• State Resolved
  • Locked Locked
  • Replies 7 replies
  • Answers 1 answer
  • Subscribers 63 subscribers
  • Views 934 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

UCC2813 - Cannot Follow the Application Note in the Datasheet

Emre Bahtiyar
Prodigy 155 points
Other Parts Discussed in Thread: TL431

I am trying to follow the application note at the end of the datasheet and then I will design an isolated flyback converter with my own parameters. I am using Mathcad 15 to make the calculations. These are some parts I couldn't understand:

1. After calculating Lpcrit (page 26), do we use it on any place else?

2. Do Lm (page 24) and Lp (page 27) are the same thing (transformer primary side magnetizing inductance)? I calculated that as 1715uH. Should we take it as 1500uH in the following equations (page 24)?

3. 120mV ripple of 12V is %1 ripple isn't it (page 23)? Or is it %0.1 (page 25)? When I calculate Cout for %0.1, I found the result as 1865uF not 2015uF, why?

4. What is the value of Pin? If Pout=12V*4A=48W. Or is it 12V*4,33A=52W. (4,33A is out current max, page 23). Now Pin=48/0.85=56.47W or Pin=52/0.85=61.13W? This affects the mosfet rms current...

5. In equation 30, what is the 100kHz expression? Is it our switching frequency which should be 110kHz? This affects the Rcsf calculation...

6. I calculate Go (DC open loop gain) as 5.59db, datasheet calculates 14.95db. We will use it on Ho(s) later. My bode plots are not the same with the ones in the datasheet. Maybe it is related with the Go calculation difference maybe not. Can someone explain this part?

7. I see in the schematic Rb1 and Rfb2 (page 22) but in page 29 there is a Rfb mention. Is it Rfb2? If so how do we calculate Rfb1 which is 4.99k in the schematic?

8. Can someone show me how to calculate Rled from compensation loop transfer function as explained in text (loop gain must be 1 at crossover freq.)?

9. For final bode plots (page 30) do I multiply G(s) with Ho(s)?

2538.UCC2813-0.rar

  • 0
    TI__Expert 4015 points

    Hi Emre, see responses below, I've had to forward on one or two of your questions to the author, who should respond to them after the thanksgiving holiday in the US probably.

    Hope this helps.

    Thanks

    Billy 

    Emre Bahtiyar said:

    I am trying to follow the application note at the end of the datasheet and then I will design an isolated flyback converter with my own parameters. I am using Mathcad 15 to make the calculations. These are some parts I couldn't understand:

    1. After calculating Lpcrit (page 26), do we use it on any place else?

    Lpcrit is used to determine whether the design will need to cope with CCM operation or not as this changes the control loop design. Particularly DCM operation does not require slope compensation of the peak current.

    2. Do Lm (page 24) and Lp (page 27) are the same thing (transformer primary side magnetizing inductance)? I calculated that as 1715uH. Should we take it as 1500uH in the following equations (page 24)?

    Yes, Lm is the same as Lp. The datasheet mentions that a value of 1.5mH is used, presumably as this is a more standard value. 

    3. 120mV ripple of 12V is %1 ripple isn't it (page 23)? Or is it %0.1 (page 25)? When I calculate Cout for %0.1, I found the result as 1865uF not 2015uF, why?

    120mV is 1% ripple yes. I think there may be a finger fumble here somewhere and am checking this.

    4. What is the value of Pin? If Pout=12V*4A=48W. Or is it 12V*4,33A=52W. (4,33A is out current max, page 23). Now Pin=48/0.85=56.47W or Pin=52/0.85=61.13W? This affects the mosfet rms current...

    I would calculate Pin as Pin=52/0.85=61.13W. In this way, the unit is design to be able to supply ~8% more than the required output current. If you design it to supply exactly Iout nominal, there is no margin and over a spread of designs some may nore be able to support full load.

    5. In equation 30, what is the 100kHz expression? Is it our switching frequency which should be 110kHz? This affects the Rcsf calculation...

    Not sure on this one

    6. I calculate Go (DC open loop gain) as 5.59db, datasheet calculates 14.95db. We will use it on Ho(s) later. My bode plots are not the same with the ones in the datasheet. Maybe it is related with the Go calculation difference maybe not. Can someone explain this part?

    5.59 is the absolute value of Go. 14.95dB is the value of Go in dB.  14.95db = 20log(10) 5.59 

    7. I see in the schematic Rb1 and Rfb2 (page 22) but in page 29 there is a Rfb mention. Is it Rfb2? If so how do we calculate Rfb1 which is 4.99k in the schematic?

    Again, I will have to check on this. 

    8. Can someone show me how to calculate Rled from compensation loop transfer function as explained in text (loop gain must be 1 at crossover freq.)?

    I believe Rled is simply calculated by plugging in all the other values (which are known) into equation 38 and setting G(s) = 1 where s is the crossover frequency

    9. For final bode plots (page 30) do I multiply G(s) with Ho(s)?

    The closed loop response is give in this case by A = H/(1+GH) 

    There are many publications and articles which deal with this such as 

    www.electronics-tutorials.ws/.../feedback-systems.html

    www.electronics-tutorials.ws/.../closed-loop-system.html

    ecee.colorado.edu/.../Ch9slide.pdf

    (Please visit the site to view this file)

  • 0 in reply to Billy Long
    Prodigy 155 points
    Thank you for your fast response, Billy. We will wait for the author for some of my questions, then...
    For my question 6, I missed the dB part, my mistake. But I will not use it in dB form right? I should use Go as 5.59 in the equation 32 but my open loop bode plots are still not the same with the datasheet...
    As for your response to question 8, I did what you proposed and end up with Rled=27.3k which should be 1.62k according to datasheet....
    For my question 9, I plot the Ho(s)/(1+G(s)*Ho(s)) but result is not similar with datasheet as expected since I am mot sure about neither Ho(s) nor G(s) :(
  • 0 in reply to Emre Bahtiyar
    TI__Mastermind 26770 points

    Hello Emre,

    Thank you for uncovering several typographical errors in the application section of our UCC2813-x datasheet. We will begin the process to correct these errors, but this may take many weeks due to internal procedures.  Meanwhile, I’ll address those items that Billy was unsure of.

    For #3, all of the calculations are based on 0.1% output ripple voltage, so we will be changing the specification table to 12 mVpp instead of 120 mVpp.  Your capacitance of 1865uF results from using the nominal 4 A for Iout, whereas 2015uF results from using the maximum output current of 4.33 A.

    For #5, equation 30 should use 110 kHz, not 100 kHz.  Since this does affect Rcsf, we will introduce a margin factor of 1.1 in equation 29 to compensate the increase of Src in equation 31.  This way Rcsf remains the same value.  Increasing the calculated slope Se (equation 29) by 1.1 is a reasonable technique to allow for circuit and component value variations, and probably should have been done anyway.  A higher slope for Se is improves the margin for stability.

    For #6, Go = 5.59, and the Bode gain and phase plots in Figures 34 and 35 are of the open-loop power-stage transfer function only. I’m not sure why your plots differ from those of the datasheet.  Be sure that the 20log(10) function is being applied to the entire Ho(S) equation, not just to the Go term up front.  

    For #7, the mention of Rfb in the text above equation 37 should be Rfb2. The author of the application section had no explicit equation to determine Rfb1, but had already arbitrarily selected 10K for Rfb2 and then selected Rfb1 to limit the value of the FB voltage when the opto-coupler is saturated on and COMP is 0 V.  The value of 4.99K for Rfb1 was chosen to meet that requirement.  

    For #8, one must set the loop-gain = 1 to find Rled at the cross-over frequency, but the loop gain = G(S) x Ho(S), not just G(S). Using G(S) alone gives you the 27.3K, but using G(S) x Ho(S) results in a value close to 1.62K, which was chosen for Rled.

    For #9, the final Bode plots on page 30 are indeed based on G(S) x Ho(S). The captions for Figures 36 and 37 are miss-labeled as “Closed-Loop” gain and phase.  They are, in fact, the open-loop gain and phase of GH, not of H/(1+GH).  Plots of GH are usually used to ascertain gain = 1 (0 dB) cross-over frequency and phase margin.   We will make corrections to the text and figure captions.

    I hope these answer your questions.  If so, please check the Verify Answer check-box in the posting window.

    Regards,
     Ulrich

  • 0 in reply to Ulrich Goerke
    Prodigy 155 points
    Hello Ulrich,

    Thank you for your answers. It explained a lot but I still have some problems. Sorry about that.

    *Output capacitance equation is ok now. But now input capacitance equation should be 137uF (not 127uF) and magnetizing inductance should be about 1585uH (not 1700uH) which is good since 1500uH is selected. I mean there are two different P_in values are used in the equations, I think. One is based on 4A I_out, the other is 4.33A....

    *My peak mosfet current is 1.464A which is close enough but my mosfet RMS current is 1.04A (not 0.75A)

    *I have added that margin factor in order to have a R_csf value of 5.95k ohm.

    *Problem of my open loop transfer function is w (omega) and f expressions. So,
    f_rhp_z= 7.65kHz but w_rhp_z is 48Khz which one will I use in TF?
    f_lhp_esr_z=6kHz but w_lhp_esr_z is 37.7kHz which one will I use in TF?
    f_p1= 43.4Hz is it correct? will I multiply this with 2*Pi in TF?
    f_p2=55kHz is it correct? will I multiply this with 2*Pi in TF?
    and is Q_p=1?

    *So we are limiting FB pin voltage to 2.5V with rfb1 as 4.99k ohm when Comp pin is 0V and opto is saturated (5V at the opto's emitter).

    *I still couldn't calculate Rled. with Ho(s)*G(s) = 1, what is the crossover frequency? Is it 2kHz? What about the Rze, Rac, Rj resistances? Rac is totally unknown, Rze and Rj are 1kohm in the schematics. Can you explain how to determine these values? Another thing is for example when I want a 65VDC output what should I do for safe operation of TL431 (another resistor divider)?
  • 0 in reply to Emre Bahtiyar
    TI__Mastermind 26770 points

    Hello Emre,

    I'll answer your new questions in the order of your bullets.
    1. Yes, unfortunately, two different values of Pin are used, as you indicate. 4.33/4 * 127uF = 137.5uF, for example. When we correct the datasheet, we'll make it plain which Pin is used for which calculation and why.
    2. I calculated 1.01A, too, so 0.75A is an error. I don't know how it came about, but we'll correct it.
    3. Okay.
    4. Transfer functions are typically written as functions of "s" (s = jw = j2pif), but the Bode plots are based an sweeping "f".
    I think that as long as the "s" in the TF is expanded to j2pif and the "wx" is expanded to 2pifx, (where x is a specific subscript), then frequency (f) can be used for the poles and zeros, instead of radians (w).
    Qp is set equal to 1 in order to first determine Mc, and from that to find the value for Se needed ultimately to calculate Rcsf.
    5. The voltage on FB is limited to 3.33V when COMP = 0V and Emitter = 5V. This is well within the maximum rating of FB.
    6. The application procedure targets 1.9kHz as the cross-over frequency at equation (33) based on the RHPZ. This frequency is used when setting Ho(s)*G(s) = 1 to find the value for Rled. The value of Ho(1.9kHz) can be taken from the Gain curve in Figure 34, which appears to be about -18dB, or 0.1259. There is no convenient gain plot for G(s) since we are still looking for Rled, a factor in G(s). But if GH(1.9kHz) = 1, then G = 1/0.1259 = 7.943 at 1.9kHz. Now all the values and poles and zeros must be entered into equation (38) to solve for Rled. Note: the poles and zeros terms are calculated as magnitudes. Plugging in, I calculated Rled = 1311ohm. I'm not sure why 1.62K was chosen, but I guess it is close enough.
    Rac is a place holder, normally set = 0, but can be changed to 50~100ohm to have a location to attach a network analyzer to measure the actual system's frequency response.
    Rj is chosen to obtain the proper bias current for the TL431 when the opto should be cut off.
    Rze is chosen to provide sufficient bias current for the TL431 + opto network + bias current for the 10-V zener.
    The zener shunt supply is added to regulate the 48-V output down to a level below the TL431 maximum rating. I think the 10-V level is an arbitrary voltage. It could have been 5V or 12V or 24V... The values of Rj and Rze would change based on the chosen voltage to get the proper currents.
    Another option for voltages higher than the TL431 rating, such as 48V or 65V, is to insert a dropping zener in series with Rze to drop most of the output voltage. This dropping zener must dissipate the total bias current. A third option involves adding a low-voltage bias output winding to the transformer, when the regulated Vout is so high that the bias power of the previous methods becomes unacceptable.

    I hope this answers all of your questions and concerns.
    Regards,
    Ulrich

  • 0 in reply to Ulrich Goerke
    Prodigy 155 points
    Thank you for your answers. With your explanations in the 4th and 6th, now I am able to plot close enough bode plots of H and GH. I still calculate Rled differently, though. with cross-over frequency=1913Hz, Cfb= 2.65nF, Rz = 83.77 kohm and with all other standard values, I calculate Rled as 2.32k ohm. With Rled as 2.32k ohm, my final bode plots tell me 69.6° phase margin and 1.9kHz cross-over frequency. So, I believe my calculation is not all wrong.

    I actually have one more question if you still open for questions :)

    I would like a second output. How can I insert it into all these equations? Especially, into inductance and mosfet current calculations. Besides It should effect rhp zero calculations. Can you or someone help me to transform the equations? I think people rarely need only one output from a power supply design.
  • 0 in reply to Emre Bahtiyar
    TI__Mastermind 26770 points

    Hello Emre,

    That is a good question. The transfer function and stability analysis is based on small signal perturbations during steady-state operation, so multiple outputs can be combined into a single output to perform the calculations.   The same with the inductance and MOSFET current calculations (although these can also be done in a different way, too).

    Impedances are transformed from one winding to another by multiplying by the turns-ratio squared. Choose the main winding (with N1 turns) as the one being directly regulated.  A second output has a second winding with some number of turns (N2), a second output capacitance (C2), and a second external load resistance (R2) (in parallel with the output capacitance).  R2 is referred to the main winding as a resistance in parallel with the main load resistance (R1), with a transformed value of R2_1 = R2*(N1/N2)^2. The impedance of C2 is transformed the same way and put in parallel with C1, but since impedance Z is based on 1/C, the transformed capacitance C2_1 = C2*(N2/N1)^2.  So when referring the impedances of a lower voltage output to that of a higher voltage winding (more turns), resistance values increase and capacitance values decrease.  Vice versa, when referring the components on a higher-turn winding to a lower-turn winding.  This method holds for any number of windings on a single transformer.

    Once all of the other windings have their impedances referred to the main output winding, all the capacitances and resistances are combined in parallel. The parallel-R becomes the equivalent total load and the parallel-C becomes the equivalent output capacitance.  These two equivalent values are used in the overall design equations.

    Regards,
     Ulrich