Opamp circuit maximum allowable gain

[Pedro]

Hi there!

I am working on some op amp stages at work and I am facing with some highish gain stages, in the thousands range. Some time ago, I remember reading that you shouldn't use an amplifier with signals that exceed your closed loop bandwidth to avoid the risk of oscillations. But, I cannot find any reference to this.

I know that with high gains, closed loop bandwidth will be decreased. But, what other effects will high closed loop gains have? I am used to seeing circuits with op amps with gains no higher than 100's. Is this just to preserve BW? Do you guys have some good app note on the subject?

Furthermore, googling for "op amp high gain" or anything similar leads to lots of pages talking about the high open loop gains that op amps have...

Thanks a lot to everyone

[IanB]

I'm no EE, but I think a general design principle is to vary the gain with the frequency, either by using capacitors to bypass higher frequencies away from the inputs, or to use capacitance in the op amp feedback making the gain low at high frequencies and high at low frequencies, or combinations of such techniques. When considering gain and stability you have to look across the whole frequency spectrum and not consider gain as if it is a fixed number.

[Alex]

The phase of the output signal of an op-amp configuration shifts as the frequency increases. There is a point where the phase shift becomes 180 degrees i.e. the output is inverted. Opamps are usually used with negative feedback derived from the output so this means that the feedback becomes positive.

If at the frequency that the phase shift is 180 degrees your gain is more than 1 then the output will oscillate (Barkhausen criteria). A circuit inside the vast majority of op-amps reduces the gain of the op-amp as the frequency increases so that at 180 degrees phase shift you have much less than 1 gain and you avoid oscillation. This is the compensation circuit and it makes the op-amp behave in a predictable way.

The product of the circuit's gain and bandwidth is nearly constant and given in the datasheet as the gain-bandwidth product (GBP). The reason this remains constant is because of the frequency response of the compensation circuit (Google GBP). So if the GBP for the 741 is 1MHz, at 1MHz without feedback the gain will be 1. If now you set a feedback network for a gain of say 100, the gain will drop to 1 at 10kHz. For a gain of 1 (buffer) the gain will remain 1 until 1MHz.

I hope this makes sense. There are many terms that you can google here for more info.

[ejeffrey]

When you have input signals that exceed the closed loop bandwidth of your amplifier, the opamp is basically operating open-loop.  The gain will be limited, unpredictable, and have large phase shifts.  In the limit where you exceed the open-loop gain, the amplifier does nothing at all, and the transfer function is basically governed by whatever gets through the passive network.  This is particularly troublesome in active filters.  An active low-pass filter may have insufficient attenuation at high frequency due to the absence of feedback.  Furthermore, the op-amp relies on feedback to present a high impedance.  At high frequency, the input simply looks like a diode.  This can cause nonlinear distortion.  This is a common cause of GSM interference in audio circuits.  If the input does not sufficiently filter RF, it can get rectified by the op-amp input, forming a crude AM radio.  GSM generates lots of RF bursts, which demodulated cause a lot of noise.  This is how a transmitter at 1.8 GHz can interfere with an audio signal at 1 kHz.   It is best to use passive filters to eliminate any signals beyond the bandwidth of your op-amp.

There is a second problem with high gain and oscillation, but this problem is generic, not specific to op-amps or operating above their rated bandwidth.  The problem is stray coupling from the output back to the input, which can cause oscillation if the phase is such that you have positive feedback.  With high gain amplifiers, the forward gain is so high it only takes a small amount of feedback to oscillate.  The feedback can come from several sources. but typically gets worse with increasing frequency.  This is a circuit layout problem, and the way to avoid this is to split your gain into multiple stages, and to keep the high level output physically separated and/or shielded from the low level input.  Don't make a U shaped amplifier chain with a volt output next to a nanovolt input...

[Pedro]

Thanks for your answers,

I think Ejeffrey has touched the key points. My question could be shortened to "why are there so few opamp circuitswith gains in the thousands range". Today's opamps have pgbs that allow for this with enough BW for many applications. The output-input coupling effect mentioned by ejeffrey could justify the fear of using high gain circuits, and it is good to be aware of this in the pcb routing stage.

See'ya

[vk6zgo]

Op Amps driving capacitive loads can oscillate,despite having internal correction.

I had a problem where a differential amplifier which was supposedly handling dc would burst into oscillation at several hundred KHz.
The worst thing about this is that the resultant signal would amplitude modulate an RF amplifier,causing it to draw excessive current on positive modulation peaks,& cook the power supply---Ouch!!!
A 200 Ohm resistor between the output of the Op Amp & the capacitive load tamed the beast.

VK6ZGO

[Alex]

I had a problem where a differential amplifier which was supposedly handling dc would burst into oscillation at several hundred KHz.

I have read of an audio power amplifier oscillating at 90+kHz. Couldn't tell from the speakers, only on the spectrum analyser.

Op Amps driving capacitive loads can oscillate,despite having internal correction.

Yeah, additional phase shift from the capacitor, a reactive element might cause oscillation. Adding the resistor pushes the pole introduced by the capacitor to the left hand side of the z-plane. This in practise translates to a decaying signal, not growing. These are just different ways of viewing the same problem.

At the end of the day if the gain is <1 when the phase shift from the opamp (and the feedback network) is 180 degrees then it won't oscillate. For linear circuits that is. Of course, physical prototypes have stray elements so it is best to leave adequate margin (phase margin) on paper.