и
частота среза интегратора - равна одной десятой частоты колебаний f_I = f_S/10.. Это приводит, в целом,
следующие направления для проектирования ПИД-регулятора:
A_P~A_S T_D=T_S T_I = 10T_S

[Davor]

the integration cutoff frequency - equal to one tenth the vibration frequency f_I = f_S/10.

Precisely. In other words, you can't use a commensurate time constant to the one a system's integrator already has because it leads to oscillations, but instead you'll use a 10 times slower integrator to impose some sort of regulation. In a process your loop becomes slower than the self-regulated one. Why not just switch the integrator off completely and obtain a more damped response envelope instead? Provided you are into dampening overshoots or flutter of the oscillator.

I'm really not into bursting your bubble. In fact my high school "title" was an automatician so I know precisely what you are talking about, and I admire your energy. What you most probably figured out by now is that an oscillator handles it's envelope in the exact manner the rail limited integrator handles it's output, so instead of fighting it you better embrace this handy feature as a part of a PID regulator, not something to overcome.

[mikebg]

Davor, Please read this chapter in the mentioned above textbook:

http://books.google.bg/books?id=h6A0...page&q&f=false

The chapter is very helpful for experts with high school "title" automatician. There is several English editions of this book.

To Russian colleagues. See circuit diagram in post #18.
Resistor R8 involves derivative action. The action is disabled by using 1 ohm resistance.
Resistor R5 makes HF filtering of derivative action. The action is disabled by using 1 ohm resistance.
Resistor R7 is used to reduce P-gain. To obtain ringing, the action is disabled by using high resistance.
Voltage divider R3-R7 can increase integrating timeconstant.
Note that unlike original circuit, the output has symmetry and distortion is reduced.

[Derx]

Need to help make model CD4016 CD4066 CD4053 for LTSpice?
Thanks.

[Davor]

These models are available at LTspice yahoo group. I think you won't find 4016, and the models of 4053 and the other more complex switches are built from SPST switches. I use these models regularly and they are fine.

Your simulations would finish much sooner if you use a switch model that is supplied with LTspice as it does not compute the mosfets behaviour as 4066 model does. You have to define the parameters in a .model statement where you put the switch on resistance etc., but otherwise it is a faster method to accomplish things. I use this switch instead of comparators as they are notoriously slow to converge, but with switches all is done in a jiffy.

Here is the thing I'm playing with at the moment. I also configured the LTspice switch for this task so you can compare the results.

[Derx]

ok 4066 model
you helped me, and how to find 4053 ?

[Davor]

You won't. 4053 is basically the same kind of switch as 4066, except that two gates are used, and one is driven by an inverter. You can make a macro model with 4066 as a starting point. You'll find HC etc. variants of 4066 at the yahoo group, but not the pre-digested 4053

[porkluvr]

Derx,

CD4053 is a triple SPDT switch.

I pieced together a single SPDT switch (model) from individual MOS transistors like what you would find in a CD4007. Here is a single SPDT switch model/symbol combination that works in LTspice.

It works about like 1/3 CD4053, except that my model has no enable input.

I did not compare transistor specifications from the CD4053 datasheet against what is here because I am not concerned with such details. I just wanted something that works.

[mikebg]

LET WE DESIGN WIEN BRIDGE OSCILLATOR

Why SPICE can not make the above circuit diagram of Wien bridge oscillator to work better?

http://www.geotech1.com/forums/showt...464#post168464

The reason is not the too bad circuit diagram, but the very bad block diagram. The correct design method requires to start with an analysis of the most suitable block diagram. The SPICE can not show what should be the appropriate block diagram. See how the revision was carried out and the redesign was made for Relic Hawk TX circuit.

http://www.geotech1.com/forums/showt...756#post155756

After the appropriate block diagram is clear, make the implementation of components of each block and obtain the complete circuit diagram by connecting the blocks. Only then apply SPICE to adjust certain values.

Below is shown the appropriate block diagram for Wien bridge oscillator. The task of the controller 7 is to maintain a constant amplitude at the output of amplifier 2. The opamp used for amplifier operates with little distortion when the output amplitude is significantly less than the supply voltage. In this case, supply rails are 5V each, so we can choose an automatically maintained amplitude 2V. The voltage generated by the amplitude detector 4 is compared by block 7 with reference voltage 6 and the difference is submitted for processing in controller 5.
The controller 5 must contain an amplifier with very high gain that is missing in the original circuit diagram. The additional amplifier is not a problem because the integrated circuit contains an unused amplifier. It remains to connect to it two resistors and two capacitors only.
Those who like to construct, let implement this block diagram. The SPICE will show that your design works better than all other Wien bridge oscillators published on the INTERNET because it is designed in the right way.

[Derx]

thanx to all
i need ltspice model to IR4428??
help

sheet is http://www.irf.com/product-info/data...ata/ir4426.pdf