OUTSTANDING DC GAIN
ACCURACY AND STABILITY
DEMONSTRATED IN A 100A MEASURING SYSTEM
The turn-on transient in the average dc gain is measured when 100A dc is suddenly
applied to the complete current measuring system described below. After a 12 mA/A downward movement in the current during
the first ten minutes there is a "noise like" variation of about ±3 mA/A from 10 minutes to 30 minutes. These
variations with time and temperature are ten to one hundred times less than those that can
be obtained using most "precision" 100 A shunts or most other commercial 100 A
Current Sources or Transconductance Amplifiers.
Figure 1. Movement of the relative error in the average dc gain
of a 100A system. Readings every 1.1 minutes. The "zero" point reading was
started 30 seconds after the 100A current was turned on.
SYSTEM DESCRIPTION
The 100A current is produced by a Fluke Model 5100B Calibrator as a ±1 V Voltage
Source which drives a Clarke-Hess Model 8100 Transconductance Amplifier operated on the
100A range. The output current is stepped down with a single turn input Danfysik Model
866-150 Zero-Flux, dc measuring, Current Transducer with an output of 133.3333 mA for a
100 A input. This current is sensed with a Clarke-Hess 7.500 W , very low power coefficient, burden resistor. The voltage across the
burden resistor is sensed with a hp3458A dc Voltmeter.
The system is operated via a computer that finds the average current output when the
system is driven alternately with a +1 V and then with a -1V input. Each plus and each
minus reading is itself the average of four separate readings. Two "dummy"
readings are taken before each plus or minus set of readings. The overall cycle time for
an output average value is about 66 seconds or 1.1 minute. When the measurement is started
all of the equipment has been on over twelve hours, however the Transconductance Amplifier
has been in STANDBY for this whole period. Before a measurement, and with the Amplifier
still in STANDBY, the system is operated through several cycles with the Calibrator
running straight into the meter. This determines the error in the average dc Calibrator
output. (The hp3458A is used as the REFERENCE instrument for the whole measurement
process.)
Then with measurement program set-up and ready to go and with the hp3458A input
connected across the 7.5 W
resistor the Transconductance Amplifier is switched ON so that the 100A flows through the
Danfysik primary cable and 133.33 mA flows through the burden resistor. Readings are then
available every 1.1 minute and the program continues either until stopped or for a
predetermined number of cycles.
ANALYSIS OF RESULTS
In addition to the variation in the system gain, with respect to the first reading the
system also produces absolute average dc gains, and dc offsets. At the end of a 30 minute
period the absolute dc current error was measured at -10 mA/A. As outlined below there is a possible uncertainty of 10-15 mA/A in this absolute measurement.
AVERAGE DC GAIN
Possible contributors to dc gain errors include the input dc Calibrator, the
Transconductance Amplifier, the Current Transducer, the Current Transducer Burden
Resistor, and the output Voltmeter. The Voltmeter was verified by Hewlett Packard within
twenty days before the measurement. At that time the +1V dc error was found to be -1.7 mV and the -1V dc error was found to be +4.4 mV
for an average error of +1.4 mV. This Voltmeter error is
ignored for the rest of this discussion, that is the hp3458A is assumed to be perfect. The
average Calibrator voltage error, at the one volt level, was measured as -34 mV with a spread among several readings of ±0.2 mV.
The specified ratio error for the Current Transducer is 1 mA/A.
The hp3458A measured resistance error at the 10 W level, when verified by Hewlett-Packard, was less than 0.1 mW. When this meter was used to measure the
Clarke-Hess nominal 7.5 W the
measured value was found to be about 6 mW/W low. The
measured error for the 100 A current out of the system at the end of the 30 minute period
was about -44 mA/A. If 34 mA/A were contributed by the input voltage
source and 6 mA/A were contributed
by the transducer burden resistor then the Transconductance Amplifier gain would appear to
have been set to within ±5 mA/A
of the expected value.(The actual resolution of the Front Panel Gain adjustment for each
range of the Model 8100 is only 10 mA/A.)
The Transducer burden resistor must dissipate 133 mW when carrying 133.33 mA hence for
stable outputs its power coefficient must be less than 7.5 ppM/W. This value is much
smaller than many so called "standard resistors". The achievement of the desired
stability in the Clarke-Hess burden resistor is shown by heating up the resistor with a
150 mA current and then cooling it with a large volume of moving air. The output voltage
movement from this exercise is generally less than 1-2 ppM.
DC OFFSET
The variation of the dc offset in the output current over the complete 30 minute period
(first reading to last reading) was from +7 mA/A to + 21 mA/A. The 7 occurred on the first reading. If the first term is
ignored then the total dc offset spread was from +11 mA/A to
+21 mA/A with a slight upward drift over the 30 minute period.
(While the short term dc offset of the system is very small and apparently almost
independent of short term internal temperature effects there are longer term variations in
some part of the system that are not yet fully understood.)
Possible sources of dc offset, in addition to the Transconductance Amplifier, include
the input Voltage Source and the Current Transducer and its electronics. In this case the
input source was measured as having a dc offset of about -5 mV/V . The Current Transducer has a specified initial dc offset of less
than ±8 mA. On the basis of a
133.333 mA output this amounts to ±60 mA/A. Measured values - in the past, not at the time of this measurement -
have shown values below ±40 mA/A
for this device. The generally very conservative specification for the Model 8100 dc
offset is ±500 mA/A of the range.
The Model 8100 has Front Panel digital control capability to set the dc offset on each
range. No special adjustments were made on the production line instrument (Serial Number
153) used to produce this test data.
CONCLUSIONS
A 100A current measuring system has been demonstrated with overall gain stability, from
a "cold" start to on 30 minutes with 100 A flowing, of better than 20 mA/A. Absolute average dc gain error and dc
offsets for the whole system are also measured and found to be very low. The existence of
such a commercially available system allows the easy measurement or verification of other
components such as shunts or current transducers or current sources or amplifiers by
simply substituting the new device to be measured for its counterpart in the
"good" system. This procedure is followed in testing all Clarke-Hess Model 8100
Transconductance Amplifiers. The results shown in Figure 1. are typical of every Model
8100 in production. |
