The "Chopper-Comparator" or "Null-Type" Readout System

Definition

In this system, a "null" means the absence of an audible tone. The tone comes from a fixed-frequency oscillator that operates a switch; this switch "chops" a difference signal. An audio amplifier presents the off-null difference signal to the user through a loudspeaker. In operation, the user makes an adjustment until the tone fades to the inaudible (null) condition.

An Illustrative Anecdote from an Airline's Maintenance Shop

The RERC was asked to design an equivalent accessible pressure gauge with which a blind employee could pressure test seals in small turbines. The task was to pressurize a bearing with 300 PSI of air, then watch the pressure drop as air leaked through the bearing.

A null-type audible gauge was built and connected to the air system along with the original visual gauge -- a very large analog visual gauge which could be glanced at from a distance while other projects were being worked on.

Some time after the blind mechanic had moved on to perform other functions, our audible gauge failed. Their instrument shop was asked to repair it; it seemed that the audible sound was seen as indispensable to sighted workers who, when hearing it, were reminded to attend to the visual gauge.

Current Importance of this Configuration

Although this system is the oldest, its importance goes far beyond historical significance. It is the blind person's galvanometer, and it is many times more durable, and repeatably more sensitive than the visual counterpart. It is an audible galvanometer which detects "balance" of our resistance bridges. Coupled with a Braille-calibrated voltage standard, it was the basis of auditory meters clear back to days of early experimentation.

It is an understatement to declare that blind technicians use electrical measuring instruments more than their sighted counterparts. Examples follow:

Rather than looking at colored bands and tiny numerals on the bodies of components, we "read component 'labels'" (identifying characteristics and values) with resistance, capacitance, and inductance meters.
We identify active components with so-called "black-box" tests.
We trace circuits with test meters (tasks otherwise performed visually by using the colors of wiring).
We assess the quality of our soldering by listening for rapid changes in our auditory analog-readout devices.
If a known reading is to be attained by "fine tuning", an audible analog indicator that shows, "Oops, you passed it"; "You have arrived," is the proper tool; this may prove to be more efficient than a host of visual equivalents.

Flirtation with speech-output meters leads to a "right tool for the right job" discussion:

A talking capacitance meter may speak the value of a capacitor, but the "hum" of an incomplete null of my capacitance bridge tells me if a capacitor is leaky.
Quick comparison of voltages or resistances allows me to trace circuits efficiently; hoping to find comparable voltages or conductance by listening to a succession of announcements is frustrating.
Chasing intermittent connections with untimely periodic announcements is impractical; thus, barring good visual inspection, getting something working that may have only minor problems is often out of reach using talking meters.
Finely tuning an adjustment to a desired quantity, while listening to announcements of your progress, is downright tedious.

To restate the premise: blind technicians have expansive uses for our test instruments -- uses which substitute fact-finding for visual inspection.

The "null-type" readout affords accurate quantified readings.
Its output is intuitively obvious; noting or remembering the pitch of a VCO is not an issue.
The noticeability of the off-null condition could be set by an optional volume control on the audio amplifier; the user may change the loudness (hence, the sensitivity to the off-null tone) in accordance with the task at hand.
Once adjusted for quiescence, the null-type meter patiently waits for a transient or change in reading to occur; the user is promptly alerted, even from across the room.
Because the user can hear activity (sometimes considered to be "noise") around the null, a null-type readout can serve as a tool to enhance the use of a simple talking meter. The talking meter may take, then store for presentation, an instantaneous reading; therefore, successive readings may change, one from the next, as these instantaneous samples are taken. Where a potentiometer is available to seek a null on a chopper-comparator system, the technician can find an "average" by listening to the activity at the null, then measure the pure DC voltage from the pot with the talking meter.

Of Historical Interest

The earliest of this technique employed mechanical contacts in series with high-impedance earphones as the "comparator". A general-purpose meter reader, soon followed by a multimeter, used war-surplus mechanical vibrators; the so-called "synchronous vibrator" had an isolated pair of contacts which could readily serve as the chopper.

Various solid-state choppers were on the horizon, desirable because the mechanical vibrators presented reliability problems.

A series of volt/ohm/milliamp meters was marketed by The American Foundation for the Blind. Robert Gunderson of the New York School for the Blind designed and made the mechanical ones. A water-shed improvement came with a design by David Hevner, a protege of Gunderson's.

Hevner's design used two JFET's, alternatively controlled by an a-stable multivibrator; the result was to achieve a very, very quiet null. Moreover, a high amplification could now be tolerated, leading to a more accurate, higher-impedance multimeter.

In the first decade of the Smith-Kettlewell RERC, many adaptive devices were created using the David Hevner chopper-comparator, with the only difference being use of an integrated-circuit audio amplifier. Since 1980, most of our null-type instruments have used an opto-isolated photo FET, operated by a 555 oscillator. Even though this is a single-ended chopper (a single-pole single-throw switch), isolation of the FET affords a pretty quiet null. But hats off to Mr. Hevner; his single-pole double-throw JFET solution should remain extant as an option.

Technical Notes

Two circuit diagrams are given here; the first is the Hevner JFET SPDT chopper, while the second is our opto-isolator SPST chopper. In addition, preliminary experiments showed good results from an analog switch chip, the 4053 triple SPDT IC.

These chopper-comparator systems are high-resistance DC voltage comparators at their null point. Off null, they have an admittance that increases as a function of the difference voltage. With the aforementioned vibrator in series with headphones, the off-null condition caused an interrupted current to flow through the windings of the phones. Our SPST chopper imposes the difference voltage across a 47,000 ohm resistor. The modified Hevner circuit shown here presents the DC difference with a chopped capacitive impedance of that amplifier.

It can be argued, however, that the impedance of all these schemes is very, very, very high when there is no difference voltage -- when a null is attained. (T.V. Cranmer published a two-range voltmeter for testing transistor circuits in which no multiplier network was used, just selectable voltages applied to the Braille-calibrated pot. He named it "The Cranmer Volt Balance", and without the usual voltmeter input network, it had no loading effect when adjusted for a null.)

Another feature is that there are no op-amps with input currents and internal offset voltages to compromise accuracy of DC readings. The limiting factors on measurement accuracy are:

The stability of the voltage standard supplying a calibrated pot could be at issue, although the settable range makes fine calibration possible.
Traditionally, wire-wound pots are used because of their linearity. Their failing is that the coarseness of their windings can prevent the user from adjusting for a perfect null.
Because the "impedance" of any such comparator is high at the null point, there is often hum from stray fields that muddies the null, making it hard to find with precision.
Choppers using an SPST switch leave themselves open to noise during portions of the cycle when the switch is open.
These readouts work best where the DC signal -- either a voltage balanced against a pot, or in a Wheatstone bridge -- is free of noise. For example, a pressure sensor with internal amplification may introduce a hiss of "pink noise" which becomes significant at the null.

In Summary

Chopper-comparators compare DC voltages. Unlike visual galvanometers, they are higher impedance, and there is no friction to overcome for small deviations. Galvanometers have a constant resistance, no matter how far they are deflected; these null-type circuits have a progressive effect for departure from the null.

As a final word of praise:

The rated accuracy of d'Arsenval visual meters is with respect to a full-scale reading; that is, a 2% accuracy rating of a 100 milliammeter becomes as bad as +-20% at 10 milliamps. In contrast, specified linearity of a calibrated pot is consistent throughout; use of a pot with 1% linearity to read 100 milliamps means an accuracy of 1% anywhere within range.

When making a "Braille" scale for a potentiometer, it must be remembered that mechanical extremes of the pot's rotation are often slightly beyond ends of the resistance element; most high-quality pots specify "electrical rotation", which is less than that from one stop to the other. Moreover, so-called "360 degree pots" usually have a small dead spot between end points; this exists to prevent burning out the resistance element should the wiper be allowed to dally between the end points.

We have adopted the practice of custom-making each scale for its pot. This partially corrects for imprecision in linearity of the pot, and end points of the resistance element are easy to find before "zero" and "full-scale" marks are made. Such a custom scale can be generated by connecting a digital meter to the pot, then fixing the scale blank to its shaft and sandwiching the blank in a Braille slate.

"Braille" on a circular dial is a misnomer; actual characters of the Braille system depend upon their orientation and cannot be rotated. Braille watches and dials use 1- 2- and 3-dot markings to serve as "tick marks" around the circle. For example, the graduations on a typical face of a "Braille watch" are double-dot marks at 3:00, 6:00 and 9:00. Single dot marks are typically used at in-between hours, and a 3-dot mark is often placed at 12:00. A mistake sometimes made on these watch dials is to place the outside of these tick marks on a constant radius; this means that the hands are farther away from single dots than they are from major markings. Greater accuracy in reading is gotten when all marks are on a radius which is just beyond, not under, the end of the pointer.

Most of our meters that are built from scratch have ten-division dials (ten major divisions with single dots sparingly placed in between). If a device needs to have a resolution which demands 200 graduations, a ten-turn pot with a crank that can afford easy counting of turns is most appropriate, rather than making the user tactually read a hopelessly crowded scale. (This is how we designed machinists' gauges where fractions of 1/10,000 were to be read.)

Scales related by factors of 10 may not be appropriate. A meter to be adapted could have a 0-to-180 milliamp scale, for example; a 10-division scale with graduations every 18 milliamps is hard to interpret. Zero plus 12 major graduations reduces the burden of arithmetic; this would put major marks at 15, 30, 45, 60, etc. Placing marks every 5 milliamps -- single dots at 20 and 25 milliamps between double-dots -- is a sensible compromise.

Attention must be paid to minimize hum pickup which would partially obscure the null. Shielded cabling should always be used for connection to outside equipment. As much as possible, separation of power transformers from the null-detector can reduce hum. Typical modifications to VTVM's include 10 microfarads placed across the adapter's input.

There are cases where neither meter terminal is common to a visual instrument's circuit ground -- VTVM'S must be accommodated for this possibility. In such a case, the readout needs its own isolated power supply. If the adapter is housed in a separate cabinet, a cable with two wires plus the shield is required for connecting to off-ground meter terminals; the cabinets must then be connected together, perhaps even by a fourth dedicated conductor.

Of course, where a visual device is to have its meter removed, perhaps to build the audible readout into its cabinet, a resistor of value equivalent to the meter's resistance must be installed. It is dangerously reckless to attempt measuring the meter's resistance with an ohmmeter; the meter might burn out before you get the answer. A standard procedure using two rheostats is recommended:

With it set to its maximum resistance, connect a 100,000 ohm rheostat in series with the meter movement, connecting this series combination to a 9-volt battery. Decrease the rheostat until the meter reads full scale. Next, place a 5,000 ohm rheostat across the meter's terminals and adjust this so that the meter reads half-scale. At this point, the latter rheostat has been set to the same ohmic value of the meter's internal resistance; the rheostat can safely be read with an ohmmeter of any sort.

The substitute resistor may ride on board of the readout, along with an appropriate filter capacitor as required.

Circuit Considerations

Duplication of the Hevner system is still possible; P-channel JFET's are still available, namely 2N5460, 2N5461, and 2N5462 type numbers. Recent experiments with enhancement-mode FET's proved discouraging; protection diodes within 2N7000's prevented their "sources" from being brought higher than their gates.

Examination of the Hevner JFET chopper will reveal a 0.022 microfarad cap between the bases of the multivibrator. This was done to take the "edge" off the waveform, thus giving the instrument a pleasant tone.

The "pinch-off" gate voltage ratings vary widely between JFET units. If luck is against choosing a reasonable match of the two JFET's, a pitch of double the frequency may be heard at the null. Other than changing one or both of the JFET's, some improvement might be had by connecting a 500K rheostat between the gates in order to force their cooperation as an SPDT switch.

Good results were obtained by using one section of a 4053 16-pin analog switch:

Pin 16 goes to VCC. Pins 6, 7 and 8 ("inhibit", VEE, and VSS, respectively) are grounded. Pin 11, the "control pin", goes to the oscillator; this could be to one collector of the multivibrator, or to pin 3 of the 555 oscillator shown in our opto-isolator diagram.

Pin 14, the "swinger" of one switch goes to the amplifier (including the filter show in the Hevner diagram). Pin 12 goes to the positive terminal of the meter; pin 13 goes to the arm of the calibrated pot.

The 2K rheostat associated with pin 1 of the LM386 is a "gain adjustment"; it has limited range. A volume control can be installed at pin 3 of the LM386 audio amplifier. Make this a high-resistance pot -- perhaps 100K or greater -- so as to preserve the high-impedance of the chopper for the off-null condition.

An embellishment first used on our machinists' gauges was to use op-amps sensing the setting of the calibrated pot to pull the frequency of the 555 oscillator. In this way, the tone would pull higher and higher for one off-null polarity, and drop lower and lower for the other off-null polarity.