Standalone using Fluke 5700A REF-DAC module

Contents

What is the Ref-Amp?

Ref-Amp consists of an NPN transistor in series with a zener diode. When biased properly, the combination has a extremely low temperature coefficient. The reference voltage that the can vary from +6.5 VDC to +7 VDC, depends on actual production bath. Also since both zener and transistor are located on same substrate and enclosed in hermetic package, they are tightly thermally coupled and protected from ambient humidity. This allows to improve stability over long time spans.

There are two well-known examples of this device family, Ref-Amps used by Fluke in high-end instrumentation equipment. These are Motorola SZA263 and Linear LTFLU-1H/LTFLU-1AH.

The Motorola SZA263 is a “Ref-Amp” that was made by Motorola, and later discontinued after getting out of high-volume semiconductor business. This forced Fluke to find a partner to design a replacement for the obsolete SZA263, as Motorola would not sell the design files and masks of original design. Linear Technology was happy to help big customer like Fluke with the design, and that’s how LTFLU-1AH Ref-Amp emerged. It’s pin-2-pin and function compatible with the original SZA263, but the LTFLU-1AH use bit aluminum alloy for the interconnects, which is not the same one Motorola used. This rendered in different long-term stability, visible by positive drift over time on SZA263, but negative in LTFLU-1AH (same as LTZ1000/LTZ1000A and the LMx99 ICs).

If you have a bank of 4 × 732A’s and a bank 4 x recent 732B’s that you get calibrated once a year, you will clearly see these drift patterns. Care need to be taken though, as early batches of Fluke 732B were still using left-over SZA263’s, but later all production 732B were updated to LTFLU-1AH.

Main differences summary between the SZA263/LTFLU-1AH IC and the market-available Linear LTZ1000 are:

  • Different package. LTFLU-1AH is 4-pin hermetic can, the LTZ1000 in an 8-pin TO-99. They are far from drop-in compatible.
  • SZA263/LTFLU-1AH has the transistor for temperature compensation in series with the zener, not parallel as LTZ1000 design.
  • Due to different manufacturing process, Motorola SZA263 have positive long-term drift, while Linear LTFLU-1AH has negative long-term drift.
  • Opposite to LTZ1000A, there is no on-die heater in SZA263/LTFLU-1AH
  • SZA263/LTFLU-1AH require much more time (vs LTZ1000 design) and care for support resistor matching and tempco testing

Last item is due to different tempco compensation transistor arrangement in SZA263/LTFLU-1AH circuit, which needs more attention for temperature and current compensation than LTZ1000 circuits. LTZ1000 design is fine even with standard datasheet reference schematics, without any analog black magic or voodoo.

All this above of course less a problem than zero availability of LTFLU-1AH, as Fluke have exclusive rights for this design and chip, and Linear is not allowed to sell Ref-Amp on open market. So Motorola SZA263/Linear LTFLU-1AH are less friendly solution to implement a stable DC reference, as even if you get LTFLU-1AH chip, lot of time and money for resistor matching and temperature testing is required.

Die photos of LTFLU-1AH and other references

LTFLU-1AH


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.

LTZ1000ACH

Here’s die photograph of Linear LTZ1000ACH chip:

LM399


National Semi LM399 die photo. Courtesy branadic (Dipl.-Ing. A. Bülau) from the EEVBlog

LTC6655

Instruments list using Ref-Amp

Known instruments to implement SZA263/LTFLU-1AH as primary DC reference:

There is also great thread here on EEVBlog created by lymex, revealing guts of various DC voltage standards.

Overview of A11 DAC PCBA from Fluke 5700A MFC

Fluke 5700A Service manual has schematics and short description of DC reference and DAC module operation theory, so it’s worth to refresh memory reading related sections to get familiar with DAC module design.

PCB with blue soldermask is 6-layer FR4, with most of the routing buried on inner layers. There are only few ground mesh fill polygons, around digital components, such as MCU, clock generator and ADC. Stripline routing might serve two purposes here – act as additional protection sensitive analog traces from environment and surface contamination. Book Basic Linear Design from Analog Devices covers some of ideas how PCB leakage and bad layout could introduce errors and issues in sensitive analog design.

DAC assembly provide adjustable stable DC voltage output, from 0 to +11VDC and features multiple subassembly units.

  • DC Reference hybrid (HR5)
  • DC Amplifier hybrid (HR6)
  • DAC Filter SIP
  • DAC mainboard

Fluke used PWM-controlled method to generate adjustable voltages. There are also few auxiliary support circuits on-board, such as sense-current cancellation block, linearity tune control, negative offset control. Block diagram below on Image 4 can help to understand overall function of circuitry on A11 DAC PCBA.

Output DC voltage is generated by 5-pole discrete filter, which has two precision square waves of different amplitude as the input. First input channel of the filter is CH1, and it’s amplitude is from 0V to reference voltage, which is around +13 VDC. This is coarse adjustment channel. Second CH2 channel is operated similar way but goes up only to attenuated reference voltage at +0.78 mVDC. This is fine tune adjustment channel. Filter is designed as LPF with bandwidth 30 Hz, and input square wave frequency is 190 Hz. So the output is clean and filtered DC voltage, derived from controlled PWM CH1 + CH2.

Filter output does not have capability to drive large currents, so output stage on separate ceramic hybrid takes filter output and provide driving capability for DAC output. Hence after all, circuit output voltage can be predicted and calculated by simple formula:

*VOUT = DutyCH1 * VREF13V + DutyCH2 * VREF0.00078V*

Here’s realtime example to try some values to generate precise 10V:

DC output VDC = VREF V * Duty_CH1 REF % * VREFuV uV * Duty_CH2 REFuV %

Use of this combined PWM scheme allows us to have efficient way to generate arbitrary voltage levels without use of very expensive resistor networks and expensive complex multi-bit DAC ICs. PWM duty cycle resolution of PWM generator used in Fluke calibrator is 0.0024%, which provides resolution of CH1 = 309 µV/bit and CH2 = 18.5nV/bit. PWM signal also electrically isolated by optocouplers.

Dark color covers around HR5 and HR6 hybrid assembly are not metal, but metallized plastic. Main purpose of these covers is to prevent stray airflow around reference circuity and DAC chopper amplifier. This is important due to parasitic thermocouple EMF-generation effect present with any thermal gradient within the board. Also since hybrids are actively heated, enclosure helps with thermal stability of inner thermostat area. Don’t forget, Fluke 5700A dissipate plenty of power during operation and has two large fans to provide airflow around PAs and high-power components and boards.

ADC section and PWM generator/digital controller are separately enclosed in metal cage shield on the top side. This serve dual purpose, to keep generated RFI/EMI enclosed and localized and to provide additional shielding from external fields. These shields are grounded to module power ground plane.

Main DC reference hybrid is built using sandwich of ceramic substrates. Main thin substrate has two Ref-Amp, hybrid resistor network, few opamps, temperature resistors and circuit tracks. Back side of HR5 has large area 27 Ω resistor acting as a heater. Then there is glued spacer to act as a heat-spreader and coupled to it hermetic hybrid resistor with clear quartz window for laser trim access.

All parts except Ref-Amps, which as Motorola SZA263 are SMT-mount. Two Ref-amps are used to provide higher +13 VDC (6.5 + 6.5) reference level to further reduce amount of noise and improve stability of the DAC. Excellent temperature coefficient of used Ref-Amps is achieved by using stable collector bias current of their transistors provided from stable thin-film resistor hybrid between Ref-Amps. This design allow to have very small output voltage impact from circuit component errors, so it’s output stability directed almost entirely from Motorola SZA263 performance.

To prevent output drift with respect to ambient temperature whole assembly is heated to constant +62 °C by external circuit block on main A11 PCBA. Temperature feedback element is thermistor RT1, located right near the Ref-Amp packages. As ceramic is a good thermal conductor, any change in temperature of hybrid will drive correction signal to Q2 on main board and have circuit adjust power to 27 Ω film resistor to adjust temperature back to set-point. There is also thermal runaway protection, implemented with second thermistor RT2, which activates Q9 to bypass base current of Q1 to avoid overheating. This protection kicks in once substrate temperature reaches +67 °C, and normally should be never used.

Interesting to note that improved Fluke 5720A also has changes in this HR5 DC reference hybrid, as we can see thanks to lymex from bbs.38hot.net chinese forums, who released photo of HR5 used in Fluke 5720A A11:

Pair of Motorola SZA263 chips replaced by Linear LTFLU-1ACH, and LF351 opamps are replaced with Linear LT1006 and TL071C.

Hybrid laser-trimmed resistor network is smaller and have different configuration as well.

Very similar dual Ref-Amp assembly used also in Fluke’s AC measurement standard, Model 5790A on A16 DAC board.

Bit simpler version with only one Ref-Amp is used in Fluke 5500A/5520A and system calibrator 57LFC. There is thermal control over ref-amp assembly in these calibrators either. So as a result DCV performance of these calibrators is “only” 11 ppm annual, with daily stability 2 ppm. Fluke 5700A has 8 ppm annual, and updated 5720A/5730A are half of that, 4 ppm.

Of course if better temperature stability is achieved with removal of all airflow, actual Ref-Amp can provide much better performance. This is proven by ± 2.0 ppm/year and 0.3 ppm/day specifications and actual performance of double-oven Fluke 732B reference assembly.

Let’s take a brief look on 732B reference assembly block diagram:

Power supply requirements of A11 PCBA

In this section we determine what requirement are to get A11 PCBA operational and working, without calibrator mainframe itself.

CAD dimensions of A11

This section will cover physical measurements and dimensions to design a suitable enclosure to fit A11 and other support electronic parts.

Overview of A16 PA PCBA from Fluke 5700A MFC

The A16 PCBA PA PCB outputs DC voltages from ±22V to ±219.99999V and AC voltages from 22V to 219.99999VRMS. The frequency limit for 220V AC output is 100 kHz. Output voltage limits are derated at frequencies above 100 kHz. At 1 MHz, the maximum output voltage is 22 VRMS. The PA drives the High Voltage assemblies (A14, A15) in all high voltage and high current functions. This assembly also contains calibration circuitry that enables the internal calibration system to determine exact PA AC and DC gain, offsets and frequency response.

The main sections of this assembly are the input stage, mid stage, output stage, sense current cancellation circuit, the DC and AC gain calibration circuits, and the PA Digital Control SIP assembly (A16A1), which is mounted on the PA PCB vertically.

Digital control for the PA PCB is contained on the SIP assembly (A16A1) mounted at the bottom of the PA. This assembly configures the PA PCB for its various modes of operation. The heart of the Digital Control assembly is an 82C55 Programmable Peripheral Interface IC (U11) operating under software control via the guarded digital bus. This IC has three ports that generate 24 outputs. These outputs control two 5801 relay driver ICs (U10, U12), two LM339 Comparators (U13, U15) and an analog multiplexer (U14) used for diagnostics.

PA Common Circuitry

Common circuitry consists of the +PA and -PA supplies, input stage, mid stage, and the output stage. PA input node, gain, and feedback are different for DC and AC operation. PA gain is -20 in the DC function, determined by the ratio of resistor network bonded to the HR8 assembly (500 kΩ/25 kΩ). Gain in the AC function is -10, which is determined by the ratio resistors [(R11 + R12 + R13)/R17].

+PA and -PA Supplies

The ±PA supplies are high voltage supplies generated by the Filter/PA Supply assembly (A18). These supplies can be controlled by the Digital Control SIP assembly (A16A1) and are switched between the two modes, either ±185V or ±365V. Theory of operation for the Filter/PA Supply assembly (A18) in service manual describes how these voltages are generated and selected.

PA Input Stage

The input stage consists of a heater-controlled hybrid HR8, Linear LTC1052 opamp U7, transistor Q6, and JFET Q2. The HR8 assembly consists of an opamp mounted on a heated-substrate hybrid, with a resistor network bonded to it. Hybrid HR8 provides the input stage with excellent DC characteristics of low offset, noise and drift. The hybrid heater-control circuit (on sheet 3 of the schematic in service manual) adjusts the base voltage of Q38 to deliver the correct current to the heater resistor. This maintains the hybrid assembly at a constant temperature in spite of environmental temperature variations. Transistor Q35 protects the hybrid in case Q38 fails. Input of the hybrid opamp is protected by CR13 and CR14.

Output of the hybrid opamp is connected to the input of a faster opamp (U7), which provides additional DC gain and a higher slew rate. JFET Q2 and transistor Q6 combined with these two opamps complete the input stage. Q2 is a very low-bias-current, high-frequency JFET.

In mid to high-frequency operation, Q2 is effectively the only path for the input stage signal. HR8 and the U7 opamps are bypassed at these frequencies by R89, C42, R24, and C12. As a result, the base of Q6 is at AC ground. In DC to mid-frequency operation, the gate of Q2 is at ground potential. At any frequency, the potential difference between the gate of Q2 and the base of Q6 results in a current through Q6 as determined by R22, and by the transconductance of Q2 and Q6. The input stage is called a transconductance stage because an input voltage results in a current output at the collector of Q6. This current output is coupled to the mid stage (Q12, Q14, and Q16) by Q8, Q9, Q13 and C15, where it results in a voltage across the base-emitter of Q16 (the input of the mid stage). Current source Q9 determines bias current in Q2 and Q6. Variations of Q6 output current become voltage variations at the base of Q16. This transfer is through Q8 and Q13 at DC and low frequencies, and through C15 at high frequencies. The input stage operates with low voltage supplies (±17V) whereas Q16 of the mid stage is connected to the -PA supply, which can be as high as -365V. This potential difference is dropped across level shifter Q13.

PA Mid Stage

The mid stage (Q12, Q14, and Q16), biased by the 8 mA current source (CR53, Q31, Q32 and R87 on sheet 2 of the schematic), is a voltage amplifier providing additional gain. The base of transistor Q16 is the input to the mid stage. MOSFETs Q12 and Q14 are biased by R41 and R53 respectively. Components CR21, CR23, and VR22 protect Q12 from excessive source-to-gate voltage, and R112 prevents Q12 from oscillating. Components CR25, CR29, VR28, and R113 perform the same function for Q14. A signal at the base of Q16 appears amplified at the drain of Q12. Total impedance from the drain of Q12 to ground, divided by R58, determines gain at DC and low frequencies.

At high frequencies, the effective drain to ground impedance is R53. Relay K12A parallels C18 and C57 during DC operation for a lower bandwidth. Capacitors C18 and C57 provide the Miller capacitance for the amplifier. Transconductance gain of the input stage and the Miller capacitance determine Power Amplifier frequency response at high frequencies. All the voltage gain of the Power Amplifier comes from the input and mid stages.

PA Output Stage

The Output Stage is an emitter follower that provides current gain but no voltage gain. It is needed because the mid stage cannot drive the rated load by itself. Voltage across R74 and R35 determines the bias current through the output stage. This voltage equals the voltage across Q7, minus the value (4 x Vbe) (for each transistor Q4, Q5, Q10, and Q11). Transistor Q7 is configured as a Vbe multiplier, the voltage across which (and thus the output stage bias current) is the value (1 + (R23+R26)/R32). The output bias current is 50 mA. NMOSFETs Q1, Q3, and transistor Q5 source current, while PMOSFETs Q15, Q17, and transistor Q10 sink current from the load. This output stage can drive up to 50 mA of load current as determined by the current limit circuit on ±PA supplies on the Filter/PA Supply assembly (A18). Zener diodes VR15 and VR18 bootstrap MOSFETs Q3 and Q15 respectively, and provide the power supplies SC+ and SC- to opamp U1 in the sense current cancellation circuit. Two stacked NMOSFETs (Q1, Q3) on the top end (+PA side), and two stacked PMOSFETs (Q15, Q17) on the bottom end (-PA side) withstand the high voltage drops between ±PA supplies and output. NMOSFETs Q1 and Q3 are biased by R15 and R19 respectively. PMOSFETs Q15 and Q17 are biased by R52 and R57 respectively. Components CR5, CR7 and VR6 protect Q1 from excessive source-to-gate voltage and R108 prevents Q1 from oscillating. Protection is also provided for remaining MOSFETs in the output stage. Output of this stage, called PA OUT HI, is the output of the Power Amplifier assembly. Components R120 and L10 isolate capacitive loads.

PA in Standby

The PA schematic shows all relays and DG211 FET switches in the standby condition. The PA 25 kΩ input resistor and R17 are tied to OS COM through Q39 and R118. PA output is close to zero and the whole loop is stabilized. To better understand PA configuration in the ac/dc 220V range, refer to block diagram below.

PA Operation: 220V DC Range

During DC operation, PA gain is -20, as determined by the 500 kΩ/25 kΩ resistor network on the HR8 assembly. Control line SW3, inverted by U8, turns on Q51. This references the +input of the precision opamp in the input stage to R COM. The DAC assembly is set to the negative 11V range and its outputs, DAC OUT HI and DAC SENSE HI, are connected to pin 2 of the resistor network on the HR8 assembly by relay K2. The sense current cancellation circuit is active during DC operation. Its output, SIG1, is connected to the resistor network feedback resistor pin 1 by relay K15A. The amplifier has a much lower bandwidth in this mode because of the much higher Miller capacitance in C57. Lower bandwidth results in lower amplifier noise.

The output signal, PA OUT HI, is routed to the High Voltage Control assembly (A14), where it goes through relay K10 and becomes PA OUT DC. PA OUT DC is routed to the Switch Matrix for connection to the OUTPUT HI binding post. The sense signal, PA SENSE DC from the Sense Current Cancellation circuit, is routed to the Switch Matrix assembly (A8) for connection to the OUT/SENSE HI or SENSE HI binding posts, thus making the binding post the sense point in internal sense and allowing for external sense through the SENSE HI binding post.

PA Operation: 220V AC Range

During AC operation, PA gain is -10 as determined by the 4.99 kΩ input resistor R17, and 49.9 kΩ feedback resistors (R11 + R12 + R13). Control line SW3 turns Q50 on, which references the +input of the precision opamp in the input stage to OS COM. The Oscillator assembly (A13) is set to the 22V range and its output OSC OUT HI is connected to the input resistor R17 by relay K10A. The PA output is connected to the feedback resistors R11-R13 by relay K12B. The PA output is attenuated by a precise 1/100 by 220V range ac attenuator. The attenuated signal is connected to OSC SENSE HI, where it is sent to the Oscillator Control assembly (A12).

The Oscillator Control assembly regulates the Oscillator Output so that an exact calibrated AC signal appears at OSC SENSE HI. Since the 220V range AC attenuator is completely characterized (as explained below), the exact desired signal appears at PA SENSE AC and hence at the appropriate sense point at the output.

The 220V range AC attenuator circuit contains opamp U4, a 400 kΩ/4 kΩ resistor network Z1, and transistor Q54. PA SENSE AC, which is connected to PA OUT HI at the load, is connected to the 400 kΩ input resistor (pin 1) of Z1 by relay K16. The 400 kΩ/4 kΩ node (pin 3) of Z1 is connected to the inverting input of U4. During AC operation, control line C0* is inverted by U8, which turns on Q58 to connect the non-inverting input to OSC RCOM. Transistor Q54 supplies current gain for the output of U4 to drive the capacitance of the OSC SENSE HI line. This voltage is connected to the 4 kΩ feedback resistor (pin 4) of Z1. The output is connected to OSC SENSE HI by relays K10B and K11.

The DC feedback 500 kΩ/25 kΩ resistor network and the sense-current cancellation circuitry are disconnected by energizing K15. The sense signal, PA SENSE AC, and the output signal, PA OUT HI, are routed to the High Voltage Control assembly (A14), where relays K10, K13, and K3 connect them to HV SENSE and HV OUT. HV SENSE and HV OUT are connected to the binding posts by the motherboard relays in the same manner as in the 1100V high voltage mode. Refer to the High Voltage assembly theory of operation in service manual for more information.

220V DC Internal Calibration Network

The 220V DC internal calibration network determines the exact gains and offsets of the PA. This circuit uses part of the resistor network HR8 as the input attenuator, and uses opamp U9, and zener diodes VR57 and VR58. Relay K4 connects the output of this circuit to the RCL line. Zener diodes VR57 and VR58 reduce the power supplies for chopper-stabilized amplifier U9, which is used as a voltage follower.

Backplane design and signaling

This section will cover electrical and physical design of interconnect backplane board.

Power supply design for A11

This section will cover design of power supply board to power up A11.

Bring-up for DAC/REF

This section will first tests.

Comparison with LTZ1000 reference


Author: Illya Tsemenko
Published: Oct. 14, 2016, 9:19 a.m.
Modified: Feb. 3, 2017, 10:38 a.m.

Comments