Feedback problem using TL431 and opt-coupler (updated)

[hsklkk]

loop compensation tl431

To be able to shut down the PWM in idle operation in case of minimum 20 % CTR, R6 must be dimensioned above 2.5k, very important.

How to determine R6? why 2.5K
My analysis:
For Min, If min = 1.5V/4.3K=0.35mA, for min CTR20%, Ic=0.07mA
For Max, If max= (30-2.5-1.5)/4.3k=6mA, CTR about 1.1, Ic= 6.6mA
Is it right for min and max Ic?

 

[FvM]

tl431 failure

My calculation is different:
R6,min = 5V/(If,max*CTR,min)=4.16k
This is the condition for complete PWM shutdown. If,min is zero anyway, so there's no additional condition to observe.

 

[hsklkk]

tl431 feedback

here's my design, I got some problems.....
My topology is full bridge phase shift
Vin=48-96V, N1/N2=5:26, Vo=200V,
When I boost my input to about 36V, the feebback path starts work. But the circuit become odd. The Vo can keep 200V but this is a on off control, which means phase shift on off alternatively. When I boost the Vin further to about 40V, the feedback is no problem (phase shift is fixed).
Also, when R6=1k, it become on off, which is PWM on off and fails to boost Vin further.....
My guess is that the problem of ramp signal or R6......

One more, if I don't add C2, the control can't work......is this value proper?

 

[FvM]

feedback compensation tl431

"On/off" control basically means a too high loop gain and/or inappropriate loop compensation. Cause I don't know the power stage part of the loop gain, as a first step, I would perform an empirical adjustment of loop gain with C1 and R5. C2 seems dubious to me, you should rather increase C1 and possibly reduce R5.

 

[hsklkk]

pc817

I encounter PWM on/off and phase shift on/off.....
Now, I fixed R6, it is a phase shift on/off

Added after 1 minutes:

is it better if I add a C parallel to C1 and R5?

 

[FvM]

high voltage tl431 feedback

There are two possible ways: trying to understand the effect of individual parts on loop gain and vary the dimensioning specificly or just try to vary arbitrary part dimensioning.

Both methods can be successful. Although the first method is supposed to achieve a good result faster, the latter can show surprizing results, sometimes. You have to find your way.

 

[hsklkk]

tl431 supply -datasheet -shunt

I added C=0.05u parallel to to C1 and R5, the feedback is successful! Thx

 

[eem2am]

how do yuo know its successful in all cases....have you used gain/phase analyzer?

what is your phase margin...and gan margin...and Xover frequency.......and transient response to full load to no load step?

 

[hsklkk]

I mean I successful the initial test, not test full load and transient yet

 

[eem2am]

ok thankyou, but you have read Ray Ridley articles switchingpowermagazine.com

transient testing is no good on its own....you must pull out your gain/phase analyser.

otherwise you and your customers are on a wing and a prayer

 

[hsklkk]

any way or software to analyse the feedback control such as phase and gain, because I need it for my final year project report, thanks!

 

[FvM]

The overall system operation can be analysed by usual analog simulators, e.g. PSPICE. You can either include switching operation in the analysis, resulting in large run times for rather short operation intervalls or use hybrid models with time-continuous equivalent circuits for the switching stage in analysis of the control loop. There are complete text books covering the topic of SMPS simulation with PSPICE.

The free simulators provided by SMPS IC vendors, e.g. Linear or National can perform rather complex simulations, as an advantage, they have example simulations circuits with their chips, a good tutorial to start with.

 

[hsklkk]

yes, I use type 2 one, how to determine the(angluar frq) zero and pole for 100kHz operation?