Report of Tests on Joseph Newman's Device

4. Consistency Checks

BACK to Contents

4.1 Introduction

In the development of measurement techniques, it is necessary to provide certain checks to verify that unanticipated errors have not corrupted the data. For conventional measurement systems, an experienced metrologist will identify any deviations from the normal behavior for a particular class of device. For one-of-a-kind devices, it is customary to make some explicit checks to provide additional assurance that the measurement results are correct within the stated uncertainties. The following paragraphs describe such consistency checks for measurements made on the Newman device.

4.2 Assessment of the Frequency Dependence of the Input Power

As mentioned previously, the input power was measured using a low-pass filter on the input current channel. To determine the effect of the filter, matched sets of filters were constructed. One set was a pair of high-pass filters; the other set was a pair of low-pass filters. Both sets had 3-dB points (50% reduction) at 8000 hertz. Thus if one were measuring the power using the low-pass filters, one would be measuring approximately the power at frequencies below 8000 hertz and when using the high-pass filters, the power above 8000 hertz. The results of such measurements are summarized in Table 2.

Table 2

Effect of high-pass and low-pass filters on input power measurements using the sampling wattmeter on various ranges for the current channel

FILTER (Arbitrary Units) (Watts)

High Pass 10 66 x 10-3
  20 84 x 10-3
  50 108 x 10-3
  100 110 x 10-3

Low Pass 20 3.7
  50 3.9

Two observations can be drawn from these data. First, the power measured using the high-pass filters is less than 3% of the power measured using the low-pass filters. This result shows that the measured efficiency should be less than 3% larger than the true efficiency of the device under test. Second, the measured power using the high-pass filters increased as the current range increased. This trend suggests that, in this frequency range, there are high amplitude signals which saturate the input circuit and produce measurement errors at the lower ranges. It is not feasible to measure power using the higher current ranges because most of the signal is at lower


amplitude and large measurement errors can occur if low amplitude signals are measured using a high amplitude range. Thus, to minimize and to control measurement error, the data were taken using a low-pass filter.

4.3 Approximate Determination of the Input Power

Because the battery voltage is nearly constant (implying that the average voltage is nearly equal to the rms voltage), the product of the average current and the voltage should nearly equal the input power. To test this relationship, measurements were made using the sampling wattmeter to determine the input power, the rms voltage, and the average current. The results, which show that the product of the voltage and the current approximates the measured power, are summarized in Table 3.

Table 3

Comparison of power measurement to the current-voltage product in the determination of the input power

(Arbitrary Units) (Arbitrary Units) (Volts) (Amperes) (Watts) (Volt-Amperes)

2 2 806 5.2 x 10-3 3.9 4.2
2 5 805 5.3 x 10-3 4.1 4.3
2 10 802 5.3 x 10-3 4.3 4.3
2 20 801 5.6 x 10-3 4.4 4.5

4.4 Measurement of the Input and the Output Power Using the Sampling Wattmeter

In addition to the various tests, calibrations, and consistency checks performed, it is good practice to demonstrate that no systematic offsets occurred between the instrumentation used to measure the input power and that used to measure the output power. Measurements, therefore, were made to identify any significant offset. These consisted of input measurements using the sampling wattmeter and the analog-multiplier wattmeter followed by measurements of the output using the active attenuator. The signal from the attenuator was measured using the digital voltmeter (DVM) which was used in subsequent measurements and also by the sampling wattmeter. These measurements, summarized in Table 4, demonstrate that the various instruments gave about the same indication under the same conditions.

Table 4

Measurement of the input and the output using the sampling wattmeter



(Ohms) Analog Multiplier Sampling Wattmeter DVM Sampling Wattmeter

oo 3.5 3.5 - 4.2 -- --
500,000 -- -- 0.7 0.7
50,000 8.2 7.7 4.2 4.2


4.5 Spectrum Analysis

A commercially available spectrum analyzer was used to measure the frequency content of the output voltage. This measurement was made using an NBS-developed divider which had a submicrosecond response time. The divider was attached to one terminal of the coil and the spectrum of the divider output voltage was measured. This measurement showed that most of the energy in the signal was in the frequency region below 50 hertz. At 500 hertz, the signal level was 1% of the low-frequency level.

These results were not used directly in the analysis of the data, but taken in conjunction with the other measurement results help to corroborate the observation that the power flow in the device is primarily a low frequency phenomenon.

4.6 Direction of Power Flow

The sign of the measured input power indicated that the battery pack was supplying net power to the device or, equivalently, that the device was not charging the battery pack. That is to say, power was flowing from the battery to the device-under-test and not from the device-under-test to the battery. To demonstrate that the signs were correct, a nominal 1-megohm resistor was placed in parallel with the device-under-test and the appropriate commutator brush was lifted so that no power was able to flow into the device. In this configuration the power was clearly flowing from the battery to the resistor and the measured polarity of the power flow was the same as it was when the battery powered the device-under-test.

4.7 In Situ Measurement of the Output from a Signal Generator

To assure that there were no unanticipated errors due to the arrangement and interconnection of the various measurement systems in the laboratory, an in situ calibration check was performed. All of the measurement systems were connected with cable runs in the same position used for measurement, except that no connections were made to the device-under-test. A signal generator was used to apply a signal to a parallel combination of the voltage divider with the current shunt, a shunt used with the thermal element and a 200,000-ohm load resistor, all connected in series. The power delivered to this combination was measured using the sampling wattmeter, the analog wattmeter, the thermal element, and the voltmeter used to measure the output from the active attenuator. Measurements using these systems differed by less than 1%.


BACK to Contents