A Quarterly Publication of
The Smith-Kettlewell Eye Research Institute’s
Rehabilitation Engineering Research Center
William Gerrey, Editor
Original support provided by:
The Smith-Kettlewell Eye Research Institute
and the National Institute on Disability and Rehabilitation Research
Note: This archive is provided as a historical resource. Details regarding products, suppliers, and other contact information are original and may be outdated.
Questions about this archive can be sent to
Table of contents
By Albert Alden
Last issue in the first article in this series we described the
ideal errorless op-amp and gave some applications. This issue we will treat the limitations of practical op-amps and show a number
of additional applications.
We will first discuss the static errors and then the dynamic
Input Offset Voltage
With an ideal op-amp, shorting the input
terminals together and connecting them to ground results in an
output of zero volts. In a real op-amp, a small voltage must be
applied between the input terminals to obtain zero volts out.
This externally applied voltage cancels an internal offset
voltage error of the opposite polarity. The source of this error
is a mismatch between the base-emitter voltages of the input
transistors of the inverting and non-inverting inputs. Eos is typically 2 millivolts for a 741. For both the inverting and
non-inverting amplifier circuits, the error due to E offset is given
Eo = Eos (1 + (Ro/Ri))
or the error is equal to the quantity one plus the gain all times
the offset, for an inverting circuit; and for the non-inverting
circuit, the gain times the offset.
For example, an inverting amplifier with a gain of 10 using an
op-amp with an input offset voltage of 2 millivolts will have an
output error of 22 millivolts.
Note that the input offset voltage specified may be of either
In many op-amps the offset voltage may be nulled with an
external potentiometer. For the 741, a 10K pot is connected
between pins 1 and 5, and the wiper is connected to the negative
power supply (-V). What this does is give one side or the
other of the differential input stage more or less current, thus
causing a base-emitter voltage change to cancel the initial
Precision op-amps with low offset voltages are available which
can be used where trimming isn't desired.
If in your application you don't need gain at DC, then
capacitively coupling the output eliminates the offset voltage
Offset voltage changes with temperature; a change of 5uV per
degree centigrade is typical for general purpose op-amps such as
Input Bias and Input Offset Current
All op-amps require that a
certain DC current be supplied at each input. This current
ranges from picoAmps to microAmps, depending on the device. This
is called the input bias current (Ib). The value given in
specifications for op-amps is the mean of the currents for the
inverting and non-inverting inputs. The 741 has a typical bias
current of 80 nanoAmps. (Note: Don't confuse this parameter
with input impedance; they are not the same.)
The error due to Ib for both the inverting and non-inverting
amplifier circuits is given by Eo = IbRo, where Ib is
the input bias at the inverting input of the op-amp.
For example, with Ib equals 80 nanoAmps and Ro equals 100K, the
output error will be equal to 8 millivolts.
This error can be reduced by taking advantage of the bias
current into the non-inverting input of the op-amp. By placing a
resistor in series with the connection to the non-inverting input
(i.e., between the non-inverting input and ground for the
inverting amplifier circuit, or between the non-inverting input
and the signal source for the non-inverting amplifier), the
output error then becomes a function of the difference in the two
bias currents. This difference is called the input offset
current, Ios. The value of the series resistor is selected
to be equal to Ro in parallel with Ri.
R3 = (RoRi/(Ro + Ri)
The output error is then
Eo = IosRo.
For the 741, Ios is typically 20 nanoAmps. This will give
an error of 2 millivolts in the above example with the proper
value of R3 inserted in the circuit.
For each application the user should check to see if the offset
is low enough without R3. If too high, calculate the offset with
R3 to see if it meets your requirements. If not, consider using
lower value resistors in your circuit or an FET input op-amp
where Ib is typically 20 picoAmps.
Input bias and input offset current change with temperature.
Typical values for the 741 are:
Input bias current--0.25 nanoAmps per deg. C.
Input offset current--0.025 nanoAmps per deg. C.
The gain of our ideal op-amp was assumed to be infinite.
In actuality it is quite high, ranging typically from 100,000 to
one million. In feedback terminology, this is called the open
We will digress a bit at this point to say a few words about
feedback systems in order to understand the effect of gain on op-amp circuits.
A feedback system is one in which some function or portion of
the output is fed back to, and combined with, the input to the
system. In a negative feedback system, the portion of the
output fed back is subtracted from the input of the system before
being amplified. Positive feedback systems add the feedback to
the input. Positive feedback usually results in oscillation.
In general, negative feedback does everything good for a
system--it increases bandwidth, improves linearity, reduces
distortion, reduces output impedance, increases input impedance,
and stabilizes the gain. The main detrimental effect of negative
feedback is a reduction in the gain of a system.
The gain of an amplifier with negative feedback is G equals A
divided by the quantity l plus A times B, where
G is the closed loop gain (Eo over Ei)
A is the open loop or forward gain
B is the fraction of the output fed back
This equation can be applied to op-amp circuits.
For our non-inverting amplifier circuit, B = Ri/(Ro + Ri) (i.e., the input and output resistors form an attenuator).
Note that if the product A times B is much larger than one, the
equation for G simplifies to G = l/B, which becomes
G = Ei((Ro/Ri) +1)
which is the same expression derived in last quarter's article.
If we put in some actual values in the equation for Eo/Ei,
we can see the effect of not having infinite gain in the op-amp.
Let: A = 200,000
Ri = 1K
Ro = 9K
Then: B equals 0.l
We get G equals 9.9995--not much of an error!
The quantity AB is called the loop gain of the system
and can be considered a figure of merit. It is approximately the
factor by which the open loop gain is reduced and its reciprocal
(l/AB) is approximately the error made in assuming
infinite gain. In the above example, l/AB = 0.005%.
Analyzing the inverting amplifier op-amp circuit as a feedback
system is not as obvious as for the non-inverting amplifier
circuit. Skipping the proof we get:
A' = (Ro/(Ro + Ri))A
B = Ri/Ro
Loop gain = ARi/(Ro + Ri)
A' is used to indicate that the forward gain in this case is
not the same as the open loop gain A. [Editor's Note: "A"
followed by an apostrophe is standard mathematical notation; it
is pronounced "A prime."]
End of digression. Conclusion: Infinite gain is a pretty good
Input and Output Impedance
Although the input impedance of an
op-amp is not infinite and the output impedance is not zero,
these parameters may be considered to be such in op-amp
applications using negative feedback. They are each improved by
a factor approximately equal to the loop gain.
In any multistage amplifier such as an op-amp, each stage has a
finite frequency response which is usually determined by its
output resistance and a capacitance. This R and C form a low-pass filter resulting in a rolloff of 6dB per octave above the
cutoff frequency. Associated with the rolloff is a phase lag of
the signal approaching 90dg, or one quarter period. As the
rolloff for each section comes into play at higher frequencies,
the open loop gain rolls off faster at an additional 6dB per
octave and an eventual 90dg lag for each section. This presents
At some frequency the open loop phase lag will reach 180dg or,
in other words, a sign change. If we were to apply feedback
around the op-amp, we would have positive feedback; and if the
loop gain were greater than one at the 180dg phase lag frequency,
we would have an oscillator. To make an op-amp a versatile
device which can be used in circuits with wide ranges of closed
loop gains, we must assure an open loop phase lag of less than
180dg for all frequencies where the open loop gain is greater
This is done by adding enough capacitance to that stage of the
op-amp circuit with the lowest frequency rolloff so that the open
loop gain drops to below one at the frequency of the start of the
rolloff of the second lowest frequency section. We now have an
op-amp which has 90dg (or less) phase lag up to the frequency
where the gain has rolled off to one. This scheme is called
dominant pole compensation. This gives us a device which is
stable for a wide range of closed loop gains. The unity gain
frequency is called Ft. For the 741, Ft is typically 1.2
The compensation capacitor comes built-in for most op-amps,
including the 741. Others, called uncompensated op-amps, require
an external capacitor, usually around 30 picoFarads. These units
are versatile in that different compensations may be used to
increase or decrease bandwidth to match particular applications.
The uncompensated 74l is a 748; the external capacitor goes
between pins l and 8.
With the open loop gain falling off with increase in frequency,
there will be some point where the loop gain is not sufficient to
keep the closed loop gain (G) dependent only on Ri and Ro. For
dominant pole compensation op-amps, the closed loop corner
frequency is given by the following equations:
Non-inverting amplifier: Fc = Ft/G
Inverting amplifier: Fc = Ft/(G + 1)
where Ft is the unity gain frequency. At Fc, the gain is down 3dB, or 0.707 of the low-frequency value, and the output signal has picked up a 45dg lag. Above Fc, the gain falls off at an initial rate approaching 6dB per octave.
An example: What will Fc be for a non-inverting amplifier with
a gain of 100 using a 741?
Fc = 1.2 megaHertz/100 = 12,000 Hertz
Not too good for hi-fi systems.
- Use a faster op-amp--one with Ft greater than 2 megaHertz.
- Divide the gain between two amplifiers, each with a gain of
Using a 741 in the second solution, each section will have a
bandwidth of 120,000 Hertz and a combined bandwidth (down 3dB) of
64% of 120,000, or 76,800 Hertz.
Input and Output Impedance vs. Frequency
It should be noted
that the input and output impedances change with frequency. The
input impedance of the 741 drops to about 400K ohms at one
megaHertz, and the output impedance goes up to about 300 ohms at
one megaHertz. Also, since the open loop gain is falling off
with increase in frequency, the effect of these parameters on
closed loop performance cannot be neglected, as was the case at
Slew Rate Limit
The slew rate limit of an op-amp is the maximum rate at which
the output voltage can change. This limit is caused because the
internal compensation capacitor has available only finite
currents for charging or discharging. For the 741, the limit is
0.5 volts per microsecond. If a signal is applied to an
amplifier circuit at an amplitude and frequency requiring the
output to change faster than the limit, the output waveform
becomes distorted. A slew rate limited sinewave takes on a
The maximum rate of change for a sinewave occurs at the zero
crossing and is a function of frequency and peak amplitude.
The maximum rate of change is given by the formula
Sr = 2π FVp
(or Sr equals 2 times pi times F times Vp),
where Sr is the maximum slew rate (which occurs at zero crossing).
F is the frequency of the sinewave
Vp is the peak voltage of the sinewave
By rearranging the above formula, we can calculate the maximum
undistorted frequency of a sinewave output of amplitude Vp using
an op-amp with a slew rate limit of Sr.
F = Sr/(2πV)p
(or F equals Sr divided by the product of 2 times pi times Vp)
For our 741 and Vp of 4 volts, F equals 20,000 Hertz--marginal
for hi-fi applications.
A Better Op-Amp
Some of the limitations of the 74l have become obvious during
the discussion of op-amp errors. While there are numerous op-amps of superior performance to the 74l, we will mention just one
of them in this installment. Next issue we will list several and
their important features.
The Texas Instruments TL07l
This unit is my present choice as
an all-round, inexpensive op-amp for the vast majority of
applications. It is an internally compensated FET input device
with standard (74l) pin connections. The recommended input
offset voltage-nulling circuit is slightly different from the
74l. A l00K pot is connected between pins l and 5, with the
wiper going through l.5K ohms to the negative supply.
Here is a summary of specifications of the TL07l. The values
given are typical.
Sr--------l3 volts per microsecond
Some More Op-Amp Circuits
Here are descriptions of three circuits using an op-amp and a
diode to simulate an ideal diode. Power supply values are
assumed to be the standard plus and minus l5V.
Active Peak Detector
The input to the circuit is the non-inverting input of an op-amp. Connected to the output of the op-amp is the anode of a diode, a lN4l48, for example. The cathode
is connected to a capacitor whose other lead is grounded. The
diode-capacitor junction is connected to the inverting input
terminal of the op-amp. Also, a connection is made from this
point to the non-inverting input of a second op-amp. The output
of this op-amp is connected to its inverting input; i.e., it is
connected as a follower. The output of the second op-amp is the
output of the circuit.
Here is an application where an FET input op-amp with its low
input current is required to minimize charge or discharge of the
capacitor when holding the peak signal. The choice of capacitor
value is a compromise between "droop" error and the ability of
the op-amp to supply the current required for changing the
capacitor's charge quickly enough to follow the input signal.
Absolute Value Circuit
This circuit appeared in the November
25, l982 issue of Electronic Design and was submitted to that
magazine by Stan Rubin of Ragen Data Systems.
The output of the circuit is a positive voltage equal to the
absolute value of the input signal.
The input of the circuit goes through a 50K ohm l% resistor to
the inverting input of the first (of two) op-amps. The output of
this op-amp is connected to the anode of a diode (lN4l48) whose
cathode goes to the non-inverting input of the second op-amp.
From the non-inverting input of the second op-amp, there is l0K
going back to the circuit's input. The output of the second op-amp is connected to its inverting input, and there is a 50K l%
resistor going from the output of the second op-amp to the
inverting input of the first op-amp. The circuit's output is the
output of the second op-amp.
This circuit consists, in its simplest form,
of one resistor, one diode, and one op-amp. The input is on one
side of a resistor (l0K is a good value). The other side of the
resistor is the output, and it is connected to the inverting
input of an op-amp. For positive limiting, the diode is
connected from the inverting input to the output of the op-amp,
with the anode at the input. To limit in the negative direction,
turn the diode around.
The non-inverting input is connected to the voltage level we
wish the signal to limit at. This may be a Zener diode, a
voltage regulator, a resistor divider across a regulated supply,
etc. A voltage follower on the output is needed if a low output
impedance is required.
In the three above circuits, the speed is limited by the time
needed for the op-amp to recover from saturation and the slew
Bandwidth Control on the Inverting Amplifier
It is often desirable to limit the bandwidth of an amplifier
more so than the limit imposed by the op-amp. A single-pole,
low-frequency and high-frequency limit can be implemented, each
with a capacitor. The low-frequency limit is set with a
capacitor in series with the input resistor, Ri, and the high-frequency limit by a capacitor in parallel with Ro.
In both cases, the equation for the limiting frequency is
F = l/(2πRC)
(or F = 1 divided by the product of 2 times pi times R times C)
This is the frequency where the response (or gain) is 0.707 of
the bandpass value (the -3dB point).
(Note: Cascading a pair of identical amplifiers will raise the
lower limit frequency by l.56 and lower the upper frequency limit
A great deal of the information contained in this article is
covered in the Application Notes published by National
Semiconductor, available together in the Linear Application
Also, I highly recommend the following book as a very readable
text on electronics; it is excellent.
Horowitz, P. and Hill, W.
The Art of Electronics
Cambridge University Press, l980.
Next month: A few more words about op-amp errors, a
"Chipography," and some more applications.
The LM3l7 (positive) and LM337
(negative) voltage regulators lend themselves
to "bootstrapping" in order to make adjustable
power supplies. Various circuit arrangements
are described, including a high-current supply
using pass transistors, and a high-voltage
supply using the LM3l7 in conjunction with a
triode vacuum tube.
Theory of Operation
[Editor's Note: Don't give up on these
devices if you do not immediately understand
this section; this brief ? explanation is
geared to the student of op-amps. The
circuits described won't know if you've read
this part or not.]
These units are basically high-current op-amps in the voltage-follower connection.
They maintain a constant output voltage by
comparing the IR drop across a "sensing
resistor" in the boot-strapping circuit with
an internal voltage standard. Since the
positive and negative units (LM3l7 and LM337,
respectively) are complementary, analysis
need only be done on one of them, the 3l7.
These animals are very similar to their
earlier counterparts, the "Three-Terminal
Monolithic Voltage Regulators" discussed in
SKTF, Fall l980. There are two important
- Their "Adjustment" pin, analogous
to the "common" terminal on their
earlier counterparts, introduces
very little current of its own into
a boot-strapping circuit--typically
50uA as compared to 5mA of the older
- They are all 1.25 volt regulators
which are intended for bootstrapping,
their basic function being that of
"series-pass regulators" for use at
voltages above this value.
Their connection is as follows: The
"Input" terminal goes to an unregulated
voltage source. The "Output" terminal is
the hot output of the supply, with the cold
output of the supply being ground. This
"Output" also goes through a "sensing resistor," R1 (perhaps 240 ohms), to the
"Adjust" pin; this "Adjust" pin then goes
through a rheostat, R2 (perhaps 5K), to
The internal circuitry of the device (which
is complicated past all understanding) is not
of use to us; however, an "equivalent
circuit" will be of value, two iterations of
which are given below:
- In its simplest form, the internal
workings can be viewed as an op-amp
which is connected as a voltage
follower--its output is tied directly
to its inverting input. The non- inverting input (not available on the
outside of the chip) goes through a
1.25V "battery" to the Adjust pin; in
this way, the voltage follower is
displaced from the Adjust pin by
- When the above fictitious device is
connected in the boot-strapping
configuration, it can be seen that
the output of the op-amp will pull up
on the external resistors until 1.25
volts develops across the "sensing
resistor," balancing this against the
1.25V battery in the regulator. An
attempt to pull down on the output
with a load would, at the same time,
decrease the voltage across the
external sensing resistor, thus
forcing the voltage follower upward
as determined by the internal
battery. As a result, the output
of the op-amp would correct for a
drop in voltage.
- The above equivalent circuit does
not account for the current which
exists in the Adjust pin, nor does
this circuit depict the true nature
of the regulator's "pass transistor."
The following is a more complete
- Instead of a battery, a 1.25V Zener
diode is used with its cathode on the
non-inverting input and its anode at
the "Adjust" pin. Supplying the
Zener is a 50uA current source.
- The output of the op-amp is no longer
the output of the regulator, nor is
it tied to the inverting input.
Instead, this output runs into a
Darlington amplifier; the op-amp
output goes to the base of a medium- power NPN transistor, with the
emitter of this unit going to the
base of an NPN power transistor.
tor. The collectors of both transistors go to the unregulated Input.
The emitter of the power transistor
is the Output and is tied to the op- amp's inverting input (thus including
the Darlington amplifier in the nega-
tive feedback loop.) (This Darlington
amplifier explains why you cannot
connect a load so as to pull up on
the output of the supply. Current
must always be drawn with respect to
Strictly from an external point of view,
the operation of these devices can be restated as follows:
The Output terminal goes through Rl, the
"sensing resistor" of perhaps 240 ohms, then
through R2, the rheostat, to ground. The
junction of Rl and R2 goes to the Adjust pin.
No matter where the rheostat is set, the
Output will pull on this resistor string
until 1.25V is developed across Rl. If the
setting of R2 is changed, the Output changes
accordingly to re-establish this 1.25V condition. The desired condition occurs when a
current of VREF over Rl (1.25V over 240 ohms)
is established through this "sensing resistor."
As promised by the manufacturer, putting a
load on the Output causes no change in the
Output voltage, VOUT. We can assume, then,
that a load has no effect on the currents in
the boot-strapping circuit. Therefore, the
following line of reasoning can be used to
derive a formula for the Output voltage as a
function of the resistors in the boot-strapping network.
Kirkhoff's Law says that our VOUT will be
equal to the sum of the voltages around the
resistor string of Rl and R2. (VOUT equals
the V across Rl, plus V across R2.) We will
consider these two elements separately as
- The manufacturer assures us that the
voltage across Rl will be maintained
at VREF, 1.25V. The resultant
current, VREF over Rl, is the main
component of the current in R2, and
we can use it in the calculations to
- There are actually two suppliers of
current in R2, VREF over Rl, plus
IADJ (the latter being current
supplied by the Adjust pin). The
voltage across R2 then becomes R2
times the fraction VREF over Rl,
plus R2 times IADJ.
Thank you, Mr. Kirkhoff. The grand total
is VOUT equals VREF, plus R2 times VREF over
Rl, plus R2 times IADJ.
Comparing the two currents through R2, we
see that IADJ is only 50uA, whereas VREF over
Rl (in our example) is 1.25V over 240 ohms or
5.2mA. IADJ is only about 1 percent of the
current through R2; it affects VOUT even less
than this. Therefore, we can drop its term
in the formula so as to get the expression:
VOUT approximately equals VREF,
plus R2 times VREF over Rl.
VOUT is 1.25 plus R2 times l.25 over Rl
[Zealous arithmetists use the associative
property of numbers to obscure the meaning
and generate the following formula given in
the book: VOUT equals VREF times the quantity 1 plus the fraction R2 over Rl, all the
above plus IADJ times R2. This expression is
These units have short-circuit protection
and thermal shutdown. Their internal voltage
standard, VREF, is rated to be within the
limits of 1.2V minimum and 1.3V maximum, with
1.25V being typical. The "drop-out" voltage
below which these regulators cease to function is about 2 volts above VOUT. The
current in the Adjust terminal is typically
50uA, 100uA maximum. Unless the device
carries an "H" or "HV" in its suffix, the
maximum unregulated voltage allowed between
Input and Adjust pins is 40V; units of the
"H" persuasion can accept up to 60V.
A wide variety of case styles is available,
each of which carries its own restrictions on
current handling and power dissipation. The
following list explains the suffixes (which
denote the case styles) and their ratings:
- K--TO3 power-transistor can; 1.5A, 20W.
- HVK--TO3 can; same as above, but good
- T--TO220; 1.5A, 15W.
- MP--TO202 (narrower than TO220); 0.5A,
- H--TO39 (sort of a low-profile TO5 can);
- HVH--TO39 as above, but good for 60V.
As an example of how these figures are
used, let us consider using an LM3l7T (the
popular TO220 package) in a variable supply
whose unregulated voltage is 40V. The above
table tells us that our regulator is good for
l.5 amps, l5 watts--whichever comes first.
With the regulator adjusted for 30V output,
the pass transistor will have across it a l0V
drop; under full load, it will be asked to
dissipate l5 watts (l0V times l.5A). However, if the supply were set for l0V output,
30V would appear across the pass transistor;
an attempt to draw l.5 amps would be asking
the unit to dissipate l.5 times 30, or 45
watts--this is not possible.
Circuits and Design Considerations
In this article, no great emphasis will be
placed on the input system--the unregulated
supply preceding the regulator. Sample circuits can be found in the original article,
"Three-Terminal Monolithic Voltage Regulators," SKTF, Fall l980, and in "A Talking
Meter . . .," SKTF, Fall l98l. Another
example will be presented in the "Split
Supply Circuit" presented in this section.
Meanwhile, in the ham radio operator's bag of
tricks are the following rules of thumb:
If a center-tapped secondary is used into a
full-wave rectifier (using two diodes), the
current rating of the transformer should be
about 1-1/2 times the desired DC load current.
If a full-wave bridge is used across the
whole secondary winding, the transformer
should be rated for twice the DC load current.
The secondary voltage should be chosen so
that the "peak" output of the filter (l.4
times the RMS secondary voltage) is a few
volts above the "drop-out" point of the
A general rule in picking filter capacitors
is to use 2,000uF per ampere of load current.
The "working voltage" of the capacitors
should be higher than the voltages which they
expect to see.
Basic Adjustable Positive Supply Circuit
suitable unregulated voltage goes to the
Input terminal of the regulator (LM3l7T, for
example). To suppress oscillations, this
Input is bypassed to ground with 0.luF
(located near the regulator). The Output of
the regulator goes through Rl (240 ohms),
then through R2 (5K rheostat) to ground; the
junction of these resistors goes to the
Adjust terminal. The supply's output
terminals are the regulator's output and
Improved transient response can be obtained
with a bypass capacitor from Output to
ground. However, if this capacitor is very
large, the possibility exists for it to
discharge back through the regulator and
damage the device. Therefore, the literature
suggests that if this unit is larger than
25uF, a protection diode should be connected
between the regulator's Input and Output
(cathode on the Input).
Improved ripple rejection can be gotten by
bypassing the Adjust terminal with l0uF.
Once again, a protection diode is required;
it is connected across Rl with its cathode on
Confusion exists throughout the literature
about picking values of the boot-strapping
resistors. The basic circuit for the 3l7
shows the above values, while others show Rl
as l20 ohms and R2 as 2K. The following can
be said regarding these choices:
Rl should be chosen so as to draw a certain
minimum load current; otherwise the Output
will float about aimlessly. The specifications list this required load current as
ranging from 3.5mA minimum to 7mA maximum.
Given a minimum VREF of l.2V and a maximum
required load current of 7mA, it would seem
that a safe value for Rl (which would work in
all cases) should be l70 ohms, l60 ohms being
a standard resistor value. For the sake of
argument, all the rest of the circuits will
use l20 ohms, as is often done in the Applications Notes.
Picking an appropriate R2 means deciding on
what range you want your adjustment to cover.
(The upper limit had better be 2 volts below
the "ripples" in the unregulated source. The
lower limit can be no lower than l.25V, as
determined by the regulator.) We can calculate R2 based on the upper limit by using our
approximate VOUT equation--rearranging the
terms as follows:
VOUT minus VREF equals R2 times VREF
over Rl, and therefore:
R2 equals the quantity VOUT minus VREF,
all times the fraction Rl over VREF,
R2 equals the quantity containing the
ratio of VOUT over VREF, minus l, this
quantity times Rl.
Suppose we want an adjustable supply with
an upper limit of 25V. Suppose also that we
have chosen an Rl of l20 ohms. Plugging
these values into the above, we have: 25
over l.25 (equals 20) minus l, all times l20
ohms. The answer is 2280 ohms, 2.5K being a
Adjustable Split Supply Circuit
tailor-made for those who wish to experiment
with op-amps.) This circuit uses both a 3l7
and a 337, running off complementary full-wave rectifiers to get plus and minus
voltages about ground. Each half is designed
to operate from l.25 to l5V at one-half amp.
A Stancor 36V lAmp center-tapped transformer is used (Stancor P867l). One side of
the primary winding goes to one side of the
AC line, while the other side goes through a
l/4Amp fuse, then through an on-off switch to
the other side of the AC line. The center
tap of the secondary is grounded.
For the positive supply, each end of the
secondary goes to the anode of a diode
(lN4003); the cathodes are tied together and
go through l000uF to ground (50V electrolytic, negative at ground).
These cathodes also go to the Input of an
LM3l7T--this Input being bypassed to ground
by 0.luF. The Output of the 3l7 goes through
l20 ohms, then through a l.5K rheostat to
ground. The junction of these two resistors
goes to the Adjust terminal; this is also
bypassed to ground by l0uF (25V electrolytic,
negative at ground). A diode (lN400l, 4003,
etc.) has its cathode on the Output and its
anode on the Adjust pin.
The 3l7 Output, which is the positive output of the supply, is bypassed by anything
from luF on up. This Output also goes to the
anode of a diode (lN400l, 4003, etc.), with
the cathode of this diode going back to the
regulator's Input and to the cathodes of the
For the negative supply, each end of the
transformer secondary goes to the cathode of
a diode (lN4003); the anodes of these diodes
go through l000uF to ground (50V electrolytic, positive at ground).
These anodes go to the Input pin of an
LM337T--this Input being bypassed to ground
by 0.luF. The Output of the 337 goes through
l20 ohms, then through a l.5K rheostat to
ground. The junction of these resistors goes
to the Adjust pin; this terminal is also
bypassed to ground by l0uF (25V electrolytic,
positive at ground). A diode has its anode
on the Output and its cathode on the Adjust
The 337 Output, which is the negative
supply output, is bypassed by anything from
luF on up (25V electrolytic, positive at
ground). This Output goes to the cathode of
a diode (lN400l, 4003, etc.), with the anode
of this diode going to the 337 Input and to
the anodes of the rectifier diodes.
High-Current Adjustable Positive Supply
This circuit uses a 3l7 in conjunction with three "monolithic power
transistors" (LMl95, 295, 395) which are good
for over l amp each. [More on these transistors later.] The connection of these
transistors is that of parallel emitter
followers; their collectors are tied together
and go to the unregulated supply, while their
emitters are tied together and become the
output of the supply. The bases are tied
together and go through 500 ohms to the
emitters. These bases go to the collector of
a PNP transistor (2N2905).
The emitter of the 2N2905 goes to the unregulated supply and to the collectors of the
pass transistors. The 2905 base goes through
5.lK to the Input of an LM3l7, with this
Input also going back through 22 ohms to the
The emitters of the pass transistors, which
are tied together, go to the Output of the
3l7. This Output also goes through l20 ohms,
then through a 5K rheostat to ground. The
rheostat is bypassed by l0uF (negative at
ground), while the l20 ohm resistor is
shunted by a diode (cathode on the output).
The output is bypassed by 47uF (negative at
ground). This Output also requires a 30mA
load--from Output to ground in order to turn
on the PNP transistor (forward-biasing it
through the 22 ohm Input resistor). At the
supply's lower limit of l.25V, this load
resistor will have to be 39 ohms.
[Editor's Note: It is unclear to me why
they don't affix this 30mA load permanently
into the system by changing the l20-ohm
sensing resistor to 39 ohms, then changing
the rheostat to a l.5K 2-watt unit. It's
worth a try, since a 39-ohm Output resistor
will have to dissipate 30 watts when the
supply is turned up. Also in their circuit,
the rheostat is generous by about 35 percent;
as it stands, there will be a dead spot over
the top third of its range.]
The pass transistors (LMl95, LM295, LM395)
are not simple transistors, but are monolithic IC's. A figure for "Beta" is not
meaningful, since their base current is a
relatively constant 3uA over a wide range of
base-emitter voltages; they are perhaps best
described as voltage-controlled, high-current
devices. As stated in the literature, they
lend themselves to being paralleled such that
one could use this same circuit to control a
hatful of them (the additional "base current"
of added units would not be significant).
Whether l95, 295, or 395, they are all
rated at "over l amp"; the l95 and 295 are
good for 42V, while the 395 is good for only
36V. (The principal difference between these
three is in temperature specifications, the
l95 being the most tolerant of hot and cold.)
LM395's with the "T" suffix are in a TO220
package and cost about $2.50 each. Units
with the "K" suffix are in a TO3 can, and
cost about $6.50 each. It seems that their
cases are not common to the collector, as is
usual; the case is the emitter in these units.
Regulation of High Voltages
As long as the
Input-to-Adjust voltage is restricted to 40
volts, these devices can be used at any
voltage above ground. The previously
described circuits can be used by lifting the
grounded end of the rheostat and putting it
atop a high-voltage Zener diode or gas VR
tube. However, the 40V functional range of
these devices is something to keep in mind;
this imposes severe restrictions at voltages
where 40V is not a significant percentage.
At 400V, for example, a 40V restriction on
the Input range represents a total of l0
percent--only variations of plus or minus 5
percent can be tolerated in the unregulated
The literature contains a circuit for a
l00V adjustable supply at unspecified current
which uses an LM3l7 in conjunction with an
unspecified triode vacuum tube.
Adjustable l00V Positive Supply Circuit
The unregulated input (unspecified) goes to
the plate of a triode tube; its cathode goes
to the Input of a 3l7. The Output of the 3l7
(which is also the output of the supply) goes
to the grid of the tube. From Input to
Output of the 3l7 is a 40V Zener diode, its
cathode on the Input. From the Output of the
3l7 to its Adjust pin is 600 ohms; a 50K
rheostat goes from Adjust to ground. [One
wonders what happened to the minimum load
restriction on the 3l7, in addition to which,
600 ohms is hardly a standard resistor
The Input of the 3l7 is bypassed to ground
by 0.luF. The Output is bypassed to ground by
luF (200V electrolytic, negative at ground).
Of course, the insulating materials used in
mounting the regulator must withstand the
voltages being considered. Capacitors at
appropriate voltage ratings must also be
chosen for use around these circuits.
- LM3l7T, LM3l7MP--TO220 and TO202, respectively. With the mounting surface toward you
and the pins facing upwards, the three leads
are: Adjust, Output, Input. The case is
common to the Output.
- LM337T, LM337MP--TO220 and TO202, respectively. With the mounting surface toward you
and the pins facing upward, the three leads
are: Adjust, Input, Output.
- LM3l7K, LM3l7HVK--TO3 can. Viewing the
unit from the bottom with the mounting holes
on the extreme left and the extreme right,
and with the pins being closer to the left-hand mounting hole, the connections are:
Input at the top, Adjust at the bottom, with
the case being the Output.
- LM337K, LM337HVK--TO3 can. Viewing the
unit from the bottom with the mounting holes
at the extreme left and extreme right, and
with the pins closer to the left-hand
mounting hole, the connections are: Output
at the top, Adjust at the bottom, with the
case being the Input.
- LM3l7H, LM3l7HVH--TO39 can. Has a triangular
lead configuration similar to a transistor
with a tab near the left-hand lead when
viewed from the bottom. The three leads are,
from left to right: Input, Adjust, Output
(Input nearest the tab).
- LM337H, LM337HVH--TO39 can. Has a triangular
lead configuration similar to a transistor
with a tab near the left-hand lead when
viewed from the bottom. The three leads are,
from left to right: Adjust, Output, Input
(Adjust nearest the tab).
A survey of stereo amplifier IC's
capable of delivering 2, 4, and 6 watts per
channel is presented. Example circuits are
given, including one with a transistor final
output stage, one having a "tone control,"
and monaural amplifiers using both halves of
the chip in combination to get double the
power. Their uses would include building a
portable power amplifier and loudspeaker
system for "Walkman-type" equipment, building
a monitor amplifier for tape decks being used
in on-site recording, or perhaps for supplying sufficient power in their monaural
connection to drive low-efficiency tactile
The devices discussed here are "power op-amps" capable of delivering generously high
currents to loudspeakers (primarily to 8- or
l6-ohm loads). An overall survey of several
IC's imposes practical limits on our discussion as follows:
Fine points such as individual distortion
characteristics will not be tabulated here;
it is hoped that general summary statements
will suffice. Likewise, the issue of picking
the smallest heat sink to serve the designer's
need for power dissipation will not be tabulated; it is hoped that a couple of examples
will suffice. Finally, these beefy op-amps
have applications other than audio--for
example, the LM379 is shown in its Applications Note as a driver for 2-phase motors.
We shall not fill your bookshelves with all
this material, but appropriate literature
will be listed with the pin diagrams at the
end of this article. The following chips
will be discussed in this survey:
- LM377, Radio Shack 276-702--Two watts
per channel in a l4-pin DIP (although
supplanted by the improved LMl877, it is
included because of its availability from
- LMl877--Two watts per channel in a l4-pin DIP (a pin-for-pin replacement for the
above 377, it has superior distortion
- LM378--Four watts per channel in a l4-pin DIP (very similar to the above, this chip
can operate on higher supply voltage).
- LM379--Six watts per channel in a l4-pin
DIP (this package carries a partial heat sink
which can be bolted down to a larger sink).
- LM2877--Four watts per channel in an ll-pin, single in-line package (has a mounting
tab for attachment to external heat sink).
Summary of Specifications
All units have current limiting and thermal
shutdown. Other than the 2877, which is
shown as driving a 4-ohm load, the other
units are recommended for 8- or l6-ohm loads.
All units require heat sinks if their rated
power is to be realized; the l4-pin packages
have multiple grounding pins for connection
to a heat sink, while the 379 and 2877 have
provisions for being bolted to heat sinks.
So as not to require a dual (split) power
supply (a plus and a minus supply as seen in
traditional op-amp circuits), these chips
contain a built-in voltage divider (made up
of two l5K resistors across the single VCC
supply). The output of this divider is a
VREF of l/2 VCC, available at a "bias pin."
(This pin is not present on the outside of
the familiar LM386 amplifier, but it is very
much a part of its inner workings.)
At medium power levels and at low frequencies, the total harmonic distortion (THD)
is less than 0.2% for all the units. Also in
every case, this distortion goes up markedly
(in some cases exceeding l percent) above
l0kHz; who cares, we can't hear the second
harmonic of l0kHz.
At higher power, these devices are not so
impressive. The 379 gives you a THD of
nearly 3% as the output approaches clipping.
The 2877 has a THD of l0% when it delivers
its 4 watts into a 4-ohm load.
The old 377 and the high-powered units have
"cross-over" distortion which becomes more
significant at low power levels. (This is
reminiscent of cross-over distortion in early
transistor hi-fi amplifiers whose output
transistors were biased near cut-off at
quiescence.) The distortion is a bit higher
than 0.2%, even at low frequencies, when
running at a power level of 0.05 watts.
Facts About Power
First of all, the realizable power is
limited by the usable voltage swing of the
amplifier's output; this limit being imposed
by the supply voltage which runs the chip (as
well as current limiting which will occur if
the load impedance is low). For example, two
watts of power can be delivered to an 8-ohm
load only if an RMS voltage of 4V can be
imposed across it. The power "P," the
voltage "E," and the load resistance "R" are
P = E2/R
Plugging in an R of 8 ohms and a power of 2
watts, we have: 2 = E2/8 -- E2 = 2(8) or l6 -- E = 4V.
Using a sine wave for our analysis, we can
define the necessary voltage swing as
follows: The peak value of a sinusoid is
l.4l4 times the RMS voltage; this gives us
the distance which the output must swing in
either direction from quiescence in order to
give us our needed 4V RMS signal. Peak
swings of equal amplitude in either direction
must be made, thus requiring that the output
be functional over a so-called "peak-to-peak"
voltage range of 2 times l.4l4 times our 4V
RMS value, or ll.3V.
The op-amps in these packages can only
bring their outputs to within about 2.5 volts
of the power supply extremes--from 2.5V short
of VCC down to 2.5V above ground. In order
to get a good, comfortable peak-to-peak swing
of ll.3V, the supply must be 5 or 6 volts
higher than this, perhaps l7 or l8 volts.
Therefore, in order to get 2 watts per
channel from any of these devices, an l8V
supply is required.
From the above, it can be seen that we
won't get full power by running these chips
from a l2-volt battery. Just for fun, we'll
perform the above calculations backwards to
see just watt can be realized.
From a l2V supply, we must throw away at
least 5 volts to allow for the boundaries of
the output range, thus giving us a functional
peak-to-peak range of 7V. Our attainable RMS
output voltage is gotten by dividing 7V by
2.828 (twice l.4l4), giving us 2.475V RMS.
"P" equals 2.475 squared over 8, which is
only 3/4 watt per channel.
It is interesting to note that the main
discernible difference between the 4-watt and
2-watt chips (the 378 vs. the 377 and l877)
is in maximum supply voltage ratings; the 378
has a rated maximum of 35V, while the low-power units specify a maximum of 26V. Their
rated "package dissipation" is the same, 4
Heat sinks are essential if these devices
are expected to operate at full power. While
no damage will occur if the heat sink is
insufficient, partial thermal shutdown tends
to cause gross "cross-over distortion" and
"flat-topping" as the chip temperature rises.
These problems disappear as the chip is
With a small heat sink, such as a 2-1/2
inch square of foil on the top side of the PC
board, the dissipation of heat will seriously
limit the audio power output as the supply
voltage is increased. For example, using a
377, the maximum power output which you can
expect from such a system will be l.l watts
per channel at a VCC of l8V; this will
decrease to 0.l5 watts per channel at a VCC
of 22V. Attaching an external heat sink to
this same chip will solve this problem to the
extent that the power output will increase
with a higher VCC.
Maximum power for any given setup will be
gotten when the choice of VCC (so as to
define an available voltage swing) is
balanced against the heat sink's ability to
handle the resultant power dissipation (4
watts in the best of cases for the smaller
units). (Notes for the smaller units
recommend a specific piggyback heat sink, a
Staver V7-l, but they also admit that this
system will not meet the 4-watt dissipation
spec of the ideal situation.) In short, the
rated power of these devices is available as
advertised, only if you run them off of
Hoover Dam and listen to your stereo in the
On the other hand, you need far less power
than you might expect to get a good listening
volume, especially from efficient loudspeakers. For example, I wired up the
monaural "bridge" amplifier described here,
using a 377 and a l2-volt power supply. (As
a heat sink, I mounted a piece of copper-clad
Vector board atop the chip.) The resultant
system, amounting to slightly over l/2 watt,
had a microphone on its input and a l2-inch
TV speaker on its output. As soon as I did
my Bing Crosby imitation, the degree to which
nearby office doors were slamming was
By the way, loudspeaker efficiency as a
function of size is badly misunderstood. For
any given baffle design, "air suspension,"
"bass reflex," or "infinite baffle," the
efficiency goes up as the speaker gets
larger. Those who say "Well, I don't need
very large speakers because my amplifier is
quite small," have their logic backwards; the
bigger the diaphragm, the more air it can
move. This does not apply when switching to
a different type of baffle; the little
speaker with the pseudo "infinite baffle" in
my transistor radio is more efficient than
someone's "air suspension" Wharfedales,
Practical Building Considerations
The 377, the l877, and the 378 have 6
grounding pins for connection to a heat sink.
Outboard units, such as the Staver V7-l heat
sink,* are recommended in the literature,
although these units will not dissipate
sufficient heat to bring you up to full power
output. I recommend the following schemes
for the home builder.
*Staver Company, Inc., 47 North Saxon Ave.,
Bay Shore, NY ll706.
Perforated Vector Board can be gotten which
is clad with copper on one side (such as the
l69P44Cl, 4-l/2 by l7 inches). Vector also
sells a tool which cuts through the foil so
as to isolate individual holes from the rest
of this "ground plane" (Pl38A). Point-to-point wiring can be done as usual on the non-foil side, using this tool to isolate each
and every hole receiving an ungrounded
component lead. Before a component is
inserted, the tool is used to isolate this
hole from the foil; this should be done for
the active pins of the chip as well. The
grounding pins of the chip can then be
soldered to the foil on the top (component)
In testing a couple of these circuits, I
used a piece of copper-clad Vector Board
which I attached to the top of the chip.
I bent 5 of the 6 grounding pins backward so
as to be sticking straight upward. Then a
piece of the copper-clad board (foil side up)
was set onto these pins and soldered.
A small copper plate could be used as a
heat sink by placing the chip on its back and
bending 5 of its 6 grounding pins down
against the plate. When soldering these pins
to the plate, you would do well to insulate
the top of the chip from the hot copper with
a little slip of braille paper. Also,
soldering will go a lot faster if you pre-tin
Remember which grounding pin you leave
sticking down to fit into the socket or PC
board; with a heat sink mounted atop the
chip, no other indication will be present as
to the location of pin l.
The 379 carries with it a partial heat sink
which extends beyond the ends of the chip,
bending down so as to meet the mounting
surface plane. Tapped holes in these end
flanges permit bolting of the unit to a heat
The 2877 is a long package, the upper edge
of which has a tab with a mounting hole by
which it can be secured to an external heat
The following circuits are taken from the
National Semiconductor literature; they are
typically shown for all of the chips. Noting
the similarity among circuits for these
various chip numbers, it is obvious that the
main differences between chips are their
packaging and/or maximum supply voltage
Only one channel of each case is given
here, since connections for both are
identical. In all cases, the grounding pins
are tied together and go to the negative side
of the supply. The VCC pin goes through a
switch to the positive supply. Except where
this supply is regulated (and to some extent,
even then), the VCC pin should be bypassed
with a healthy capacitor of 500uF or greater
(rated at greater than VCC and with its negative end grounded). Also in every case, the
bias pin is bypassed with a capacitor whose
rated voltage must be greater than l/2 VCC
(negative at ground). (Values of the latter
unit will be given individually in the
As we know from
"Operational Amplifiers," Winter l983, non-inverting amplifiers have a characteristically
high input impedance (in this case, l00K as
set by a resistor). If the bias pin were
low-impedance, one could return the inverting
input's resistor directly to it, thereby
leaving the non-inverting input to serve as
an "infinite-impedance" input. This is not
possible here; the bias pin is taken from a
voltage divider inside the chip, which is
made up of two l5K resistors, hence dictating
the bias arrangement shown. Aside from this
exception, this non-inverting amplifier is a
The cold side of the
input signal is grounded. The hot signal lead
goes through 0.luF to the non-inverting input
of the op-amp. This non-inverting input also
goes through l00K to the bias pin, with this
bias pin being bypassed to ground by 250uF.
There is a feedback resistor of l00K going
from output to inverting input. This
inverting input also goes through an "input
resistor," then through 5uF to ground (good
for l/2 VCC, negative at ground). Two values
of "input resistor" are shown; 2K gives you a
gain of 5l (l plus l00K over 2K), while the
other is 5l0 ohms, giving you a gain of about
200 (l plus l00K over 0.5l0K equals l97).
The output goes to the positive end of a
coupling capacitor (200uF or greater, rated
at greater than VCC). The negative end of
this capacitor goes through the speaker to
this circuit is straight out of the bastions
of Academe. It has one major disadvantage;
its input impedance is that of its input
resistor--this impedance gets low as the gain
is increased. Two circuits are given, one
whose gain is unity and whose input impedance
is reasonable, and the other whose gain is
50, having an input impedance of 2K. As we
know, the gain is the feedback resistor
divided by the input resistor.
Unity Gain Inverting Circuit
point, which is bypassed to ground by l0uF,
goes to the non-inverting input. The output
goes through l00K to the inverting input,
with this inverting input going through l00K,
then through 0.luF to the hot input signal
lead. The cold side of the input signal is
grounded. As in the non-inverting circuit,
the output goes through a large coupling
capacitor, then through the speaker to
High-Gain Low-Impedance Inverting Circuit
The bias pin, which is bypassed to ground by
l0uF, goes to the non-inverting input. The
output goes through a l00K feedback resistor
to the inverting input. This inverting input
goes through 2K to the positive end of a 5uF
input coupling capacitor (rated above l/2
VCC), with the negative end going to the hot
signal lead. The cold side of the input
signal is grounded.
As before, the output goes to the positive
end of a large coupling capacitor (200uF or
greater, rated above VCC), with the negative
end of this capacitor going through the
speaker to ground.
Stereo Amplifier with "Bass Tone Control"
for Ceramic Phono Pickups
Set for a flat
response, this amplifier has a bass roll-off
of about 75Hz. With a cross-over point at
about lkHz, the control can be used to cut
the bass l5dB and to boost it slightly more
than l5dB (this measurement being done at
The cold side of the ceramic cartridge is
grounded. The hot side goes to the top of a
lmeg audio-taper volume control, the bottom
of which is grounded. The arm of this
control goes through 0.luF to the non-inverting input; this non-inverting input
also goes through lmeg to the bias pin, which
is bypassed by 50uF.
The inverting input goes through 5lK, then
through l00uF to ground (negative at ground,
rated at greater than l/2 VCC). This
inverting input also goes through a 5l0K
feedback resistor to the arm of a l00K pot
(bass control). The top of this pot goes
through l0K to the output of the op-amp;
between the arm and the top of the pot is
connected 0.033uF. The bottom of this pot
goes through lK to the junction of the 5lK
input resistor and the l00uF capacitor;
between the arm and the bottom of the pot is
connected a 0.33uF capacitor.
The output goes to the positive end of a
very healthy coupling capacitor (250 or 500uF,
rated at greater than VCC), the negative end
of which goes through the speaker to ground.
Inclusion of an Add-On Transistor Amplifier
Listed as a "l0-watt amplifier," this
circuit can be added to any of the chips.
However, given the supply constraints as
discussed above, as well as recognizing that
an additional voltage will be lost across the
transistors, the only time you would get to
see this "l0 watts" is into a low-impedance
load (4 or 8 ohms). There's no way around
it, the supply voltage must be as high as you
can muster for this setup to be advantageous.
The circuit uses a "complementary pair" of
power transistors to make the op-amp system
beefy indeed. As can be seen, they are
included in the feedback network, and thus
they become part of the operational amplifier. (A complementary pair is made up of an
NPN and a PNP emitter follower connected
Circuit with Transistor Amplifier
side of the input is grounded. The hot input
lead goes through 0.luF to the non-inverting
input. This non-inverting input goes through
l00K to the bias pin; this bias pin is by-passed to ground by 250uF (negative at
ground). The inverting input goes through
2K, then through 5uF to ground (negative at
The collector of a PNP transistor is
grounded; the collector of an NPN transistor
goes to VCC. (The PNP unit is a 2N5l94 or
National NSP5l94. The NPN is a 2N5l9l or
National NSP5l9l.) The VCC line is bypassed
by l000uF (negative at ground, and rated at
greater than VCC).
The two bases are tied together and go to
the output of the chip. These bases also go
through 5 ohms to the emitters, which are
tied together. These emitters are the output
of the system; they go through a l00K feedback resistor to the inverting input on the
chip. To suppress oscillations, this feedback resistor is shunted by 82pF in series
with 27K. The emitters also go to the
positive side of a l000uF capacitor, the
negative side of which goes through the
speaker to ground.
Monaural "Bridge" Amplifier
The two op-amps in the package can be
connected in opposition to get a monaural
amplifier of twice the power. This configuration is called a "bridge" amplifier.
The speaker is connected between the two op-amp outputs; any jack for the speaker or
earphones must be appropriately insulated
One op-amp of the chip is set up as a non-inverting amplifier, with the other being
connected as an inverting amplifier. The
non-inverting section has the input signal
capacitively coupled directly into its non-inverting input. As was discussed in "Op-Amps," Winter l983, an op-amp in the non-inverting connection has a replica of the
input information on its inverting input--a
"virtual input signal." It is this virtual
input signal that drives the other half of
the chip, an inverting amplifier which
necessarily has a low impedance.
Another trick used in this circuit is that
both op-amps share a common input resistor--in this example, l0K. (A coupling capacitor
is placed in series with this resistor to
eliminate offset effects from direct
coupling. For the purpose of analysis, let
us pretend that this 0.47uF capacitor is not
With respect to the inverting amplifier,
everything looks straightforward and
standard. Its non-inverting input is fixed
to a bias point, VREF. It has a feedback
resistor of l megOhm going from its output
back to its inverting input. Its inverting
input goes through a l0K input resistor to a
signal source (in this case, a "virtual
signal point" on the other op-amp).
With respect to the non-inverting amplifier, an external signal is applied to its
non-inverting input. Its gain-determining
network consists of a l megOhm feedback
resistor (from output to inverting input) and
a l0K resistor from its inverting input to a
fixed "common" (in this case, "common" is a
"virtual bias point" on the other op-amp).
The l0K resistor mentioned in both cases is
the same unit, common to both sections. The
gain of each stage is calculated in the usual
way; for the inverting amplifier, the gain is
l meg over l0K equals l00, and for the non-inverting amplifier, the gain is l plus the
ratio of l meg over l0K equals l0l. The
speaker, straddling the outputs like a wishbone, sees both gains--the resultant bridge
amplifier has a gain of 20l.
Bridge Amplifier Circuit with Gain of 20l
The grounding pins on the chip are tied
together and go to the negative side of the
power supply. The positive side of the
supply goes through a switch to the VCC line,
this line being bypassed to ground by l000 or
2000uF (negative at ground). The VCC pin on
the chip goes to the VCC line. The bias pin
on the chip (VREF) is bypassed to ground by
250uF (negative at ground).
The input signal (either from the hot side
of an input jack or from the arm of a volume
control) goes through a Mylar 0.luF capacitor
to the non-inverting input of A1. This non-inverting input also goes through l00K to the
bias pin. There is a lmeg feedback resistor
from output to inverting input on A1.
The non-inverting input of A2 goes to the
bias pin. The output of A2 goes through a
lmeg feedback resistor to its inverting
input. The A2 inverting input goes through
0.47uF (Mylar), then through l0K to the
inverting input of A1. The speaker is
connected between the two outputs.
(Note: Catering to technicians who are
not, as we are, students of op-amps, the
manufacturer has chosen more simplistic
labels for the input pins--"Input" for the
non-inverting input and "Feedback" for the
inverting input. We sophisticates will get
more use out of these devices if we call
their pins by their true colors.)
LM377, LMl877, LM378, LM379:
(These and other circuits are given in National's Applications Note AN-l25.)
- Pins 3, 4, 5, l0, ll, l2--Ground
- Pin l4--VCC
- Pin l--Bias
- Pin 7--Inverting input, Al
- Pin 6--Non-inverting input, Al
- Pin 2--Output, Al
- Pin 8--Inverting input, A2
- Pin 9--Non-inverting input, A2
- Pin l3--Output, A2
(Circuits discussed in National's book, "The Audio/Radio Handbook, l980.)
- Pins 3, 6, 9--Ground (6 is actually the case; it can be grounded or left open, but it must not be used for circuit ground.)
- Pin ll--VCC
- Pin l--Bias
- Pin 2--Output, Al
- Pin 4--Non-inverting input, Al
- Pin 5--Inverting input, Al
- Pin 7--Inverting input, A2
- Pin 8--Non-inverting input, A2
- Pin l0--Output, A2
On the above single in-line package, the
pins are labelled l through ll, from left to
right. The pin 1 end of the package is
marked in two ways; there's a small cutout on
the end adjacent to pin 1, and a diagonal cut
on the corner of the mounting tab is larger
than at the end of pin 11.
[Editor's Note: I have just been informed
that the brand new Radio Shack listing for
the 276-702 is LMl877, the improved version.
However, this change is so recent that stores
with old stock may still sell you a 377 under
that same number.]
This discussion of the DRI
Industries'* "One-Hand Soldering Shop" iron
is in two parts. The first is an honest
attempt to objectively assess the merits of
the device as a tool. The second outlines
reasons why the Editor feels that thumbwheel
solder-feeding systems are not advantageous
to the blind user; these points are questions
for debate, and rebuttals for publication
here are encouraged.
*DRI Industries, Inc., lll00 Hampshire
Avenue South, Bloomington, MN 55438
Features of the Tool
The "Soldering Shop" comes with
a rather complete collection of worthwhile
accessories, which are listed below:
- A well-organized carrying case. (A
piece of the foam packing material is
intended for insertion into the lid,
so do not discard it offhandedly.)
- The iron itself, all loaded with
solder and ready to go. (A large 1/4-inch wide, chisel-shaped tip comes
- A conical tip to replace the large one
mentioned above (not quite small
enough for soldering IC's).
- A small sheet metal cradle to serve as
a rest stand.
- Three extra 50-inch rolls of 0.045
inch rosin-core solder, along with two
extra feeder spools.
- Three "soldering aid" probes, a
Phillips screwdriver (for performing
surgery on the iron), a clip-on heat
sink, a pair of self-closing tweezers,
and a putty knife.
- A can of "Nocorode" zinc chloride flux
paste (not for electrical work), and a
small brush for applying it. (Residue
from this flux should be rinsed away
- Assorted short lengths of "heat
- A packet of assorted hardware--everything from large cotter keys to wood
- Brochures on other DRI Industries'
products, primarily cabinets, work
benches, and hardware assortments.
The device is a 60-watt iron
of the "constant-power" type (not temperature
controlled). It has a pistol grip, the upper
left corner of which has a thumbwheel for
feeding solder to the work. The bottom end
of the handle contains a spool of solder, the
free end of which is threaded up through the
handle; it emerges from a tube which is
underneath and parallel to the heating
element. The tip is bent downward so as to
meet the solder about 5/8 of an inch beyond
the forward end of the feeder tube.
The unit takes 2-l/2 minutes to heat up to
soldering temperature. It remains hot enough
to melt solder up to 5 minutes after being
unplugged, and it should not be touched for
more than l0 minutes after being disconnected.
I do not view the unit's relatively high
power (60 watts) and its high heat capacity
(thus the long cooling time) as faults of the
tool. Those of you who have read "Soldering," Parts I and II, know why irons of high
power and high heat capacity promote quick
action in soldering and do less damage than
By luck or by design, the solder spool in
the handle is insulated from the feeder tube,
and the feeder tube is attached by a wire
clip to the heating element. A continuity
tester can be connected between the solder as
it emerges from the spool and the feeder
tube; this tester will then beep to indicate
that the solder has reached the tip. (A
spade lug can be mounted under one of the
screws which holds the heating element to the
handle; this makes a very sturdy permanent
connection to the feeder tube. The other
tester lead must be attached to the solder
via a clip lead.)
strike against this instrument is its long
barrel (heating element), and the fact that
set screws for the tip protrude significantly
from either side. The long barrel puts the
tip 6 inches beyond the user's hand, thus
greatly increasing the error in reaching the
target and magnifying the effects of tremor
in his hand. The basic diameter of the
barrel, l/2 inch, is not unreasonable, but
set screws securing the tip (which also
support the forward end of the feeder tube)
protrude to make the width of this assembly
7/8 of an inch. The resultant size and shape
of the iron is something to be reckoned with
--it is relatively commonplace that protrusions and/or the feeder system snag on wiring
and holding devices. Inadvertent contact
with the user's free hand is also hard to
The system used for securing the tip is not
one to ensure intimate contact between the
tip and the heating element; this intimate
contact is desired for efficient heat transfer. Ideally, where a set screw system is
used, the screws should be placed so as to
force the shank of the tip to one side. In
this unit, however, the two set screws are
directly opposing each other; they hold the
tip firmly between them, but serve as a pivot
about which the tip can rock up and down in
Neither the system holding the tip nor the
mounting of the chuck to the heating element
is rigid. Up and down motion of the tip can
always be accomplished, even accounting for
expansion of these components when the iron
is heated. (As received, 3 of the 4 mounting
screws holding the barrel to the handle were
very loose, allowing even more freedom of
movement. This is something to check for
before turning it on.)
The inherent mobility of the tip with
respect to the handle leads to instability in
holding the iron against the work, as well as
raising questions as to how well its heating
efficiency can be maintained (oxides may
eventually develop between the various components of the device).
All the above factors, as well as the
available tip sizes, define the size of work
for which this iron is appropriate. Its use
would be difficult for anything smaller than
terminal lugs and sparsely populated PC
If I may be permitted to briefly discard my
cloak of scientifism, I would like to say
that aside from the issue of self-contained
solder feeders, this iron is ill-suited for a
blind user. I find it very hard to hit a
target with the instrument, and they could
hear my yelps clear out to the Islands of
Langerhans every time I inadvertently brought
a knuckle up against that barrel assembly.
Thumbwheel Solder Feeders
and the Blind User
I have personal experience with three such
devices--a universal attachment (of unknown
make, made perhaps two decades ago) for any
soldering iron, a "Free-Hand Solder Feeder"
designed for the standard Weller soldering
gun, and the DRI Industries device described
here. I have not tried the Wahl "Isotip
Cordless Iron" (see Winter l98l), which is no
longer being manufactured.
The devices I have tried are generally of
sound mechanical design and do the job as
advertised; they permit the soldering process
to be executed with one hand while work
pieces are supported with the other. My
quarrel with them as a blind user has been
that they indirectly address the true problem
of how to place the iron on the connection,
at the same time stripping away several key
bits of information which a blind person uses
during the soldering process. In addition,
they lead to instability of the hand when
their thumbwheel is operated.
If everything were to go right, one could
do as the instructions say--hold the iron
against the work long enough for it to heat
up, then use the thumbwheel to feed solder to
the work. There are several variables here
which the blind person must account for.
Heating time at the connection depends on
several factors. One is the physical size
(heat capacity) of the work pieces. Another
is the effectiveness with which the iron is
held against them. Another is the presence
or absence of oxides on the materials and on
the tip of the iron. The only way to assess
the heating process is to monitor the temperature of the work pieces with the free
hand (touching a component lead or the like); as long as your "free" hand is already down
there, it might as well have solder in it,
Waiting too short a time before feeding the
solder will often result in the solder being
used up in cleaning and tinning the tip.
Sometimes the unmelted solder will be deflected off to one side by the work pieces, thus
no longer finding itself in contact with the
materials. Waiting too long before feeding
allows rapid oxide buildup at the connection,
which prevents soldering.
Because of the solder's proximity to the
tip of the iron, initial melting occurs more
often than not. However, this does not
assure that all components of the work will
experience "wetting"; the solder is never fed
to the work pieces alone, at a point not in
contact with the iron. Localized "wetting"
may occur on only one of the components of
the work. This, by the way, is quite
apparent visually, but is not so obvious from
tactile feedback through the iron.
Flux-core solder tends to kink and wrinkle
as it is rolled and unrolled; slight irregularities in it often cause binding in the
feeder system. Furthermore, since solder is
a soft material, one skid of the thumbwheel
against its surface creates a "dead spot"
from which no motion can be initiated. When
this happens, the solder must be pulled forward slightly from the forward end of the
feeder tube so as to re-engage with the
wheel. None of this is apparent through
tactile feedback from the thumbwheel; the
wheel turns stiffly whether it is feeding the
solder or spinning against it. Therefore,
you can never be absolutely sure that you
have fed solder to the work, and the amount
you feed is always open to question.
The final flaw in all of these devices is
that you must move your thumb or a finger to
operate the feeder. This causes muscles in
the arm to pull tendons through your wrist, a
very destabilizing action. It happens frequently to the sighted user that the tool
slips off the connection--something for which
he says "Oops," and puts it back into position. Re-establishing contact with the
target is not so trivial for the blind
I stand by my discussion of "Tactile Feedback" in "Soldering," Part II, Winter l98l,
wherein the iron and the solder become
"canes" in each hand. With this two-handed
system, you can find out where the iron is
with the solder, you can tell which items
are hot enough to melt the solder, and you
rarely overheat and oxidize a connection
before initial melting occurs.
Yes, placing the iron on the target is a
problem, but I stand by my discussion of
"Landmarking," also in Part II, as a viable
alternative. For those instances where a
probe is needed to guide the iron down to the
work, I urge you to try the JA3TBW Solder
Guide as another alternative.
It would indeed be wonderful to have an
iron, all the operations of which are done by
one hand. This is desirable because of the
many instances where holding the work pieces
in position is necessary. However, although
I am philosophically slow to admit it, tools
with a thumbwheel feeder system are ones
which are more accessible to the sighted user
than to the blind.
As mentioned in our "Hints and Kinks,"
SKTF, Spring l982, Mike Bhagwandas (JA3TBW)
of Kobe, Japan, invented this guide. Since
that time, our lab has done considerable
experimenting with its use as an aid to
soldering short-legged components which have
poor "landmarking" features into printed
boards. Blind subjects who have built our
prototype "kit sets" have found this instrument indispensable when soldering the IC
sockets and trim pot. So that you can enjoy
the fruits of this research, we have attached
the described version to each copy of the
magazine; please wear it in good health.
The material we settled on is stainless
steel, and we did so for the following
- It will never char and contaminate the iron.
- If "acid" or "stainless steel" flux is not used, it is unlikely that solder will adhere to the tube.
- Stainless steel has poor heat conductivity.
- The resultant device can be of small diameter; it is available having a wall thickness of 0.009 inches.
- This tubing is commonly available, sometimes being known as "hypodermic tubing."
Specifically, we purchased "Hypoflex" seamless thin-walled stainless steel tubing of
Grade 304; its dimensions are 0.072 inches
outside diameter (o.d.), 0.009 inch wall
thickness, giving us what they call a
"theoretical inside diameter" (t.i.d.) of
0.054 inches. Unfortunately, this product
only comes in bundles of l20 feet at 62 cents
per foot. However, a similar product can be
gotten from "Small Parts,"* in 6, l2, and 24
inch lengths. They have two types available:
HTX-l5 Hypodermic Tubing with a 0.009 inch
wall, and HTX-l5TW "Thin Wall" tubing with
the same outside diameter of 0.072 inches and
a wall thickness of 0.006 inches. The former
is $2 per foot; the latter is $l.20.
The size was chosen so as to accommodate
solder of approximately 0.03 inches; the tube
has plenty of room to spare for wrinkles in
the solder to pass freely through it.
Two brands of solder can readily be gotten
with these dimensions. The first is Kester,
604A33-lSN6066.03l44, available from Marshall
Industries.* This Kester solder, which has a
diameter of 0.03l inches, is also available
from Mouser Electronics* under their number
533-24-6040-27. Finally, 0.028 inch "Multicore" solder is available as Ersin number
SN60 22SWG. The Kester is preferable, since
"Multicore" solder tends to contain an overabundance of flux, thus causing gumminess and
stickiness in the tube.
The tubing has such a thin wall as to
afford being "parted" with a file, especially
knife-shaped or three-cornered files. After
sectioning off a piece (4 inches being a
popular length in our trials), square off the
ends with a file; then de-burr the inner
edges with the tip of a file or with a drill
bit. (We have done all of this to the 4-inch
piece you've received.)
For unclogging these units when solder gets
lodged inside, a No. 55 drill bit is a handy
addition. This bit is 0.050 inches in diameter, and will comfortably ream it out without galling.
One final note on making your own customized version: You may wish to devise a
handle so as to afford better control. We
would be very interested in the schemes you
come up with along this line. A small "shaft
stop" with a set screw is included with your
guide. Used as an adjustable collar, the
shaft stop can be secured to the tube just
below your hand so as to promote stability in
holding the tube against the work. (The
shaft stop is available from Player Piano
Company.* It has an inside dimension of
0.098 inches, listed as their catalog number
889 at 20 cents each.)
In operation, solder is passed through the
tube so as to just make itself known at the
business end, this end being rested on the
connection. Solder can be fed to the work
with the thumb and forefinger, while some
arrangement of the remaining three fingers is
chosen to support the tube. Good stability
will be promoted if the connection has a
feature against which the inner edge of the
tube can be braced. For example, I bend
socket pins outward once they have passed
through the board; I actually set the tube
over the intended pin, at which point I lean
the tube out to the side so as to permit the
iron to reach the pin.
Once you are in position, find the lower
half of the tube with the iron and slide it
down onto the connection, all the while
pressing the solder down against the work.
I find this technique to be so easy that glee
overtakes me, giving me a tendency to apply
too much solder. Remember in using this
technique that all the solder will be applied
in the right place, for a change, and very
little of it will be used up exploring the
If left stationary in the tube after the
connection is made, the solder will seize up
in the bottom end. This is not due to actual
bonding with the stainless steel, but is due
to the flux solidifying, acting to glue the
solder in place. Reheating the end of the
tube with the iron will often loosen things
up; otherwise your No. 55 drill can be used
to ream the tube clear.
The way to prevent adhesion is to create
relative motion between the tube and the
solder while the flux is cooling. Some of us
pull the tube away from the connection and
run the solder through so as to protrude l/4
inch or so. Others withdraw the solder into
the tube about l/2 inch or so. I spin the
tube while it is cooling. All of this is
done after the connection has been soldered;
you have plenty of time because the flux must
cool considerably before it solidifies.
Like anything, this takes a little practice; you will at first, by applying far too
much solder or by heating up adjacent connections with the iron, inevitably bridge
pins here and there. However, comparing my
success with and without this instrument
would be ludicrous--there is no comparison.
Thanks to Mike, JA3TBW, there is a way to
take the heat off our target practice when
the going gets tough.
* Address List of Suppliers
- Marshall Industries, 788 Palomar Ave.,
Sunnyvale, CA 94086; phone (408) 732-ll00.
- Mouser Electronics, ll433 Woodside Ave.,
Lakeside, CA 92040; phone (7l4) 449-2222.
- Player Piano Company, 704 East Douglas,
Wichita, KS 67202; phone (3l6) 263-324l.
- Small Parts, Inc., 690l N.E. Third Ave.,
Miami, FL 33l38; phone (305) 75l-0856.