Need gain control without phase shift

[surkris]

Hi,

I have a 1KHz sine wave synchronized to a 1KHz TTL square wave. I need to control the peak-to-peak voltage-level of the 1KHz sinewave. I require two gain settings 1.5 and 0.5. Would it be possible for me to use an OPAMP for this function? The problem might be that at 1KHz and at the required gain settings , there would be some amount of phase shift, however small. I need to maintain the original phase with an error of about 1 degree or so.


Can anyone suggest a suitable approach?

 

[Vamsi Mocherla]

If you want to have a gain of 0.5, it cannot be realized with an opamp. I mean there will be stability problems. Well, if you want a 1.5 gain stage, you can surely try it with an opamp in a non-inverting configuration.

 

[JPR]

Here is how I would approach the solution:

1- Attenuate the input. Use a resistor divider to generate Vout = 1/2 Vin. (This assumes that the sine wave input can tolerate the resistance load) Keep the resistor values smaller than 500kohms with a load capacitance of 10pF at the next stage. (Keep the bandwith greater than 115kHz for less than 0.5deg of phase change at 1kHz.

2- Use a noninverting amplifier with a feedback resistance from output to inverting input of R, and a resistance of 2*R from gound that can be connected to the inverting terminal or not connected. This will provide a gain of 3 when the resistor is connected and a gain of 1 when the resistor is not connected. Agian the bandwith must be greater than 115kHz for less than 0.5deg of phase change at 1kHz.

The overall effect will be gains of 3*0.5 = 1.5 and 1*0.5 = 0.5. The overall phase change will be the sum of the two phase changes and will be less than 1degree.

If you can handle high output resistance and have a low capacitance load at the output, and your power supplies can always handle the 3x gain, you could put the resistor divider at the output of the gain stage, rather than before. If you can not handle the load conditions at either point, use a buffer (again with sufficient bandwidth).

 

[10kangstroms]

2- Use a noninverting amplifier with a feedback resistance from output to inverting input of R, and a resistance of 2*R from gound that can be connected to the inverting terminal or not connected. This will provide a gain of 3 when the resistor is connected and a gain of 1 when the resistor is not connected.

You got the resistor values reversed. Rgnd=R, Rfeedback=2*R gives gain=3.

 

[surkris]

Yes, I was worried about stability for a gain of 0.5.

Thank you JPR. That's what I had thought I ought to do. Actually I ran a simulation using 5spice (you can get a free evaluation at www.5spice.com) and I seem to get a phase shift of less than 0.02 degree at 1KHz. This was for an OPAMP from Linear Technology.

JPR, how did you arrive at the B.W figure of 115KHz for .5 degree phase shift?

 

[lastdance]

i couldn't quite figure why one should use a monstrous gain device in the form of opamp and make an attenuator.

why can't a simple resistive divider do the job?

u r working on kHz range, parasitics should not affect ur linearity. how linear and up to what range of freq. do ur app call for?

 

[JPR]

10kA - Yes, I did have the resistors reversed in my description.

lastdance- The reason for the op-amp is the need for gain in one of the two settings(1.5x). If the needed gains were 1x and 0.5x, I would opt just for the resistor divider (if there was no need for buffering any of the signals)

surkris- The BW calculation assumes a single pole. The phase change at 1kHz due to a [single] pole at BW will be arctan(1kHz/BW). By rearranging the equation, the BW for 0.5degrees at 1kHz will be BW = 1kHz/tan(0.5deg)

I used 0.5deg for each to ensure that the total phase change is less than 1deg. Higher bandwidth will provide even less phase change. Opamps will typically provide a GBW of a MHz or more. With the gain of 3 configuration, this will be a BW of 330kHz. To obtain less than 0.02deg of phase margin, I would guess that your amplifier has a GBW about 8MHz?

 

[surkris]

JPR: GBW is 15MHz. Thank you very much.

 

[Vamsi Mocherla]

It would be very risky to have an OPAMP used in attenuator configuration. The phase will be really bad in attenuator conditions.

 

[JPR]

In this case, the attenuator is OUTSIDE the feedback loop, so, as long as the amp is unity gain stable, it will be okay.

 

[Vamsi Mocherla]

Then the amp is only working as a buffer. The attenuation is not done using an active device.

 

[10kangstroms]

What is all this concern about stability?
First of all - you can obviously not get less than unity gain (noninverting) in a single op amp stage (unless you use an attenuator in front of it).
You CAN get less than unity gain in inverting mode - all the way down to zero gain when Rf=0. If the op amp is stable in the unity gain, non-inverting configuration, it will be stable for ANY inverting gain (even zero), since the loop gain will always be less than or equal to the loop gain in the unity gain, noninverting configuration.
I'm not saying that fractional inverting gain is desirable (although it might be in some cases), only that stability is not a problem.