Smith-Kettlewell TECHNICAL FILE

Published by Rehabilitation Engineering Center Smith-Kettlewell Institute of Visual Sciences

Bill Gerrey, Editor

Supported, in part, by Smith-Kettlewell Eye Research Foundation and National Institute of Handicapped Research

Produced by

Table of Contents

OPERATIONAL AMPLIFIERS III

THE RETICON R5620 PROGRAMMABLE FILTER IC

A TUNABLE AUDIO NOTCH FILTER USING THE RETICON R5620

THE SMITH-KETTLEWELL TRAINING AID FOR PIANO TUNERS

THE SMITH-KETTLEWELL AUDIBLE STROBE TUNER

EDITORS CORNER

OPERATIONAL AMPLIFIERS III

By Albert Alden

Introduction

So far, we have described op-amp circuits which, with the exception of the "active diode" circuits of the last installment, have only used resistors as scaling elements. Capacitors and inductors may also be used. Inductors tend to be bulky, expensive, and often depart from the ideal. This, and the fact that an equivalent circuit can be designed using capacitors, essentially eliminates inductors from consideration in op-amp circuits. The op-amp circuits described in this installment utilize capacitors and resistors to make a class of circuits referred to as active RC networks.

Capacitors

The relationship between the voltage across and current through a capacitor is I equals C times the derivative of the voltage with respect to time. The derivative is the instantaneous rate of change of the voltage in units of volts per second. Thus, the current through a capacitor equals the capacitance times the rate of change of the applied voltage.

The above relationship can be rearranged to give voltage as a function of current. This becomes

E = (1/C) ∫Idt

(or E equals 1 over C, that fraction times the integral of the current over time).

This means that the voltage across the capacitor is equal to the reciprocal of the capacitance times the integral of the current with respect to time. The integral function may be thought of as a running summation of the current. This works out to be equal at any instant to the charge "q" accumulated in the capacitor since the start of the integration period.

The above two relationships represent two operations which may be performed on a signal to obtain additional useful information. For instance, if we have a signal proportional to the velocity of a device, the derivative of the velocity signal yields a signal proportional to its acceleration, and the integral gives the displacement from the starting of the integration operation. Note that integration has to have a starting point.

Using just a capacitor for these two operations requires a voltage- to-current or current-to-voltage conversion to have the input and output as voltage signals. The op-amp will do this for us.

Op-Amp Integrator

The circuit is as follows. An input resistor (Ri) is connected to the inverting input of the op-amp. The non- inverting input is connected to ground directly (or through a resistor equal to Ri to eliminate bias current error). An output capacitor (Co) is connected from the output of the op-amp back to the inverting input. If we apply a positive voltage (Ei) to the input resistor, a current I = Ei/Ri flows, since the inverting input is at virtual ground. The op-amp, in order to keep the inverting input at 0V, has its output go in a negative direction as required to accept the current flowing through the input resistor. The relationship between output and input is

Eo = -(1/(RiCo))∫Eidt

(or Eo equals the quotient of one divided by the product of Ri times Co, the negative of that quotient times the integral of Ei times dt.)

For example, if Ri is l megOhm and Co is 0.5 microFarads, then for an input (Ei) of 1 volt, the output will be a negative-going ramp which will be minus 2 volts after one second, minus 4 volts after two seconds, etc.

Integrators usually have a reset switch across the capacitor which is opened at the start of the integration. If one desires to hold the output signal at the end of an integration period, the input should be opened and/or shorted to ground with a switch. Mechanical switches or relays may be used where speed is not important; otherwise, analog switches such as the CD4053 are excellent for this application. The CD4053 is controlled with digital signals and can switch analog signals of either polarity.

Both input offset voltage and input bias and offset currents cause the output of the integrator to drift. Bias current error is eliminated, as explained above. Using an FET input op-amp and nulling the input offset, as described in the last installment, will reduce the drift.

The input-output relationship for an inverting amplifier, Eo = -Ei (Ro/Ri), is a special case of the general expression Eo=-Ei (Zo/Zi). Zi and Zo are the impedances of the input and output elements. These may be resistors, capacitors, inductors, and any combination thereof. The output element of the integrator is a capacitor, a device whose impedance is inversely proportional to frequency. Therefore, the frequency response of the integrator circuit is the same, since the impedance Zi of the input resistor is constant, leaving Zo as the only frequency-dependent term in the above equation.

Op-Amp Differentiator

By interchanging the input resistor and output capacitor in the integrator circuit, we get a differentiator. When applying a time-varying voltage to the input capacitor, a current proportional to the instantaneous rate of change (expressed in volts per second) flows to the virtual ground. The op-amp output voltage goes negative to accept this current, such that Eo = -I times Ro. This expression, combined with the voltage-current relationship for the input capacitor, gives Eo = -Ro Ci (dEi/dt).

Given a Ci of 0.5 microFarads and an Ro of 1 megOhm, applying a ramp input of 1 volt per second will result in an output of -0.5 volts.

Op-amp offset voltage and current errors don't cause drift as with the integrator, but only a constant offset error in the output.

Zi for the differentiator is that of the capacitor. This results in a circuit whose frequency response increases proportional frequency. This causes two problems for the op-amp differentiator. First, high- frequency noise is emphasized (like having a treble boost); second, the phase shifts of the external components and the internal op-amp frequency compensation add so as to approach l80 degrees at a frequency where there is sufficient gain to cause oscillation. By putting a resistor in series with the input capacitor, the input impedance Zi is prevented from going below the value of the resistor. The frequency at which this occurs is equal to

1/(2π RCi)

(or 1 divided by the product of 2 times pi times R times Ci).

Above this frequency the differentiator turns into an inverting amplifier. R should be chosen to match the bandwidth of the signal you wish to differentiate.

Putting a capacitor across Ro will give additional filtering. The formula is the same, with the R and C now being the components comprising Zo.

Filters

We will describe two low-pass and two high-pass filter circuits without delving too deeply into the general subject of filters. (It is anticipated that there will be a future series of articles on filters.) [Note: The single-order filters were described briefly in the last installment. More detail will be given here.]

Single-Order Low-Pass

This filter passes signals of frequencies below a cut-off frequency and attenuates signals above that frequency. The attenuation of a single-order filter is such that the ratio (Eo/Ei) is proportional to the frequency ratio (F0/F), where F0 is the cut-off frequency. This is equivalent to -6dB per octave. The response (Eo/Ei) at F0 is -3dB, or 0.707 times the low- frequency gain. This is a common definition of the cut-off or "corner" frequency of a filter.

The low-pass filter consists of an input resistor Ri going to the inverting input, and an output resistor Ro with a capacitor Co in parallel from the inverting input to the output of the op-amp.

F0 = 1/(2πRoCo),

(or F0 equals 1 divided by the product 2 times pi times Ro times Co)

and the low-frequency gain is equal to Ro over Ri. In terms of Zi and Zo, Zi = Ri, and Zo = Ro in parallel with Co.

Single-Order High-Pass

The high-pass filter passes signals of frequencies above the cut-off frequency and attenuates signals below the cut-off frequency. For signals below F0, Eo/Ei is proportional to F/F0 (the inverse of that for the low-pass filter). The response at F0 is -3dB, or 0.707 times that of the high-frequency gain.

The circuit consists of an input resistor Ri and input capacitor Ci in series going to the inverting input of an op-amp. An output resistor Ro goes from the inverting input to the output of the op-amp. Zi = Ri in series with Ci, and Zo = Ro.

F0 = 1/(2πRiCi)

(or FO equals 1 divided by the product of 2 times pi times Ri times Ci).

The high-frequency gain is equal to (Ro/Ri).

When capacitively coupling op-amp amplifier stages, the coupling capacitor constitutes Ci of a high-pass filter. Its value should be picked using the above formula.

To limit the bandwidth of an op-amp amplifier stage in a circuit, use the Zi of the high-pass filter and the Zo of the low-pass filter, picking the resistors for the band-pass gain required and picking the capacitors for the frequency limits desired.

Both the low-pass and high-pass filters can be made with only passive components (a resistor and a capacitor). The use of the op- amp gives a filter which has several advantages: the filter can be designed with gain, the output impedance is low, the input impedance can be made high, and several stages can be cascaded without each stage loading and affecting the performance of the others.

Second-Order Filters

We will describe here a low-pass and a high- pass second-order filter. The operation of a second-order filter differs from the single-order filter in a couple of ways.

First, there are two parameters to be chosen: (l) a cut-off frequency F0, and (2) a parameter which controls the shape of the frequency response for frequencies near F0. This latter parameter is either expressed as the quality factor Q or the damping ratio. The values of these are related such that either is equal to one-half of the reciprocal of the other.

Second, the response of a second-order low-pass filter (Eo/Ei) above the cut-off frequency is proportional to (F0/F) to the second power. This is equivalent to -l2dB per octave.

The filter chosen to be discussed here has for Q a value of 0.707. (The damping ratio happens to also be equal to 0.707.) This value of Q for the second-order filter results in a response which is referred to as being "maximally flat." Also, its gain at F0 is minus 3dB, or 0.707. It is called a Butterworth filter, after the author who first described such a filter characteristic in l930. The actual circuit we will use is called a voltage-controlled, voltage-source (VCVS) filter. Its attributes are simplicity of construction (few parts) and simplicity of design (easy calculation of component values).

Low-Pass Circuit

The circuit is as follows. Two equal resistors (R) in series go from the input to the non-inverting input of an op- amp. The op-amp is configured as a follower; i.e., the output and the inverting input are tied together. A capacitor (C) goes from the non- inverting input to ground, and a capacitor with a value of 2 times C goes from the op-amp output back to the junction of the two resistors. The output of the filter is the output of the op-amp. The correct value of Q (0.707) is taken care of by the ratios of the component values and the closed loop gain of the op-amp (plus l). The formula for F0 is

F0 = 1/(2πRC √2)

(or F0 equals 1 divided by the product 2 times pi times R times C times the square root of 2).

High-Pass Circuit

The operation of the second-order high-pass filter has a similar relationship to the second-order low-pass as the single-order filters had to each other. The response below F0 is proportional to (F/F0) to the second power. The gain at F0 is -3dB, or 0.707 times the high-frequency gain.

The circuit is like that used for the low-pass, but with the R's and C's rearranged. Two capacitors in series, each with a value of C, go from the input to the non-inverting input of the op-amp. A resistor with a value of R goes from the output of the op-amp back to the junction of the two capacitors. A resistor with a value of 2 times R is connected from the non-inverting input to ground. The op-amp is again configured as a follower with a gain of plus l. The formula for F0 is the same as for the low-pass filter.

General Comments

Here are a few general comments which apply to the circuits described in this article.

Op-Amps

While not mentioned explicitly, it is assumed that you will make the power connections and add the frequency compensation capacitor if required for all of the above circuits. For op-amps used in filters, the open loop gain in the frequency range of interest must be sufficiently large to get performance approaching theoretical. Op- amp filters are best suited for audio frequencies.

Capacitors

Film capacitors (polyester, polypropylene, polystyrene, and polycarbonate) should be used for the best performance from the circuits we have discussed. Ceramic capacitors are OK for less critical applications. Avoid using tantalum or electrolytic capacitors.

Scaling

Filters and other circuits using capacitors and resistors as frequency-determining components may be scaled in the frequency domain by changing all of the resistors and/or all of the capacitors by the same ratio. The frequency is changed by a ratio equal to the reciprocal of the resistor and/or capacitor value ratio. For example, if we design a l00Hz filter with l00K resistors, changing to l megOhm resistors gives us a l0Hz filter with the same characteristics. This principle can be followed for first order, second order Butterworth (or not), high-pass, low-pass, etc. Also, the time domain response of the circuit will change by a factor proportional to the ratio of the R or C change. For example, by increasing the resistors by a factor of l0, the circuit will take l0 times as long to respond to a given input.

Asymptotes

The description of the frequency responses given for the various filters are those of the asymptotes--that is, these are the limits which are approached below and above F0. At one or two octaves (F/F0 = 2 or 4) above or below F0, the actual response is quite close to that described in this article. In between, the response moves smoothly from one asymptote to the other. For a single-order filter, this is fixed. For the second-order filter, this is determined by Q.

THE RETICON R5620 PROGRAMMABLE FILTER IC

Abstract

This IC contains a 2nd-order audio-frequency filter of the "switched- capacitor" type. Its critical frequency is easily tunable over a wide range; not only is it determined by the frequency of an applied clock signal, but the ratio of the filter's frequency to that of the clock is set by imposing a binary number on five input pins. In addition, the "Q" can be set by imposing a binary number on another set of inputs. Two example circuits (a notch filter and a band- pass filter) can be found in the two articles to follow, "A Tunable Notch Filter," and "A Training Aid for Piano Tuners."

There are other filter chips using the "switched capacitance" technique; Reticon makes a dual 2nd-order filter in their R562l and a quad filter in the R5622. National Semiconductor also has a line of such devices. However, the R5620 is unique in that it is digitally programmable; the other devices use external resistors to set the Q and the ratio of clock to the filter's critical frequencies.

A full discussion of filter theory (including the necessary math for cascading filters, etc.) will not be attempted here. A complete package of "all you'll ever need to know" would indeed be an undertaking, and the discussion of this nifty little filter chip would get lost within it.

Heavy Theory

Use of the R5620 is so simple as to be trivial--hook it up as shown, and it will do as an obedient black box should. The discussion of its innards is a moral obligation on my part to you; you deserve nothing less than having access to the full story. There's no way around it, however, this is university level stuff--including some calculus--and you have my permission to take it or leave it as you wish. [My profound thanks to Al Alden, John Brabyn, and Jules Madey for working through this section with me, and for making it possible to re-live my analog computer class with vivid realism.]

Part I
The Analog Simulation of a 2nd-Order System

Analog computers (heavily used from the l950's on into the early l970's) were the scientist's Tinker Toy set by which electrical "analogs" of physical systems could be built and tried. Automotive suspension systems, airplane wings, and earthquake-proof buildings were all simulated on analog computers first before actual models were built. In this way, necessary changes in the actual designs could be anticipated by adjusting potentiometers on the analagous electrical version, after which the tried and true physical system could be built on the basis of its computer simulation.

Once a proposed physical system is expressed in terms of its mathematics, it can be simulated electrically on an analog computer. An analog computer set up to solve a basic 2nd-order differential equation was contained in the 5620 chip. The operation of the filter can be explained by describing two simulated systems: the first is a weighted spring and a dashpot, while the second is an RLC circuit.

A Weight, a Spring, and a Dashpot

The system we are modelling consists of a weight, one side of which goes through a spring to an inertial reference (ground), with the other side of the weight going through a dashpot to ground. (Because we are here on earth, the weight is supported by frictionless wheels so that we needn't consider the effects of gravity.) A "dashpot" is a disc or paddle submerged in oil; it is used to "damp" or lower the Q of the system. A force from the outside (an input signal) is then imposed on the above system, with the weight's behavior being observed. (The weight's displacement, velocity, and acceleration correlate directly with the low-pass, band-pass, and high-pass functions, respectively.)

First, we must derive a mathematical model of the above mechanical system so that we can build its "analog" electronically. The basic equation can be gotten by equating forces-- the forces being mass times acceleration, the oil's viscosity coefficient times velocity, the spring constant times the displacement, all balanced against the force imposed from the outside.

We will call the input signal F(t) [read F of t], a force as a function of time. M is the mass of the weight, B is the viscosity coefficient of the oil, and K is the spring constant. Displacement of the weight will be denoted X. Students of physics will recall that acceleration is the 2nd derivative of X with respect to time, while the velocity is the lst derivative of X with respect to time.

The basic mathematical expression for the system is:
F(t)=M(d2X/dt2)+B(dX/dt)+KX

Our next step is to build one, not with rubber bands, a toy truck, and a spoon dipped in cooking oil, but with our electronic Tinker Toys of operational amplifiers. As customary, we will use two integrators and a summing amplifier as the main components of our system. As students of Al Alden's series on Op-Amps, you will recall that integrators are op-amps with capacitors as their feedback element (see Part III), and that a summing amplifier (one with multiple inputs) is gotten simply by adding more input resistors onto an inverting amplifier.

We will make the following compromises to enhance the intuitive understanding of the analog simulation: (l) We will ignore the fact that all of the op-amps invert (attaching the negative sign) the signal as they perform their function; we will put necessary inversions in black boxes outside the main elements where we can see them. (2) The gains of the main building blocks-- the integrators and the summers--are taken as one. Often, coefficients of the X terms in the equation are included as gains in the main building blocks; we will bring these out separately where we can see them.

Since we are using integrators to solve the basic equation, we could rewrite the equation in terms of its integrals. This is more cumbersome and, as it turns out, we don't have to do so if we are willing to traverse our integrators backwards. For example, an integrator whose output is X as a function of time has the first derivative of X with respect to time on its input. When we create our analog simulation, we will draw the integrators so that their outputs point to the left.

The building blocks and their positions will be easier to imagine if we rewrite the equation as follows:
M(d2X/dt2)=F(t)-B(dX/dt)-KX

Dividing both sides by M, we have:
D2X/dt2=(1/M)[F(t)-B(dX/dt)-KX]

Analog Simulation Circuit

Two integrators, their outputs pointing to the left, are connected in cascade. The output of the leftmost one, A1, gives us the displacement X, while its input is the velocity, the lst derivative of X with respect to time. This input is being fed by another integrator, A2, whose input becomes the acceleration, the 2nd derivative.

Below this cascade of integrators is a summing amplifier which has three inputs. One of the three is the input to the system, F(t). Another is being fed from a -K black box, with the input of this -K unit going back to X, the output of A1. Another input to the summer is from a -B black box, with the input of this block being gotten from between the integrators, the lst derivative. The output of the summer goes through a black box of 1/M, the output of this latter unit going to the input of A2, the 2nd derivative.

The input of A2, the 2nd derivative, is the high-pass output. Between the integrators, the lst derivative, is the band-pass output. The output of A1, the displacement X, is the low-pass output.

In the above, the input of A2, the 2nd derivative, gets the sum of terms in the right half of our equation, all passed through a 1/M black box. As mentioned earlier, efficient analog simulation would have the scaling elements such as 1/M, B, and K incorporated into the gains of the various stages. Also, since all the stages are inverters, minus signs in the black boxes need not be included.

Intuitive Explanation

In the spring, mass dashpot system with a sinusoidal force applied as an input, the displacement velocity and acceleration of the mass correspond to the low-pass, band-pass, and high-pass outputs of the filter.

If we apply a low-frequency signal to the system, the displacement is proportional to the force. This is true up to frequencies where the inertial force of the mass (its mechanical impedance) becomes equal to the stiffness force of the spring. This is the critical frequency. For higher frequencies, the displacement decreases as the input frequency increases; i.e., a low-pass "filter."

At high frequencies where the inertia of the mass is predominant, the acceleration is proportional to the input force. As the frequency drops below the critical frequency, the acceleration decreases; i.e., a high-pass filter.

At the critical frequency, the velocity of the mass reaches a peak and decreases for both higher and lower frequencies; i.e., a band-pass filter.

In the low-frequency range, the performance of a system of this type is characterized as being spring- or stiffness-controlled, the high-frequency performance as being mass- or inertia-controlled, and the mid-frequency (near the critical frequency) as being resistance- or viscosity-controlled.

The Q of the System

With a given mass and a given spring constant, the Q can be varied by changing the viscosity of the oil; in fact, the Q and the damping factor B are inversely proportional.

Q = ωcriticalM/B

(or Q equals the product of the angular frequency times M, divided by B)

(ωcritical (or lowercase omegacritical) is the angular frequency and is 2 times pi times the frequency in Hertz.)

ωcritical = √(K/M)

(or the angular frequency equals the square root of the quotient of K divided by M)

In some analog simulations, the damping factor B is accomplished by putting a damping resistor (one of very high value) across the feedback capacitor in the A1 integrator.

Series RLC Simulation

On the advice of the good Mr. Alden, this analagous problem has been included. A voltage of V(t) has across it the series combination of an inductor, a capacitor, and a resistor. Kirkhoff's Law tells us that we can equate voltages around the circuit.

V(t)=L(dI/dt)+(1/C)∫VCdt+RI

(or V as a function of time equals the product of L times the quotient of dI divided by dt, that product plus the reciprocal of C times the integral of VCdt, that product plus RI)

Rather than solve the RLC problem in this form, we can make the problem look exactly like our other one by rewriting the equation in terms of the flow of charge q. Three bits of trivia are necessary tools to get us there:

  1. I=dq/dt with respect to time.
  2. VL=L(dI/dt); this now becomes VL = L(d2q/dt2).
  3. The voltage across the capacitor will be q/C.

We can now write the basic equation:
V(t)=L(d2q/dt2)+R(dq/dt)+(1/C)q

Solving for the 2nd derivative as we did before:
d2q/dt2=(1/L)[V(t)-R(dq/dt)-(1/C)q]

RLC Simulation Circuit

Two integrators in cascade have their outputs pointing to the left. The left-most one, A1, has its output labelled q, with its input being the lst derivative of q. The input of the other integrator, A2, bears the label of the 2nd derivative of q with respect to time.

Below these integrators is the summing amplifier with the following three inputs: (1) V(t), the input signal. (2) The q output of A1 through the black box, -1/C. (3) The points between the integrators going through the black box -R. The output of the summer goes through a box labelled 1/L, its output going into A2.

The output of A1, just plain q, is low pass. From between the integrators we get the first derivative of q, the band-pass output. The input of A2, the 2nd derivative of q, is the high-pass output.

The Q of the System

With a given L and C, the damping factor which determines Q is R; R and Q are inversely proportional.

Q= ωcriticalL/R

(or Q equals the product of angular frequency times L, divided by R)

ωcritical (or lowercase omega critical) is the angular frequency and is equal to 2 times pi times the critical frequency in Hertz.)

ωcritical =1/√LC

(or lowercase omegacritical equals 1 divided by the square root of the product of L times C)

The above systems are tunable, if you don't mind varying all the coefficients at once (quite a tracking problem). Wouldn't it be just dreamy if we found a way to gang all of these gain adjustments together? This brings us to the next section.

Part II
The "Switched Capacitor" Technique

The 2nd-order analog simulation within the 5620 filter chip is done entirely without resistors. Resistors are "simulated" by switching capacitors (at a very high sampling rate) between two voltages. The resultant "equivalent" resistors are frequency- dependent; their value changes in accordance with the clock signal controlling the switches. Let us by example make such a resistor out of a capacitor and a high-speed switch.

A clock-driven, single-pole double-throw switch has its arm going through a capacitor to ground. A voltage V1 is on position 1, while another voltage V2 is on position 2. The charge that flows from V1 to V2 is C times the difference, V1 - V2. If the switch is operated fast enough, an average current equal to the 1st derivative of q will flow:

Iaverage=C(V1-V2)fclock

(V1-V2)/I is a resistor:

Requivalent= 1/Cfclock

The above gimmicks only replace resistors in the filter system; feedback capacitors on the integrators are not twitched or tampered with. A couple of sample circuits for the main building blocks of analog computation are shown below:

Integrator

An op-amp has a feedback capacitor from output to inverting input. Instead of a resistor from the inverting input to the input signal, the inverting input goes to position 2 of an SPDT switch. Input to the integrator is applied at position l. The swinger of the switch goes through a capacitor to ground.

Summing Junction

A two-input version is fairly easy to visualize. Two switches are used. Position 2 of the first one goes to the inverting input of an op-amp (this could be a summing amplifier or even an integrator); position 1 becomes one of the inputs of the summer. The arm of this first switch goes through a capacitor to the arm of the second switch. Position l of the latter is grounded, while position 2 is the second input to the summer. Inversion of the second input signal is accomplished.

Actual Switching Arrangement

MOS/FET's (single-pole single-throw switches) are actually used. The gates of these MOS/FET's are operated by a flip-flop; the gate of one goes to the Q output of the flip-flop, while the gate of the other goes to the flip-flop's Q bar output. In this way, the SPST switches are closed alternately. The drain of one goes to the source of the other, and this common connection becomes the "swinger" of the equivalent SPDT switch.

As stated in the literature, the "resistors" have a delay equal to one cycle of the controlling flip-flop (this is twice the period and half the frequency of the clock). Delay problems are offset by trading positions of the FET gate leads on adjacent stages; the switches of adjacent stages are out of phase with each other. It is important to note that the main component of high- frequency noise introduced by the switched capacitor technique will be at the "sampling rate frequency," one-half the clock frequency. This noise will be farthest away from the audio spectrum when a high ratio of clock-to-filter frequency is chosen. However, if the filter is to be used at high frequencies, limitations of the op-amps dictate that lower clock-to-critical frequency ratios must be used.

In the literature, they do not show an equivalent circuit for the analog simulation which would explain the fact that the filter has multiple inputs and one output, as opposed to having one input and three outputs (see our simulation in Part I). Derivation of this analog simulation is left up to the reader--anybody else but me.

Using the R5620 Filter Chip

"Good Gravy, do I have to know all that!?" No, and a skeletal description of the chip will do much to point the way to hooking it up.

Description

The R5620 is an l8-pin DIP package. Three of these are supply pins (ground, along with plus and minus voltages). The frequency-determining scheme, the Q selection scheme, and the audio input and output connections are discussed below:

  1. The chip has a clock input pin into which a high-frequency pulse is fed. Two clock circuits will be discussed later, one using a 555 timer and a crystal oscillator using CMOS inverters. A 5-bit binary number imposed on 5 input pins selects the ratio of clock to F0. Counting down from 11111 to 00000 brings this ratio up logarithmically (in steps of about 3 percent) from 50 to 200.
  2. The Q is selected by choosing a 5-bit binary number to be imposed on another set of 5 pins. The range of available Q's is from 0.57 to l50, as you count up from 00000 to 11111.
  3. The filter has one output pin from which the filtered audio signal is taken. Three input pins are provided for the three functions, low pass, high pass, and band pass. Being a 2nd-order filter, the roll-off for the low- and high-pass sections is l2dB per octave; the band-pass roll-off approaches 6dB per octave for frequencies either side of F0.

Basic Hook-up

Pin 8 goes to ground, pin l goes to the minus supply, and pin ll goes to the plus supply. A signal from a clock oscillator goes to pin 7.

Pins l3 through l7 accept the binary code for setting the frequency divider (pin l7 is the least significant bit, and pin l3 is the most significant bit). The number 00000 causes the chip to divide the clock frequency by 200; 11111 divides it by 50.

Counting downward from pin 6 through pin 2 (pin 2 being the least significant bit, and pin 6 being the most significant), we have the 5 pins which accept the Q-determining 5- bit code. An address code of 00000 sets Q at 0.57; 11111 sets it for its maximum, l50.

One can either hard-wire these inputs to their respective logic levels, or use switches to make them selectable from the front panel. By connecting each pin through a pull-up resistor to logic 1 (or a pull-down resistor to logic 0), one can then use SPST switches to pull them in the other direction (to 0 if pull-up resistors are used, or to 1 if pull-down resistors are used). One nifty scheme would be to use a hexadecimal switch (a l6-position switch whose output is a 4-bit binary number) to control the 4 least significant bits; then a toggle switch could be added to control the most significant bits, sort of a "range" switch. With this arrangement, you can dial in any number you like.*

Selection of low-, high-, and band-pass functions is done by grounding the unused inputs and feeding the signal into the desired one. A notch filter can be gotten by tying the low- and high-pass inputs together --of course, grounding the band-pass input.

In all cases, unused inputs should be committed somewhere, and/or protected by resistors where appropriate. The binary inputs should be committed to ground or either supply line at all times. Any signal from the outside world (not from previous stages in the project) should be connected to its input through a resistor (4.7K or higher). If you are tampering with any inputs on a protoboard, commit these pins to some reference with resistors before you disconnect and reconnect them. Failure to follow these suggestions will cost you; I have burned out several units with my blase attitude.

Clock Circuits

The following two clock oscillator circuits were provided courtesy of Jules Madey, owner of Applied Inventions and supplier of these chips. The filter chip will tolerate these circuits being referenced either to ground or to -V. This is nice to know, since CMOS chips will not tolerate voltages higher than l8 volts, whereas the filter chip can be operated at 22 volts (+ and - 11 volts).

The first is a 555 timer set up to work around 100kHz. Pin l goes to ground or to -V, as the case may be. Pins 4 and 8 go to +V. Pins 2 and 6 are tied together and go through 680pF to pin l. Pins 2 and 6 also go through 7.5K, then through a 5K rheostat to plus V. Between pins 6 and 7 is 5.6K. The output, pin 3, goes to pin 7 on the 5620. (If a CMOS 555 is not used, there should be 0.luF connected between pins l and 8.)

A l MHz crystal oscillator is shown using CD4049 inverters. The output of the first goes through 680pF to the input of the second. A l MHz crystal goes from the output of the second to the input of the first. Across each inverter (from its output to its input) is a high-value resistor (l0 or 20 megOhms). The output of the second goes through another section (through another inverter) which serves as a buffer.

The above crystal oscillator can be fed into a CD4520 counter (a chip containing 2 sets of 4 flip-flops in cascade). In this way, you can also get crystal-controlled frequencies of 500, 250, l25, and 62.5kHz. Pin 8 of the 4520 goes to -V (or to ground), while pin l6 goes to +V. One section of this chip is disabled by connecting pins 9, l0, and l5 to pin 8. Pin l, the clock input, goes to the output of the crystal oscillator. Pin 2, the enable, goes to +V. Pin 7, the reset, is tied low (tied to pin 8). The outputs, in order of descending frequency, are pins 3, 4, 5, and 6.

Sample Circuit (Very Sample)

Suppose we want a low-pass filter with a "maximally flat" response and a cut-off frequency of 7.5kHz. First, let us pick a divisor and a clock frequency suitable for this F0--say a ratio of 50 (a binary number of 11111), thus a clock frequency of 375kHz. Second, the Q desirable for a maximally flat response is 0.707, a binary input of 00010. The circuit follows:

Pin l goes to -V, pin ll goes to +V, and pin 8 is grounded. The unused inputs, pins 9 and l8 (band-pass and high-pass, respectively), are grounded. Pin l0, the low-pass input, goes through 47K to the signal source. Pin l2 is the output. (Load capacitance should not exceed 50pF; otherwise, feed pin l2 through a resistor of perhaps l0K to the load.)

Pin 7 goes to a clock of 375kHz. Pins l3 through l7 go to +V. For our Q selection, pins 2, 4, 5, and 6 can go either to ground or to -V. Pin 3 goes to +V.

Notes on Q

For the band-pass and notch filter, Q is expressed in terms of bandwidth: Q equals the critical frequency divided by the bandwidth.

For the band-pass filter, the critical frequency will be at the peak of the response curve. The bandwidth is the difference in frequency between the upper -3dB frequency and the lower -3dB frequency. For example, if a band-pass filter is tuned for maximum output at l000Hz, and the -3dB points are found to be 1050Hz and 950Hz, the bandwidth is 100Hz; the Q will be 1000 over 100, or a Q of 10. If we know the Q to be 20, the bandwidth will be 1000 over 20, equals 50Hz; the -3dB points will be at 975Hz and 1025Hz.

This relationship holds true for the notch filter as well. In this case, the bandwidth is the difference in frequencies between the -3dB points at which the signal falls off as you approach the notch (not the points 3dB up in response from the bottom of the notch).

For high- and low-pass filters, the response will be "maximally flat" at Q equals 0.707 (an address code of 00010). The response will become "peaky" at Q's higher than this; a hump will develop before cut-off is reached. In fact, the gain of the filter at the critical frequency will be equal to Q. Therefore, you can see that with a Q of 10 (thus a boost in response of 10 times the gain at other desired frequencies) the filter will have what would usually constitute an undesirable effect on the signal. (Actually, the peak of this hump will be offset slightly from F0, and so there will be a point in our above example in which the boost in gain is greater than 10 times.)

Specifications

Supply Requirements

A split supply is needed; from + and - 4V to + and - 11V. Powered from + and - 10V, the current drain is 4.5mA, typical. Being CMOS and given this low-current drain, a single supply can be used if a suitably husky voltage divider is put across the single supply, the junction of the divider going to the filter's ground pin. (Jules Madey of Applied Invention recommends using two 220 ohm resistors for supplies of 8 to 15 volts.)

Clock Input:

Logic low--0.8V (above ground) or lower, not to go more than 0.5V below VSS.

Logic high--2V (above ground) or higher, not to exceed VDD.

Minimum trigger pulse width--200 nanoseconds

Range of acceptable clock frequencies--10Hz to 1.15MHz, can be extended to 2MHz for low values of Q.

Clock input capacitance--10pF maximum

Data Inputs (Both Q and Frequency Selection):

Logic low--0.4V (above ground) or lower, not to go beyond 0.5V below VSS.

Logic high--2V (above ground) or higher, not to exceed VDD.

(These are very high impedance inputs and are subject to damage when moving them from one logic level to another on a protoboard. I recommend committing them somewhere with resistors of perhaps 47K.)

Audio Filter Specs:

(These figures are given for a supply of + and - 10V, 20V total. Note that the noise and dynamic range figures are for very large signals, taking advantage of the filter's full output swing.)

Range of acceptable F0--0.05Hz to 25kHz.

Insertion loss--0dB.

Output swing--14V peak-to-peak (for Q equals 1).

Output noise--270uV (for Q equals 1; noise is broadband up to 1/4 of the clock frequency).

Dynamic range--94dB (for Q equals 1), 84dB (for Q equals 40).

Total harmonic distortion--0.2% (for 14V peak-to-peak swing).

Input impedance--1 megOhm, minimum.

Input capacitance--20pF, maximum.

Dynamic output impedance--10 ohms, typical.

Maximum load capacitance--50pF.

Maximum output load current--4mA.

Table I
Q Selection

Listed below are the binary input codes, followed by their respective Q. While the manufacturer says that these Q values will typically be within 5% of those given, they only guarantee that the actual Q will be 2/3 of those values given. Gee, thanks.

Table II
Programmable Frequency Divider

Listed below are the binary codes followed by their respective frequency divisor, the clock frequency over F0. As stated in the literature, you can pick a number which will get you within 3% of a desired critical frequency (the actual ratio from one to another is the 31st root of 4). The ratio of clock to critical frequency should be kept high at lower F0's to avoid the switching noise (which becomes more significant above 1/4 the clock frequency). For high filter frequencies, however, this ratio will have to be lower because of limitations of the op-amps (they don't say what will be the evidence of these limitations).

Table III
Pin Connections, R5620

*Addresses and Miscellaneous

For having gobs of fun selecting frequency ratios and Q's, hexadecimal switches can be used. These have l6 positions and generate a 4-bit binary number (a common terminal or arm is tied to a logic level). One such switch could be assigned to the four least significant bits on a set of inputs, while a toggle switch could be used to control the 5th bit. The arms of the switches can go to ground or to VSS when used with this filter chip.

Jameco Electronics, l355 Shoreway Rd., Belmont, CA 94002, phone (4l5) 592-8097, sells a thumb-well hex switch, catalog No. SF-53. In addition to the switch unit, "end plates" are required for mounting, SF-EP (come by the pair). You can stack up as many switches as you wish by using the end plates on either side of the stack and by putting "divider plates" between them (SF-DP).

The R5620 and other Reticon items are available from Applied Invention, R.D. 2, Route 2l, Hillsdale, NY l2529, phone (5l8) 325-39ll. The man who owns this company is Mr. Jules Madey; he was for many years the Chief Engineer on the staff of Smith- Kettlewell. Mention my name and you're "in like Flynn."

A TUNABLE AUDIO NOTCH FILTER USING THE RETICON R5620

The original name for this article was going to be "A Tunable Notch Filter for the Short Wave Listener." What elusive things are electronics projects--everything works on the bench and you proudly bring your SWL receiver into the shop hoping to hear Radio Nederlands without a piercing 5kHz interference tone. Your elation is short-lived as you turn on your filter only to find the harmonics of your clock signal put Radio Nederlands to shame as far as signal strength is concerned. Well--hmmm--you could do wonders for a tape of Radio Nederlands. I was able to turn this filter into a PA system on which I could notch out the dominant feedback frequency, and thus I could sing louder.

I am not ready to give up on the idea of a short wave filter, especially where a receiver with good shielding and an antenna with a shielded lead-in is used. The next obvious thing to try will be to enclose the filter circuit in a grounded metal box, enclosing an RF filter to "bottle up" the clock signal. The unit described illustrates what will be my first attempt along these lines.

The filter uses a Reticon R5620 programmable filter in the "notch" configuration (low-pass and high-pass inputs tied together). I chose to include an audio amplifier and speaker as part of my project. The filter could be built with the intention of feeding it into another amplifier and speaker system, provided the constraints on the filter's output circuit are observed (load capacitance must be under 50pF, and peak output current should not exceed 4mA).

The filter is driven by a tunable clock using a 555 timer (trying a CMOS 555 did not reduce RF interference from the clock). The clock is tunable over a range of 5 to 1. The 5 address lines of the frequency divider in the filter chip are all tied together and go to a toggle switch so as to give the user two ranges--one is 4 times the frequency of the other. In other words, one range will fall from l to 5kHz, while the other will cover from 250Hz to l2,500Hz.

A 3-position "Q" switch permits the user to select a Q which suits his purpose. My primitive Q-selection scheme is done more out of convenience than for the reasons of good science. Three independent toggle switches could be used to give the user more combinations of the Q inputs.

Background noise from the filter is evident with the gains chosen. The ambitious builder might consider preceding the filter chip with a preamplifier (an op-amp with a gain of perhaps 5), then reducing the gain into the audio output stage by increasing its 47K input resistor to perhaps 220K. This would better exploit the full dynamic range of the filter chip and put the signal well above the noise. However, as it stands now, the filter is much less noisy than the signals I was interested in processing with it.

As required by the filter chip, a split supply (plus and minus voltages about ground) must be used. I used two groups of 4 penlight cells to get plus and minus 6 volts. The audio amplifier, on the other hand, operates across the entire supply of l2 volts. Just to keep all input and output commons the same, I used the minus 6V line as signal ground.

Circuit

Two 6V batteries are connected in series, with their common connection going to the 5620 filter's ground pin, pin 8. Pin 9 of the filter, the unused bandpass input, is also tied to pin 8. A double-pole, single- throw switch must be used to turn the filter on and off. The positive side of the plus 6V battery goes through one pole to the plus 6V line, while the negative side of the minus 6V battery goes through the other pole on the switch to the minus 6V line. Pin l of the filter goes to the minus 6V line, while pin ll goes to the plus 6V line.

Pin l of a 555 goes to minus 6V, while pins 4 and 8 are tied together and go to plus 6V. A disc capacitor of 0.luF is connected between pins l and 8. Pins 2 and 6 are tied together and go through 0.00luF (mica) to the minus 6V line. Pins 2 and 6 also go through l.8K, then through a l0K rheostat to the output, pin 3. (This crazy scheme of driving a 555 with its own output gives nice broad pulses with which to clock the filter, as well as allowing for considerable range.)

Pin 3 of the 555, the clock output, goes to the clock input of the filter chip, pin 7 of the 5620. Pins l3 through l7 are tied together (these are the programmable frequency divider inputs) and go through 47K to the minus 6V line. These pins also go through a single-pole, single-throw toggle switch to the plus 6V line. (With the switch open, the divisor is 200; with the switch closed, the divisor is 50.)

The least significant bits of the Q selection pins, pins 2 and 3, are tied together and go to minus 6V. Of these address lines, pins 4, 5, and 6 each go through 47K to the minus 6V line. Pins 4, 5, and 6 go to positions l, 2, and 3, respectively, on a 3- position Q selector switch; the arm of this switch goes to the plus 6V line. (This gives us Q's of 0.87, l.35, and 5.0, in accordance with the binary numbers 00l00, 0l000, and l0000, respectively.)

The sleeve of the audio input jack is common to the metal cabinet housing the filter. This cold audio lead also goes through an RF choke (2.5mH) to the minus 6V line. The tip of the jack goes through another RF choke (2.5mH), then through 0.luF, then through 47K to both pins l0 and l8 of the filter which are tied together (these are low-pass and high-pass inputs, respectively). The input jack is shunted by 0.0luF; between the far ends of the RF choke is connected another 0.0luF capacitor.

The output of the filter, pin l2, goes through 47K, then through l0K to the minus 6V line (this l0K resistor could very well be a volume control). The junction of these resistors (or the arm of the volume control) goes through 0.luF to pin 3 of an LM386. Pins 2 and 4 of the 386 go to the minus 6V line. Pin 6 goes through l0 ohms to the plus 6V line; pin 6 also goes through 250uF to minus 6V (negative at the minus 6V line). Pin 7 goes through 25uF to minus 6V (negative at the minus 6V line). Between pins 4 and 5 of the 386 is connected 0.22uF.

Pin 5, the output of the 386, goes to the positive side of a l00uF capacitor, with its negative end going through a speaker to the minus 6V line. If an earphone jack is to be included, the output leads must also be isolated for RF. The aforementioned 2.5mH RF chokes will have too much resistance to permit duplicating this filter on the output. However, output leads from the board can be twisted together in a "twisted pair," after which these leads should be wrapped around a hunk of ferrite (preferably wrapped in a toroid through a ferrite toroidal coil form).

Parts List

Resistors:

Capacitors:

Switches:

Semiconductors:

THE SMITH-KETTLEWELL TRAINING AID FOR PIANO TUNERS

Abstract

This is primarily a tool to facilitate training of prospective tuners; however, its small size permits its being included in the tuning kit of veterans. The aid is an electronic filter which emphasizes the subtle beating of the harmonics of a pair of notes (4th and 5th intervals) which are played together in laying the temperament octave. It is not an "electronic tuning aid" (strobe tuner), the use of which is controversial in the piano tuning industry. Rather, it is designed to remove the subtlety of the sound of beats from intervals as they are tuned in the regular way. For those not in the piano business, this project contains a couple of circuits of interest; it well illustrates a use for the Reticon R5620 filter chip, and it contains an elemental PLL frequency synthesizer.

Introduction

The easy part in tuning a piano is matching up high and low notes of the keyboard to a previously tuned "temperament octave" in the middle register. The task which apprentices find hard to learn and for which even old salts must stay in practice is laying the temperament octave. As was discussed in "Singing Chips," Winter l982, the equally tempered scale is not based on ratios of whole numbers. Except for octaves, which are perfect, compromises are made throughout the temperament; intervals which are nearly whole number ratios are tuned slightly off in order to create equally spaced ratios within the l2 tones as we know them.

Purposefully creating the above imperfections within the temperament octave requires listening to subtle "beats" in the interval's "4th's" and "5th's." The "beats" of interest are not simply the difference frequency between two notes. The tuner is listening for the near coincidence of harmonics of each note; for example, the second harmonic of middle C and the third harmonic of the F below it both land on C above middle C. It is academic to say that these two notes should be tuned so as to beat once every l.7 seconds--it is quite another matter to pull this information out of all that noise. The notes themselves are louder than the harmonics you are listening for, the harmonics are not of equal intensity and no perfect cancellations occur, and higher harmonics are always there to confuse the issue. For training purposes, and perhaps on the job, a simple bandpass filter can sector out the "beep note" of interest so as to make it more audible. The Reticon R5620 is a filter which can be easily tuned to the desired beat note frequencies. [Tim Cranmer tells me that Helmholtz tried this scheme using tuned cavities (Helmholtz resonators), one for each interval. These and other tidbits can be found in his arduous work, "The Sensations of Tone."]

Design Considerations

One could make a very simple version of this device by driving the Reticon filter with a tunable clock; the user would adjust the filter to the desired harmonic every time an interval was to be tuned. I wanted the frequencies to be available on a selector switch, both for convenience and to afford rapid comparisons between adjacent intervals. What an engineering bonus it would be if the filter's programmable divider moved in steps of the equally tempered scale (the l2th root of 2 instead of the 3lst root of 4). Alas, this is not the case, and it was necessary to build a frequency synthesizer which follows the traditions of the equally tempered scale.

A "musical clock" for the filter was made using an MK50240N top-octave generator (organ chip) was used in conjunction with a CD4046 phase lock loop IC to make an equally tempered frequency synthesizer. This whole PLL system is then driven by a low frequency clock which need be adjusted only once. Because tuners sometimes wish to tune the piano at other than standard pitch (if a clunker is old enough, it won't stand the strain), the main low-frequency clock is variable over a small range.

A "Q" of about l7 was chosen experimentally; it was sufficient for attenuating the unwanted signals, yet leaving adjustment of the filter non-critical. The rest of the design was very straightforward; a microphone preamp picks up the signal through an electret condenser microphone in the tuning aid box, whereupon it is fed through the filter to an LM386 audio amplifier which drives an earphone.

Synthesizer

A PLL was set up with its main input signal being a low-frequency oscillator tunable over a small range (using a 555). Between the PLL's VCO and the VCO phase detector input is inserted a frequency divider--in this case, the frequency divider is selectable and is one of the set within the 50240 organ chip. In other words, the PLL VCO output goes to the clock terminal of the organ chip, and a desired output from the organ chip goes to the phase detector (pin 3 of the CD4046). On the other phase detector input (pin l4 of the 4046) is a main clock of 437.9Hz (this is not a musical pitch, as you will soon see).

Selection of the various outputs (divisors) on the organ chip causes the PLL VCO to run at an appropriate multiple which puts the two phase detector inputs in synch. Whichever output of the top-octave chip is chosen, it will be driven so as to duplicate the 437.9Hz tone of the main clock. The organ chip's clock input will be driven in accordance with the equally tempered scale as these outputs are presented to the phase detector.

As per the table of divisors in "Singing Chips," the smallest divisor on the 50240 is 239. Therefore, the PLL VCO will be running at a high frequency. It makes sense, then, to set the filter's programmable divider to a high value, 200 being convenient. The lowest "f0" (filter's critical frequency) we will need for tuning the temperament is C above middle C, 523.3Hz. The filter must be clocked at 200 times this frequency, l04.66kHz. After this, the frequency of the main clock into the PLL system can be found by dividing l04.66 by 239, 437.9Hz.

The Q of our filter is loose enough, however, to permit this 437.9Hz tone to be used as a reference pitch; the filter can be tuned by matching this clock frequency to 440Hz. A pushbutton is provided for this purpose.

Picking a higher divisor on the organ chip will make the PLL VCO run faster. For example, the largest divisor on the organ chip is 478; this would make the PLL VCO run at 209.32kHz, putting the f0 of the filter at 6th-octave C, l046.6Hz. Note that the outputs of the organ chip which before gave us its low notes (large divisors) give us high clock frequencies and high f0's; musically speaking, the sequence of its pins are reversed. A table of filtered "notes" versus chip pins is given later in the discussion of piano tuning.

I built my prototype in a bakelite box measuring 2 by 3 by 6 inches. Though I would sooner endure a flogging as do it the same way twice, I built my circuit on a piece of perforated Vector board measuring 3 inches square. The size of my circuit board was inappropriate, but my layout bears considering. The board, which should be perhaps 3 by 4-l/2 inches, can be divided lengthwise by the supply lines (both plus and minus 6 volts are required). A bus line for ground (the common connection between the two 6-volt supplies) runs along the top edge. The board then has a top and a bottom section separated by the supply buses.

With the component side toward you and the ground bus along the top edge, it makes sense to use the top section for the audio circuitry and the bottom section for the synthesizer. From left to right in the top section, first install the two-stage mike preamp, then the filter, and finally the LM386 audio output stage. Similarly across the bottom it makes sense to put the PLL in the middle, the organ chip to its left, with the main clock being at the lower right.

The microphone, the earphone jack, the volume and pitch controls, the on-off switch, the l2-position synthesizer selector, and the pitch reference pushbotton were all mounted on the top panel of my bakelite box. The Radio Shack microphone element is omnidirectional, so it does not appreciably affect pickup if it is not aiming directly at the piano. This mike element was cemented into a 7/l6 inch hole.

Constraints on the battery supply are rather severe; two packs of four AA cells were used in series to get the plus and minus 6 volts about ground. I would have preferred to use 9-volt batteries, but the organ chip is constrained to work under l6 volts, and the LM386 does not work on voltages much over l2 volts.

"Synthesizer noise" proved to be quite a problem; I never succeeded in eliminating it from the LM386. No doubt this is partly due to my poor choice of circuit boards, since some of the parts are literally sitting on top of each other. However, there are many devices generating audio frequencies contained in this project, and a constant din of organ noises and hum from the main clock can be heard, even with the volume turned down. Some reduction in noise from the main clock was achieved by switching to a CMOS 555. This device, along with the other specialty item, the organ chip, can be gotten from Jameco Electronics, l355 Shoreway Rd., Belmont, CA 94002; phone: (4l5) 592-8097.

[Since this paper was written, I was able to reduce the noise by l5 or 20dB simply by cabling (bundling together) the wires from the selector switch--a remarkable effect. In my original unit, I got further noise reduc tion by connecting 500uF between the plus and minus supply busses. Units built since then have had the advantage of better layout, and this capacitor proved to be unnecessary.]

Circuit

The positive side of the minus 6- volt battery and the negative side of the plus 6-volt battery are grounded. The minus 6-volt battery lead goes through one pole of a double-pole, single-throw toggle switch to the minus 6V line, while the plus 6-volt battery lead goes through the other pole on this on-off switch to the plus 6V line. The plus 6V line is bypassed to ground by l00uF (negative at ground); the minus 6V line is bypassed to ground by l00uF (positive at ground). A decoupling network for the electret microphone is included; the plus 6V line goes through 330 ohms, then through l0uF to ground (negative at ground). The junction of this resistor and capacitor goes to the supply lead of the microphone.

The cold microphone lead is grounded; the center conductor of its shielded cable goes through 0.luF (Mylar), then through l0K (metal film) to pin 2 of an LM358 dual op- amp. Pin 3, the non-inverting input of A1, is grounded. Pin 2, the inverting input, goes through 243K (metal film, value not critical) to the output of A1, pin 1.

The non-inverting input of A2, pin 5, is grounded. The inverting input, pin 6, goes through l0K (carbon composition) to pin 1. A 220K feedback resistor (carbon composition) is connected from output to inverting input on A2 (from 7 to 6). Pin 4 of this 358 goes to minus 6V, while pin 8 goes to plus 6V.

Pin 7 of the 358, the output of the two- stage mike preamp, goes directly to the bandpass input, pin 9, of the Reticon R5620 filter chip. The high pass and low pass inputs, pins l0 and l8, are grounded. Pin l2, the output, goes through 47K to the top of a l0K audio-taper volume control, with the bottom of this control being grounded.

Pin 1 of the 5620 goes to minus 6V, pin 8 is grounded, and pin 11 goes to plus 6V. Pins l3 through l7, the binary inputs to the programmable counter, are tied low (this can be either to minus 6V or to ground; I used minus 6V). (An address of 00000 sets the frequency divider to 200.) A number of 11000 is impressed on the Q-selection inputs (giving us a Q of about l7); pins 5 and 6 go to plus 6V, while pins 2, 3, and 4 can either go to ground or to minus 6V (I used minus 6V).

The arm of the volume control goes through 0.luF to pin 3 of an LM386 audio amplifier. Pins 2 and 4 go to minus 6V, while pin 6 goes through l0 ohms to plus 6V. Pin 6 is bypassed to minus 6V by l00uF (negative at the minus 6V line). A 0.22uF capacitor goes between pins 4 and 5; pin 5, the output, goes to the positive end of a 47uF capacitor, the negative end of which goes through the earphone to minus 6V.

A 555 or CMOS ICM-7555 is used as the main low-frequency clock. Pin l goes to minus 6V while pins 4 and 8 are tied together and go to plus 6V. A 0.luF capacitor goes between pins l and 8 located close to the chip. Pins 2 and 6 are tied together and go through 0.luF (Mylar) to the minus 6V line. Pin 6 goes through l0K to pin 7, while pin 7 goes through l0K, then through a 5K rheostat (pitch control) to the plus 6V line. Pin 3, the output, goes to pin l4 of the PLL.

To create a reference pitch for tuning the filter, pin 3 of the 555 also goes through 1.5 megOhms to the arm of an SPDT pushbutton switch. The normally open position of this switch goes to the top of the volume control; the normally closed contact goes to the bottom of this control and to ground.

Pin 8 of the PLL (CD4046) goes to minus 6V, while pin l6 goes to plus 6V. Located close to the chip and connected between pins 8 and l6 is 0.luF. The oscillator Enable, pin 5, also goes to minus 6V.

As a loop filter, pin l3 (the output of the wide-band phase detector) goes through 470K, then through l20K, then through 0.47uF (Mylar) to the minus 6V line. The junction of the two resistors goes to pin 9, the VCO input.

Pin 11 goes through 5.6K to the minus 6V line. Between pins 6 and 7 is 0.00luF (l000pF). Pin 4, the output of the VCO, goes to the clock input (pin 2) of the 50240 organ chip, as well as going to pin 7 (the clock input) of the 5620 filter.

Pin l of the 50240 organ chip goes to the plus 6V line, while pin 3 goes to the minus 6V line. Connected between pins l and 3 and located next to the chip is 0.luF.

As mentioned, the pin l4 input of the phase detector goes back to the main clock (the 555). The pin 3 input of the phase detector goes to the arm of a l2-position, single-pole switch. The various positions on this switch go to outputs of the organ chip as follows (the order in which they are connected depends very much on the type of temperament the tuner uses):

The above system only allows tuning of intervals within one octave in the middle register. There are many tuners who would choose to expand the range over which this filter is usable. The general approach for adding an octave switch is outlined below.

The filter can easily be raised an octave by programming its frequency divider differently; an input address of 10000 changes the filter's divisor from 200 to 97.8. Given our chosen "Q," 97.8 can be considered l00 without retuning of the filter being necessary.

The simplest way of cutting the filter's frequency in half is to insert a flip-flop between pin 4 of the PLL and the filter's clock input, pin 7. Other schemes, such as dividing the low-frequency clock or tampering with the PLL system in other ways, may necessitate readjustment of components of the loop filter.

We leave it up to the ambitious builder to design switching networks for these purposes. Remember in changing the address lines, it is perfectly acceptable to tie them low with pull-down resistors and use a single-throw switch to pull desired ones high.

By changing any of the components around the 555, the placement of this filter in a different range can be accomplished. The frequency of the main clock can be calculated by choosing the lowest harmonic involved in your temperament, and multiplying this frequency by 200/239. By trying different timing capacitors on the 555, a new range for the system can be chosen experimentally.

Besides depending on layout, interference of synthesizer noise to the user will depend heavily on the efficiency of the earphones. If this noise is unbearable, try inserting a resistor in series with the earphones; the value of this resistor will have to be chosen taking the impedance of the phone into account--for 8-ohm phones, try l00 ohms.

Piano Tuning

There are many different schemes for tuning a temperament; most primarily involve tuning 4th and 5th intervals alternately around one octave. There are, however, even variations within this scheme, and it will necessarily be up to you to build the tuning aid to your own specifications. It is hoped that the three tables to follow can help you in doing so. Following these, I will briefly describe the temperament scheme which I learned from my papa.

Table I
Synthesizer Connections

After each pin number of the organ chip is listed the "note" of the filter's critical frequency, followed by the interval for which this frequency has been chosen. My temperament scheme includes the notes from 3rd octave F through 4th octave E. These octave signs will not be included in the table. Note that the filter's critical frequencies all fall within the 5th octave.

5ths:

4ths:

Table II
Beat Rates

Within the range of 3rd octave F to 4th octave E, the beat rates for 5th and 4th intervals are listed. Two ways of expressing these beat rates are given; the first is generated by an Al-Alden computer program, with the second being approximate "seat of the pants" figures gotten from handbooks of the l920's.

5ths (these are "contracted intervals"; they are tuned so as to be narrower than perfect):

4ths (these are "expanded intervals"; they are tuned so as to be wider than perfect):

Table III
Temperament a la Williams of Gerrey

This lists the sequence of intervals which I use, and accounts for the connections to the synthesizer switch from the pins of the organ chip as listed in the circuit. The first note listed in the interval is the one being tuned. At this time, the pitch control of the filter can be adjusted; push the reference tone button and tune it so that the reference tone matches 4th octave A, either from your fork or that on the piano.

Rather than tune a straightforward circle of 4ths and 5ths (which would result in an accumulated error over the whole sequence), the second half of the circle is tuned in reverse, using 3rd octave F as a secondary standard. As a result, notes during one half of the tuning are tuned just short of being perfect (flat), while notes in the second half are tuned just beyond being perfect (sharp).

  1. 5th octave C to a tuning fork, 523.3Hz.
  2. Middle C to 5th octave C.
  3. 3rd octave F, a secondary standard, to middle C (contracted interval, tuned sharp). Beats once in l-7/l0 seconds.

The following notes are tuned flat, short of being perfect:

  1. G to middle C, beats once in l-l/8 seconds.
  2. D to G, beats once in l-5/l0 seconds.
  3. A to D, beats once a second.
  4. E to A, beats once in l-l/3 seconds.
  5. B to E, beats twice in l-7/9 seconds.

The following notes are tuned sharp, just beyond the point of being perfect:

  1. A sharp to F, beats once in l-l/4 seconds.
  2. D sharp to A sharp, beats twice in l-7/8 seconds.
  3. G sharp to D sharp, beats once in 1-3/7 seconds.
  4. C sharp to G sharp, beats once in 1-l/l5 seconds.
  5. F sharp to C sharp, beats once in 1-6/l0 seconds.
  6. B to F sharp is a test; it should beat once every l-l/5 seconds.

Testing the Temperament

First of all, it is essential that after the circle of intervals is completed, the beat of the last interval (step l4) should agree with that specified for the interval (in this case, a 4th from F sharp to B). Other intervals can be tested along the way, such as major and minor 3rds. Because beat rates for these intervals are much faster, they are easier to hear and no filter should be needed in order to hear them.

I spot-check the sequence using major 3rds as I go along. For example, as soon as step 6 is completed, the major 3rd from F to A can be tested--it should flutter at 7 beats per second--with its beat rate being compared to the phrase "from Chicago to New York, from Chicago to New York, . . ." A general progression of the 3rds' beat rates should be looked for as the sequence of steps is followed. At the very end, ascending major 3rds should have smoothly increasing beat rates as you check on up through middle C to E.

Parts List

Resistors (carbon composition, l/4 watt, 5%)

Resistors (metal film, exact values not critical)

Capacitors (disc ceramic)

Capacitors (Mylar)

Capacitors (electrolytic, l2V)

Integrated Circuits

(all but the Reticon filter are available from Jameco. The Reticon filter is available from Applied Invention, R.D. 2, Route 2l, Hillsdale, NY l2529; phone 5l8-325-39ll).

Front Panel Items

THE SMITH-KETTLEWELL AUDIBLE STROBE TUNER

Abstract

In collaboration with Dr. T. V. Cranmer, Director of Technical Services at the Bureau for the Blind in Kentucky, we have devised a tuning aid whose auditory feedback is temporal in nature--the device "beeps" or "buzzes" each time the unknown note from the musical instrument drifts through a complete cycle of phase relationships with a frequency standard in the device. Three octaves of standard pitches are selectable with switches on the front panel. A microphone is placed near or directly against the musical instrument. Auditory feedback from the tuning aid is obtained either from its internal loudspeaker or through an earphone. If the musical aspect of this project does not interest you, portions of the circuit may very well be of use; there is a dynamite circuit for a crystal oscillator using NOR gates, and the project well illustrates the use of phase lock loops.

Introduction and Operation

It is a recognized fact that one's talent in playing a musical instrument need not be accompanied by the ability to service, maintain, or tune instruments. For this reason, visual "strobe-type" tuning aids are commercially available; sighted musicians who need them have access to devices which can provide visual feedback as to when their instrument is emitting proper pitches. Heretofore, an equivalent tuning aid has not existed for use by the blind; blind musicians have had the matching of pitches by ear as their only recourse. Since Tim Cranmer provided the inspiration for developing this device, our laboratory has nicknamed it the "Timmophone."

Our prototype was built into a so-called "amplifier cabinet" such as the Radio Shack 270-269 measuring about 3-l/2 by 8 by 6 inches. This cabinet is large enough to accommodate the rather large circuit board (9 IC's mounted on perforated board and wired point-to-point), as well as leaving room for a l2-volt battery supply made of 8 C cells.

The front panel of our unit contains the following controls:

  1. A l3-position switch selects the note to which an instrument is to be tuned. (This switch selects l2 notes of the equally tempered scale including the octave; for example, A-440 through A-880.)
  2. A 3-position octave switch puts the selected note into the appropriate octave; i.e., A-ll0 through A-220, A-220 through A-440, and A-440 through A-880.
  3. A knob for controlling listening volume is provided; this control includes the on-off switch.
  4. A toggle switch for squelch (to be discussed later) was included.
  5. A "double mike jack" such as those found on tape recorders, was included; a l/8 inch jack accepts the microphone signal, while a l/l6 inch jack accommodates the microphone switch (the mike switch becomes a remote squelch).
  6. Although our unit contained a loudspeaker, an earphone jack was provided; use of an earphone is recommended so as to avoid the problem of the device hearing itself.

In operation, the phase detector section of a phase lock loop integrated circuit (PLL, RCA type CD4046) is used to compare two independent signals; one is the pitch of the unknown note, and the other is a selected note from the internal standard. As the two input signals drift past each other in phase, a clearly discernible signal is obtained from a "lock-detection circuit" using a pair of external NOR gates as recommended in the RCA application notes on the PLL. To allow the phase detection circuit to pass the internal standard pitch when no signal is present at the microphone, deliberate degradation of performance of the "lock-detection circuit" is assured by a low-value filter capacitance.

The internal pitch generator uses an organ integrated circuit, which is clocked by a crystal oscillator followed by two D flip- flops as selectable octave dividers. This organ IC is the "top octave chip" discussed in David Plumlee's article, "Singing Chips," SKTF, Winter l982.

The microphone signal goes through a signal processing section which is best called a "pitch replicator"--it assures that the fundamental frequency is presented as a clean square wave to the phase detector. This pitch replicator is composed of four sections described as follows:

  1. The microphone signal is first amplified by a two-stage preamplifier.
  2. The input wave is then "squared up" in a Schmitt Trigger (using an RCA CA3l30). A Schmitt Trigger is a fancy term for a comparator using an op-amp with a slight amount of positive feedback to assure that it is bi-stable. An input signal has to overcome a certain amount of hysteresis as determined by the ratio of feedback to non- inverting input resistors.
  3. A nonretriggerable tuned one-shot (using a 555 timer chip) assures that no doubling of input frequency can occur; zero crossings of the input wave at other than the fundamental period are ignored.
  4. Finally, a PLL replicates input frequency and produces the necessary square wave for clean comparison in the phase detector to come later.

Because the phase detection circuit passes the internal standard pitch when no signal is present at the microphone, an optional squelch circuit is included to interrupt the audio output of the phase detector, and a rectified signal from the microphone preamplifier is necessary to open the squelch. A front panel toggle switch is provided to defeat the squelch circuit, as well as making this option available at the remote switch on the microphone. (Defeating the squelch permits the tuning aid to emit a reference pitch that the user can "aim for" in the initial adjustment of his instrument.)

Because of the project's complexity, some advice as to component layout may be warranted. (I certainly did it all wrong; what follows is how I would proceed were I to do it all over again.) The perforated "Vector" board should be made as large as possible; for example, a board whose dimensions are a generous 3 inches wide by 7-3/4 inches long can be secured to the rear panel of the cabinet (exact dimensions are not critical).

It is probably wise to run bus strips along each edge, which are tied together and considered ground. Another bus strip running lengthwise along the middle of the board can be the VCC line.

With the component side toward you, my inclination would be to put the microphone preamp and squelch rectifier at the upper right corner, with the 3l30 Schmitt trigger placed just below the VCC strip. The 3l30 circuit is simple enough to leave room for the squelch transistor below it in the lower right corner. To the left of the 3l30 can be placed a 555 tuned one-shot, leaving room below it for at least one of its timing capacitors. Above the 555 and above the VCC line (to the left of the mike preamp), the PLL for the pitch replicator can be placed.

To the left of the 555, I would then mount the PLL chip which is used only for its phase detector. Directly above it and on the other side of the VCC line, the 400l NOR gate chip can be positioned. Components of the crystal oscillator associated with this NOR gate package will take up some acreage; these can be fit into the space immediately to its left and above the VCC line.

An efficient use of space might then dictate that the LM386 audio amplifier be placed directly below the components of the crystal oscillator and to the left of the phase detector PLL. Finally, the 40l3 flip- flops can then be placed to the left of the crystal oscillator, with the organ chip put on the other side of the VCC line directly below the flip-flops.

The above suggestions will work best if the chips are oriented at right angles to the length of the board, as opposed to placing their in-line pins along the board's length.

These suggestions are offered for convenience, and not because the layout is critical; I had crystal oscillator and clock signals running lengthwise from end to end with such abandon as to be embarrassing.

The switches used were Centralab "Universal" types whose number of usable positions is mechanically set by moving metal tabs. For example, the l3-position unit is actually capable of working through l7 positions, and the 3-position unit can be expanded to 5 positions. When mechanically restricted to the number of positions noted, free "unused" positions can be used as tie points for components of the tuned one-shot. For example, position 5 on one of the poles of the octave switch was used to return two of the timing capacitors to ground, and positions l6 and l7 on one deck of the note selector were tied together and used as a common connection for the six 555 charging resistors.

The wafers of the Centralab switches are very, very fragile. In cutting and filing the shaft of the note selector to the right length, I broke one of its wafers simply holding it in my hand. Be careful!

While the squelch switch can be a simple on-off toggle unit, a fancier 3-position switch was used in our prototype. This unit had a center "off" position, a "momentary" (spring-return) action in one direction, and a standard snap action toggle operation in the other direction. The two positions were jumpered together so that moving in either direction closed the switch. Whether using the microphone remote switch or the toggle, closing the switch disables the squelch and allows the note to be heard.

The double mike jack can be made by mounting a l/8 inch unit and a l/l6 inch unit close together. A standard spacing for such microphone plugs is l centimeter between centers of the holes. If metric measuring instruments are not part of your tool collection, it is a handy thing to note that l centimeter is very close to 4/l0 of an inch. I made my pair of holes using l/l0 inch perforated Vector board as a guide.

I actually had trouble with noise due to ground loops when I secured both the microphone and earphone jacks to the cabinet. (As soon as I touched them both to the chassis, the organ tones became burbly.) The only way I could cure the problem was to mount the speaker jack with insulating "shoulder washers" to isolate it from chassis ground.

Circuit

The negative side of the l2-volt battery is grounded, while its positive terminal goes through the on-off switch (contained on the volume control) to the VCC line. The VCC line is bypassed to ground by 250uF (negative at ground). The sleeve and the switch contact of the microphone jack (l/8 inch closed-circuit) are tied together and go to ground.

The hot mike lead goes through luF (negative toward the mike), then through 2.21K (metal film) to the inverting input, pin 6, of an LM358 dual op-amp. The non-inverting input, pin 5, goes to VREF (to be discussed later). A feedback resistor of 332K (metal film) goes from pin 7 to pin 6 of this op- amp. The output, pin 7, goes through 2.2K to the inverting input of the second op-amp, pin 2 of the LM358. The non-inverting input, pin 3, goes to VREF. This second stage has a feedback resistor of 330K from pin l to pin 2. (These latter resistors are carbon composition.) Pin 4 of the LM358 is grounded, while pin 8 goes to the VCC line.

The output of the second op-amp, pin l of the LM358, goes through l00K to the inverting input, pin 2, of a CA3l30 used as a comparator (Schmitt trigger). The non-inverting input, pin 3, goes through l00K to VREF. This comparator has positive feedback; a 470K resistor goes from output to non-inverting input, from pin 6 to pin 3 of the 3l30. Pin 4 of the 3l30 is grounded, while pin 7 goes to the VCC line.

The output of the Schmitt trigger, pin 6 of the 3l30, goes to pin 2 of a 555. Pin l of the 555 is grounded, while pins 4 and 8 both go to the VCC line. A 0.luF capacitor goes between pins l and 8, located close to the chip.

Pin 6 of the 555 goes through 0.022uF (Mylar) to ground. Pin 6 also goes to the arm of one pole on the octave switch. Position 3 of this pole is left open, position 2 goes through another 0.022uF Mylar capacitor to ground, and position 1 goes through a 0.068uF Mylar capacitor to ground. (On the switch wafer, the unused position 5 is grounded and is used to secure these latter two capacitors.)

Pin 7 of the 555 goes through 2.2K to pin 6. Pin 6 also goes to the arm of a pole on the note selector. The following resistors go from their respective positions on this pole to VCC:

(The unused positions l6 and l7 are tied together and go to VCC--they are used to secure one end of each of these resistors.)

Pin 3 of the 555 goes to pin l4 of the PLL (RCA CD4046). Pin 8 of the PLL is grounded, while pin l6 goes to VCC. Pin l5, a Zener diode in the chip, goes through 47K to VCC, as well as going through l0uF to ground (negative at ground). This pin l5 is the aforementioned VREF.

Pins 3 and 4 of the PLL are tied together. Pin ll goes through 47K to ground. Between pins 6 and 7 is 0.033uF. Pin 5 is grounded.

Pin l3 of the PLL goes through 330K, then through 56K, then through 0.47uF to ground. The junction of these two resistors goes to pin 9. Pin 4 of this chip is the output of the pitch replicator.

Another PLL chip (4046) is used strictly for its phase detector. Pin 8 is grounded, while pin l6 goes to VCC. To disable the oscillator, pin 5 goes to VCC, while pin 9 is grounded. Pin 4 of the pitch replicator (the previous PLL) goes to pin 3, one phase detector input, of this latter IC. Pin l4, the other phase detector input, goes to the arm on the main deck of the note selector.

The audible signal is gotten off a modified lock detection circuit on the latter PLL chip. CD400l NOR gates are used; pin 7 of the 400l is grounded, while pin l4 goes to VCC.

Pin l of the latter PLL goes to pin 8 of the 400l. Pin 2 of the PLL goes to pin 9 of the 400l. Pin l0, the NOR gate's output, goes through l00K, then through 0.005uF to ground. This l00K resistor is shunted by a lN9l4 diode (anode toward the 400l). The junction of the diode cathode, the resistor and the capacitor, goes to pin l2 of the NOR gate. (While this second NOR gate is simply a buffer, its pin l3 input is operated by the squelch.)

Pin ll, the output of the second NOR gate, goes through 0.05uF, then through 470K to the top of a l0K volume control. The bottom of this control is grounded. The arm goes to pin 3 of an LM386. Pins 2 and 4 are grounded, while pin 6 goes through l0 ohms to VCC. Pin 6 is also bypassed to ground by 250uF (negative at ground). Pin 7 goes through 25uF to ground (negative at ground).

To suppress oscillations, 0.22uF is connected between pins 4 and 5, very close to the chip. Pin 5 goes through l00uF to the tip of the earphone jack (positive toward pin 5). The sleeve of the earphone jack is grounded. One side of the speaker goes to the sleeve, while the other side goes to the switch contact on the jack.

Two NOR gates are used for the crystal oscillator; one is used for the oscillator, while the other is simply a buffer. Pins l and 2 of the 400l are tied together and go through 22 megohms to pin 3. Pins l and 2 also go through a 24pF mica capacitor to ground. Pins l and 2 also go to one side of the crystal. Pin 3, the output of the first gate, goes through 22K, then through a l50pF mica capacitor to ground. The junction of the 22K and l50pF goes to the other side of the crystal. (The crystal is designed for 32pF shunt capacitance, and has a fundamental frequency of 2l0.3kHz.)

Pin 3 of the 400l goes to pins 5 and 6 on the same chip, with pin 4 being the buffered output of the crystal oscillator. This pin 4 goes to position 3 on the second pole of the octave switch, as well as going to pin 3 of a dual-D flip-flop (CD40l3).

Pin 7 of the 40l3 is grounded, while pin l4 goes to VCC. The various "sets" and "pre- sets" are all grounded, pins 4, 6, 8, and l0. To make the flip-flops toggle, each section has its D terminal going to its own Q bar; pin 5 to pin 2, and pin 9 to pin l2. Finally, the flip-flops are connected in cascade; pin l, the output of the first, goes to pin ll, the clock of the second.

Pin l of the 40l3, the output of the first flip-flop, goes to position 2 on the octave switch. Pin l3, the output of the second flip-flop, goes to position l of the octave switch. As mentioned, position 3 of this pole on the octave switch goes to the output of the crystal oscillator.

The arm of this pole on the octave switch goes to the clock input, pin 2, of the MK50240 organ chip. Pin 3 of the organ chip is grounded, while pin l goes to VCC. The following pins on the organ chip go to their respective positions on the note selector switch:

As mentioned earlier, the arm of this pole on the selector switch goes to the pin l4 input of the phase detector PLL.

Pin l3 of the 400l goes through 47K to VCC, and pin l3 also goes to the collector of a 2N2222. The emitter of the 2222 is grounded. Its base goes through 22K to ground.

Pin l of the LM358 (the output of the dual stage mike preamp) goes through 2.2K, then through luF (positive toward the resistor), then to the cathode of a diode (lN9l4), the anode of this diode being grounded. The junction of the capacitor and the diode's cathode goes to the anode of another lN9l4 diode; the cathode of this diode goes through luF to ground (negative at ground). The positive end of this latter capacitor goes through l00K to the base of the 2N2222.

The collector of the 2N2222 goes to the tip of the "remote" jack, with the sleeve of this jack being grounded. This jack is shunted by a toggle switch on the front panel.

Model II

Two improvements to the design of this instrument bear considering. First of all, the squelch rectifier off the output of the mike preamp is somewhat particular about the wave shape of the musical instrument. Secondly, the device could be made more sensitive to the "out of tune" condition. Embellishments on the circuit which might improve performance are listed below:

Squelch

As it stands now, some instruments require that the microphone be placed very close to or in direct contact with them in order to open the squelch; this is true because of asymmetry in their wave shape (at present, the rectification scheme is half wave).

One could try using a full-wave rectifier to detect the audio signal. A good start would be to try the "absolute value circuit" in Al Alden's "Op-Amps" article in the Spring l983 issue.

Another idea might be to get the squelch voltage elsewhere in the circuit. For example, it might be possible to use the VCO signal from the first PLL to open the squelch; when this PLL replicates a frequency, a voltage will appear on pin l0 (a buffered version of the VCO input). (Pin l0 needs a load resistor to ground in order to function.) First try the simple idea of running pin l0 of the PLL through l0K to the base of the 2222. A problem to anticipate is that at low frequencies (with the octave switch in position l), there might not be enough voltage to bias the transistor on.

Instrument Sensitivity

As it stands now, this device beeps once every time the musical instrument goes through a complete cycle of phase crossing with the internal standard. It would be a simple matter to double this rate of "beeping" for a given out-of-tune condition.

This modification would simply require that the phase comparisons be done an octave high. Both the standard and the pitch replicator must be changed to accomplish this end. First of all, double the crystal frequency so as to run the organ chip and its octave dividers at the higher frequency. Second, it is a simple matter to make the pitch replicator PLL into a frequency multiplier by inserting a flip-flop (divider) between pins 4 and 3. (Of course, the value of either the VCO's resistor or its capacitor must be cut in half so as to permit the PLL to cover this higher range.)

In principle, this approach could be carried to extremes by multiplying the two frequencies by any number, but the beeps from the phase detector would soon be out of audible range. (You might have to reduce the value of the 0.005uF capacitor in the lock detection circuit in order to pass the higher frequencies of the beeps.) The main disadvantage in this scheme is that with the squelch disabled the note you hear will be that of the organ chip an octave higher than the musical instrument in the above example.

Parts List

Capacitors (mica):

Capacitors (Mylar):

Capacitors (disc ceramic):

Capacitors (electrolytic, l2V):

Resistors (metal film)

Resistors (l/4 watt, carbon composition, 5%):

Semiconductors

(available from Jameco Electronics, l355 Shoreway Rd., Belmont, CA 94002; phone: 4l5-592-8097):

Crystal

(available from your favorite crystal company; for example, International Crystal Company, P.O. Box 26330, Oklahoma City, OK 73l26; phone: 405-236-374l).

Jacks and Switches

(available from Newark Electronics, 500 N. Pulaski Rd., Chicago, IL 60624; phone: 3l2-638-44ll)

Cabinet:

EDITOR'S CORNER

A Debt Paid to Heritage

Having grown up in a piano store, the option of pursuing this field as my career was always open. As I reached adolescence and began to seek career advice, however, there were those among my counselors who actively discouraged me from taking that road, partly because the profits made are small and partly because it is such a traditional occupation for the blind as to be a stereotype. Though my ultimate career choice was not the piano business, I am proud of my engineering projects in support of it (an audible vacuum gauge for repair of player pianos and the piano tuning training aid of this issue). Financially speaking (even for the moonlighting amateur like me), the practice of the craft is flexible enough so as to save our bacon every once in a while, and I would like to speak a moment in defense of it.

No doubt the emergence of low-cost portable keyboard instruments will cut into the piano's ultimate role in the music business. However, although I cannot verify it, I have heard that more pianos were being made in this decade than at any other time in history. In any case, there are millions of them, 98 percent are out of tune, and well over half need the touch of a craftsman to put them in good working order. Being a qualified piano technician is a craft which is almost depression-proof; canvassing churches, schools, and residential neighborhoods will generally find you work, and in the worst of times you can always barter tunings for meals. (By the same token, you will never get rich; the profits are low and you will burn up at least a third of them in paying a driver.) Finally, making something sound good and work properly is always rewarding.

My sentimental attachment to the field is very personal. It goes back to a father teaching his small child--not out of patient indulgence, but out of genuine enthusiasm-- the principles of bearings and springs, the relationships of levers, the properties of various materials, and the elements of harmonic physics. To that shop I owe my basic mechanical aptitudes which were nurtured at an early age. To that shop I owe my basic understanding of running a small business. To that shop I owe my college education and the financing of my house. I have said before that I owe my career choice to Braille Technical Press and to its constituents. To the piano shop I owe the fact that I grew up in what they now call an "enriched environment" (known to be an advantage in child development); from that environment I got the basic skills of my employability.

A lot goes into an invention, such as my piano tuning filter. It took Al Alden to find for me the filter chip. Al Alden taught me everything I know about phase lock loops. By reading my own magazine ("Singing Chips" by David Plumlee, Winter l982), I learned of the organ chip with which I could clock the filter. It turns out that the basic experiment of using such a device was done a century ago by Helmholtz. (Buckminster Fuller once said that all of his better ideas had a mustiness about them.)

Yet, in a little black box called the "Smith-Kettlewell Training Aid for Piano Tuners" is something undeniably mine, a device to ease the training of new apprentices.

Of utmost importance is my work for this magazine; it is, no doubt, my contribution of highest impact. However, as far as engineering projects are concerned, my instruments for piano work are those by which I would most like to be remembered.

* * * * * * * * * * * * * *

Confidential Memorandum to Braille and Talking Book Readers:

The reason this issue is ever-so-slightly delayed is that I shattered an ankle in late Spring, and am dictating it over phone lines from my wheelchair. By the time you receive this, I will no doubt be walking again. Sssssshhhhhh--the official story around work is, "The doctor says he might/can hitch himself up on one elbow by Christmas!" In any case, this issue is a good one; what's worth having is worth waiting for.

Sincerely,

Your Monopeditor