TL431 is commonly used to create a voltage reference or small linear shunt regulator. It is often considered to be a “programmable zener”. Though, there is a big difference between TL431 and a zener diode. As many of other feedback systems, TL431 can oscillate if inappropriately decoupled.
I was measuring performance of linear voltage regulator based on one of TL431 clones. Very soon I realized that this regulator has big ringing, while responding to transient step of current. This kind of response usually corresponds to stability problems and small phase margin. I looked into datasheet of this part vendor and found no information, related to stability evaluation. Fortunately, this part is very common. So, I pulled out datasheet for original TI TL431 and realized that there are limitations on capacitors that should be used to decouple this chip. It is specified in the form of “stability boundary conditions” and depends on TL431 voltage, current and capacitance used.
While “stability boundary conditions” are specifying region of stable operation, it says nothing of what TI considers being “stable operation”. Is it just no oscillation? What is phase margin and system bandwidth within these conditions? Without this information you can’t really estimate transient response of your system.
I decided to make SPICE simulations of TL431 to evaluate stability. What I realized was-available models do not represent TL431 very well.
What is available?
Very similar model is included in Spectrum Software Micro-Cap 9.
You can’t really use it to simulate anything other than basic DC performance. I fact, you can substitute this model by just proper Voltage Source and get almost the same quality of simulation.
I’ll reference to this model as “OLD”.
Very similar model is mentioned in “I n t u s o f t N e w s l e t t e r “ #43 August 1995 Issue.
As it states in its header, it is for transient and AC simulations.
This model represents gain/phase of TL431 pretty accurate. This means that you can evaluate stability and transient response more or less accurate while using this model.
Output impedance of this model is not realistic. This means that it should not be used to evaluate input signal rejection ratio. Also, it does not simulate noise performance of TL431. Last two parameters are critical for some applications.
I will reference to this model as “TI New”.
The most detailed publically available model, allows to simulate DC, AC, transient, noise and impedance. It also claims realistic temperature behavior.
This model is based on schematic drawing from TI datasheet. It shows good match for DC performance. AC performance is relatively good but lacks some accuracy for maximum gain (by approximately 6 dB). Output impedance is also not very accurate. Noise performance is relatively realistic, but shows less flicker noise at low frequency’s than TI data. Temperature drift accuracy was not evaluated as soon as I was not interested in this kind of simulations.
I will reference to this model as “Helmut”.
While “Helmut’s” model is the only one that shows more or less acceptable results, I was not satisfied with its accuracy. This convinced me to create 4th and the most accurate model for TL431.
I will reference to this model as “Eugene”.
We will continue into “Eugene’s” TL431 model development, analysis and evaluation.