UCD3138: UCD3138 CLA coefficient calculation

Jack Tan80

Part Number: UCD3138

Hi UCD expert,

Currently I'm trying to tuning loop parameter through GUI, but found some of the parameter calculated in GUI can't match the value calculated in mathcad. Can you please do me a favour on figuring out what the root cause is?

1. I refer to the "UCD3138 Control Theory" documentation, I take the equations in Table 18(please see the right table in below picture), and write the same equations in mathcad to calculate KP, KI, KD, alpha, and found that the value calculated in mathcad are:

KP = 15.954 ,

KI = 1690,

Kd = error( this is due to unit error in mathcad),

alpha = -255.746.

This is not match the result from GUI, for the same condition( fs = 250Khz, K = 1, fz1 = 1khz, fz2 = 1khz, fp = 10Khz), the GUI calculation results are:

// Coefficient set A(0)

Filter0Regs.FILTERKPCOEF0.bit.KP_COEF_0 = 7969;

Filter0Regs.FILTERKICOEF0.bit.KI_COEF_0 = 53;

Filter0Regs.FILTERKDCOEF0.bit.KD_COEF_0 = 30181;  

Filter0Regs.FILTERKDALPHA.bit.KD_ALPHA_0 = 199;


- Mathcad Calculation:

- The calculation result in GUI, please see below picture, and I upload the gui project here:CLA_PARAM_CALCUALTION_DC-DC HSFB @ Address 88d Project.zip

over 4 years ago

0 Jack Tan80 over 4 years ago

Mastermind 7100 points

In addition, I have some other question of the parameters in equations:

1. For the parameter K(Real Zearos: K, Fz1, Fz2)in GUI, what's the unit of K? I think the unit is db, isn't it?

2. The table 18, equations, what the unit of TDPWM? I thought it's seconds in mathcad.

3. NOS seems meaning number of oversampling in front end, isn't it?

0 Bosheng Sun over 4 years ago in reply to Jack Tan80

TI__Expert 6960 points

Hi Jack,

I am working on this, please allow me some time to get back to you.

Regards,

Bosheng

0 Jack Tan80 over 4 years ago in reply to Bosheng Sun

Mastermind 7100 points

Hi Bosheng,

Thanks for supporting here, looking forward to your answer. Thanks again.

0 Bosheng Sun over 4 years ago in reply to Jack Tan80

TI__Expert 6960 points

Jack,

1, K is the DC gain, no unit

2, TDPWM = 250e-12 (250ps)

3, NOS is oversampling rate, if you only sample once in each switching cycle, NOS =1

regards,

Bosheng

0 Jack Tan80 over 4 years ago in reply to Bosheng Sun

Mastermind 7100 points

Hi Bosheng,

After changing the TDPWM to 250*e-12, the KP, KI, KD, and alpha calculated in Mathcad still can't match the GUI calculation. Please see below result in mathcad.

0 Bosheng Sun over 4 years ago in reply to Jack Tan80

TI__Expert 6960 points

jack,

here is what we used in our GUI to translate real zeros to PID:

0 Jack Tan80 over 4 years ago in reply to Bosheng Sun

Mastermind 7100 points

Hi Bosheng,

Can you please help to attach the screen shot again which shows the equations? I can't see anything here. Thanks...

0 Bosheng Sun over 4 years ago in reply to Jack Tan80

TI__Expert 6960 points

Jack, if you let me know you email address then I can send it to you through email

0 Jack Tan80 over 4 years ago in reply to Bosheng Sun

Mastermind 7100 points

Hi Bosheng,

Here's my email address: tanjack1988@gmail.com. Thanks...

0 Jack Tan80 over 4 years ago in reply to Jack Tan80

Mastermind 7100 points

Hi Bosheng,

After fill the new equation in mathcad, I see that KP, KI, and KD in mathcad requires to multiply 1000 to match the result in GUI, alpha is the same with mathcad and GUI. Can you please help me know the differences?

0 Jack Tan80 over 4 years ago in reply to Jack Tan80

Mastermind 7100 points

Hi Bosheng,

I found that the transfer function of CLA(s domain) in GUI should be shown as below:

But there is no 1000 parameter in GUI model.

And I have used the above transfer func in matlab, and the bode plot result in matlab could meet the magnitude curve in GUI. Please see below bode plot result in matlab and GUI. Can you please help double confirm this? If my assumption is correct, where the 1000 parameter comes from?

0 Bosheng Sun over 4 years ago in reply to Jack Tan80

TI__Expert 6960 points

Jack, I asked our GUI expert to help you on this question.

0 Jack Tan80 over 4 years ago in reply to Bosheng Sun

Mastermind 7100 points

Hi Amiel,

I'm not sure if you've noticed about this, may I have your comments on this? What's the excat transfer function of CLA model in GUI?

Please see my post above, I thought the s domain transfer func of CLA shoule be as shown below, this is because I have compared the bode plots in matlab and GUI, the magnitude can be matched. If my assumption is correct, where does 1000 param comes from? Thanks...

0 Bosheng Sun over 4 years ago in reply to Jack Tan80

TI__Expert 6960 points

Jack, the 1000 is because the unit of fp, fz1, fz2 are in Hz, not KHz.

For PRD:

Fs in Hz and TDPWM is 250e-12

public double PRD

{

get { return 1 / Fs / 16 / TDPWM - 1; }

}

0 Jack Tan80 over 4 years ago in reply to Bosheng Sun

Mastermind 7100 points

Hi Bosheng,

I don't think this related with the unit of Khz and Hz, I've changed the unit to Hz in mathcad, and found that KP, KI, KD in mathcad still needs multiply 1000 compared with GUI result. Actully the calculation in mathcad will change the unit before calculations. Please see the calculations in mathcad as below picture.

And in matlab s domain bode plot,all frequency units are in Hz. Please see below s domain transfer function. If there is 1000 param, the bode plots can be matched with GUI generated.

0 Bosheng Sun over 4 years ago in reply to Jack Tan80

TI__Expert 6960 points

Jack, Amiel will help you on this.

0 Jack Tan80 over 4 years ago in reply to Bosheng Sun

Mastermind 7100 points

Hi Amiel,

May I have your comments here? Thanks...

0 Jack Tan80 over 4 years ago in reply to Jack Tan80

Mastermind 7100 points

Hi Amiel and Bosheng,

Thanks for your effort on suporting here.

From the docs that you shared with me, the z domain transfer function has a 1000 parameter. From my point of view, this 1000 should be the gain of front end. This is because the unit of EADC's output is mV, and input is V, so the gain is 1000, which is fixed value. But it shouldn't affact the value of Kp, Ki, Kd due to it's a independent part of filter, I'm not sure if my understanding is correct.

I have using the conversion of RZ to CZ, and CZ to PID, the calculate result is the same with previous, the KP, KI, KD have to multiply 1000 to match GUI calculate result.

When I have free time, I will try to do a matlab simulation on my side to see if I could fIgure out what's going on.

Thanks again...

Power management

Power management forum

UCD3138: UCD3138 CLA coefficient calculation