LM4040 model issues

The TI LM4040 is a shunt voltage reference with excellent output and thermal stability. It was first used one to replace a 10V zener diode as the reference voltage in a linear power supply. The circuit had a multi-turn preset for precisely adjusting the output voltage but it was found that over a short period of time even over small temperature variations such as  when it was first powered up the output voltage of the circuit varied by a significant number of millivolts.

The problem was traced to the zener diode so it replaced it with a high precision 0.1% variant of the 10V LM4040. The drift completely disappeared and justified the use of a precision multi-turn preset to set an exact voltage with a 6.5 digit DVM. The design was complete but the zener diode was left in the simulation circuit because basic transient simulation did not reproduce the drift problems caused by temperature change in the diode so it always gave completely stable measurements.

The Problem

More recently a new circuit design was being worked on for a linear power supply but this time there was a need to improve the simulation accuracy so it was decided to see if a model for the 10V LM4040 existed. It was found on the TI website so it was download and used to replace the 10V zener diode in the simulation circuit.

The first observable issue was that the model just did not work as a direct replacement. A simple test circuit was created using a 20V DC supply (VDD) and a 10K resistor (RS) in the same configuration shown in the LM4040 datasheet where pretty much exactly 10V should have been measured at the cathode (VOUT) The voltage however was much closer to 16V.


After looking at the SPICE model for the device, which is a sub-circuit due to the complexity and internal circuitry of the device (it’s not really working like a basic zener diode at all but the symbol is reasonably representative of its function), pin 1 was labelled ‘V+’ and pin 2 was labelled ‘V-‘. This implied that the cathode was pin 1 and the anode was pin 2 however a native diode model in SPICE connects the anode to pin 1 and the cathode to pin 2. The LTSpiceIV software used, like the representation in the TI datasheet, allows a basic zener diode symbol to be used in the schematic for this type of custom component so this seemed to be the best way of using the custom TI model.

Once these pins in the model were swapped over the output settled to exactly 10V which confirmed that the pins were indeed reversed as far as a default SPICE ‘diode’ model was concerned. (Reversing the diode in the circuit also has the same effect).

Below is a fragment from the original sub-circuit model text: –

And the revised circuit with the pin labels reversed: –

Having achieved a working result the diode was copied to an identical but negative voltage regulator circuit. The potential difference between the anode and cathode remained exactly the same at 10V however the positive (Cathode) pin was now connected to 0V and the negative (Anode) pin was connected to a resistor off of a -20V source. The voltage measured on the anode should have been exactly -10V but again it was closer to -16V.

The model was opened up in an editor and a mathematical computation was found that looked as if it might be sensitive to the absolute polarity of the voltages across the devices terminals.

Below is the statement that computes the E_Estart value from the LM4040 simulation model: –

The result of this formula was converted to an absolute value to see if it would solve the negative Anode potential problem: –

Once the modified circuit was loaded into the design the shunt voltage reference worked equally well in both positive and negative circuit topologies and all that remained was a small offset voltage on the negative side connected reference so that the final voltage measured -9.863V. This difference is probably due to one remaining issue in the models calculations when used in a negative circuit but all that mattered at the time was to get this circuit working so that more important design issues could be checked out.


The positive and negative terminal issue can be attributed to the way in which a sub-circuit can be used as if it was a native diode model in the simulation software but for the sake of compatibility it might be a good idea for all such modeled circuits to use the standard model pin-out convention. The polarity error in the model must have existed for some time and it shows that less commonly tested scenarios can result in problems years down the line. It might be a good idea for all simulation modellers to test their devices in both positive and negative circuit topologies before releasing them to the general public.

About Open Technologies Limited

OTL specialise in the design and development of real-time software and hardware applications. We strive to innovate and bring our customers both cutting edge technologies and efficient solutions to their problems. With many years of experience working in industrial and consumer sectors we are always conscious of the need for reliable, maintainable and sustainable products. We are always looking to work with new clients and welcome the opportunity to discuss how we may be able to help you.

Contact OTL

Leave a Reply