TL431 Questions

[sean0118]

Hi everyone 

I'm attempting to design a compensator for the feedback loop of a converter using the TL431. However there are a few aspects I'm having trouble with.

Firstly, I used the formulas in this document to find values for R1 and R2. However their example uses I_FB equal to 0.2mA, but I have seen other examples where it's in the order of 2mA. I know there is a minimum value of I_REF but I don't understand what the method is for an appropriate I_FB value?


The other values I have based on the example of the Type 1 compensator from ON Semiconductor. Which leads me to my other question of the correct calculation of C2. My knowledge of control theory is dubious at best but now is probably a good opportunity for me to improve that.

In their example they use C2 = CTR/(2*pi*G_fc*f_c*R_LED)

Now, I don't really know what the G_fc and f_c represent or the correct way to calculate them. Come to think of it, I don't really understand the purpose of C2 in general. 


Also, I hoping to build this circuit very soon in order to understand it better. Which means ordering components asap and I'm overwhelmed by the choices of optocouplers available. What sets the optocoupler used in this application apart? Or is there even a standard model that's used?


This post turned out longer than I intended, thanks if you read it through. 

edit: attached schematic, I new I was forgetting something important

[T3sl4co1l]

Use >1mA, or use a TLV431 (1.24V, 7V max, 0.1mA min.).  Note that's 1mA even when "off", so you should have that Rbias so it can draw bias without lighting the opto.

What's the converter?  Flyback?

Most of the time, type 1 compensation can give you a stable result, but, you are *guaranteed* overshoot.  Type 2 is better (R+C across the TL431).

You can estimate the value of R, C from the circuit around, but I always end up poking in values anyway.  Oh, and beware of simulators, if you want to "try before building": TL431 models are notoriously awful, and rarely capture the correct frequency response or impedances of a real device.  The E.D. version is probably the best:
http://www.audio-perfection.com/voltage-regulators/linear-regulators/accurate-tl431-model-development-analysis-and-evaluation.html
but always play around with values in the real thing to make sure it's right.

Tim

[sean0118]

Thanks Tim

Use >1mA, or use a TLV431 (1.24V, 7V max, 0.1mA min.).  Note that's 1mA even when "off", so you should have that Rbias so it can draw bias without lighting the opto.

Your referring to the cathode current right? I think my R_bias will allow that, but I'll double check once it's built. The current I'm not sure about I_FB is the one through R₁/R_upper

What's the converter?  Flyback?

LLC Resonant, and the IC is the FAN7631 

Most of the time, type 1 compensation can give you a stable result, but, you are *guaranteed* overshoot.  Type 2 is better (R+C across the TL431).

You can estimate the value of R, C from the circuit around, but I always end up poking in values anyway.  Oh, and beware of simulators, if you want to "try before building": TL431 models are notoriously awful, and rarely capture the correct frequency response or impedances of a real device.  The E.D. version is probably the best:
http://www.audio-perfection.com/voltage-regulators/linear-regulators/accurate-tl431-model-development-analysis-and-evaluation.html
but always play around with values in the real thing to make sure it's right.

Thanks I'll look into that. I'm using PSIM to simulate it, but I don't really trust it, but the TL431 setup I have now does seem to regulate fairly well, but with some overshoot with load changes.

[T3sl4co1l]

Your referring to the cathode current right? I think my R_bias will allow that, but I'll double check once it's built. The current I'm not sure about I_FB is the one through R₁/R_upper

The divider, that is?  Pick resistors such that the Thevenin equivalent resistance at the "REF" pin gives less than rated error (<2%?) at maximum rated input bias current (~100nA or something like that?).  Usually 1-10k is suitable.

Remember, TL431 isn't a zener, it's an opamp with +in = 2.50V (or whatever variant), -in = REF, and an open collector output.  With spare bias current pulling against the open collector output, the output voltage range is a reasonable approximation of a normal op-amp, and you can hang feedback bits between input and output to set frequency response and all that.  Real handy.

LLC Resonant, and the IC is the FAN7631 

Hmms, don't know offhand what the transfer function is going to look like.  Probably a little slower than flyback, but not massively so, like a forward converter.  So, it probably shouldn't be too bad?  You can compensate a lot with the R+C feedback, so yeah, when you get something breadboarded, give it a try, it should be able to do something.

If you're really worried, you can put an R+C in parallel with R1 in the feedback divider -- this R+C works against the divider and the feedback R+C, and lets you set up extra phase margin at a certain breakpoint.  Very handy for voltage mode forward converters, where the LC filter necessarily incurs 180 degree phase shift above the cutoff frequency -- which is probably still a frequency of interest as far as regulation and control!

Tim

[sean0118]

The divider, that is?  Pick resistors such that the Thevenin equivalent resistance at the "REF" pin gives less than rated error (<2%?) at maximum rated input bias current (~100nA or something like that?).  Usually 1-10k is suitable.

Thanks I think I'm beginning to understand. The error your referring to is the change in Vref? So if the deviation of Vref over the rated temperature range is 25mV, rated error = 25/2495 = 1%?

So when choosing R2 I have to make sure the voltage over it (Vref) doesn't change more than 1% as Iref changes? Is that right? 

If you're really worried, you can put an R+C in parallel with R1 in the feedback divider -- this R+C works against the divider and the feedback R+C, and lets you set up extra phase margin at a certain breakpoint.  Very handy for voltage mode forward converters, where the LC filter necessarily incurs 180 degree phase shift above the cutoff frequency -- which is probably still a frequency of interest as far as regulation and control!

Yeah this is when my lack of control knowledge starts to show, I don't understand the definition of terms such as phase margin, phase boost, cutoff frequency and breakpoint...        You don't happen to know of a resource that explains it all in terms beginners can understand? I've had a look around but it all gets over my head fairly quickly. That would probably answer my questions about C1 & C2 as well. 

[sean0118]

Yeah this is when my lack of control knowledge starts to show, I don't understand the definition of terms such as phase margin, phase boost, cutoff frequency and breakpoint...   

I have started looking at this problem again and I am beginning to understand the function of C1 and C2 in a type 1 configuration.

As far as I can tell C1 and C2 are there to position a pole to cancel a zero? But in order to place it correctly I would need to plot the frequency response of my converter first? Is that correct?

[T3sl4co1l]

Yes.

Tim

[sean0118]

Sorry one more question, to plot the frequency response of the LLC would I run an AC sweep on the converters input or its switching frequency? 

[T3sl4co1l]

Impossible to do in standard SPICE.  You need a cycle-averaged-equivalent model to perform that type of analysis.

Tim

[sean0118]

Does that include PSIM/SmartCtrl? It looks like it might be possible?

Or am I missing something? Is it better to find the frequency response using equipment once the hardware is built?

[T3sl4co1l]

Hmm, sounds like it.  I'd want to read a lot more about it to be sure it's doing what it's supposed to (stock SPICE is ornery enough as it is).

Tim

[David Hess]

Yeah this is when my lack of control knowledge starts to show, I don't understand the definition of terms such as phase margin, phase boost, cutoff frequency and breakpoint...   

I have started looking at this problem again and I am beginning to understand the function of C1 and C2 in a type 1 configuration.

As far as I can tell C1 and C2 are there to position a pole to cancel a zero? But in order to place it correctly I would need to plot the frequency response of my converter first? Is that correct?

It is usually possible to calculate this ahead of time but sometimes the necessary data is not available or component variation makes it impractical.  It is often easier to make a rough estimate and then either use a network analyser or measure the transient response while adjusting the compensation network.  Then knowing the proper compensation values, it is possible to work back to find the uncompensated response and check that against the circuit for sanity.