Power Operational Amplifier LM675

Hi
If I want to build a unity gain non-inverting amplifier using LM675
https://www.electro-tech-online.com/custompdfs/2012/11/lm675.pdf (figure 4 page 7)
Data sheet tells me that unit gain bandwidth is equal to 50KHz



So can I change the unit gain bandwidth by selecting/changing op amp noise gain?
Or simply unit gain bandwidth is constant and equal to 50KHz. And all this additional component are needed because LM675 is not unity gain stable amplifier and this component don't have any effect on unit gain bandwidth.

I'm not an expert in op-amps, but I know a bit. I'll chime in with what I do know if nothing more than just to kill some time.

As far as I know, most op-amps have in them a bypass capacitor that is more and more closed circuit to higher and higher frequency's. It's a high pass element is intended to tune them to be more linear for all frequency's. Again, just to stress this point, that part is internal to the IC and can not be changed.

That being said, looking at the schematic, R[SUB]1[/SUB] and C appear to form just such a high pass element from the (+) input to the (-) input. If my assumption is correct then you may just be able to raise R[SUB]1[/SUB] or lower C to increase the frequency limit. HOWEVER. . . Those parts could be there to promote stability. Meaning that if you change them a significant amount the thing may oscillate and/or draw massive amounts of power.

Bottom line, IMO the manufacturers schematic most likely give you as much room as the thing can safely and accurately amplify, as that should be win-win for all party's. But you should wait for one of our wise elders to confirm this before you start chopping things up... OK?

Edit: I just realized that the before mentioned parts may mess with phase, making my hypothesis about them less likely.

Hi Joni,

you are right in assuming that R1-C is necessary for stability reasons. This scheme can be used to operate an opamp that is not unity-gain stable with a gain of 0 dB.
This input compensation effectively reduces the loop gain in the "critical" frequency region and, thus, provides stability.
However, as a consequence, the loop gain reaches 0 dB at a relatively small frequency (in your case 50 kHz).
This frequency determines the bandwidth of the whole amplifier with feedback.
Of course, you can try to somewhat increase this frequency by increasing the value of R1 - but at the same time the stability margin will be reduced.
Thus, watch the step response which is a good indication for the stability properties.

W.

So the unit gain bandwidth is constant and always equal to 50KHz? Even if I choose R1 = (Rs+R2)/20 and R1*C = 1/2pi250KHz the unit gain will not be change to 25KHz?

You can go downhill from where you're at I would think, just not up hill. (i.e. Lower freq, but not higher.)

Jony130 said:

So the unit gain bandwidth is constant and always equal to 50KHz? Even if I choose R1 = (Rs+R2)/20 and R1*C = 1/2pi250KHz the unit gain will not be change to 25KHz?

Click to expand...

I don`t know how much stability margin was envisaged for a loop gain of 0 dB at the frequency of 50 kHz. And I don`t know your stability requirements.
Why don`t you simply try another value for R1 (play a bit with this value via simulation) and watch the result of the step response.

But which op-amp I should use instead LM675. Because I do not have LM675 spice model.

Jony130 said:

But which op-amp I should use instead LM675. Because I do not have LM675 spice model.

Click to expand...

What really is your problem? Do you need more than 50 kHz bandwidth?

I need less than 50 KHz. So can I select components as I show in post 4 to get 25 Khz?

Do you need (at least) 25 kHz or are you required to restrict the bandwidth to 25 kHz?

In the mean time, I have simulated a unity gain amp with the LM675.
However, the results are not very interesting because the phase margin of the model without compensation network R1_C is still more than 50 dB.
That means the model does not reflect the data sheet information (gain larger than 10).

I simply want to know whether I can use this additional component to restrict the bandwidth to 25 kHz.

Yes, of course. As mentioned before, you can set the frequency at which the loop gain is 0 db - equivalent to a closed-loop bandwidth of same value.
On the other hand, there are some other methods to restrict the bandwidth of am amplifier. What about a first order lowpass ?

Ok so this calculation show in post 4 give me the right solution?

Also why this "technique" don't work for "normal" op- amp? LTspice show no effect on unit gain frequancy.

No, I don`t think so.
What is the source of this calculation? I think, it is a bit more involved. You need the opamps open-loop gain response.
As to your 2nd question: This technique of input phase compensation works for each opamp.
I don`t know what you have done with LTSpice - so I cannot comment.

I try use equation from data sheet.
If for R1 * C > 1/(2pi500Khz) and R1< (Rs+R2)/10 we have UGB = 50KHz so for R1 * C > 1/(2pi 250Khz) and R1<Rs+R2)/20 we should get UGB = 20KHz.
But this approach will not work. So can you show me how it should be done?

Jony130 said:

I try use equation from data sheet.

Click to expand...

The symbol ">" does not indicate an equation!
Give me some time - I have to consult the data sheet.

Winterstone said:

The symbol ">" does not indicate an equation!

Click to expand...

I know that.

Winterstone said:

Give me some time - I have to consult the data sheet.

Click to expand...

No problem I can wait.

Hi Joni,

I did a rough calculation - based on the information from the data sheet.
Unfortunately, there is no gain-vs-frequency diagram and no mentioning of the first pol frequency.
Thus I´ve made some guessing.

From the data sheet:
DC gain approx. 90 dB (31600) and gain-bandwidth-product (G=60 dB) of approx. 5.5 MHz
From this I have deduced that the first pole frequency is approx. fp1=175 Hz.

With a gain slope of -20 dB/dec. we have at 17.5 kHz an open-loop gain of approx. 90-40=50 dB.
In the following I use this frequency of 17.5 kHz as a design goal - you easily can adapt the calculation to 20 kHz or 25 kHz.
Let`s go one decade back - and we require that the feedback network Hf(s) has a pole at fp=1.75 kHz.

According to your drawing the feedback function is

Hf(s)=(R1+Xc)/(R1+Xc+R) with R=Rs+R2 and Xc=1/sC

and: 1/Hf=1+R/(R1+Xc).

For large frequencies (s approaching infinity, Xc=0) we now require

1/Hf=1+R/R1=50 dB

with a pole frequency wp=1/R1*C=2*Pi*1.75 kHz.

This leads to

R1*C=1E-4 sec.

Both equations in bold can be used for finding the component values.

Finally, let me state that I really don`t know for what purpose you need a unity gain amplifier with a fixed upper frequency limit, which is realized using the compensation scheme under discussion.
There are other methods to limit the frequency band. Nevertheless, I have tried to show you the way you have chosen.

Thank you Winterstone.
I understand almost everything except why the pole is at 1.75 kHz ??

Joni, perhaps you are interested in another - simpler - alternative to wire your opamp as a unity-gain amplifier with a bandwidth of approx. 20 kHz.
The method is as follows:
* Use a resistive feedback network as it is normally used for an inverting gain of approx. -100 (R2/R1=100).
* Connect the input signal to this inverting amplifier (R1) and - at the same time - also directly to the non-inv. input.
* This scheme gives you a gain of +1 for all resistor ratios R2/R1.
That means, you now have the possibility to select the amount of feedback with R2/R1 without influence on the resulting gain of +1.
* Modifying the feedback network changes the loop gain and, thus, the bandwidth of the closed-loop gain.
* For my opinion: A better solution than the one as discussed before.
Try it.