It is rarely that you need a very specific model opamp in a circuit; usually you just need certain characteristics that can fulfilled by a number of devices. Once you understand what the needs are in various circuits, you can identify what types of opamps can go there. Here are some opamp characteristics to take into consideration
|
First I'll identify several classes of opamps and their major characteristics:
Class | Characteristics | Devices |
General purpose "741" type | Slower speed, average dc specs | uA741, 1448, LM324 |
JFET input types | Higher slew rate and bandwidth, slightly worse dc specs, very low input bias current | TL07x, TL08x, LM356, LM357 |
Decompensated general purpose | Very fast speed used open loop, average dc specs | 748, uA301 |
CMOS opamps | Higher slew rate and bandwidth, slightly worse dc specs, very very low input bias current, lower supply voltage and output current | CA3130, CA3140 |
High precision | Very good dc specs, low speed | OP07 |
High precision, low noise, high speed | Audio capable slew rate and bandwidth, excellent dc specs, average input bias current, low noise | OP27, OPA275 |
Audio opamp | Audio capable slew rate and bandwidth, slightly worse dc specs, higher input bias current, low noise | NE3324, NE3332 |
High speed opamp | High slew rate and bandwidth, worse dc specs, higher input bias current | LM318 |
DC Specs | |
Vos, input offset voltage | Output voltage caused by component mismatch, referenced to the input. This will be multiplied by the gain. For example, an
opamp with a 2mV Vos connected as an inverter with a a gain of 1000 and the input grounded, will have the output sitting at -2v.
Normally synth circuits aren't as concerned with Vos because they tend to be used with low gains. |
Ibias, Input bias current | The inputs of opamps are either the base of a pnp/npn transistor, or the gate of a JFET/MOS transistor. Typically this is a few
nanoAmps for bipolar inputs, or picoAmps for JFET/MOS types. The input bias current produces an apparent input voltage when flowing
through input resistors.
In a synth circuit, use JFET or MOS input opamps for circuits that deal with very small currents - like integrators on VCO's (the input bias current will determine the low frequency), or Sample and Hold buffers. |
Ios, input offset current | Ios with matched input resistances causes the equivalent of a Vos. Typically the Ios is much smaller than
the Ibias.
If you have input bias currents that exactly match, you can compensate for them by using matched resistance on the inputs. For example, on an inverter, have a resistance of Rf in parallel with Rin on the + input to ground (this is the equivalent resistance seen looking "out" the - input). Don't bother doing this on JFET/MOS input opamps - the error is too small to make any difference. |
CMRR, Common Mode Rejection Ratio | This is a measure of an opamps' ability to act as a true differential amplifier. Ideally, if you apply the same signal to both inputs,
you should get zero output. Such an opamp would have an infinite CMRR. CMRR is defined as the ratio of the change in common mode voltage
to the change in output voltage, expressed in decibels. The spec is usually at dc, because it is very sensitive to frequency (it
gets worse as frequency goes up). An opamp spec sheet will usually have a CMRR vs. Frequency graph.
Related to CMRR is the Common Mode Range, or how close to the supply rails an opamp can handle common mode input voltages. Typically it's within a couple of volts of the supplies. Common mode range can be important because some opamps do weird things when their common mode range is exceeeded. The TL07x series can change phase (+ input becomes - input, causing latchup). The OP27 can oscillate at 5MHz or so. The voltage follower is the most common circuit where you will have to be concerned with common mode range. |
PSRR, Power Supply Rejection Ratio | This is a measure of an opamps' ability to ignore changes in the power supply voltages. Ideally, if you change the power supply voltages,
you should no change in the output. Such an opamp would have an infinite PSRR. PSRR is defined as the ratio of the change in power supply voltage
to the change in output voltage, expressed in decibels.
Since we normally do not change power supply voltages on purpose, the ripple on the supply lines becomes important. The frequency of this will be 120Hz on a full wave rectified linear supply. If you use a switching power supply, it is much much worse because the PSRR will be very low at the ultrasonic frequencies of the switching - the signal will get into everything. That's whay switching power supplies are not recommended for analog circuits. |
Output swing | This is a measure of how close to the power supply voltages the output can swing. Typical opamps can reach within 1v to 2v of the
supplies. Special purpose "rail to rail" opamps can reach within a few millivolts of the supplies.
The output swing will determine what type of signal levels you can suppoprt in your synth circuits. For example, to support +/-10v (20vp-p) signals, you probably should use +/-15v supplies. If you are using +/-12v supplies, you probably want to standardize on +/-5v signal levels. A side note: the LM324 quad opamp does not perform like a quad 741. It suffers from "crossover distortion". I've had control voltage circuits using an LM324 that didn't work right until I swapped it out for a TL074. I've heard that adding a resistor from the output to the negative supply (10K or so) improves this. |
AC Specs | |
GBW, Gain-Bandwidth Product | Negative feedback differential amplifiers such as opamps depend on high open loop gains to support the feedback. The open loop gain of
an opamp is high at dc, but decreases rapidly with frequency, usually at a one to one logarithmic rate i.e., Aol (open loop gain, not America
OnLine) decreases by a factor of two every time the frequency doubles. The Gain Bandwidth Product is the frequency where the open loop gain
has dropped to unity.
GBW gives a measure of how high a frequency the opamp can handle for small signals. |
Slew Rate | Slew Rate, usually expressed in volts per microsecond (V/uSec), is a measure of the large signal frequency handling capability
of an opamp. It's actually a limitation of the amount of current available to charge capacitances internal to the opamp.
The graphs for large signal handling capability you see in opamp datasheets are based on the opamp's ability to output an undistorted sine wave. But in synths we have signals like saw and pulse waveforms that should have very straight up and down slopes - which requires an even higher slew rate. For example, to handle large audio signals (say 10v p-p) requires a minimum slew rate of about 1.5V/uSec. This rules out the 741 types that only have 0.5V/uSec. It's why the JFET input opamps like the TL07x or TL08x series are so popular for synth circuits - they have slew rates of around 13V/uSec. |
Noise | All electronic components produce noise, and every noise source adds up and is amplified by whatever gains you have in your circuits.
It's the hiss you hear when there is no signal present. The noise specs of an opamp become important when you are using a lot of gain to
amplify small signals.
Signals in synth circuits tend to be fairly large, so using low noise opamps are not that important. The TL07x used so commonly in synth circuits actually have pretty high noise. You might want to use low noise opamps like the NE5532 or the OP27 types to amplify low level signals like microphones. Mixers are another good candidate. |