Linear Regulators

A couple of test bench schematics were created in order to evaluate TL431 model performance and make a comparison between suggested model and existing ones. Most of test benches correspond to ones described in the datasheet to take chip measurement data and can be used to tweak model parameters to datasheet numbers. **TL431 datasheet** was used as a reference.

First Test Bench** “TL431_AC_TestBench.asc”** is used to evaluate frequency response of all models.

TL431 AC TestBench Schematic

AC Simulation Resuts

As you can see from result plots, both “Eugene” and “New TI” models shows close match between test results and datasheet. “Eugene’s” plot seems to be slightly closer to the datasheet numbers. “Helmut” plot is lacking 6 dB gain in low frequency range, while keeping functionally correct slope. And as expected, “OLD” model does not show AC performance – it has flat frequency response.

Next Test Bench **“TL431_Z_TestBench.asc”** is used to evaluate TL431 output impedance.

TL431 Output Impedance Test Bench

TL431 Output Impedance Simulation Results

You can see that “Eugene’s” plot shows very good match in low and middle frequency range and reasonably good match at high frequency. “Helmut” plot shows 2.5 times higher impedance at low frequency and reasonably good match at middle and high frequency. This is probably related to missing 6 dB of gain at low frequency in “Helmut” model. Both “TI New” and “OLD” models show unrealistic impedance plots.

The last Test Bench **“TL431_Noise_TestBench.asc”** is used to evaluate TL431 noise performance. There is no exact description in the datasheet, how this measurement was taken. I’m expecting that my testbench corresponds to test methodology used to evaluate real chip performance.

TL431_Noise_TestBench

Noise Performance of Eugene's TL431 model

Noise Performance of Helmut's TL431 model

Both “Eugene” and “Helmut” models showed good match in mid and high frequency range, while “Eugene” model has better matching in low frequency flicker noise area.

Two other models, “TI New” and “OLD”, can’t simulate noise performance of TL431.

### Comments

Apparently you were comparing the models tested to the TI data sheet (some other vendors, such as ON, show similar, but significantly different characteristic curves).

Here is a behavioral model without any transistors at all that you might find interesting to compare. I recently added noise (not normally needed for switching power supplies) and tweaked the parameters to better match the TI version of the part. I think it matches better than the other models, although all the newer ones are probably good enough for most purposes. — a.s.

.subckt TL431AS K A R ; by analogspiceman

D1 R K Dc

R1 A R 1e6

G1 A 5 R 4 1

C1 5 A 1n Rpar=700

D2 7 R Dc

D3 7 4 Dn

R2 7 4 900k

R3 6 5 700k

D4 A 6 Di

C2 K 6 40p Rser=6k Rpar=2e6

G2 K 7 6 A 90m

D5 A 7 2V5

D6 A K Dc

.model Dc d Ron=7 Vfwd=0.6 Vrev=36 epsilon=50m

.model Dn d Is=5p Kf=55f Cjo=0p3

.model Di d Ron=10m epsilon=1m

.model 2V5 d Ron=0.13 Vfwd=0.6 Vrev=2.5 epsilon=10m

+ Roff=7k revepsilon=10m

.ends TL431AS

Thanks, I’ll try it and post results

You absolutely right, there is a difference in parameters of similar chips from different vendors. I just aligned parameters of mine to TI.