OP AMP Stability

[eladta]

Hi all,
I want to use the AD8638 OP Amp as a simple buffer.
I know that according to theory we need to check the phase of the open loop gain where the magnitude of it crosses 0 dB. According the datasheet page 8 it comes to 1 MHz, and the phase is aprox. 65 degrees (meaning stable). But what does it mean exactly ?
say i wanna use another freq. what than ? how do i check the stability ?
Are there any other parameters i need to check in terms of stability ?

Thanks, Elad.

 

[MrAl]

Hi,

This is all about how waves add together, usually just two sine waves.

The reason for all the concern is because the feedback goes right back to the input, so it adds to the input. The feedback comes from the output, so if you have the output go back to the input and it adds up to more than what went in the first time, then you get a higher output, and if then THAT new output goes back to the input, it adds up to even MORE output, so the output keeps growing and growing until the output starts to reach the power supply level and then you either get real bad distortion or you get an output that remains high or low, so the output becomes useless for most applications when this happens.

Now if it was DC, DC just adds to the input in a given ratio, but since we are concerned with frequencies there are two sine waves: the input and the output coming back to add to the input as feedback. If the two sine waves add up to a level that is too high, the output goes into saturation and locks up or else produces a huge amount of distortion.

Sine waves are a little more tricky than DC because they have a phase associated with them. So we have to think about what happens when two sine waves add together. The phase now has a lot to do with it as well as the gain, because two sine waves that are in phase add just like DC, but two sine waves that are out of phase by 90 degrees do not add the same way. For example, if we had a DC level of 1v added to another DC level of 1v we would get 2v. If we had two sine waves with a peak of 1v and they were in phase we would get 2v peak output, but if they were 90 degrees out of phase we'd get a peak again of 1v, just with a shifted phase. So in the 90 degree case we dont see the amplitude increase, while in the 0 degree case we see it increase by two times the value of one sine peak, and that is what could cause oscillation which could lead to saturation. That's basically why there has to be a phase criterion as well as amplitude criterion.

 

[RCinFLA]

The spec refers to 'gain/phase' margin. For oscillation, (which you don't typically want in an amplifier) the closed loop gain (forward gain and feedback) has to be greater then unity and phase of shift of 360 degress. The invertering amp has 180 degrees so any phase shift that adds another 180 degs to total loop phase shift will cause oscillation if close loop gain is greater then 1.

Gain/phase usually refers to what is phase shift at open loop gain of 1 and what is open loop gain at phase shift of 180 degs. General practice is to ensure at least 45 degrees of phase margin (no more then 135 degrees open loop shift).

An op amp setup for unity gain with feedback is usually the toughest situation since feedback is full output. An op set with a gain of 10 will have 0.1 times output applied to input so the gain around the loop is reduced by factor of 10.

Many low frequency op amps have internal compensation that makes them almost unconditionally stable with just resistor feedback. Higher frequency op amps are only partially compensenated or no compensation. For these op amps you have to be careful about stray capacitance of PCB layout and don't use too high of values for feedback resistors. For example if you use a 10k resistor for output to input feedback and there is 5 pF of node capacitance at the op amp negative input you have a low pass pole in the feedback of about 3 MHz. This 10k resistor feedback adds up to 45 degs of shift at 3 MHz to feedback. For high frequency op amps you generally want to use lower value resistors (which also take more drive capability from op amp output) in the feedback network so you can stand some stray capacitance in the hook up.

 

[simonbramble]

Have a look at this article. I could never understand gain and phase margin until I thought about how a real op amp worked. I then analysed it in LTspice and the whole concept started to make sense. Starting with a model of the ideal op amp is where most people go wrong. it all falls apart after that. You have to think of an op amp with limited gain.. Have a read:

Gain and Phase Margin in Op Amps:
http://www.simonbramble.co.uk/techarticles/advanced_op_amps/advanced_op_amps.htm

 

[eladta]

The spec refers to 'gain/phase' margin. For oscillation, (which you don't typically want in an amplifier) the closed loop gain (forward gain and feedback) has to be greater then unity and phase of shift of 360 degress. The invertering amp has 180 degrees so any phase shift that adds another 180 degs to total loop phase shift will cause oscillation if close loop gain is greater then 1.

Thanks everybody for your comments, i just got back to work on this subject so sorry for the late response..

What you are saying makes a lot of sense Intuitively, but the equation for the transfer func is Aol/1+LG (where LG is the loop gain, beta*Aol), Hence the circuit is unstable when 1+LG=0, meaning when the amplitude of LG is 1 and the phase is 180 deg . not 360 deg !! as we also know we always check the difference between the phase shift of the LG and 180 deg, even when we use non. inv. op amp. so there is a contradiction here !!

 

[simonbramble]

The equation Aol/(1+BAol) assumes the feedback fraction, B, is negative. If you want to neglect phase, the correct equation to use is Aol/(1-BAol). Then B is a negative value (as it is fed back to the inverting input), thus making the denominator (1+BAol).

So with the denominator equal to (1+BAol), system is always stable with negative feedback since (1+BAol) is always greater than 1. However, if the feedback fraction is inverted, the denominator becomes (1-BAol) and when the loop gain, BAol, equals unity, the system will oscillate