The Smith-Kettlewell Technical File

A Quarterly Publication of
The Smith-Kettlewell Eye Research Institute’s
Rehabilitation Engineering Research Center

William Gerrey, Editor

Issue: sktf-Spring-1993

Original support provided by:
The Smith-Kettlewell Eye Research Institute
and the National Institute on Disability and Rehabilitation Research

Note: This archive is provided as a historical resource. Details regarding products, suppliers, and other contact information are original and may be outdated.

Questions about this archive can be sent to







Frictionless position sensors comprise a class of mechanical sensors. For years we have searched for low-cost inclinometers (tilt sensors) to make sensitive carpenter's levels. There are linear position sensors as well as rotary pots. Seventeen years ago, I made gauging instruments for machinists out of those (although I used LVDT's), and since then I have adapted a refrigerant scale that used such stuff.

Midori has a new kind of frictionless "pot" which uses "magneto-resistive semiconductors" -- a permanent magnet being carried by the shaft. This survey includes the basic pots (with shafts) -- spring-loaded plunger-type devices which could be used as "dial gauges," and inclinometers.


No mechanical assembly is frictionless; bearings all rub and bind somewhat. What can be said is that there are position-sensing techniques which do not contribute friction to the system.

A standard potentiometer has three components of friction: the wiper rubs against a resistance element, some sort of "slip ring" connects the wiper to a terminal on the device, and stress on the shaft causes bearing friction. Without a wiper and without a slip ring, two components are eliminated, and bearing friction is reduced, since spring-loaded attachments are not contributing to stress on the bearing.

[There are two competing techniques which will find their way into this magazine: One is optical encoding; interrupters or reflectance pickups sense lines on a wheel or linear strip, and these operate a counter. The other technology uses transformers -- the "linear variable differential transformer (LVDT) and the rotary variable differential transformer (RVDT). The transformer types pass a core through a system of coils -- one in the center being driven and two others feeding a differential detector. LVDT's are very linear over a specified range, with errors less than 0.2% being achievable. Optical encoders have coarse resolution and would not do for a carpenter's level. RVDT's are hundreds of dollars and their associated circuitry is complex.]

Midori makes a selection of conventional conductive-plastic pots; they call these their "green pots." They speak of their magneto-resistive products as "potentiometric devices," and they dub those as their "blue pots."

The "blue pots" contain two strips of semiconductor, each bent in a semicircle and positioned opposite one another. They have one common connection (the output), while the free ends go to plus and minus supply terminals. The shaft carries a semicircular permanent magnet (they don't show where the poles of the magnet are). Presumably, the magnetic field shoves charge carriers over to one edge of the strip, thus changing the current flow (similar to an electric field restricting the current flow in the "channel" of an FET, I suppose).

In most cases, the shafts are free to turn 360 degrees. A graph of the output versus shaft position shows a curve that is somewhere between a triangle wave and a sinusoid. The negative and positive peaks are rounded off, but the lines going through 1/2 VCC are quite linear -- good to within 2% of the recommended range (for the unit plotted). The peaks of the wave are typically at 32% and 68% of VCC.

The Midori literature shows a comparison between conductive-plastic (green) and "blue pots." The resolution of blue and green can be as minute as 0.0001%. As for linearity, conductive plastic units can achieve tolerances from 0.025% to about 1%, whereas tolerances for blue pots can range from 0.6% to as high as 6%.

The key to good linearity is restricting the range. For many units, the recommended range is plus/minus 40 degrees of rotation, with the center position being the point where the output is one-half the supply voltage.

One more important difference between "blue pots" and regular pots is that their input impedance is no indicator as to how their output should be loaded. The specs are shy about telling you this, but the tilt sensor we are using specifies a minimum load resistance of 2.5 megohms, even though the input impedance is about 7K. Therefore, always run the output into a follower; use an op-amp with output tied to inverting input, with the non-inverting input going to the output of the "blue pot."

Potentiometric Devices

The editor has chosen only a few of the Midori line to feature here. I was particularly interested in those with quarter-inch (6mm) shafts and convenient mounting schemes. They have one with a lever coming out the side -- still a rotary device -- and there are many which are said to be "servo mount" that require a machined cavity to put them in. If you want to know of odd-ball configurations, their catalog is worth having.

All their products list a rated insulation resistance of 100 megohms at 500V DC. They specify a dielectric strength that can withstand 500V AC for one minute.


The body of this unit has a diameter of 23mm, but there is a 38mm diameter flange with three screw holes used for mounting. Its shaft diameter is 6mm and it is 27mm long.

The mechanical travel is 360dg. Maximum torque is 0.5 gram-centimeters (gcm). The maximum radial play is 0.02mm, and the maximum shaft end play is 0.02mm.

The maximum input voltage is 10V; the input impedance is 15K, plus/minus 30%. The output sensitivity is greater than 2.5% of the input voltage per 10dg of rotation. The effective linear range is plus/minus 45dg with an "independent linearity" (whatever that is) of plus/minus 1.5% of full scale. The operating temperature range is minus 40dg C to plus 125dg C.


The specs are the same as above. This most basic one has a 3mm diameter shaft, and mounts with a 9.2mm diameter bushing. The body is 22mm in diameter. Without a suffix, it has wire leads; the CP-2UT has terminals.


This has the so-called "signal-processing" circuitry built in. Its overall diameter is 37mm; it has two mounting ears. Its shaft is 6mm in diameter and 31mm long. Its effective linear range is plus/minus 40dg, and its independent linearity is 1.7%. It is amusing that its linearity is worse than the bare pot; it is probably more temperature stable, however, since it has a temperature sensor as part of its circuitry. Its operating temperature range is minus 40dg C to plus 100dg C.

The input voltage can be from 4.5V to 10.5V. The output voltage range is approximately from 10% to 90% of V-in. The load resistance must be greater than 100K.

The CP-2UK-R is specifically designed as a "swash-plate sensor" in hydraulic pumps. (I'd hold a contest for anyone who can tell me what a "swash-plate" is, but I wouldn't know who won.) It is water-proof; what this means to us is that the torque needed to operate the shaft is much higher (probably due to an O-ring at the bearing). This torque is 20gcm, unfortunately.

Gauging Heads

These sense linear motion. All but one have a plunger that moves in and out; the one of different design has a shaft, protruding through a slot, which slides sideways.

These are particularly suitable for machine indicators. The disadvantage of these over LVDT's is that their linearity is inferior -- 1.5% compared to 0.2% attainable with the expensive LVDT counterpart. The repeatability and sensitivity of these are good, however, so they still have applications.


This unit is 22mm in diameter and 27mm long. It has two mounting ears.

Its mechanical stroke is plus/minus 5mm, being electrically useful over a 3mm range in the middle of this mechanical travel. The linearity is 1.5% of full scale. It sensitivity is 8% of the input voltage per millimeter of travel.

LP-UF-R Series

These contain electronics like the rotary one listed earlier, so they will be expensive. However, their plungers have return springs, so they are no doubt intended for machinist's gauges.

The mechanical travel on these is long -- 12mm, 22mm, or 33mm; the electrical ranges are 10mm, 20mm or 28mm, respectively. Their linearity is 2% of full scale.

The bodies are 22mm in diameter and 60mm long. They have a mounting flange on the front.

The usable range of the unit is specified by a number in the designation; LP-10UF-R, LP-20UF-R, and LP-30UF-R (this last being good for a 28mm travel). Their input voltage range is 8V to 11V. The load resistance should be 100K or higher. The electrical range is from 10% to 90% over the specified travel of 10mm 20mm and 28mm.

LP-15UL Series

These are square and are mounted on their side; they have mounting ears on opposite corners, one on the front and one at the rear. Their cross section is 12mm by 12mm, and they are 39mm long.

Their mechanical stroke is 18mm; the effective electrical travel is 15mm. Their sensitivity is 1.6% of the input voltage per millimeter. Their linearity is plus/minus 1%.

Their input impedance is 8K, plus/minus 30%. The maximum DC input is 7V.

The LP-15UL-S is the most basic; it has a shaft coming out the end.

The LP-15UL-K has a shaft coming through a slot in the top; this moves sideways in the slot.

The LP-15UL-B has a shaft coming out the end; it is spring loaded, which makes it an ideal head for a machinist's "dial gauge."


While there are several configurations, there is a line that the editor likes very much -- the PMP-S series. The only disadvantage in this series is the very inconvenient mounting arrangement.

There are two 2.6mm tapped holes in the top of the unit, which is a small round can. Two things are wrong with their placement: First, the three terminal pins also emerge from the top, as if they intended this precision instrument to be suspended from a rickety circuit board. Second, the two screw holes are placed along the axis of tilt sensitivity instead of either side of it; thus, the screws would not prevent it from rocking in case the extruded end cap is not flat.

PMP-S Series

There are four units in the series: good for plus/minus 5dg of tilt, plus/minus 10dg, plus/minus 20dg, and plus/minus 30dg of tilt. A number in the suffix specifies the unit; for example, the carpenter's level in this issue uses the PMP-S10T -- good for plus/minus 10dg.

These units are 20mm in diameter and 40mm tall. Their linearity is 1% of full scale, with the exception of the 30dg unit whose linearity is 1.3%.

The "output sensitivities" are as follows: For the S5T, the output changes 1.5% of V-in per degree, plus/minus 0.3%. The output of the S10T changes 0.75% of V-in per degree, plus/minus 0.15%. For the S20T, the output changes 0.39% of V-in per degree, plus/minus 0.11%. The output of the S30T changes 0.22% of V-in per degree, plus/minus 0.07%.

They give a spec which they call the "tilt sensibility." It turns out to be 0.1% of the range; thus, the PMP-S10T has a "tilt sensibility" of 0.01dg maximum, and the PMP-S30T has a tilt sensibility of 0.03dg max.

Their response time is approximately 0.3 seconds. For damping, they are filled with a silicone oil having a kinematic viscosity of 20 centistokes. (Molasses has a viscosity of several-hundred centistokes.)

They have an input "resistance" of 7K. The maximum input voltage is 8V. The load resistance should be higher than 2.5 megohms.

The terminals are in a row. With them pointing upward and positioned to the right of the screw holes: the one toward you is ground, the middle is output, and the farthest is plus V-in. With the terminals to the right, the output voltage goes up when the sensor is tilted to the right.


Midori America Corporation, 2555 E. Chapman Ave., Suite 400, Fullerton, CA 92631; Phone: (714) 449-0997.



Commercial electronic levels come and go; it has been our misfortune to adapt one, only to see the maker disappear and render the design obsolete. We can be safe in saying that "tilt sensors" of various sorts will be available, and so we offer this design as a generalized solution to satisfy the need for an audible level. As described in the previous article, the Midori company has a class of tilt sensors that serves this purpose very well. The instrument described here is capable of repeatedly sensing a 0.004-inch incline in a 1-foot beam.


One could start completely from scratch and make a level from well-cured straight-grain cherry wood. This would require drilling a 20 millimeter hole -- about 35 millimeters deep; this hole would then accept the Midori tilt sensor.

Instead, we chose to add our apparatus to a commercial level. While Stanley makes many models, we found one among the bunch which is much more suitable than others. Our choice is the Stanley "Professional" (TM) 24-inch Heavy-Duty Model 42-240. (Other Stanley models have ridges and things that make mounting inconvenient.)

The Stanley unit is an aluminum I-beam with three visual levels mounted in it -- two for "plumb" (vertical measurement) and one for "level" (horizontal measurement). To serve as handles, there are two roughly oval-shaped cutouts in the I-beam; these are 4-1/4 inches long.

The ends of the cutouts are vertical, so we milled away a window which is about one inch wide, one side bordering on the end of the original cutout, taking away all the material to the bottom flange of the I-beam. In other words, we created a flat surface onto which the bottom of the tilt sensor could be cemented. (The mounting scheme provided by Midori is unwieldy; we used "Bondo" to glue the bottom of the sensor in place, and filled in the vertical cracks with silicone rubber.)

Next, we milled away some of the curvature of that same cutout so that the 9-volt battery could be nestled edgewise in place. Brackets to hold it in place will be left up to you.

While we used a milling machine to make these cuts, careful work with a file might be good enough. The calibration of the level does not depend on the sensor's absolute position; it is only important that its mounting be rigid, and that the incline of the sensor with respect to the bubble indicator put it near the middle of its range. A degree or two error in gluing the tilt sensor in place will make no difference.

Two Phillips screws hold the bubble levels in place; we removed one of the vertical vials, which left a round hole that is ideal for accommodating the loudspeaker. Finally, the 2- by 3-1/2-inch perforated board carrying the circuit was mounted over the other oval cutout, and the four bolts holding it also carry plastic cover plates on both sides of the board.

How It Works

When turned on, a tone is heard. The pitch of this tone changes as the level is rocked end for end. To one side of "level" (in the direction that causes the pitch to rise), the tone is "chopped" -- interrupted about ten times a second. On the other side of level, the tone is smooth -- uninterrupted. The incline of the surface is adjusted so that the audible output is just on the border between pulsating versus smooth; you may notice an intermittency of pulsations, since very small motions make a comparator in the circuit turn the pulsation oscillator on and off.

Circuit Operation

The Midori "tilt sensor," a PMP-S10T, which can be described as a potentiometer, is in parallel with a screwdriver-adjustable calibration pot; this comprises a wheatstone bridge. The output of the sensor, as stated by the specifications, changes 0.75% of the voltage across it per degree of tilt (with an overall accuracy of plus/minus 1% of full scale). Given this specification, and allowing for the possibility of error in the absolute position when the sensor is mounted, a calibration range of 5% was chosen -- determined by 100K resistors on either end of the 10K calibration pot.

The "readout" of this level is the "Fowle Gimmique" (SKTf, Summer 1982). This circuit has two parts: a comparator senses polarity of the minute signal of the wheatstone bridge at its zero crossing, and a VCO responds to the voltage at the output of the sensor.

Besides adjustment of the wheatstone bridge (which calibrates the angle at which the device is "level"), a 100K rheostat in the emitter of the 2N2222 off pin 2 of the LM358 determines the sensitivity of the VCO, and this is adjusted according to your preference. (Increasing the sensitivity of the VCO also raises its pitch throughout the sensor's output range. In order to offset this -- to bring the pitch down to a pleasant tone -- some of the charging current for the oscillator's capacitor is shunted away by a 910K resistor.)

Although not necessary to maintain calibration, the wheatstone bridge is supplied from a zener diode. This is done so that the pitch of the VCO will be the same throughout the life of the battery -- a useful accommodation for anyone with good pitch memory.


The negative side of the 9V battery is grounded; the positive terminal goes through an on/off switch to the VCC line. A 556 dual timer is powered from a decoupling point: pin 7 is grounded, while pins 4, 10 and 14 go through a 10-ohm 1/2-watt resistor to VCC. Pin 14 is bypassed to ground by the parallel combination of 0.1uF and 100uF (negative of the latter at ground).

The output of the VCO section, pin 5 of the 556, goes through a 47-ohm 1/2-watt resistor, then through a small loudspeaker to pin 14.

Pins 2 and 6 of the 556 are tied together and go through the parallel combination of 0.0047uF (Mylar) and 910K to ground. Pins 2 and 6 go through 22K to pin 1. Pin 1 goes through 22K to the collector of a 2N2907. The base of this transistor goes to both the base and collector of another 2N2907. The emitters of the 2907's are tied together and go to pin 9 of the 556 (the output of the pulsation oscillator).

The bases of both 2907's also go to the collector of a 2N2222. The emitter of the 2222 goes through 10K, then through a 100K rheostat to ground.

The base of the 2222 goes to pin 1 (an output) of an LM358 dual op-amp. Pin 2, the inverting input, goes to the 2222 emitter.

Pin 4 of this op-amp is grounded; pin 8 goes to VCC. Pin 7 of the 358, the output of the other section, goes through 22K to pins 8 and 12 of the 556, which are tied together. Pins 8 and 12 of the 556 also go to the positive end of a 2.7uF electrolytic capacitor, the negative end of which is grounded. Pins 8 and 12 also go through 10K to pin 13 of the 556.

On the LM358, pins 3 and 5, both non-inverting inputs, are tied together; this junction is the positive input to the readout system. Pin 6, an inverting input, is the negative input of the readout.

A 6.8V zener (1N754) has its anode grounded, while its cathode goes through 680 ohms to VCC. The cathode, the 6.8V point, goes through a 100K 1% resistor to the top of the calibration trimmer (10K 10-turn with screwdriver adjustment); the bottom of this trimmer goes through another 100K 1% resistor to ground.

The arm of the calibration trimmer goes to pin 6 of the LM358. A Midori PMP-S10T inclinometer has its negative terminal grounded, while its positive terminal goes to the cathode of the zener. The "output" terminal of the sensor goes to pins 3 and 5 of the 358.

Circuit for Employing Two Sensors

This allows you to construct an instrument for adjusting vertical studs, as well as using it for leveling horizontal surfaces. The simple on/off switch is replaced with a 3-pole double-throw switch with a center-off position. On pole A, positions 1 and 2 are jumpered together and go to the positive battery lead; the swinger goes to the VCC line. Two sensors and separate calibration adjustments are powered from an emitter follower connected as follows:

A 2N2222 has its collector going to VCC; its base goes to the cathode of the zener, which still has its 680-ohm resistor to VCC.

Two Midori sensors, one for "level" and one for "plumb," have their negative terminals grounded. Their positive supply terminals go to the emitter of this 2N2222. Two 10K ten-turn trim pots, one marked "plumb" and one marked "level" are connected in parallel. The junction of their clockwise ends goes through a 49.9K 1% resistor to the emitter of this 2222; the junction of their counterclockwise ends goes through another 49.9K 1% resistor to ground.

On pole B of the switch, position 1 goes to the arm of the trimmer marked "level"; position 2 goes to the arm of the trimmer marked "plumb." The swinger of this pole goes to pin 6 of the LM358.

The swinger of pole C on the switch goes to pins 3 and 5 of the LM358. The output of the "level sensor" goes to position 1 of this pole, while the output of the "plumb sensor" goes to position 2 of pole C.


We chose to trust the accuracy of Stanley's bubble as our standard. If you make your own beam, secure another level to it in some way, being sure that the edges of your beam are exactly parallel.

Two adjustments have to be made. The first is non-critical; adjust the 100K rheostat in the emitter of the 2N2222 (the one off pin 2 of the LM358) for good sensitivity, making a trade-off between achieving a pleasant mid-range tone and a sensitive rate of change as the level is tilted. Next, place a few sheets from a memo pad under each end of the beam; at the higher end, remove sheets one at a time until the bubble indicates level. (You may never get an exact level indication, since these sheets are typically 0.004 inches thick, but by pressing on the higher end of the beam, you can juggle the system a bit.) Adjust the 10K pot associated with the tilt sensor so as to find the border between pulsating and smooth outputs from the VCO.

Another way to calibrate the level is completely independent of the visual bubble. First, adjust the 100K rheostat of the VCO for a sensitivity and pitch of your liking.

Next, place memo pads -- as adjustable shims -- on the table so that the ends of the beam can rest on them. Ignoring the pulsations for the moment, remove or add sheets to the piles so that the same pitch emanates from the VCO when the beam is rotated horizontally end for end. In other words, you are creating a "level playing field," testing this by noting that the pitch is the same whether the loudspeaker faces toward you or away from you.

Finally, adjust the 10K calibration pot so that the boundary between pulsed and smooth tones can be found in either orientation.

The Generalized Solution

Given that many types of sensors exist that translate physical measurement into electrical output, a flexible readout which can accommodate their differences is all you need to "grow your own" instrument. Tom Fowle has developed such a readout. Although the Fowle Gimmique was invented for tuning radio circuits, it has proven to be an effective readout for many instruments -- blood-pressure gauges and thermometers. We are very happy with the readout scheme of this level.

This readout can be fit to any inclinometer whose output voltage varies with tilt. Such sensors would include true potentiometers, RVDT (rotary-variable differential transformer) tilt sensors (with their required driving and detector circuits), strain-gauge accelerometers, etc.

The only circuit changes necessary in using the "Fowle Gimmique" readout system involve adjusting the VCO to respond properly over the output range of the sensor. The comparator inputs that determine where pulsation starts, pins 5 and 6 of the LM358, can be anywhere from 0.5V above ground to VCC minus 2V, so this part of the system need never be considered.

The operating range of the VCO is determined by three components: The emitter resistor of its 2N2222 is selected knowing that the voltage seen by the VCO input will appear across this resistance and that the resultant current drawn by it will be the charging current sourced to pins 2 and 6 of the 556. The oscillation caused by this current can also be changed by the timing capacitor from pins 2 and 6 to ground. Finally, for voltages that never get near ground (such as those seen from the Midori sensor), getting adequate sensitivity will mean excessively high frequencies from the VCO, unless some current is shunted away from the timing capacitor; that is accomplished here by the 910K resistor in parallel with it.

The original 1982 circuit, by Mr. Fowle, lists a very flexible offset compensator as follows: Pins 2 and 6 of the 556 go through 1 megohm to the arm of a 1 meg pot. The bottom of this pot is grounded, while the top end goes to VCC. With this arrangement, the frequency of the VCO can be adjusted for something other than 0Hz for very low-level input signals. In other words, compensation in either direction can be accomplished with this circuit. (Naturally, the 910K resistor listed in this level is not present with the adjustable offset circuit just described.)

Parts List

Resistors (1/4-watt 5%, unless otherwise specified):

  • 1--10 ohms, 1/2-watt
  • 1--47 ohms, 1/2-watt
  • 1--680 ohms
  • 2--10K
  • 2--22K
  • 2--100K, 1%
  • 1--910K


  • 1--10K 10- or 20-turn trim pot
  • 1--100K 1- or 10-turn pot (connected as a rheostat)


  • 1--0.0047uF plastic or mica
  • 1--0.1uF disc ceramic
  • 1--100uF 10V electrolytic


  • 1--1N754 6.8V zener
  • 1--2N2222
  • 2--2N2907
  • 1--LM358
  • 1--556 dual timer
  • 1--Midori PMP-S10T inclinometer


  • 1--Loudspeaker
  • 1--9V battery connector and holding assembly
  • 1--On/off toggle


Midori America Corporation, 2555 E. Chapman Ave., Suite 400, Fullerton, CA 92631; Phone: (714) 449-0997.



We were asked by the Arizona State School for the Deaf and Blind to build a hearing-aid battery tester for hearing impaired students who cannot see well enough to read visual testers. This, our first prototype, is a tactile adaptation of our Smith-Kettlewell Auditory Battery Tester, published in Vol. 11, No. 1 of SKTF (Winter, 1990). For the experimental prototype, we replaced the auditory output with two vibrotactile outputs: one uses the Star Micronics buzzer as a vibrator (even though it does make noise), and the other sends a 200Hz signal to a small loudspeaker whose vibration can be felt by touching the cone. For those not using hearing aids, consider this a basis for a general-purpose battery tester; it's very good, and if connected to the Radio Shack "Handy Checker," No. 22-096, 9-volt and common flashlight sizes can be tested quite effectively (see SKTF, Winter 1990, for details).


As with the auditory version, two configurations are possible. You could, as described in the original article, provide your own cell holders to which proper load resistors would be affixed. For the purpose of making a button-cell tester for hearing-aid batteries, we chose to mount a commercial "Activair Button-Battery Tester" to the top of a cabinet which houses our circuit. This solved the problem of designing a battery holder.

The adaptor is high-impedance. Thus, the Activair gauge works whether or not our circuit is turned on.

The Activair tester has a d'Arsenval meter on it. The cell holder is rather clever; there is a long slot containing two facing metal plates. By pressing a spring-loaded "key" on the side of the tester, the positive plate can be moved toward the negative one. The idea is to sandwich the button cell in between those contact plates, and the cell can be inserted anywhere along the length of the slot. (The "can" of the cell -- the face having the largest diameter -- is the positive terminal; this stems from the construction of mercury cells and is retained by tradition.)

Some of these cells are small enough to fall flat in the bottom of the slot so as not to make proper contact. The way I load it is to close the slot first, stand the cell on edge and gradually open the slot until it accepts the cell.

The box containing our circuit measures 1-1/4 by 2-1/2 by 4-1/2 inches, and there was room to mount the Activair tester on the lid. Positioning it all the way to one end left enough space ahead of the meter to cut a square hole in the lid of our box, and the Star Micronics buzzer -- mounted on our circuit board, which in turn rests on 1/2-inch spacers -- appears almost flush with the lid.

I chose the Star Micronics buzzer because its vibrations are strong enough to feel, not for its sound. Actually, the sound of this buzzer is rather temperamental, and should probably be ignored.

Along the bottom edge of one side of our instrument case is the on/off switch and a 1/8-inch mini phone jack. A 555 oscillator feeds a 200Hz signal to this jack. We sent the prototype with a small speaker which could be plugged into this jack, the idea being to feel the signal from the cone of the speaker (as an alternative to the buzzer).

The signal from this jack is also able to drive an earphone fairly loudly. Although they cost "an ear and an eye," bone-conduction units make good vibrators (not put against the skull, but felt with the hand). This jack would certainly drive one of those.

How Does It Work?

The duration of buzzes -- the "duty cycle" -- is the indicator of battery voltage. The shorter the buzzes, the better the cell. Long buzzes, with short spaces between, means that the cell is bad. (By exchanging pins 12 and 13 of the LM324, you could reverse these, causing a solid buzz to mean good, but our experience with an auditory version showed that this mode led to users leaving the tester on and killing its internal battery.)

The cell voltage is "compared" to a calibrated triangle wave. If the voltage of a new cell exceeds the peaks of the triangle wave, the buzzer is not triggered. As the cell voltage drops, more and more of the triangle wave operates the comparator, and bursts of the buzzer get longer as a result. A dead cell, one whose voltage is below the lower peaks of the triangle wave, causes the tester to buzz continuously.

Construction Details

A piece of perforated board -- 2.2 by 3.2 inches -- was cut. Then, corners of one short side were filed away with a rat-tail file so as to accommodate the corner posts at one end of the box. The board was to reside at the right end of the box, leaving room for the 9-volt battery to lie on its edge at the left end.

Mounting holes were drilled through the uncut corners, with one more mounting screw being located at the middle of the nearest long edge. By doing this, the buzzer -- which was affixed to the upper-right corner -- could use the board as a springy cantilever, thus increasing its freedom to vibrate. (A piece of Styrofoam was placed under the buzzer to prevent the board from breaking.) One-inch 4-40 screws were used to mount the board; these made room for 1/2-inch spacers under the board.

The ground bus was run along the near edge. The 555 was located below the buzzer, the 2N2222 was placed to the 555's left, and the LM324 was positioned, more or less, in the middle of the board. The pots were placed to the left of the 324, and the left end of the board carried the standard voltages and the VMOS FET.

The Activair tester has a removable back panel. When one screw is removed, this back cover can be removed, although you may have to use a knife blade, since it is still held in place with little tabs, similar to those on the back covers of radios.

The back panel of the Activair tester was then screwed to the lid of our box. There is plenty of room for bolts and nuts in three of its corners -- one either side of the meter and one in the corner opposite the spring-loaded key. So as to leave the buzzer exposed, this tester was positioned all the way toward the battery end of our cabinet.

A hole for wires to come through was drilled between the spring and the meter. A large hole was drilled in our lid over the screw hole for the Activair back panel; this allows its screw to be put back after mounting.

The Activair tester partially occluded the screw holes in the lid of our cabinet, so we used a rat-tail file to make notches in its back plate to make room for screw heads to clear it.

Circuit Operation

An LM336 voltage standard (said to be a 2.49V standard) feeds a voltage divider so as to give us a standard at half its output value. Fortunately, by feeding its "adjust pin" with a pot, the baseline of 1.245 volts can be shifted to suit our needs. A buffered version of this voltage is created so that an open-drain comparator (called the "composite comparator") can be clamped to this voltage without affecting the standard. (Note that since an FET introduces an inversion of the output swing, the "sense," or polarity, of the composite comparator's inputs is reversed.)

Two op-amps are used to create a triangle wave; one is an integrator being directed up or down by a 2N2222 transistor. The other is part of the composite comparator.

Hysteresis of the composite comparator (adjusted with the 50K trim pot) determines the peak-to-peak amplitude of the triangle; the output from pin 8 of the LM324 goes up and down by the value of that hysteresis.

A simple comparator, the last of four op-amps, compares the voltage of the test cell with the triangle wave. The output of this final stage goes high when positive excursions of the triangle exceed the cell voltage. Both the Star Micronics buzzer and the 555 oscillator turn on when this output goes high.

Circuit for the Vibratory Battery Tester

The negative of the 9-volt battery is grounded. Its positive terminal goes through an on/off switch to the VCC line; VCC is bypassed by 220uF (negative of this cap at ground).

An LM336 2.5V standard has its "anode" grounded; its "cathode" goes through 1.6K to VCC. Its "adjust" lead goes to the arm of a 10K trim pot. The bottom of this pot goes through a 1N914 diode to ground (cathode at ground). The top of the pot goes through another 1N914 diode to the junction of the 1.6K and the LM336 (cathode toward the pot).

This junction of the 1.6K and the "cathode" of the LM336 is the adjustable 2.4V point. This point also goes through two 10K 1% resistors, in series, to ground; the junction of these resistors is bypassed by 1uF (negative of this cap at ground).

The output of the voltage divider -- the junction of the two 10K resistors -- is the "unbuffered 1.2V point." This is buffered as follows:

An LM324 quad op-amp has pin 4 going to VCC, while pin 11 is grounded. Pin 3 of the 324 goes to the above unbuffered 1.2V point. Pins 1 and 2 are tied together and go to the buffered 1.2V line.

The second op-amp in the 324 and a VN0300M VMOS power FET comprise a comparator (called the "composite comparator" here). Pin 7 of the 324 goes through 56K to the gate of the VN0300M. The source of this FET is grounded; its drain goes through 1.6K to the 2.4V point.

The triangle-wave generator uses the third op-amp in the 324. Pin 10, the non-inverting input, goes to the unbuffered 1.2V point. Between output and inverting input, pins 8 and 9, is a 0.47uF Mylar cap (non-polarized is important). The inverting input, pin 9, goes through 1.8 megohms to the 2.4V point. Pin 9 also goes through 910K to the collector of a 2N2222, the emitter of which is grounded.

The base of the 2222 goes through 56K to the drain of the FET -- the output of the composite comparator.

The drain of the FET also goes through 120K to the top of a 50K trim pot. The bottom of this pot goes to pin 1 of the LM324, the buffered 1.2V line. The arm of this pot goes to pin 6 of the 324, the composite comparator's non-inverting input. Pin 5, this comparator's inverting input, goes to pin 8, the output of the triangle-wave generator.

The fourth op-amp is a simple comparator; pins 12 and 13 are non-inverting and inverting inputs, respectively. Pin 12 goes to pin 8 of the LM324, the triangle-wave output. Pin 13 goes through 56K to ground; pin 13 is the positive tester input. The negative side of the cell being tested is grounded.

Pin 14 of the 324 is used to drive two vibratory output schemes as follows:

Pin 1 of a 555 is grounded. Pins 4 and 8 are tied together and go through 10 ohms to VCC. Between pins 1 and 8 of the 555 is the parallel combination of 0.1uF and 220uF (negative side of the latter cap at pin 1). Pins 2 and 6 are tied together and go through 0.01uF to ground. Pins 2 and 6 also go through 160K to pin 7; pin 7 also goes through another 160K resistor to pin 14 of the 324.

Pin 3, the output of the 555, goes through 43 ohms (1/2-watt), then through the speaker to pin 8 of the 555.

A Star Micronics CMB12 buzzer was also installed. It has four pins in the corner positions of a 14-pin DIP -- pins 1, 7, 8, and 14. Pin 1 is grounded, while pin 14 goes through 10 ohms to VCC. Between pins 1 and 14 is the parallel combination of 0.1uF and 220uF (negative of the latter cap at pin 1). Pin 8 goes to pin 14 of the LM324.

Tapping into the Activair Button-Battery Tester

In this tester, the positive of the cell holder (the movable contact) goes to the positive side of the meter. The stationary contact (negative) goes through 47K to the negative side of the meter. Between the stationary contact and the positive of the meter is a 510-ohm load resistor.

Thus, the meter side of this 510-ohm unit goes to pin 13 of the LM324, while the lead soldered to the cell holder goes to circuit ground of the vibratory circuit.


In the original auditory article, comments refer to how various battery types make the tester behave. For flashlight cells -- zinc-carbon, alkaline, and nicad -- the tester can be set so that its triangle wave is centered at 1.25 volts, with the peak-to-peak amplitude being set to one-half volt. Fresh zinc-carbon and alkaline cells will make the tester fall silent. Freshly charged nicads will cause the indications to have a 50% duty cycle.

In contrast, zinc-air cells are not so easy to check; their terminal voltage wanders around, making bad ones look okay and good ones look a bit soft. In fact, I disagree with the Activair meter; a battery which was declared dead by a hearing impaired colleague tests fine.

Discharge curves for zinc-air batteries suggest that a plateau of 1.2 volts will be available during most of their lifetime. My initial thought was to set the "offset" pot -- the 10K unit associated with the LM336 -- so that 1.2 volts appears on pin 1 of the LM324 quad op-amp chip. This would mean that 1.2 volts would cause the buzzer to operate at a 50% duty cycle -- half on, half off. However, the he sample cells I have don't jive with the textbook figures, so I've since set the baseline just shy of 1.18 volts (seen at pin 1).

The "range" pot -- the 50K hysteresis adjustment -- adjusts the amplitude of the triangle wave. (Unfortunately, the frequency of this wave is determined by its amplitude, which is the nature of home-made function generators. It is the duty cycle that is the important indication, not the frequency of buzzes.)

For this latter adjustment, I took a cell that was declared dead and adjusted the 50K pot until the buzzes more or less blended together. Set this way, a fresh cell caused the tester to be quiet.

Parts List

Resistors (1/4-watt 5%, unless otherwise stated):

  • 2--10 ohms
  • 1--43 ohms 1/2-watt
  • 2--1.6K
  • 2--10K 1%
  • 3--56K
  • 1--120K
  • 2--160K
  • 1--910K
  • 1--1.8meg


  • 3--220uF 16V electrolytic
  • 1--1uF 3V electrolytic
  • 1--0.47uF Mylar
  • 2--0.1uF
  • 1--0.01uF


  • 2--1N914
  • 1--2N2222
  • 1--VN0300M or VN10KM N-Channel VMOS Power FET
  • 1--LM336 voltage standard
  • 1--LM324
  • 1--555 (not CMOS)


  • 1--SPST On/Off Toggle
  • 1--CMB12 Star Micronics Buzzer (optional)
  • 1--Button-Battery Tester by Activair
  • 1--9V battery connector

Address List

Duracell/Activair Company: Contact Charles Walton, national sales representative, for distributor in your area. His number is 1-800-243-9540, ext. 4743. The company's address is Berkshire Corporate Park, Bethel, CT 06801; phone 1-800-548-5489.

Star Micronics, Inc., 200 Park Ave., Suite 2308, New York, NY 10017; phone (212) 986-6770.