[sdiy] Weird op amp features & rules of thumb

[Grant Richter]

>
> * 741, 4558, and 4136 have special circuitry to avoid the chip  
> going boom if the output is short circuited. Grant noted that the  
> 220 ohm output resistors Buchla uses on many Music Easel circuits  
> was put INSIDE the feedback loop, and that Buchla was relying on  
> this circuitry. So if another op amp is used, Grant suggests a  
> typical 1K output resistor.

With all due respect to Professor Lanterman, these are additional  
notes on the subject.

Resistor power rating and short circuits to panel faceplates. If you  
plug a patchcord into an output first and let it dangle, the other  
end can easily short to the grounded faceplate. So a resistor is  
place in series with the op-amp output to prevent damage.

How do you calculate the value of that resistor and it's power  
rating? The two are interrelated. To shorten the story we'll take the  
standard 1K. A worse case scenario is the the op-amp is at either  
supply rail and there is 15 volts across the resistor. The power  
dissipation is (15/1000)*15 = 0.225 watts. So a 1K resistor is the  
lowest value of 1/4 watt resistor that will sustain the expected  
fault condition.

In the case of a 741 op amp, the output short circuit current is  
rated at 25 ma. So using a 220 ohm resistor the fault power  
dissipation is 220 * 0.025 = 5.5 volts * 0.025 = 0.137 watts, fine  
for a 1/4 watt resistor.

For the Serge, the output current limit for an LM3900 at 12 volts is  
20 ma. Calculating the power dissipation of a 330 ohm resistor, 330 *  
0.020 = 6.6 volts * 0.020 = 0.132 watts. Also fine for a 1/4 watt  
resistor.

So the values the designers selected for both the Serge and Buchla  
current limiting resistors are based on the type of op-amp used  
routinely in those systems. It is important to note that the Buchla  
uses the 741s only for the control portion of the instrument. For the  
audio portion of the instrument, the low noise 5534 op-amps are used  
and I have measured the signal to noise ratio of the audio section of  
the original Buchla 200 at better than -80dBv, which is fantastic for  
1968 electronics.

>
> * If the voltage difference on the input pins exceeds a magic  
> number, the TL07x/TL08x do a really nasty phase inversion where the  
> + acts like - and the - acts like +. For typical negative feedback  
> mode circuits where the pins try to stay at the same voltage, this  
> probably isn't a problem, but if it's being used as a comparator  
> (or in a voltage-starved feedback mode like the Buchla timbre  
> circuits, in which the op amp can't keep the voltages the same),  
> this could yield all sorts of vicious nastyness. So blindly  
> plugging in a TL07x/08x for another op amp may be dangerous and  
> should be carefully considered.

The voltage reversal is a nasty fact, but the only time I had a  
problem with it was using one as a high impedance buffer for a sample  
and hold. If the voltage got too negative, phase reversal happened.

My simple fix was to use a 10 volt zener to limit the input to the  
sample and hold circuit to above -10 volts. Most synth control ranges  
are between -10 volts and + 10 volts anyway.
>
> Aaron's possibly nonsensical rule of thumb: In terms of noise, BJT  
> input op amps are a better choice when being driven by sources with  
> low output impedances (like other op amps), and JFET input op amps  
> are a better choice when being driven by sources with high output  
> impedances (OTAs, expo converters, etc.), and when you don't want  
> the input current to be doing things like sucking electrons off  
> your integration cap. (A corrollary of this is that if you really  
> need minimal input current, you want a MOSFET input chip like the  
> CA3140.)

Op-amp low noise design is a can of worms. Here are two references  
you can read if you just like to hurt your brain or are trying to go  
to sleep. No, actually the AD797 datasheet is pretty easy to follow  
and gives a quick overview of this complex subject.

http://www.analog.com/static/imported-files/data_sheets/AD797.pdf

For even more mind numbing material, the ap note AN-940 "Low Noise  
Amplifier Selection Guide for Optimal Noise Performance" should put  
you out in no time.

http://www.analog.com/static/imported-files/application_notes/AN_940.pdf

>
> ---> Aaron's usage rule of thumb, based on reading books and stuff:
>
> BJT input amps need that little resistor from the + terminal to  
> ground when you're setting up an inverter. JFET input op amps  
> don't. If you're doing a more complicated circuit, you generally  
> just want the resistances seen at each of the input terminals to be  
> more or less the same.
>

What the resistor does is correct the output offset caused by the  
voltage drop in the resistors from it's own bias current. BJT op-amps  
have typical 500-1000 nano-ampere bias currents and can produce  
millivolt errors. JFET inputs have pico-ampere bias currents and  
produce errors in the microvolts. This error is generally smaller  
than the microvolt errors produced by the thermocouples formed where  
ever two dis-similar metals are joined, like lead and copper (solder  
joints) or copper and steel (screw terminals).

You can go down the error source rabbit hole as deep as you want, but  
you will never get anything done. Just pretend the rabbit hole isn't  
there unless management or the customer MAKE you look at it.