KineticSystems ::. Data Acquisition (DAQ)

Technical Notes

Driving Balanced Analog Inputs from Unbalanced Sources

Abstract - Most low to medium-speed, analog-to-digital channels include a balanced input in order to provide some degree of rejection of common-mode voltage. Unfortunately, most signal sources are unbalanced with one side grounded. The two most-often-asked questions regarding the connection of unbalanced sources such as transducer amplifiers to balanced inputs such as A/D converters are:

Should coaxial or balanced-pair cable be used?
Should the cable be grounded at one end or at both ends?

This technical note shows test results that indicate the optimum cable type (shielded balanced pair) as well as the best grounding practice (generally, grounding at both ends).

I. THE TEST SET-UP

The test set-up is shown in Figure 1 below.

Figure 1. Noise test setup with simulated unbalanced source.

* 10 ohm resistor simulates low-impedance output of typical transducer amplifier.
** Wideband Instrumentation Amplifier (Gain = 1).

The output impedance of a typical amplifier is generally low (100 ohms down to a fraction of an ohm). This is simulated with a 10 ohm resistor. Since no actual signal is present, any resulting voltage would be noise. The test includes 50-foot cables between the simulated signal source and the balanced input. A wideband instrumentation amplifier represents the balanced input. This amplifier is monitored by a wideband oscilloscope. The noise environment is that of a typical development laboratory. The electrical conduit grounds for the source and loads are approximately 50 feet from each other. The coaxial cable to the oscilloscope is quite short; the instrumentation amplifier and oscilloscope share the same receptacle ground to limit noise generated from that connection.

II. COAXIAL VS. BALANCED CABLE

The noise performance for coaxial as well as shielded balanced-pair cable is shown in Figure 2. Note that Figure 2(a) shows approximately ±500 mV of high-frequency noise present when coaxial cable is used and that cable is grounded only at the source end. Figure 2(b) shows the effect of grounding the coaxial cable at both ends. The high-frequency "hash" is replaced by a ±100 mV signal that is predominately 60 Hz and its harmonics. This is expected because any potential difference between the two conduit systems results in a voltage being developed across the cable shield that is also the signal return for the analog system. Figure 2(c) shows the resulting interference when a balanced-pair cable is used and it is grounded at both ends. A later measurement with higher resolution on the oscilloscope [ Figure 4(a) ] shows the noise level to be approximately ±3 mV. This is an improvement of 167:1 (or 45 dB) over Figure 2(a) .

	Figure 2(a). Coaxial cable, ground at source only

	* 100 mV/div; 5ms/div

	Figure 2(b). Coaxial cable, ground at source & load

	* 100 mV/div; 5ms/div

	Figure 2(c). Shielded balanced pair, ground at both ends

	* 100 mV/div; 5ms/div

III. GROUNDING BALANCED-PAIR CABLES

Because the grounding of coaxial cables results in shield- conducted power line noise, one is often reluctant to ground both ends of a balanced-pair or twinaxial cable. However, the balanced-pair cable should generally be grounded at both ends to minimize noise pickup. The only exception to this is when one end is "floating" (not connected to ground); this situation will be dealt with later in this note.

The shield of a balanced-pair cable can be grounded at both ends because, in a balanced system, the shield is not a signal-carrying conductor for normal-mode signals as it is in a coaxial cable. The results of grounding a balanced-pair cable at one end only are shown in Figure 3. Figure 3(a) shows the shield connected at the load only, while Figure 3(b) shows the shield connection at the source only. Both of these approaches give unacceptable results. The noise level is nearly as high as that found with the coaxial cable in Figure 2(a) .

Figure 3(a). No shield connection at source

Test Setup

* 100 mV/div; 5ms/div

Figure 3(b). No shield connection at load

* 100 mV/div; 5ms/div

IV. GROUNDED OR FLOATING SOURCES/LOADS

Thus far we have described the grounding recommendations and cable type when the source is grounded. Figure 4(a) uses the same set-up as was shown in Figure 2(c) : balanced-pair cable with the shield grounded at both ends. The oscilloscope is now set to a higher sensitivity (5 mV/division compared with 100 mV/division in Figure 2). If the signal source is "floating" (isolated from ground) as in the case of a passive source (a thermocouple without amplifier, for example) or an amplified transducer that is powered from a ground-isolated source, then the connection should be as shown in Figure 4(b) . The source-end shield is connected to the "common" of the source, not ground. Note that the "floating" source gives a noise level of ±1 mV, rather than the ±3 mV for the grounded source in Figure 4(a) .

The signal conductors associated with the balanced input must always have some path to ground in order to supply the low bias current needed by the instrumentation amplifier. In the case where a "floating" source is used, this return path is via the cable shield and through the load-end ground connection.

If the load is isolated instead of the source (when using an isolated instrumentation amplifier, for example), then the source-end shield should be connected to ground, while the load-end shield is connected to the "common" of the load circuit. If both source and load are "floating," then one end should be connected to ground in addition to the shield connection to both circuit "commons." This is primarily for safety so that capacitive coupling to a high-voltage conductor in the same cable tray will not cause the analog cable to become charged.

Figure 4(a). Grounded source

Test Setup

* 5 mV/div; 5ms/div

Figure 4(b). Floating source

* 5 mV/div; 5ms/div

V. FILTERING THE SIGNAL

Another way to reduce the noise into a balanced circuit is to limit the bandwidth of the signal presented to the instrumentation amplifier. The cutoff frequency that can be used depends upon the signal rate-of-change to be monitored. Figure 5 shows the effect of adding an 8 kHz, single- pole passive low-pass filter. The components for this filter with a 3 dB loss at 8 kHz are shown in Figure 7 . Note in Figure 5(a) that this filter reduced the noise from ±500 mV to ±3 mV for the coaxial cable case. The results are even better when combining the balanced-pair cable grounded at both ends with the filter [ Figure 5(b) ]. The noise in this case is reduced to ±500 µV. The filter should be located on the module or within the same rack (at the termination panel, for example) so that high-frequency noise is not picked up after the filter.

Reducing the cut-off frequency will decrease the noise further as well as to limit the system frequency response. For example, an option on the KSC 3512 A/D module contains input filters with a 10 Hz, single-pole rolloff to attenuate 50 or 60 Hz normal-mode noise. Similarly, the KSC V213 module supports 2-pole input filters. The 3527 A/D module integrates many readings over a one-cycle period of power-line frequency to attenuate the effects of noise from that source. Other noise will also be reduced by this averaging technique.

Figure 5(a). Coaxial cable, ground at source only

Test Setup

* 5 mV/div; 5ms/div

Figure 5(b). Shielded balanced pair, ground at both ends

* 5 mV/div; 5ms/div

VI. ACTUAL SIGNAL DISTRIBUTION DEMONSTRATED

Thus far we have shown the noise response by using a resistor to simulate the signal source. We then used a Hewlett-Packard oscillator producing a 1000 Hz signal with an amplitude of 1.5 V peak-to-peak as the signal source and transmitted this signal over the 50-foot cable. The results are shown in Figure 6. Note the high-frequency "hash" with the coaxial cable in Figure 6(a) and the "clean" signal in Figure 6(b) .

	Figure 6(a). Coaxial cable, ground at source only

	* 500 mV/div; 200 us/div

	Figure 6(b). Shielded balanced pair, ground at both ends

	* 500 mV/div; 200 us/div

VII. OTHER CONSIDERATIONS

The level of noise, even when using balanced-pair cable, is affected by the exposure and the relative noise potential between the transmitting and receiving ends of the analog path. When analog signals are transmitted between racks that are in close proximity, noise performance is generally improved by bonding the racks together with one or more large conductors. The important consideration is not how "clean" a ground is, but rather how much "noise" is present when measured from one part of the system to another part of that same system. This is the same effect that causes us not to feel that the earth's surface is rotating at nearly 1,000 miles/hour because all parts that we have contact with are rotating together.

The cable type and grounding methods shown here also apply to balanced-line binary signaling such as that provided with RS-422. The EIA RS-422 specification indicates that the shield should be grounded at both ends.

Figure 7. Details of 8 kHz lowpass filter

Test Setup

VIII. CONCLUSIONS

When receiving balanced analog signals, the following conclusions can be drawn:

Shielded balanced-pair cable should be used with both ends grounded when the source and load are ground- referenced [ Figure 4(a) ].
If the source is isolated, then the source-end shield should be connected to the "common" of the source circuit [ Figure 4(b) ].
If the load is isolated, then the load-end shield should be connected to the "common" of the load circuit.
A filter can be added to reduce the noise bandwidth of the input signal.

Author Biography

Bob Cleary co-founded KineticSystems Corporation in 1970 and is its Chief Executive Officer. He has developed many of the high performance CAMAC and VXIbus data acquisition and control products in the KineticSystems product line. He is a co-author of the book, "High Performance Data Acquisition and Control," has given many seminars on VXIbus and has presented numerous technical papers. Mr. Cleary has been granted 18 U.S. and 24 foreign patents.