Effect of feedback in amplifiers

I wanted to know what are the effects of feedback, for example what are the pros and cons of high feedback vs low feedback on the negative node of an opamp? How are the characteristics of the an amplifier change if we have only 10% feedback vs an amplifier which has 90% feedback ? What are the benefits or limitations of the two?

Since OA's open loop gain > 1 million, the amount of negative feedback actually the ratio of resistors determines the -ve input gain. if you had many input resistors of different value then they each have different ratios with the common feedback.

The same can be done with transistors with a current gain much larger than desired voltage gain, so the Voltage gain from the base would be the ratio of impedance including load on collector to impedance on emitter since they share a common current. ( well almost)

The positive input for an OA has a gain of 1 + the feedback ratio or gain on the negative side.

Positive Feedback is useful for hysteris if < 1 and becomes an oscillator if >1 which can be tuned with the input cap to ground.

Naturally with negative feedback, if you provide a LPF feedback it becomes a HPF. You can even use cap to simulate an inductor...and visa versa. a diode feedback becomes either a log amp in some modes or a clamp in other mode.

If you are referring to the feedback divider (supposing you have a certain op-amp configured as a non-inverting amplifier), then feeding a greater fraction of the output voltage to the inverting input will reduce the amplifier's gain (while increasing its accuracy); less feedback will result in greater amplifier gain, but decreased accuracy. In addition, more feedback means that it is faster, and more likely to go unstable (an op-amp is most unstable when configured as a simple follower), while less feedback means it's generally more stable but slower. I've just mentioned the obvious pro/cons, but there are many others (noise sensitivity, distortion, etc.)

To make things more complicated, I haven't even touched current-feedback amplifiers yet. :shock:

understood, but when you say accuracy what do you mean?

General op amps are made unconditionally stable (for normal amp operation) by putting internal a low frequency dominate rolloff pole. This makes relatively low freq (<100kHz) use of op amps simple and nearly fool proof.

If you want higher frequency bandwidth you have to start working with op amps that don't have the dominate low freq rolloff pole. You now have to know how to evalute gain-phase performance to ensure stability. Some high frequency op amps spec stablity for gains greater then a certain level. Too much feedback can make these op amps unstable. Too much stray PCB capacitance on the negative input can cause stability problems, along with making the feedback resistor too large in value creating a phase lag pole via the high value resistor between output of op amp and neg input with the stray capacitance on the neg input.

Related to stability, it can also degrade transient performance to sharp slew rate inputs.

If you are referring to the die hard, vacuum tube audio amp enthusiest, too much feed back degrades intermodulation distortion performance.

seamoss said:

understood, but when you say accuracy what do you mean?

Click to expand...

In any control system, it is the "loop gain" that determines the accuracy. (What is the loop gain? It is the gain that you would calculate if, somehow, you took the feedback loop, cut it open, injected a signal, waited for the signal to propagate all the way back around to its starting point, and then you measured how much it got amplified.) The greater the loop gain, the more accurate it is. (In an ideal op-amp, the loop gain is infinite — therefore, it is perfectly accurate). If the loop gain decreases (for whatever reason), then the accuracy diminishes. Adding a feedback divider decreases the loop gain (this should be intuitively obvious from my definition of loop gain; if you add an attenuator in the signal path, then the signal gain will be reduced), and therefore decreases the accuracy.

Diminishing loop gain, and the associated inaccuracy, is seen even in unity-gain buffers: if you run a sufficiently high-frequency signal through a buffer, you will find that it's not accurate anymore. This is because, at high frequency, the loop gain reduces (there are internal components in the op-amp that decrease its gain at high frequency in order to ensure stability). To be specific, at high frequency, the buffer would output an attenuated signal that is no longer in-phase. If you were to perform this same test to an amplifier with gain, you would find that this pronounced inaccuracy occurs at a lower frequency (since the loop gain reduction is compounded by the feedback divider).

I think, everything written above is correct - however, I am not sure if it answers the more general question of seamoss.
Therefore, I like to summarize in short:

Negative feedback
*reduces gain - but makes the resulting gain less dependent on tolerances and uncertainties of the active devices (transistors, opamps). For very large gain values (without feedback) the resulting gain is determined primarily by external components (feedback network), see Black's classical formula.
*stabilizes the dc operating point against tolerances (transistor current gain, opamp offset voltage) and environmental conditions (temperature, aging, power supply)
*enlarges the usable bandwidth
*reduces distortions (improves linearity)
*drastically changes input and output impedances (depending on the kind of feedback: voltage or current)
*decreases dynamic system stability (reduces stability margin) - that is the only disadvantage.

Thank you guys, your posts really helped me understand some general feedback principles. I appreciate it!

It might also broaden your horizon to know that negative feedback control systems exist everywhere in life.

https://en.wikipedia.org/wiki/Negative_feedback

The unique properties of close loop systems is that if you have < 180deg phase shift at the unity gain (=1) point on the curve, it is stable.. The more margin the better. Thus to make it unconditionally stable OA's behave like 1st order LPF with a gain bandwidth product being constant. Thus if it had 1MHz GBW product and the DC gain was 1e6 then the LPF point is 1Hz and negative feedback increases that BW such that with 100% feedback you can get 1MHz BW instead of 1Hz.

But in life phase shift from delay makes the circuit more complex to be stable. If you have 180deg phase shift with an inverter that is 180Deg shift by default. you have positive feedback and if the gain at that frequency of phase shift is >1 , you have steady ringing or an oscillator. Biological systems, drunk drivers, and any closed loop system that weaves back and forth is said to be unstable so the phase shift in the loop bandwidth determines the stability at unity gain frequency point. So special compensation is added to solve this problem. Drivers who want to be stable in the middle of the road, try to predict where they are heading. This is phase lead behavior. Those who are tired cause phase lag and if over-reactive ( too much feedback) can easily overshoot the edge of the road. So there is a direct relation between time domain and frequency domain of any closed loop system such as Operational amplifiers. Back in the 70's we could model any behavior such as the blood pressure in humans to changes in fluid intake using ANALOG COMPUTERS which were simple OA's with different feedback types and ratios connected in a closed loop with a jumper board to make easy connections to do simulations.

SunnySkyguy said:

The unique properties of close loop systems is that if you have < 180deg phase shift at the unity gain (=1) point on the curve, it is stable..

Click to expand...

Just a small correction: The closed-loop will be stable if you have <180 deg phase shift at the unity gain (=1) point of the loop gain function (transfer function of the complete open loop).