Old-School Analog Temperature Control Circuits Solved With Modern LTSpice Thermistor Dynamic Models, Part 3

Part 3: Thermistor SPICE Models Make Sense With Older ICs, but What About a Newer One, Such as the LTC1041?


It’s with a somewhat provocative title that we begin the third and last installment in this series of articles about dynamic SPICE modelling of thermistor-based temperature control circuits. I know very well that if a model works with an old IC, it is possible that it will work with new ones. But as they say, seeing is believing. So we will first pay honor to whom honor is due, and that is Linear Technology. Indeed, without the speed of computation of LTSpice and the nearly live display of the results, it would have been rather tedious to make the simulations during the time lengths needed by a temperature control application (sometimes hours, while simulation time is measured in seconds). We will thus first build a simulation involving an IC from Linear Technology, the LTC1041 (a bang-bang controller) and then show that the simulation results correspond well to our expectations and are reproducing the circuit specification.

Finally, we will focus on a circuit that could be a potential thermistor analog version circuit of the Texas Instruments’ LM56. Why the LM56? The LM56 explicitly proposes two input temperature thresholds and regulates temperature within these two limits, and the SPICE model of the LM56 even has a special external temperature pin for transient simulation. It has an internal silicon-based temperature sensor, though; and this is to its disadvantage as far as mechanical flexibility and sensitivity (lower for ICs than thermistors) are concerned. As an academic exercise (and a bit out of jealousy, I confess), I have built my own circuit based on the principles of the LM56 but with an external thermistor.

Doing this we will explore the influence of the thermistor characteristics (response time and tolerances) in the cadre of a last Monte Carlo analysis.

Simulation With LTC1041

The departure point is an excerpt of the datasheet of the LTC10411 . Figure 1 presents the ultra-low power thermostat.

Figure 1

The method of completing the (slightly modified) circuit with a system to be heated has been explained before2 , so we can draw the simulation circuit, substituting the YSI 44007 with a Vishay NTCLE100E3502_B0 offering a similar resistance temperature characteristic (5 k Ω ).

Figure 2

Figure 3

Performing two sweeps for the initial temperature (Tinit) and for the final steady cycling temperature (Rset) values, we get the expected results for the system temperature, going from the initial temperature (15o C, 20o C, 30o C, and 35o C) and oscillating around a temperature defined by the Rset value (32o C and 37o C).

One subtle but not unimportant detail can be verified: as indicated on the application drawing of Figure 1, a switched power output (VP-P) is provided to prevent self-heating of the thermistor, and it is indeed used in our simulation. I could verify this fact by setting the initial ambient temperature at 25o C and looking for the system temperature evolution in time; it’s a flat line, indicating that absolutely no self-heating occurs in the thermistor.

Let’s isolate the voltage divider containing the thermistor in the drawing and verify what happens when we apply the plain 6 V to it (see the circuit embedded in Figure 4). We see that after 100s, the voltage has dropped on the NTC, showing self-heating. Thanks to the switched power input, we avoid this, and our modeling shows it graphically.

Figure 4

The Equivalent of LM56 but With an External Thermistor

In Figure 5, we designed an analog version of the LT56, with the limit cycle of temperature control Tlow and Thigh, both determined by two virtual thermistors, as explained in Part 2 of this series2 .

The thermistor and set signals are first amplified and compared, and an RS flip-flop commutes the voltage ON/OFF on our heating system. For this simulation, the sensing thermistor characteristics (R (25o C) and B value) and the fixed resistors all have a tolerance of +/- 1 %. The result of the system temperature regulation is printed on Figure 6 for several Tlow and Thigh. Each time 10 runs have been performed in a Monte Carlo analysis.

Figure 5

Figure 6

There are overshoots and undershoots due mainly to the thermal inertia of the system, but doing my simulations I found that it is possible to reduce the overshoot and undershoots by playing the R14 and R15 in comparison to the R13. The “trick” is to stop the heating a few degrees before reaching the Thigh, and start the heating before the Tlow temperature threshold.

Figure 7 presents the two curves corresponding to the normal situation when R14 and R15 are equal to R13, and the situation with reduced over and undershoots when the R14 and R15 are reduced and increased a few percentages with respect to the R13. The percentage value can be measured form the original simulation (see parameters perc1 through perc3 in the LTSpice directives).

Figure 7

The last simulation I would like to present sweeps the response time TAU of the thermistor and here we will examine the influence of the relative response time of the thermistor to the reaction time of the system.

In Figure 8, we present the overshoots / undershoots of the system with a relative response time (thermistor / system) going from 1m to 100m. These quantities are computed from the measurement .MEAS LTSpice directives appearing in Figure 2 and stored in the log file available in the menu of LTSpice. This graph shows that it is somewhat vain to try to reduce the response time of the thermistor lower than 0.2 % to 0.3 % of the response time of the whole system.

Figure 8


This article wraps up the trilogy of the so-called old-school temperature control circuits solved with modern LTSpice NTC thermistor models. We have seen that the models accommodate old and new ICs.

Simulations with thermistor SPICE models available on the market are now possible not only in a dynamic transient mode with external temperature driving, but also with a temperature input for the thermistor, which is generated by the circuit itself.

I hope you enjoyed the simulations as much as I enjoyed performing them, and that if you face a problem in the design of temperature control in the future, you will be able to use part of the information presented.

As usual, the simulations presented in this article are available on request at


  1. Linear Technology BANG-BANG Controller, Web. Feb. 2017.
  2. Old-School Analog Temperature Control Circuits Solved with Modern LT spice Thermistor Dynamic Models, Part 2
  3. “LTSpice Nouvelles Commandes, Applications Inédites, Création et Importation de Modèles et de Sous-Circuits.” Chapter 8, Dunod, 2015.

2 comments on “Old-School Analog Temperature Control Circuits Solved With Modern LTSpice Thermistor Dynamic Models, Part 3

  1. abbott
    April 5, 2017

    Thanks for sharing 

  2. liveassist
    December 26, 2017

    yes it is correct

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.