Setting up a SZA263/LTFLU voltage reference


While voltage references based on LMx99 or LTZ1000CH and LTZ1000ACH are common and lot’s of information about them is available, there is less knowledge on how to setup a voltage references based on the Motorola SZA263 and the Fluke LTFLU fabricated by Linear Technology.

The main key differenc between the REFAMP such as SZA263 or LTFLU and more traditional LTZ1000/LTZ1000A are:

  • REFAMP has no heater built-in unlike LTZ part.
  • REFAMP has the temperature compensating transistor in series with the buried zener diode, not parallel as in the LTZ design.
  • Therefore the LTFLU circuit needs a lot more temperature and current compensation work from designer than LTZ1000 circuits do.
  • The LTFLU/SZA263 is in a 4-pin TO-can, the LTZ1000 in an 8-pin for additional signals. They are not compatible and need different bias circuit around them.

Fluke used REFAMP (Motorola SZA263 first, Linear LTFLU later) in many instruments, such as benchtop DMM, calibrators and DC Voltage Standards. Below is list of some examples.

Fluke 341A DC Voltage Calibrator Fluke 5440B DC calibrator – two SZA263 chips
Fluke 343A DC Voltage Calibrator Fluke 5500A Multi-function calibrator
Fluke 515 Portable calibrator Fluke 5502A Multi-function calibrator
Philips PM2530 7½-digit DMM Fluke 5520A Multi-function calibrator
Fluke 8840A 5½-digit DMM Fluke 5522A Multi-function calibrator
Fluke 8842A 5½-digit DMM Fluke 57LFC System Calibrator
Newer Fluke 8508A – 8½-digit DMM Fluke 5700A Multi-function calibrator – two LTFLU chips
Fluke 8588 and 8558A – 8½-digit DMMs Fluke 5720A Multi-function calibrator – two LTFLU chips
Fluke 731B DC Voltage Standard Fluke 5730A Multi-function calibrator – two LTFLU chips
Fluke 732A DC Voltage Standard Fluke 5790A AC Measurement standard – two LTFLU chips
Fluke 732B DC Voltage Standard Fluke 5790B AC Measurement standard – two LTFLU chips
Fluke 732C DC Voltage Standard Keithley DMM7510 7½-digit DMM

Also for educational purpose schematic section with REFAMP LTFLU voltage reference from Fluke calibrator shown on figure below.

LTFLU-1 reference from Linear (now ADI) and Fluke hermetical resistor networks in 8842A DMM

Fluke 57LFC calibrator using same reference REFAMP and resistor networks Z3,Z4

So it’s a special part, that nobody can buy from Digikey. Yes, however, there are reference modules around SZA263/LTFLU and voltage reference chips available from time to time on secondary market. Even though nobody can say for sure if they are fakes or genuine parts, sort out with bad specification or gray market production.

A tear-down of both parts revealed, that SZA263 is a two chip construction in one TO-package with separate silicon and zener diode, while LTFLU is a one chip design, both connected as a REFAMP with collector, base, emitter of the zener and anode of the silicon diode.

LTFLU-1CH die photo. Courtesy branadic (Dipl.-Ing. A. Bülau) from the EEVBlog

On the die photo we can see much more complex design than just diode and transistor. Different to Linear LTZ1000 which needs a separate stage to boost the zener voltage to 10V, the boost of the output on these REFAMPs is part of the zener circuit itself, realized in a bootstrap fashion. However, a 10V output requires an individually trimmed voltage divider for each device from the output to the base of the REFAMP. As found earlier the TC(Thermal Coefficient) of this divider is rather critical, as it is only dampened by a factor of ~3, while all other resistors of the circuit are less critical and their TC(Thermal Coefficient) is dampened by a factor of 150 … 500.

The typical zener current is IZ=3mA, with the zener voltage varying between VC=6.8V … 7V. Operation of this refamp based voltage reference without an oven but somewhat optimized value for R13 can lead to a temperature behavior of < 1ppm/K over 10..45°C, with a parabolic shape of the TC(Thermal coefficient) curve. However, for best performance the REFAMP should be ovenized to achieve 0ppm/K. To set the zero TC(Thermal coefficient) for a given temperature the value of R13 needs to be determined, while the range for the current through it is IC=20 … 200µA.

This article addresses the findings so far, even though no datasheet is available. Fortunately, circuit diagrams of a lot of Fluke gear can be found using these refamps, which gives a point to start with.

Circuit design and theory of operation

Figure 1: LTFLU/SZA263 module circuit

The goal is to build a battery powered, ovenized single supply 10V voltage reference based on LTFLU1-CH, similar to LTZ1047B designed by Andreas Jahn, a LTZ1000 based portable voltage reference. The oven is planed to use a ceramic or aluminum core substrate mounted on a thick-film resistor.

Figure 2: Aluminum substrate PCB, shows very thin layer of dielectric embedded to metal

Pre-regulation shall be done by LDO LT1763 supplied by a battery pack of 12× 1,2V Eneloop, a total of 14,4V. Based on circuits available a schematic as shown in Figure1 was designed. Therefore a single supply op-amp LT1006 (or OPA189, ADA4522-1) comes into play. As for the critical resistor divider R7A/B a NOMCA16035001 is used with additional resistors (R1 and R2) being integrated into the divider to trim the output voltage to 10.00000V and to reduce their influence.

The oven temperature is planed to be +45…+50 °C, which should be enough headroom even in summer. A NTC mounted to the reference board serves as a temperature sensor for the oven. NCP15 series by Murata is said to have good long-term stability, thus it is used here. Zero TC(Thermal Coefficient) temperature has to be found within the desired temperature range by adjusting R13 value respectively. To do so a breadboard was used with R7A/B pretrimmed for a nominal 10V output voltage by arranging the NOMCA resistor network to R7A = 5 kΩ and R7B = 11 kΩ. This is necessary, as the output voltage also influences the current through a given R13.

Measurements of temperature profile inside a temperature chamber while varying R13 with a decade resistor box gave the following values for the zero TC(Thermal Coefficient) temperature:

R13 25.343 kΩ ~30 °C
R13 24 kΩ ~35 °C

Zero TC(Thermal Coefficient) point is the middle highest point of the flipped parabolic shaped curve, when plotting output voltage over temperature. Assuming a linear correlation between IC and zero TC(Thermal Coefficient) temperature point a value of about 22kΩ for a temperature setpoint of 45°C can be calculated. A repeated temperature profile well agreed with the assumptions, the zero TC(Thermal Coefficient) point is well within 45 … 50°C. Based on this results a reference board was designed as shown in Figure2. It‘s 20 × 40 mm2 in size. The LTFLU is soldered in a SMT style to the board and the output is realized as a 4 wire connection.

Figure 3: LTFLU reference board

Typical resistor values used for one of the first modules are shown on Figure 4 for reference.

Figure 4: Components values for first module

Several temperature profiles were performed with R13 = 22 kΩ and by varying the additional resistors R1 and R2, giving the following results.

Figure 5: LTFLU Temperature profile 1

Figure 6: Temperature profile 2

Figure 7: Temperature profile 3

Profile 1 Profile 2 Profile 3
R7A 5 kΩ 5k Ω 5 kΩ
R7B 10.9091 kΩ 11 kΩ 11.0651 kΩ
Ratio 0.4583 0.4545 0.4519
Voltage @ zero TC(Thermal Coefficient) 10.061870 V 10.035500 V 10.016960 V

With the values given one can calculate the required ratio of 0.4494 as well as the required resistors to trim the output to 10.00000V. It turns out that for this specimen R1 = 6.2 kΩ trims the output to the required range, while R2 = 0…10 Ω adjusts the output to the final value.

Figure 8: Temperature profile 4

Figure 9: Temperature profile 5

To be continued …

Building one of the test modules in xDevs US lab for testing

Recently Illya have received and assembled one test module during livestream recording, showing process and steps involved.

And photograph of assembled module with cheap test resistors:

Next step would be trimming for precise 10V output, fine-tuning temperature stability and doing noise performance tests and measurements.

Author: Dipl.-Ing. André Bülau
Published: Dec. 2, 2019, 6:18 a.m.
Modified: Sept. 13, 2020, 8:14 a.m.