Text preview for : JW_EDN_p3.pdf part of Fluke JW EDN p3 Fluke 720A doc JW_EDN_p3.pdf
Back to : JW_EDN_p3.pdf | Home
designfeature By Jim Williams, Linear Technology Corp
UNDESIRABLE AND UNINTENDED THERMOCOUPLES ARE THE
PRIMARY SOURCES OF ERROR IN LOW-DRIFT CIRCUITS.
ATTENTION TO LAYOUT AND OTHER CONSTRUCTION
DETAILS IS THE ONLY WAY TO TACKLE THE PROBLEM.
Minimizing thermocouples
maintains 20-bit DAC precision
ubtle parasitics can have pronounced and
S seemingly inexplicable effects on the perform-
ance of low-level circuits, and a 1-ppm DAC is
certainly in this category. Part 1 of this three-part se-
COPPER
JUNCTION
BISMUTH
ries discussed the circuit design of a 20-bit
Figure 1
DAC with 0.1 ppm/ C of drift, and part 2 dis-
cussed the measurement techniques (references 1 N
and 2). This third and final part discusses how you JUNCTION
deal with cables, connections, solder, component
W E
choice, terror, and circuit arcana.
Perhaps the most prevalent detractors to micro-
S
volt-level circuitry are unintended thermocouples. COMPASS
(Reference 2 also includes considerable discussion
on dealing with thermocouples.) In 1822, Thomas Joining pieces of bismuth and copper led Thomas Seebeck
Seebeck, an Estonian physician, accidentally joined to his accidental discovery of what he called "thermomag-
semicircular pieces of bismuth and copper while netism" and what we now call the Seebeck effect.
studying thermal effects on galvanic arrangements
(Figure 1). A nearby compass indicated a magnetic
disturbance. Seebeck experimented repeatedly with
3
different metal combinations at various
Figure 2 2.8
temperatures, noting relative magnetic-field
2.6
strengths. Curiously, he did not believe that electric
2.4
current was flowing and preferred to describe the ef-
2.2
fect as "thermomagnetism." He published his results
2
in a paper (Reference 1). Subsequent investigation
1.8
showed the "Seebeck effect" to be fundamentally 1.6
electrical in nature, repeatable, and quite useful. V REFERRED 1.4
Thermocouples, by far the most common trans- TO 25 C
1.2
ducers, are Seebeck's descendants. Unfortunately, 1
unintended and unwanted thermocouples are also 0.8
Seebeck's progeny. 0.6
In low-drift circuits, unwanted thermocouples are 0.4
probably the primary source of error. Connectors, 0.2
switches, relay contacts, sockets, wire, and even sol- 0
25 30 35 40 45
der are all candidates for thermal-EMF (electromag- C
netic-field) generation. It is relatively clear that con-
nectors and sockets can form thermal junctions.
However, it is not at all obvious that junctions of cop- Two supposedly identical copper wires generate thermal EMFs due
per wire from different manufacturers can easily gen- to oxidation and impurities.
www.ednmag.com May 3, 2001 | edn 83
designfeature 20-bit DACs
erate drifts of 200 nV/ C, which is sary resistor to promote such
four times a precision amplifier's 100 thermal balancing (Figure 5).
SLOPE 1.5 V/ C
drift specification (Figure 2). LESS THAN 25 C For remote signal sources,
Even solder can become an error connectors may be unavoidable.
term at low levels, creating 50 In these cases, choose a connec-
Figure 3
a junction with copper, 64% SN/36% PB tor specified for relatively low
Kovar wires, or pc-board traces THERMALLY
thermal-EMF activity and en-
(Figure 3). PRODUCED 0 sure a similarly balanced ap-
VOLTAGE ( V)
Table 1 lists thermocouple po- SLOPE 160 nV/ C proach in routing signals
LESS THAN 25 C
tentials for some common mate- 60% CD/40% SN through the connector along the
rials in electronic assemblies. The 50
pc board and to circuitry. If
information indicates the inad- some imbalance is unavoidable,
visability of mixing materials in deliberately introduce an inten-
the signal path. The table also 100 tional counterbalancing junc-
dramatically points out that you 0 10 20 30 40 50 tion. In all cases, maintain the
must keep copper/copper (top SOLID-COPPER JUNCTION differencing junctions in prox-
DIFFERENTIAL TEMPERATURE ( C)
table entry) connections clean or imity, which will keep them at
a degradation of 5000-to-1 oc- the same temperature. Avoid
NOTE: SOURCE IS NEW ELECTRONICS 2/6/77.
curs as they oxidize (bottom drafts and temperature gradi-
table entry). The unusually en- ents, which can introduce ther-
ergetic response of the copper/- Solder-copper junctions can create thermal EMFs. Cadmium/tin has mal imbalances and cause prob-
copper-oxide combination ne- notably lower activity but is toxic, unavailable, and not recom- lems. Figure 6 shows the
cessitates cleaning digital volt- mended. LTC1150 amplifier in a test cir-
meter and Kelvin-Varley divider cuit to measure its temperature
connections with a copper deoxidant a common standards-laboratory proce- stability. The lead lengths of the resistors
(Caig Labs, "Deoxit" D100L). dure, can be effective in reducing drifts connected to the amplifier's inputs are
If you find the information in Table 1 that originate from thermal EMFs. A sim- identical. The thermal capacity each in-
to be seemingly academic, the implica- ple example uses a nominally unneces- put sees is also balanced because of the
tions in Figure 4 should wake you up.
This figure lists thermoelectric potentials THERMOELECTRIC
CONNECTION TYPE DESCRIPTION POTENTIAL ( V/ C)
for commonly employed laboratory con-
nectors. Thermocouple activity of some BNC-BNC MATE 0.4
connectors is more than 20 times greater
than other types, so be careful when us-
ing them.
BNC-BANANA ADAPTER 0.35
LAYOUT CAN REDUCE THERMAL ERRORS
Minimizing thermal-EMF-induced er-
rors is possible if you pay judicious at-
BNC-BNC "BARREL" ADAPTER 0.4
tention to pc-board layout. In general, it
is good practice to limit the number of
junctions in the signal path. Avoid as MALE/FEMALE BANANA MATE SAMPLE #1 0.35
much as possible using connectors, sock-
MALE/FEMALE BANANA MATE SAMPLE #2 1.1
ets, switches, and other potential error
sources. When avoiding the use of these MALE/FEMALE BANANA MATE SAMPLE #3
error sources is impossible, attempt to (TYPE SPECIFIED FOR LOW THERMAL ACTIVITY) 0.07
balance the number and type of junctions COPPER LUG-COPPER BANANA BINDING POST 0.08
in the signal path so that differential can-
cellation occurs. Ensuring this cancella-
tion may involve deliberately creating and COPPER LUG-STANDARD BANANA BINDING POST 0.5
introducing junctions to offset unavoid-
able junctions, which can be a tricky pro-
cedure. Repeated and deliberate temper- PLATED LUG-COPPER BANANA BINDING POST 1.7
ature excursions may be necessary to
determine the optimal number and
placement of added junctions. Ex- Measured thermoelectric potentials of some common laboratory connectors can
perimentation, tempered by a healthy Figure 4 vary widely. The pronounced difference between samples of banana connectors is
reserve of patience and abundance of due to manufacturers' materials choice; the copper-lug/copper-banana post has 20 times lower
time, is necessary. This practice, which is activity than the plated-lug/copper-banana post.
84 edn | May 3, 2001 www.ednmag.com
designfeature 20-bit DACs
symmetrical connection of the resis- close by to achieve a good grounding
tors and their identical size. Thus, TABLE 1--THERMOELECTRIC POTENTIALS scheme. An RF choke connected
thermal-EMF-induced shifts are equal FOR VARIOUS MATERIALS across a scope probe can determine
in phase and amplitude, and cancel- Materials Potential ( V/ C) the presence and relative intensity of
lation occurs. Slight air currents can Copper/copper Less than 0.2 transformer fields, aiding layout ex-
still affect this arrangement. Figure 7 Copper/silver 0.3 perimentation.
shows a strip chart of output noise Copper/gold 0.3 Another source of parasitic error is
with a small styrofoam cup covering Copper-cadmium/tin 0.3 stray leakage current.You must prevent
the circuit and with no cover in "still" Copper-lead/tin 1 to 3 such leakage currents from influencing
air. This data illustrates why it is often Copper/kovar 40 circuit operation. The simplest way is
prudent to enclose low-level circuitry Copper/silicon 400 to connect leakage-sensitive points via
inside some form of thermal baffle. Copper/copper-oxide 1000 Teflon standoffs. Then, stray leakage
Thermal EMFs are the most likely, Note: This information comes from Keithley Instru- currents do not affect sensitive points
but not the only, potential low-level ments, "Low level measurements," 1984. because they never contact the pc
error source. Electrostatic and elec- board. Although this approach is ef-
tromagnetic shielding may also be nec- not distinguish between this spurious fective, its implementation may not be ac-
essary. Power-supply-transformer fields signal and the desired input. Attempts to ceptable in production.
are notorious sources of errors that are eliminate the problem by rolling off the Guarding is another technique for min-
often mistakenly attributed to an ampli- circuit's response may work, but the fil- imizing board-leakage effects. The guard
fier's dc drift and noise. A transformer's tered version of the undesired pickup of- is a pc trace that completely encircles the
magnetic field impinging on a pc trace ten masquerades as an unstable dc term. leakage-sensitive points. You drive this
can easily generate microvolts across that The most direct approach is to use trace at a potential equal to that of the
conductor in accordance with well- shielded transformers, but careful layout point, preventing leakage to the "guard-
known magnetic theory. The circuit can- may be equally effective and less costly. ed"point. On pc boards, the guard should
The transformer's magnetic enclose the node or nodes you want to
DELIBERATE SPLICE MAY
field may disturb a circuit that protect. This guarding technique elimi-
NORMALLY UNNECESSARY BE DESIRABLE TO BALANCE requires the transformer to be nates the effects of capacitor surface leak-
RESISTOR THERMALLY OTHER JUNCTIONS
BALANCES OTHER
age in Figure 3 of part 2.
INPUT RESISTOR LEAD WIRE/SOLDER/ 50k
COPPER-TRACE JUNCTION Figure 6
+ References
RESISTOR LEAD, SOLDER,
OUTPUT 1. Williams, Jim, "20-bit DAC demon-
COPPER-TRACE JUNCTION
LTC1150 strates the art of digitizing 1 ppm, Part
+
100
1: exploring design options," EDN, April
12, 2001, pg 95.
Figure 5 EOS 1000 2. Williams, Jim, "Measurement tech-
niques help hit the 1-ppm mark," EDN,
50k April 26, 2001, pg 117.
3. Seebeck, Thomas,"Magnetische Po-
larisation der Metalle und Erze durch
Typical thermal-layout considerations emphasize minimiz- Temperatur-Differenz," Abhaandlungen
ing and compensating for parasitic thermocouples. The In an amplifier-drift test circuit, der Preussischen Akademic der Wis-
thermal mass at the amplifier inputs should be equal to thermal EMFs and the thermal senschaften, 1822 to 1823, pg 265.
allow parasitic-thermocouple outputs to arrive matched in capacity at each input must be
phase and amplitude. similar for cancellation to occur. Author's bio graphy
Jim Williams is a staff scientist at Linear
Technology Corp (Milpitas, CA, www.
#1 COVERED
linear-tech.com), where he specializes in
analog-circuit and instrumentation design.
He has served in similar capacities at
1 V National Semiconductor, Arthur D Little,
and the Instrumentation Laboratory at
#1 UNCOVERED
the Massachusetts Institute of Technology
(Cambridge, MA), where he first encoun-
tered serious 1-ppm measurement using the
10 SEC Kelvin-Varley divider. A former student at
Figure 7 Wayne State University (Detroit), Williams
Slight air movement can affect the performance of a thermal baffle for a low-frequency amplifier. enjoys art, collecting antique scientific
In the top trace, a small cup covers the amplifier, and the amplifier in the bottom trace is uncov- instruments, and restoring old Tektronix
ered. The instability worsens if air movement increases. oscilloscopes.
86 edn | May 3, 2001 www.ednmag.com