Everything You Always Wanted to Know

Semiconductor
About the ICL8038
Application Note November 199 AN013.1
Author: Bill O’Neil

Introduction Answer
The 8038 is a function generator capable of producing sine, First of all, the voltage difference need only be a few hundred
square, triangular, sawtooth and pulse waveforms (some at millivolts so there is no danger of damaging the 8038. One
the same time). Since its introduction, marketing and appli­ way to get this higher potential is to lower the supply voltage
cation engineers have been manning the phones explaining on the 8038 and external resistors. The simplest way to do
the care and feeding of the 8038 to customers worldwide. this is to include a diode in series with pin 6 and resistors RA
This experience has enabled us to form articulate responses and RB. See Figure 1. This technique should increase the
to the most frequently asked questions. So, with data sheet sweep range to 1000:1.
and breadboard in hand, read on and be enlightened.
+15V
Question 1
1N457
I want to sweep the frequency externally but can only get a
range of 100:1 (or 50:1, or 10:1). Your data sheet says 0.1µF DUTY
CYCLE
1000:1. How much sweep range can I expect?
1K 15K
Answer 4.7K
RB
Let’s look at what determines the output frequency. Start by 4.7K
RA
examining the circuit schematic at pin 8 in the upper left hand
corner. From pin 8 to pin 5 we have the emitter-base of NPN
Q1 and the emitter-base of PNP Q2. Since these two diode 5 4 6
9
drops cancel each other (approximately), the potential at pins 100K
8, 5, and 4 are the same. This means that the voltage from V+ FREQUENCY 8
3
LOG POT
to pin 8 is the same as the voltage across external resistors
7
RA and RB. This is a textbook example of a voltage across 2
10 11 12
two resistors which produce two currents to charge and dis­ 10M
charge a capacitor between two fixed voltages. This is also a
linear system. If the voltage across the resistors is dropped 0.0047µF DISTORTION
100K
from 10V to 1V, the frequency will drop by 10:1. Changing
from 1V to 0.1V will also change the frequency by 10:1. -15V
Therefore, by causing the voltage across the external resis­
FIGURE 1. VARIABLE AUDIO OSCILLATOR, 20Hz TO 20kHz
tors to change from say 10V to 10mV, the frequency can be
made to vary at least 1000:1. There are, however, several fac­
tors which make this large sweep range less than ideal. Question 4
O.K., now I can get a large frequency range, but I notice that
Question 2
the duty cycle and hence my distortion changes at the low­
You say I can vary the voltage on pin 8 (FM sweep input) to est frequencies.
get this large range, yet when I short pin 8 to V+ (pin 6), the
ratio is only around 100:1. Answer

Answer This is caused partly by a slight difference in the VBEs of Q2

and Q3. In trying to manufacture two identical transistors, it is
This is often true. With pin 8 shorted to V+, a check on the not uncommon to get VBE differences of several millivolts or
potentials across the external RA and RB will show 100mV more. In the standard 8038 connection with pins 7 and 8 con­
or more. This is due to the VBE mismatch between Q1 and nected together, there are several volts across RA and RB and
Q2 (also Q1 and Q3) because of the geometries and current this small mismatch is negligible. However, in a swept mode
levels involved. Therefore, to get smaller voltages across with the voltage at pin 8 near V+ and only tens of millivolts
these resistors, pin 8 must be raised above V+. across RA and RB, the VBE mismatch causes a larger mis­
match in charging currents, hence the duty cycle changes. For
Question 3 lowest distortion then, it is advisable to keep the minimum
How can I raise pin 8 above V+ without a separate power voltage across RA and RB around 100mV. This would of
supply? course, limit the frequency sweep range to around 100:1.

1 Copyright © Harris Corporation 1998

 Application Note 013

Question 5 Finally, trimming the various pins for lowest distortion

deserves some attention. With pins 7 and 8 connected
I have a similar duty cycle problem when I use high values of
together and the pot at pin 7 and 8 externally set at its maxi-
RA and RB. What causes this?
mum, adjust the ratio of RA and RB for 50% duty cycle. Then
Answer adjust a pot on pin 12 or both pins 1 and 12 depending on
There is another error term which becomes important at very minimum distortion desired. After these trims have been
low charge and discharge currents. This error current is the made, set the voltage on pin 8 for the lowest frequency of
emitter current of Q7. The application note on the 8038 gives interest. The principle error here is due to the excess current
a complete circuit description, but it is sufficient to know that of Q7 causing a shift in the duty cycle. This can be partially
the current charging the capacitor is the current in RA which compensated for by bleeding a small current away from pin
flows down through diode Q9 and into the external C. The 5. The simplest way to do this is to connect a high value of
discharge current is the current in RB which flows down resistance (10MΩ to 20MΩ) from pin 5 to V- to bring the duty
through diode Q8. Adding to the Q8 current is the current of cycle back to 50%. This should result in a reasonable com­
Q7 which is only a few microamperes. Normally, this Q7 cur- promise between low distortion and large sweep range.
rent is negligible, but with a small current in RB, this current
will cause a faster discharge than would be expected. This Question 7
problem will also appear in sweep circuits when the voltage This waveform generator is a piece of junk. The triangle wave
across the external resistors is small. is non-linear and has large glitches when it changes slope.
Answer
Question 6
You’re probably having trouble keeping the constant voltage
How can I get the lowest distortion over the largest
across RA and RB really constant. The pulse output on pin 9
frequency sweep range.
puts a moderate load on both supplies as it switches current on
Answer and off. Changes in the supply reflect as variations in charging
First of all, use the largest supply voltage available (±15V or current, hence non-linearity. Decoupling both power supply pins
+30V is convenient). This will minimize VBE mismatch prob­ to ground right at the device pins is a good idea. Also, pins 7
lems and allow a wide variation of voltage on pin 8. The and 8 are susceptible to picking up switching transients (this is
potential on pin 8 may be swept from VCC (and slightly especially true on printed circuit boards where pins 8 and 9 run
higher) to 2/3 VCC +2V) where VCC is the total voltage side by side). Therefore, a capacitor (0.1µF or more) from V+ to
across the 8038. Specifically for ±15V supplies (+30V), the pin 8 is often advisable. In the case when the pulse output is not
voltage across the external resistors can be varied from 0V required, leave pin 9 open to be sure of minimizing transients.
to nearly 8V before clipping of the triangle waveform occurs.
Second, keep the maximum currents relatively large (1mA or
Question 8
2mA) to minimize the error due to Q7. Higher currents could What is the best supply voltage to use for lowest frequency
be used, but the small geometry transistors used in the 8038 drift with temperature?
could give problems due to VCE(SAT) and bulk resistance, etc. Answer
Third, and this is important, use two separate resistors for RA The 8038AM, 8038AC, 8038BM and 8038BC are all temper­
and RB rather than one resistor with pins 4 and 5 connected ature drift tested at VCC = +20V (or ±10V). A curve in the
together. This is because transistors Q2 and Q3 form a differ­ lower right hand corner of Page 4 of the data sheet indicates
ential amplifier whose gain is determined by the impedance frequency versus temperature at other supply voltages. It is
between pins 4 and 5 as well as the quiescent current. There important to connect pins 7 and 8 together.
are a number of implications in the differential amplifier con­
nection (pins 4 and 5 shorted). The most obvious is that the Question 9
gain determines the way the currents split between Q2 and
Why does connecting pin 7 to pin 8 give the best temperature
Q3. Therefore, any small offset or differential voltage will
performance?
cause a marked imbalance in the charge and discharge cur-
rents and hence the duty cycle. A more subtle result of this Answer
connection is the effective capacitance at pin 10. With pins 4 There is a small temperature drift of the comparator thresh­
and 5 connected together, the “Miller Effect” as well as the olds in the 8038. To compensate for this, the voltage divider
compound transistor connection of Q3 and Q5 can produce at pin 7 uses thin film resistors plus diffused resistors. The
several hundred picofarads at pin 10, seriously limiting the different temperature coefficients of these resistors causes
highest frequency of oscillation. The effective capacitance the voltage at pins 7 and 8 to vary 0.5mV/oC to maintain
would have to be considered important in determining what overall low frequency drift at VCC = 20V. At higher supply
value of external C would result in a particular frequency of voltages, e.g., ±15V (+30V), the threshold drifts are smaller
oscillation. The single resistor connection is fine for very sim­ compared with the total supply voltage. In this case, an
ple circuits, but where performance is critical, the two sepa­ externally applied constant voltage at pin 8 will give reason-
rate resistors for RA and RB are recommended. ably low frequency drift with temperature.

2
 Application Note 013

Question 10 Question 13
Your data sheet is very confusing about the phase relationship How can I buffer the sine wave output without loading it down?
of the various waveforms. Answer
Answer The simplest circuit is a simple op amp follower as shown in Fig­
Sorry about that! The thing to remember is that the triangle ure 3A. Another circuit shown in Figure 3B allows amplitude and
and sine wave must be in phase since one is derived from offset controls without disturbing the 8038. Either circuit can be
the other. A check on the way the circuit works shows that DC or AC coupled. For AC coupling the op amp non-inverting
the pulse waveform on pin 9 will be high as the capacitor input must be returned to ground with a 100kΩ resistor.
charges (positive slope on the triangle wave) and will be low
during discharge (negative slope on the triangle wave). Question 14
The latest data sheet corrects the photograph Figure 7 on Page Your 8038 data sheet implies that all waveforms can operate up
5 of the data sheet. The 20% duty cycle square wave was to 1MHz. Is this true?
inverted, i.e., should be 80% duty cycle. Also, on that page Answer
under “Waveform Timing” the related sentences should read
“RA controls the rising portion of the triangle and sine-wave and Unfortunately, only the square wave output is useful at that
the 1 state of the square wave.” Also, “the falling portion of the frequency. As can be seen from the curves on page 4 of the
triangle and sine wave and the 0 state of the square wave is:” data sheet, distortion on the sine wave and linearity of the tri­
angle wave fall off rapidly above 200kHz.
Question 11
Question 15
Under Parameter Test Conditions on Page 3 of your 8038
data sheet, the suggested value for Min and Max duty cycle Is it normal for this device to run hot to the touch?
adjust don’t seem to work. Answer
Answer Yes. The 8038 is essentially resistive. The power dissipation
The positive charging current is determined by RA alone is then E2/R and at ±15V, the device does run hot. Extensive
since the current from RB is switched off. (See 8038 Applica­ life testing under this operating condition and maximum
tion Note AN012 for complete circuit description.) The nega­ ambient temperature has verified the reliability of this prod­
tive discharge current is the difference between the RA uct.
current and twice the RB current. Therefore, changing RB
will affect only the discharge time, while changing RA will Question 16
affect both charge and discharge times. For short negative How stable are the output amplitudes versus temperature?
going pulses (greater than 50% duty cycle) we can lower the
Answer
value of RB (e.g., RA = 50kΩ and RB = 1.6kΩ). For short
positive going pulses (duty cycles less than 50%) the limiting The amplitude of the triangle waveform decreases slightly
values are reached when the current in RA is twice that in with temperature. The typical amplitude coefficient is
RB (e.g., RB = 50kΩ). This has been corrected on the latest -0.01%/oC, giving a drop of about 1% at 125oC. The sine
data sheet. output is less sensitive and decreases only about 0.6% at
125oC. For the square wave output the VCE(SAT) goes from
Question 12 0.12V at 25oC to 0.17V at 125oC. Leakage current in the “1”
state is less than a few nanoamperes even at 125oC and is
I need to switch the waveforms off and on. What’s a good
usually negligible.
way to strobe the 8038?
Answer +15V
With a dual supply voltage (e.g., ±15V) the external capaci­ RA RB 15K
tor (pin 10) can be shorted to ground so that the sine wave
and triangle wave always begin at a zero crossing point. 4 5 6
9
Random switching has a 50/50 chance of starting on a posi­
tive or negative slope. A simple AND gate using pin 9 will 8 8038
allow the strobe to act only on one slope or the other, see 7 2 1N914
11 10
Figure 2. Using only a single supply, the capacitor (pin 10)
can be switched either to V+ or ground to force the compara­ 1N914
tor to set in either the charge or discharge mode. The disad­ 2N4392 STROBE
vantage of this technique is that the beginning cycle of the C OFF
100K +15V (>0V)
next burst will be 30% longer than the normal cycle. -15V
-15V (< -10V)
ON

FIGURE 2. STROBE-TONE BURST GENERATOR

3
 Application Note 013

V+ V+
RA RB
RA RB

4 5 6 AMPLITUDE
4 5 6 2
7
7 100K
8038 +
8038 +
- 8 -
8 2
10 11 20K
10 11

4.7K
C
C

V­
V­

FIGURE 3A. FIGURE 3B.

FIGURE 3. SINEWAVE OUTPUT BUFFER AMPLIFIERS

Schematic Diagram

V+