The Smith-Kettlewell Technical File

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

William Gerrey, Editor

Issue: [current-page:title]

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 Details regarding products, suppliers, and other contact information are original and may be outdated.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







Michael Wyatt, SSO Honeywell Inc, Clearwater, FL

The following is reprinted from EDN, copyright 1993 by Cahners Business Information, a division of Reed Publishing, USA. Permission was granted for one-time-only reprinting, herein. Any unauthorized copying of this material is prohibited.

from EDN, July 22, 1993

The modified Sallen-Key active filter in Fig 1b allows greater freedom in component selection than does the original Sallen-Key filter of Fig 1a. Capacitor values are generally available in the familiar 10% sequence of 1, 1.5, 2.2, 2.7, ...; resistor values are available in the 1% sequences of 1, 1.02, 1.05, ...

Sallen-Key filter designs based on equal capacitor values rather than the more common equal resistor values would benefit from better fidelity and lower cost. However, the original filter in Fig 1a is restricted when using equal capacitors to damping factors greater than or equal to 1. This restriction limits the filter's usefulness for practical designs. In the more common equal-resistor case, the damping is unrestricted and equals the square root of C2 over C1 . However, capacitor values that fit this equation may not be available at reasonable costs.

Fig 1b's modification lifts those component-value restrictions. Filter damping is unrestricted for the equal-capacitor or the equal-resistor case. This effect is achieved because the addition of the buffer amplifier isolates the R1C1 network from the R2C2 network and lets you independently select the two time constants.

You design the filter by selecting a standard value for C1 and C2, then compute R2 over R1, finally solving for R2 and R1 from the natural frequency equation. For example, to design a second-order low-pass Butterworth filter with a corner frequency of 1 kHz, first choose convenient values of 0.01 uF for C1 and C2. Then compute R2 over R1 equals 2, for damping of 0.707. Now solve for R1 equals 11.25 kilo ohms and R2=22.50 kilo ohms from the equation for natural frequency (omega sub N equals 2 pi f).

Fig. 1A:

An op-amp is shown with the output tied to its inverting input. The signal source goes through R1, then through R2 to the non-inverting input. This non-inverting input also goes through C2 to ground. C1 goes from the output to the junction of R1 and R2.

If "omega" is the "natural frequency" (2 pi times f):

omega squared equals 1 over R1 times R2 times C1 times C2.

damping factor equals 1/2 times the square root of C2 over C1,, times the fraction whose numerator is R1 plus R2 and whose denominator is the square root of R1 times R2.

Fig. 1B:

Another follower is inserted; its non-inverting input is on the junction of R1 and C1, with its output going to R2. In other words:

The signal source goes through R1 to the non-inverting input of an op-amp (the new one). Its output and inverting input are tied together and go through R2 to the non-inverting input of another (this latter being the original of Fig. 1A);. This latter non-inverting input also goes through C2 to ground. This, too, has its output tied to its inverting input and this is the output of the circuit. This output goes through C1 to the junction of R1 and the added op-amp's non-inverting input.

The damping factor equals 1/2 times the square root of C2 over C1 times the square root of R2 over R1.

Comments from Al Alden:

The low-pass filter described above (using two op-amps) can easily be reconfigured into a high-pass filter. This is done as follows:

Switch the positions of the resistors and capacitors. In order for the damping equation and the component selection to come out the same, re-label the resistor and capacitor connected to the first op-amp as R2 and C2; re-label the resistor and capacitor at the non-inverting input of the output op-amp as R1 and C1.

Thus, the equations for "natural frequency" and damping factor are the same for both low-pass (as in the original article) and high-pass configurations. Also note that Q equals 1 over two times the damping factor. The response of such a filter is maximally flat when a Q (and, coincidentally, the damping factor) is 0.707.



The following is reprinted from EDN, copyright 1997 by Cahners Business Information, a division of Reed Publishing, USA. Permission was granted for one-time-only reprinting, herein. Any unauthorized copying of this material is prohibited.

From EDN 18 December 1997

[Editor's note: In the interest of parts availability to the individual builder, your SKTF Editor has added a section to the end of this reprint which contains liberal parts substitution and other comments.]

The original application of the circuit in Figure 1 was to locate coins, but it applies wherever you need to locate metal objects. The circuit uses a beat-frequency technique: Whenever a metal object comes close to the search coil, the metal causes the coil's inductance to decrease. Though the inductance change is small in itself (a coin, for example, causes a small frequency shift), beating the frequency against that of another fixed oscillator at almost the same frequency produces a noticeable audio-frequency shift.

The inductance of the search loop and the input capacitance (series combination of the three input capacitors, or approximately 2 nf) of the Q1 oscillator determine the frequency of the search oscillator. In this design, the frequency is approximately 370 khz. For years, this frequency has prevailed in metal detectors, because it yields the lowest sensitivity to ground conditions, such as moisture and density.

You can change the inductance of the search coil by modifying its turns count or diameter. The loop has Faraday shielding to reduce capacitive-coupling effects. You can fabricate a simple Faraday shield by using adhesive copper tape of the sort used in EMI shielding. Simply wrap the completed loop with the foil tape. Make sure the tape has a gap at one point around the loop, or a shorted-turn transformer results. You should mechanically secure the loop to a non-metallic form to prevent microphonics.

With the prescribed 4-in. diameter, the search coil in this design lets you detect a penny buried approximately 2 inches in the ground. If you wish to detect larger objects at greater depths, you can build a larger search coil. To keep the oscillator frequency roughly the same, you should halve the number of turns for every doubling of the coil's diameter.

The Q2 beat oscillator uses a similar Colpitts design. However, instead of using inductance changes for tuning, the beat oscillator uses a varactor diode. Depending on your exact search-coil inductance (hence, oscillation frequency), you may need to adjust the value of 82 microhenries to obtain adequate tuning range in the beat oscillator.

[SKTF Editor's Note: The authors remark that you may have to change the value of the 82uH choke in the reference oscillator if you cannot tune the system for a satisfactory beat note. (Presumably this would be due to wide variations of distributed capacitance in the search coil, depending on who winds it.) Coils are harder to come by than capacitors; I suggest "padding" either that choke or the search coil with parallel capacitance to get the oscillators matched. They say the search coils capacitance is about 2 nanofarads (2000 uF), so 200 or 300 picofarads across one or the other of the oscillators' coils might fix this -- easier than procuring a different fixed inductor.]

Both oscillators have light coupling to the remaining audio section of the LM389. The amplifier, whose gain is approximately 26 db, easily drives a set of 8-ohm headphones. If you need more gain, you can add a 10uF capacitor between pins 4 and 12 of the LM389.

[SKTF Editor's Note: For reasons I describe later, I do not recommend earphones with this design.]

The prototype detector uses some polyvinyl-chloride sprinkler pipes and fittings to provide a convenient handle and chassis for the electronics.

Operating the metal detector is simple. Place the search coil on the ground, and adjust the tune potentiometer for a tone of approximately 100 Hz. Then, taking care to keep the search coil at a fixed distance from the ground, slowly sweep the coil over the area of interest.

Metal objects cause a change in the audible tone. The relative amount of tone shift indicates the object's size and depth. The greater the tone shift, the bigger or closer the object is. You should avoid "zero beating" the two oscillators, because they will lock with each other, thereby reducing the search oscillator's sensitivity.


[Note: The pin connections listed in this circuit description are for the 8-pin LM386 amplifier, not the LM389 intended in the original article. A short discussion of the 389, along with pin assignments, is given later.]

A 9V battery has its negative terminal grounded. Its positive goes through an on/off switch to VCC; VCC is bypassed to ground by 220uF (negative of this cap at ground).

Pin 4 of the LM386 is grounded. Pin 6 goes to VCC. Its decoupling pin, pin 7, is bypassed to ground by 10uF (negative of this cap at ground). The authors mention that if more gain is desired, pins 1 and 8 can be shunted by 10uF (positive of this cap at pin 1).

The inverting input, pin 2, is grounded. Pin 3 goes to the arm of a 10K volume control. The bottom of this control is grounded; its top end goes to two 2.2pF capacitors whose free ends go to the emitters of the oscillator transistors to be described below.

Two similar transistor oscillators, using NPN's such as 2N2222, are arranged to beat against one another. Each has the following circuitry:

The collector goes through 1K to 9V; this collector is bypassed to ground by 10uF (negative of this cap at ground). The emitter goes through the parallel combination of 20K and 0.0022uF to ground. From emitter to base is 0.022uF. From base to collector is a 20K resistor.

The search loop CONSISTS OF 20 turns of 22-gauge ENAMELED (MAGNET) WIRE. One end of THIS coil, IS grounded. The other end goes through 0.1uF to the base of Q1. The search coil is encased in a shield (wrapped in conductive tape); one end of this shield is grounded. (Leave a gap between where your wrapping starts and where it finnishes. In other words, the shield is a sleeve that reaches almost all the way around the circumference of the coil; almost is important, don't let the ends of the "sleeve" touch and short out.)

Q2 is the tuneable "reference" oscillator against which the Q1 search-coil oscillator is made to "beat". The base of Q2 goes through 0.1uF to the top of an 82 microhenry choke, the bottom of this choke being grounded. This 82 microhenry choke is tuned by a "varactor diode" as follows:

The anode of a varactor is grounded. Its cathode goes through 0.1uF to the top end of the 82 microhenry choke. A 10K ten-turn pot goes from VCC to ground; its wiper goes through a 10K resistor to the cathode of the varactor (NTE618, 90 to 440 picofarads).

The authors show the device with an earphone output. Thus, pin 5 of the LM386 goes through 220uF to the tip contact of an earphone jack (negative of this cap at the jack). The sleeve of this jack is grounded. By using a closed-circuit jack, both earphone and speaker options can be had: The tip contact goes to the negative of the 220uF coupling cap as before, the sleeve is grounded, and the switch contact goes through the speaker to ground.

Editor's Meddling:

The LM389, a so-called "amplifier with transistor array" has been changed to an LM386 amplifier, then describing the oscillators using two discrete transistors, 2N2222's. If one were making tons of these metal detectors, perhaps the LM389 with on-board transistors would offer a slight economical advantage -- maybe not. For those interested, the pin assignments are listed for the 389 (which is a 386 in an 18-pin DIP with three transistors).

Also, the authors chose to use a varactor diode to tune the system. (I suspect they chose that method so that a 10-turn pot could be used to make adjustment less touchy.) In case you cannot get hold of a varactor, I have added substitution of a mechanical variable capacitance as an option to their varactor solution. This discussion follows:

The base of Q2 goes through 0.1uF to the top of the 82uhf choke; the bottom of this choke is grounded. The top of the choke goes to the stator of a variable capacitor; the rotor of this variable is grounded.

We have three options as to how to comprise the variable capacitance: (1) Though making the adjustment would be fussy, we could use a single variable cap whose range is perhaps 500pF. (2) Tuning the reference oscillator with "band spread" (a small "padder" which makes a smaller change per degree of rotation) makes fine adjustment much easier. Thus, we could parallel a large-range unit (400pF or more) with a small "padder" (50pF or more). (Both rotors are grounded; their stators both go to the top of the choke.) (3) If we knew more about how the search coil turned out, we could parallel a smallish-value variable with the appropriate fixed "ballpark" capacitance. (The fixed capacitance could be chosen experimentally; a large-value variable, in parallel with the small one, could be connected temporarily, then taken out and measured with a capacitance bridge.)

When first viewing the authors' circuit, lab boys around here exclaimed, "Where's the mixer?" Sin omega1 plus sin omega2 does not yield sum and difference frequencies like a "cross product" of the two would do; some non-linearity is needed to generate the "beat note".

We found this non-linearity in the misbehavior of the amplifier. When I built the circuit with no modification, it worked fine. Looking at the output of the amplifier, however, it was self-oscillating at perhaps 2 mHz, and otherwise distorting everything fed to it. With standard techniques used to tame it and make it distortion free, the metal detector no longer was audible. I shunted the volume control with a diode (anode at ground) and put 0.001uF across it as a filter. I got a signal back, but not a very loud one. Thus, I would have to couple more tightly from the oscillators to make such a modification work.

Heqq with it, it worked in the first place; let the amplifier misbehave and do the mixing of the oscillator signals. The only thing of note is: as shown here, the operation of the volume control is erratic, and the level of the audible beat note can change wildly as different false modes of operation occur. Thus, if you try earphones, place them well in front of your ears to prevent inadvertent damage to your hearing, in case the system howls or chirps uncontrollably.

I preferred the instrument's operation with the 10uF gain-boost capacitor installed. It goes between pins 1 and 8 of the LM386 (positive toward pin 1).

Discussion of the LM389:

The transistors are not totally independent of the chip substrate (which makes sense, since they are grown onto it). They cannot be operated with their collectors below ground. The equivalent circuit shows their collectors each going to cathodes of diodes whose anodes are common to the substrate.

Next, for whatever reason, the output circuitry of the amplifier has its own ground (pin 18). Otherwise, all eight connections of the 8-pin LM386 are present, labeled with identical designations.

We suspect that the 389's existence originated from a special order. Suggested applications range from making a noise generator (using a back-biased transistor junction as the noise source), to an AM radio employing the three general-purpose transistors in very traditional circuitry (not like new receiver chips). I wonder about the longevity of the 389 as a product. Hmmmmmmm.

Pin Assignments for the LM389 "Amplifier with On-Board Transistor Array":

  • For the three transistors, the emitter, base and collector pins are, respectively: 8, 7, 6; 9, 10, 11; 15, 14, 13.
  • Pin 2--Plus Supply (4 to 12 volts)
  • Pin 17--Substrate and Ground
  • Pin 18--Ground (for amplifier's output transistors; both 17 and 18 must be grounded)
  • Pin 3--"Bypass"
  • Pin 1--Output
  • Pin 5--Inverting Input
  • Pin 16--Non-Inverting Input
  • Pin 12--Gain (most positive, like pin 1 of 386)
  • Pin 4--Gain (pin 8 of the 386)

Miscellaneous Parts:

The smallest quantity of readily available magnet wire is from Radio Shack, 278-1345 (an assortment of three sizes -- 22-, 26-, and 30-gauge).

The copper shielding is a bit of a stumbling block; whether from Newark (which has narrow and 1-inch wide), or Mouser, the available rolls contain 18 yards of the stuff -- costing about $45. The 1-inch by 18-yard roll of "Scotch copper-foil shielding tape" is Mouser No. 517-1181-1.0. You could use any metal foil; you cannot solder to aluminum foil, but a good mechanical connection will do. Just remember to leave a gap somewhere along the circumference of the loop.

Mouser has the 82 microhenry choke, No. 542-77F820. It has axial leads.

As suspected, the 440pF varactor is a specialty item. The majority of those available are "surface-mount" type; anyhow, their maximum values are often 100pF or so.

However, luck was with us; it seems that the varactor belongs to a special class, "replacement parts". Although Mouser doesn't list it as a regular catalogue item, they will order the NTE618. Their designation number is 526-NTE618.

Next, an equivalent is listed by a division of Philips in the ECG "Master Replacement Guide" This is designated ECG618.

The ECG specs state that the capacitance is 440pF at 1.2 volts, and 22pF at 8 volts. Maximum reverse voltage is 12 volts.

The case style is a TO92. The center lead of three (unspecified) is externally bridged to the cathode -- the cathode is on the left with the flat side of the package toward you and with the leads pointing up.

Address List:

Mouser Electronics: 11433 Woodside Ave., Lakiside, CA 92040; Phone: (619) 449-2222; (800) 34-MOUSE (346-6873).

NTE Electronics Inc.: 44 Farrand St., Bloomfield, NJ 07003; Phone: (800) 631-1250.

Philips ECG: 1001 Snapps Ferry Rd., Box 967, Greeneville, Tn 37744-0967; Phone: (423) 636-5693.


by Bill Gerrey and Albert Alden


A system for adding a sub-audible tone to speech information, then detecting it on a recording or received signal is described. Our purpose was to actuate a vibrator; however, a DC signal indicating the presence of the tone is available from our circuit. Possible uses would include:

Alerting a listener to index tones on a tape -- tones otherwise evident only on fast-forward or rewind.

Starting and stopping a second recorder which is copying marked notes from an unedited, but, tone-indexed, recording made in class.

Turning something on via a wireless transmitter/receiver system.

Making your walkie-talkie vibrate when a particular signal, one bearing a subaudible tone, comes on the air.

By choosing a higher filter frequency (perhaps 200Hz) and driving a tactile transducer from the filtered output, an improved CW system for the deaf-blind could be added to a general-coverage receiver whose BFO position is intended for SSB reception as well as CW.


For an experiment, as an adjunct to our Talking Signs system, we wished to flag certain messages and have the receiver detect these. The receiver would vibrate in the user's hand during those messages. (That's how we got involved with those crazy mechanical vibrators, see SKTF,Vol. 15, No. 1.)

Separating the sub-audible tone from the speech was done using the "modified Sallen-Key low-pass filter" in this issue. Albert Alden chose this filter because the roll-off above the desired frequency is 12dB per octave, rather than the 6dB per octave that a band-pass filter would give us. Being "under-damped", a peak in the response is obtained, and audio -- which we passed through a single-order high-pass filter whose corner frequency is about 300Hz -- is quite effectively rejected by the subaudible tone detector.

The sub-audible tone added to the recording was generated by the familiar 8038 function generator (see SKTF, Fall 1983, Vol. 4, No. 4). Given the ease of tuning this generator, and with the design information you now have on this Sallen-Key filter (see this issue), you can pick and choose design criteria to suit your application

Notes on Under-Damped Low-Pass/High-Pass Filters:

As mentioned in the EDN article, the response of such a filter is maximally flat with a damping factor, and also the Q of 0.707. (Q equals 1 over 2 times the damping factor; they are equal at "square root of 2 over 2.) However, for a purpose such as this, the under-damped case is very desirable.

The amplitude gain at the natural frequency is equal to Q. As the Q is increased, a peak in the response develops near the "natural frequency". For a Q of 1, the gain at the natural frequency is one; however, a peak slightly more than this can be seen at 70.7% of the low-pass filter's natural frequency. For Q higher than 2, the amplitude peak moves closer and closer to the natural frequency, being at 98% of it with a chosen Q of 5.

Thus, with the Q of 5 chosen for this detector, the gain of the filter is approximately 5 at 50Hz, the gain is 1 for all very low frequencies, below the peak in the response, and the response asymptotically approaches a roll-off at 12dB per octave above 50Hz.

Since nothing in our recorded program occurs below 50Hz and the roll-off of the filter is very steep as the band of audible program is approached, we have an ideal system for rejecting audio, and we have gain at the desired 50Hz with which to accentuate the subaudible tone for detection.

To review the calculations, we quote from the EDN article reprinted in this issue:

If "ommega" is the "natural frequency" (2 pi times f):

omega squared equals 1 over R1 times R2 times C1 times C2.

Damping factor equals 1/2 times the square root of C2 over C1, times the fraction whose numerator is R1 plus R2 and whose denominator is the square root of R1 times R2.

With the values shown in this circuit (73.2K, 2.87K, and matched 0.22uF capacitors), omega equals 313.6. Then, f equals omega over 2 pi, equals 49.93Hz.

[Note: for the tape-recorder applications noted in the abstract, you have two choices. (1) Several of our specially designed tape recorders, those made by APH for example, have tone indexers as a built-in feature. Various models will, no doubt, use frequencies higher than 50Hz. Thus, you will likely have to change the "natural frequency" of the filter to detect the index tone of your recorder. (2) The Smith-Kettlewell Tone Indexer (published in SKTF, Fall 1983, Vol. 4, No. 4) can be used with any tape recorder, and it can be tuned to the 50Hz chosen in the system here.]

General-Purpose Sub-Audible Oscillator Circuit:

An 8038 function-generator chip is used. It is powered by 10.5 volts (a 9V battery in series with an AAA cell).

Pin 11 of the 8038 is grounded. Pin 6 goes to the 10.5V line. Pins 7 and 8 are tied together (the internal reference voltage and the VCO input, respectively). For coarse "trimming" of the sinewave, pin 12 goes through 82K to ground.

Pins 4 and 5 are tied together and go through 68K, then through a 100K rheostat (frequency adjustment) to the 10.5V line. Pin 10 goes through 0.047uF to ground.

Pin 10 also goes to the collector of a 2N2222 "keying transistor". The emitter of this transistor is grounded. Its base goes through 33K to VCC, this base also going through 0.22uF to ground (this cap being a "key-click filter).

The oscillator is "keyed" (turned on) by grounding the base of the 2N2222. This can be done via a toggle switch, a normally open pushbutton, or a telegraph key.

Pin 2 of the 8038, the sinewave output, goes through 0.22uF in series with 47K to the top of a 50K output level control. The bottom of this control is grounded; its wiper is the output of the circuit.

Fifty-Hertz Detector Circuit:

[Note: A low-impedance signal source is required. The earphone output of a loudspeaker amplifier will do.]

This system can be powered from 7.5 to over 15 volts; a 9V battery was used in our system. Also, we had access to a 5-volt regulator in the Talking Signs receiver; this comes in handy for various reasons (not the least of which is isolating the vibrator motor from sensitive parts of the circuit), so the regulator is included here.

The negative terminal of the 9V battery is grounded. The "Common" terminal of a 7805 5V regulator is grounded; its "Input" terminal goes to the plus 9V line. This 9V line is bypassed by 470uF (negative of this cap at ground). The "Output" of the regulator goes to the 5V line, which is bypassed by 100uF (negative of this cap at ground).

Pin 11 of an LM324 quad op-amp is grounded. Pin 4 goes to the 9V line.

Two op-amps comprise the "modified Sallen-Key filter". (Note: the capacitors in this filter should be matched, if possible, for best results. The resistors are 1% units.)

A third op-amp serves as an active rectifier, and the fourth is a comparator whose output jumps to 8 volts when the tone is detected.

The signal source goes to the negative end of a 4.7uF electrolytic capacitor; the positive of this cap goes through 1.3K to the 5V line. The junction of the cap and the 1.3K resistor goes through a 73.2K 1% resistor to pin 12 of the 324 (a non-inverting input). Pins 14 and 13 (this op- amp's output and inverting input) are tied together and go through a 2.87K 1% resistor to pin 10 (the non-inverting input of another op-amp).

Pins 8 and 9 (output and inverting input of the latter op-amp) are tied together and go through 0.22uF to pin 12 (from the output of the filter to the first non-inverting input). Pin 10, the second non-inverting input, goes through a matching 0.22uF to ground.

Pin 8, the output of the filter, goes to pin 5 (a non-inverting input) of the rectifier op-amp. A diode has its cathode at the output, pin 7; the diode's anode goes to pin 6 (the inverting input). Pin 6 goes through the parallel combination of 15K and 4.7uF to the 5V line (positive of this cap at the 5V line).

Pin 6 is the output of the rectifier; this goes to pin 2 (an inverting input). Pin 1 (the output of the fourth op-amp, the comparator) goes through 1 megohm (positive feedback) to pin 3 (the non-inverting input). Pin 3 goes through 15K to the wiper of a 10K pot (threshold adjustment). One end of the pot goes to the 5V line; the other end goes through 47K to ground.

Pin 1 of the LM324 goes high when the 50Hz tone is detected. Pin 1 goes through 10K to the base of a 2N2222. The collector goes through 33 ohms to the 9V line. The 2222 emitter goes through a vibrator motor (such as the Motorola Part No. 59-02890W11 ) to ground.

A small loudspeaker can be used as a tactile transducer to "feel" the output of the filter directly. In this case, the rectifier and comparator sections are not required. An LM386 amplifier drives the speaker as follows:

Pins 2 and 4 of the LM386 are grounded. Pin 6 goes through 10 ohms (1/2-watt) to the 9V line; pin 6 is bypassed to ground by 1000uF (negative of this cap at ground). Pin 7 is bypassed by 22uF (negative of this cap at ground).

Pin 3 goes through 0.22uF to the output of the filter (pin 8 of the 324). Pin 5 of the amplifier goes to the positive end of a 1000uF cap, the negative end of which goes through the speaker to ground.

Note: If the above is desired as a CW filter with tactile output, 200Hz would probably be the frequency of choice. If the matched 0.22uF capacitors are changed to 0.056uF, the natural frequency of the filter comes out to be 196Hz; the damping factor, and hence the Q, is unaffected, since the ratio of capacitances is chosen as "1".

Component Notes:

The 5V regulator used here is the small one in the TO92 package, the 78L05. With the flat side toward you and the leads pointing up, the terminals are, from left to right: Input, Common, Output.

The terminals are in this same order for the 7805 in the TO220 package, held with the mounting surface toward you and the leads pointing up.

Two vibrators of different physical configurations are available (see SKTF, Vol. 15, No. 1). They are:

Motorola vibrator, No. 59-5046H03, used in their "Bravo Plus" pager (cylindrical); or, 59-02890W11 (pancake style)

Note: The above vibrators can be ordered directly from Motorola, Parts and Accessories; Phone: (800) 422-4210, menu choice 2.


by Tom Fowle


A means of guiding a portable circular saw through long cuts in large sheets of plywood is described. The system is built from commonly available aluminum stock, and does not require special tools or skills beyond those available to workers (people who are "shopworn" enough) to use the resulting device.


In carpentry, means for making cuts in large pieces of plywood and similar materials can be awkward. The usual tools, a table saw or radial-arm saw, are expensive and bulky, although they make it easy for a blind operator to safely set them up to do the job. The sighted worker often depends upon the portable circular saw for such tasks as making initial cuts in 4- by 8-foot sheets of plywood. The blind worker can find using this tool difficult or impossible because no way is generally available for keeping the saw "tracking" along the desired line of cut.

My system works by temporarily fastening a length of 3/4 inch aluminum angle stock, with clamps or screws, along the work and causing the saw to track this rail with an attachment fastened to the tool. The guide rail is straddled by a "sandwich" made up from aluminum flat stock and angle. This gives a rigid connection between the saw and the fixed guide rail.

The tool easily can slide along the intended path of cut while being pushed by the operator. Most importantly, both hands are kept on the prescribed handles of the saw, ensuring that the usual standards of safety can be maintained.

The device presented here is easily made from commonly purchased materials. Mounting the guide on the tool is through attachments intended for the standard "rip fence."

The saw chosen for the prototype is a Skill Model 5625, 6-1/2 inch 2-horsepower unit; it should be commonly available, and it features two fastening points for the rip fence allowing a rigid guide assembly to be affixed without modifying the saw. Probably, other brands or models could be similarly equipped once the builder understands the principles.

Modifying the rip fence:

The fence purchased with this saw consists of a J-shaped steel bar whose longer leg measures about a foot in length. This long end can be inserted into the saw through rectangular slots made into the base plate. This base plate has 2 sets of such holes, one at the front which allows the fence bar to extend entirely across the width of the base, and another near the back edge of the base plate which extends only a few inches in from the left edge. The fence is held in place with No. 8 bolts equipped with winged heads, which thread into built in "nuts" molded into the base plate.

The shorter leg of the J is about 5 inches long and ends in a flat piece mounted at right angles to the length of the bar. Originally, this flat piece is intended to be run along the edge of the work which is to be tracked. The amount of material to be removed is set by the penetration of the rip fence bar into the saw.

This rip fence is needed here only as an available source of bar stock, pieces of it being used as proper fittings which fit the built-in holes in the saw. Therefore, the fence is ruthlessly hacked to pieces to obtain 2 chunks of straight bar which, when inserted into the front and rear recepticals, form 2 attachment points to the left of the saw onto which the "guide sandwich" is built.

To facilitate setup of the guide strip (a long "rail" of angle stock), a "jig" -- a second sandwich -- is made up which has a single strip of flat stock extending at right angles from it to the distance required between the guide strip and the intended cut. Thus, the operator marks the work using his preferred measurement and marking tools, then orients the guide strip to the left of the intended cut using this jig to assure the strip is properly placed. C clamps or wood screws affix the guide strip firmly in place. Holes in the guide strip are drilled to accommodate woodscrews in cases where holes in the work are of no consequence.

The sandwich on the saw is made to extend about 2 inches in front of the base plate to allow fitting to the guide strip before the saw hits the edge of the work. In most cases, most of the weight of the tool is taken by the work piece soon after the cut is started, but care must be taken to see that the saw is properly aligned before starting the cut to avoid binding and "kick back". In general, at least in plywood, little pressure on the saw is needed to help it through the work. Some extra push at the start of the cut may be needed to get the blade guard started opening.

Under no circumstances should the worker be tempted to remove the blade guard or gimmick it open. There is a possibility of significant danger from improper use of a tool of this power and speed, and fooling safety features just to save a few seconds of setup time is a good way to get yourself another disability.

Some care should be taken to be sure the saw does not tilt forward as it comes off the end of the work, however a firm grip on the trigger handle and the forward handle provides good control and assures the hands stay where they belong.


  • The following materials should be obtained before building the "Straight Cut".
  • Skill Model 5625 6-1/2 inch circular saw (comes with general purpose blade).
  • Rip fence, Model No. Skill 95100.
  • Aluminum angle stock, 1-1/2 inch on a side. (Used to make the sandwiches.)
  • Aluminum angle stock, 3/4 inch on a side , at least one 4-foot section (Used as guide strip; buy it in 8-foot sections if you can get it home that way).
  • Aluminum flat stock, 1-1/2-inch wide (Used as one face of sandwich.)
  • Aluminum flat stock, 1-inch (Used as "sandwich filling" -- a "spacer"). This flat piece should have the same thickness as the guide-rail angle stock.
  • No. 8 by 1/2-inch pan-head machine screws, A dozen or so.
  • No. 8 hex nuts and star lock washers, to match screws.
  • No. 4-40 by 1/2-inch flat head machine screws, also with star lock washers and hex nuts. About a dozen or so.

Tools needed:

  • Hack saw with new sharp blade, equipped with a good tireless arm!
  • Files and emery cloth or fine sand paper.
  • Carpenters square.
  • Drill and bits.
  • Assorted hand tools.


First the rip fence is cut as near as possible to the bend at the long end of the J; this gives the longest possible straight piece. This long piece is then cut into two chunks: one measuring 6-3/8 inches and the other 3-1/2 inches long. (The reason for these specific measurements is that, when they are inserted into their receptacles, they are to protrude the same distance out to the left.) You may wish to check out the theory of these pieces for yourself; the following procedure will confirm the measurements.

The saw, with blade removed and blade guard open all the way, is placed on the bench with the blade side to the right of your viewpoint. (1) the rip-fence bar should be slid into the forward rip-fence holding slots in the base plate; the excess material will be protruding out to the left; the right end of the bar should just penetrate the appropriate slot in the right edge of the base plate. Run the supplied bolt with winged head down into the threaded hole provided to tighten the bar in place. Measure out along the bar from the left edge of the base plate about 1-3/4 inches and mark this spot. (2) Now turn the bar end for end and slide its unmeasured end into the rear rip fence guide slots as far as possible. It comes up against a protrusion which holds the blade wrench in its storage position. Bolt it in place and measure out from the left edge of the base plate along the bar, again about 1-3/4 inches. If things work out, this should be the same spot you found and marked in the measurement of step 1.

The point is to cut the bar into two pieces which, when inserted into the guides in the base plate, protrude exactly the same distance out the left edge. Hopefully this distance will be a bit more than the 1-1/2 inch inside measurement of the larger angle stock. (See where we're going yet?)

Now mark and drill 2 holes in the outer ends of each bar as follows. (A drill press is nice here to get the holes through the bar as straight as you can.)

Each bar should have a hole drilled through its lesser dimension about 1/2 inch from what will be its left most end. Another hole should go through each bar about 1-1/4 inches in from the left end. These holes should be drilled quite small for now until their matching holes are measured and setup. These will eventually pass No. 4 flathead screws.

Next cut 2 pieces of the 1-1/2 inch angle stock, one about a foot long, and the other 6 inches long. Put aside the 6-inch piece to use in the marking jig to be described later. The 1 foot length is long enough that when laid alongside the saw, it extends about 2 inches in front of the tool and just a bit beyond the rear.

With the prepared bars inserted into the saw and bolted in place, arrange the 1-foot angle (which will become the "inner side of the sandwich") with its horizontal surface under the extending bars and its vertical surface on the left. In other words, the bars should be on top of the angle, holding it down on the bench. Slide the angle so that its rear end is just in back of the saw's base and press it firmly to the right to get the ends of the bars up against the inner wall of the angle.

Here you will find one of the few limitations of this device as I prototyped it. To get the vertical side of the angle in against the bars, you have to raise the saw motor itself by loosening the "depth of cut" lock and lifting the back end of the saw with relation to the base. The depth of cut lock is a lever made of bend heavy sheet metal which can be found in back of the motor housing just to the left of the plastic blade housing. This lock is pulled up and forward to loosen it so that the saw body can be tilted up and forward. With the sandwich in place, the depth of cut is limited to about 3/4 of an inch or a bit more. Luckily this is just enough to allow for most work with larger materials, where this tool is at its best. (A possible modification to solve this problem will be discussed later.)

Now comes the most critical part of the whole project! You must assure that the new piece of angle is exactly parallel with the baseplate of the saw. Leaving everything carefully in place, check that the distance between the right edge of the horizontal angle piece and the front and back edge of the base plate is the same. If this isn't well done, the system is likely to bind when in use. The easiest adjustment to make for this alignment is slightly moving the front bar left and right in its slots. Try not to let it come free of the right-hand end slot at the front right corner of the saw, as this causes instability.

When you are sure of this setup, use a small prick punch or a small drill bit to mark through the holes drilled through the bars and create marks in the angle to match. Now take everything apart and enlarge the holes in the bars to pass the No. 4 flatheads. If you like you could tap these holes for No. 4-40 threads and top them off with nuts and lock washers for added stability, but I didn't bother.

Carefully pilot drill from the inside of the angle piece through the marks you made through the bars. From the outside of the angle, drill these 4 holes to pass the No. 4 screws and countersink them to be sure the heads of the screws will be just below the surface of the stock. This prevents the heads from binding on or damaging work surfaces. (It might be possible to use No. 6 bolts, but I didn't feel there would be enough aluminum thickness after the countersink to hold the screws securely against vibration and stresses involved.)

Next, secure the bars to the angle with No. 4 flatheads passing up through the angle and through the bars. Leave everything a bit loose and fit it all into the saw. Again check that the angle is straight with the saw by checking the distance between the right edge of the angle and the left edge of the saw base. When you're sure of this, tighten first the bolts that hold the bars into the saw, then the screws that hold the bars to the angle. If you're like me, this took just a tiny bit of wiggling, but hopefully not much filing of holes which could allow things to move under stress.

Now you've got the basic guide parts, so after lunch its time to make the sandwiches. One that holds you're angle and saw to the guide rail which is affixed to the work, another shorter sandwich is for the jig, a marking guide for setup.

Cut and file, smooth and clean, 2 pieces of each of the flat stocks, 1-foot and 6-inch lengths of each size of stock. The narrower stock forms the "filling of the sandwich" (a spacer), the wider piece forms the "left bread" while the angle installed on the saw is the "Right bread". (The guide strip, when straddled by a sandwich, also becomes part of the filling.)

With the saw flat on the bench (no blade installed), and the blade guard all the way open, install the guide thus far assembled in the rip fence holes on the saw. Place a piece of the 3/4-inch guide-rail angle stock on the bench with its vertical side up against the left side of the large angle. This leaves the horizontal portion of both angles on the bench and their vertical faces together.

Now lay, on its long edge, the 1-foot section of 1-1/2-inch flat stock inside the smaller angle, firmly up against the left face of the small-angle's vertical side. Holding things together so that the flat stock and the angles are all parallel and held together, this leaves an open space on top of the vertical edge of the smaller angle into which the chunk of 1 inch flat stock can be inserted.

The point here is to make the larger angle and the 2 pieces of flat stock just fit over the piece of guide-rail stock. With a couple of C clamps, clamp it all together near the top edge of the sandwich. . Be sure the 1-inch flat material is held firmly inside the sandwich so that it can't move down while you're fixing things together.

Now, being sure that your clamps are secure, remove this assembly from the saw and hold it in a vise where you can drill through the sandwich pieces while not allowing them to shift. Drill pilots, and then body drill, clearance holes for No. 8 bolts 1/2 inch below the top edge and about an inch in from each end of the sandwich. Be sure you're screws go through the 1-inch flat stock and don't get into the space reserved for the guide angle.

De-burr all holes and take the time to file and emery paper all the surfaces of the sandwich pieces, as smooth as you have patience for. (The nasty paper labels found on some stock can best be removed with "Googone", a wonderful relatively non-toxic solvent for gummy adhesives which is available at most large drug stores.) I finished up with No. 400 grit paper so that things slide nicely. An application of paraffin wax seems to work nicely as lubricant without being too messy.

Fit the sandwich with No. 8 screws, lock washers and nuts. You may find you want extra lock nuts and lock washers on these to keep them from shaking loose. I found that an added lock washer was necessary between the inner stock and one side of the sandwich to give enough room for the device to slide on the guide strip. A flat washer was too much; you want just enough shim so as to get a fit that slides without jiggling or binding.

The accompanying jig is what you use to position the guide rail on the work. It is a T-shaped affair; a piece of 1-inch-wide stock is attached to a 6-inch-long sandwich. The length of this flat piece is cut so that, when the sandwich is installed on the guide rail, it reaches to the position of the saw's blade. Therefore, a second sandwich is constructed around the 6-inch piece of 1-1/2 inch angle which, of course, has no rip fence bars attached to it. Again, be sure the 1-inch flat-stock spacer is positioned high enough to accommodate the guide strip's vertical wall without lifting the sandwich off the bench.

When this second sandwich is complete, take a section of 1-inch stock and carefully file one end off square. Placing the shorter sandwich on the bench, position this 1-inch stock on the midpoint, and at 90 degrees to, the angle. A clamp here will keep the end of the flat stock firmly against the inner bend of the large angle.

Mark 3 holes in a triangular pattern on top of the 1 inch piece so as to go through both pieces. Drill clearance holes for 4-40 screws. As before, countersink the holes in the outside or bottom of the sandwich so the heads do not protrude.

The next move requires that you actually make a test cut to determine the correct length of the flat strip on the marking guide. A scrap piece of plywood is clamped or screwed in place on the bench or other working surface along with a piece of the 3/4 inch guide angle. Note that this guide rail must go about 6 inches in from the area to be cut, and it has its vertical side facing the saw. Be sure things are secure and that most of the tool is well supported on the work. Just enough of your scrap should be hanging out off the edge to allow a cut.

Install the plywood blade on the saw, noting that as you stand at the back end of the tool with the blade arbor to your right, the blade turns counter clockwise, with the top of the blade moving towards the back of the saw. It is important to know this in order to install the blade correctly. Follow the instruction books recommended blade installation and safety procedures.

You might want to make a couple of passes with the saw not running, and with the depth of cut set so the blade just misses the surface of the work. This will give you some idea that the guide works smoothly and you can make certain that nothing will get in the way.

Once you've cut this piece of scrap plywood, don't remove the guide rail. Remove the saw and put it aside. Put the smaller sandwich on the guide rail and mark the flat strip just where it crosses the cut in the plywood.

Hacksaw the strip a tiny bit long for this measurement, then file it off nicely just to match the real cut. Now when you're setting up a cut, you have this marking guide to work with, making it easier to place the guide rail where it needs to be.

Some Thoughts on Guide Rails:

Up to now, I've left my 3/4 inch guide angles in 4 foot sections. In some circumstances, a single 8-foot chunk would be nice, but it presents an awkward storage and maneuvering problem. Future plans include a simple means of joining 2 guide angles together while still allowing the sandwich and saw to pass the joint smoothly. Other desirable attachments I'm planning include a protractor on one end of a guide strip for doing odd angles.

In most cases, it should be possible to attach the guide with C clamps, but in some cases a clamp with a very deep throat would be needed. There will be cases where you can put the guide on material which will be scrap after the cut, flathead wood screws could be driven through the guide strip with appropriate countersinking in the angle, and into the work. In a desperate situation, one could even try hot gluing the rail down if you're willing to hassle getting all the glue off the work. Other ideas will suggest themselves to you clever folks, and I will enjoy hearing of your modifications and additions.

Now, what about a guide strip out of flat stock that could bend, with movable attachment points, then a jig saw fitted with 2 short sandwiches that could follow the curving guide. With a marking guide like the one here, we might be able to do curves of fairly large radius! Neat what? I'll work on this unless one of you beats me to it.

Future plans also include better means of making cross cuts in narrow material which won't support the entire tool.

As previously mentioned, one disadvantage of this setup as it exists, is the sandwich getting in the way of the saw motor and limiting the depth of cut. I considered finding longer bars to move the sandwich farther leftward to clear the motor housing, but felt the increased bending moment on the sandwich, due to the long "lever" between the saw and the guide rail, would probably result in binding and jamming, which could have serious consequences.

An obvious move is to cut away some of the middle of the sandwich to clear the motor housing, and with the addition of extra bolts to hold the sandwich together just outside the area covered by the motor housing, this will probably work. At time of writing, I want a little more experience with the tool before making this change to be sure the modified sandwiches won't bind.

Words of Caution:

I strongly recommend wearing ear plugs when working with this tool, it is very noisy indeed, and I actually find that I concentrate on what I am doing better if I' not distracted by the noise. A slightly humorous aside with a good point to it: When very young, the son of a colleague called his father's saw "that big noisy machine!" and would run away from it.

Though this can be a very useful tool, it is a powerful and formidable machine and should be treated with great respect. When cutting, keep both hands on its handles at any time the blade is turning, note also that it takes quite a while for the thing to slow down even after you've released the trigger; so pay attention!

When you lift the saw free of the work, listen for the sound of the blade guard closing with a small thump, if for any reason you are not sure this has closed completely, carefully remove your hand from the front handle and unplug the tool before checking out the situation. The least bad thing that might happen, if you get careless, is that you may set the saw down with its teeth showing on a nice surface and scratch things up. Worse possibilities I'll leave to your imagination.