Opamp with fixed gain that changes.

[antknee]

I have an OPA452 operational amplifier circuit that I made up last night. The circuit works and the schematic is below with a copy of the datasheet. It is non inverting with R1 and R2 both at 39Kohms and the chip is listed as stable at unity gain or above, so the gain should be stable at x2. GBP = 7.5MHz and SR = 5V/us.

I'm finding at 10KHz the gain is 2 and at 100Khz it is 4. Why is the gain changing? I thought it was supposed to be fixed?

Thanks.

 

[Gary B]

At a guess, you have a stray capacitance in your layout that is providing a feedback path that is not in your schematic. That assumes R1 is not a wire-wound resister.

 

[antknee]

That is interesting, I will check to see if there is anything obvious on the board. I will be making another one up maybe tonight, I will use a different layout aswell. I did check the datasheet and it mentioned phase gains but I don't have a load attached, so it can't be that.

 

[unclejed613]

is it exactly as the schematic shows. or are you coupling input or output through a capacitor?... or are you driving a reactive load?

 

[antknee]

I've just rechecked the circuit, it is exactly like in the schematic and does work. I dont have a load attached as of yet, and the input and output arent coupled to a capacitor.

I've just rechecked the gain...at 10KHz with 2.5Vpp input I get an output of 5Vpp and at 100KHz it is 7.5Vpp. Perhaps it is one of the vagaries of this chip. I guess I just need to be aware of this if it happens with the next circuit I make up.

 

[BrownOut]

I doubt it is a vagarie of the chip. How are you measuring the input/output signal? Are you monitering the input as you increase frequency?

 

[MrAl]

Hi antknee,


What are your plus and minus power supply voltages? You need at least plus and minus 10 volts, and you may also need at least some resistive load to test this chip.

Also just for the record, the GBW for the 452 is not 7.5MHz, it's only 1.8MHz. You should double check to make sure you are using the 452 and not the 453 because the 453 may not be stable at a gain of only 1 or 2.

 

[MikeMl]

Also, are you building it on a strip-board with wire jumpers? You will never get the correct performance above ~100Khz unless you build it on a PC board with almost zero lead lengths, separation between inputs and the output leads, guard bands, bypassing with chip capacitors, etc, etc...

 

[antknee]

The input is from a signal generator, I have a push button that changes from 10s to 100s of KHz so is I can compare output with the push of a single button. It is unusual.

The chip is the OPA452, I mistakenly noted the GBP at 7.5MHz but it is 1.8MHz. I have an OPA 453 that I am making up right now, this led me to err. It is possible to get the OPA453 working at lower gains thus taking advantage of its higher SR but that requires a different configuration, the OPA453 is a very good chip assuming the gain doesn't drift I'm supplying +36 and -36.

Also, are you building it on a strip-board with wire jumpers? You will never get the correct performance above ~100Khz unless you build it on a PC board with almost zero lead lengths, separation between inputs and the output leads, guard bands, bypassing with chip capacitors, etc, etc...

I did build it on a strip board with jumper leads farily quickly. I am currently making up another, which will be tidier with shorter leads, I half have it in mind that the bypassing has been done in the schematic but i'm not sure? not sure what guard bands are? the output leads are apart I will make sure they stay apart on the next build.

Thanks.

 

[MikeMl]

...I did build it on a strip board with jumper leads farily quickly. I am currently making up another, which will be tidier with shorter leads, I half have it in mind that the bypassing has been done in the schematic but i'm not sure? not sure what guard bands are? the output leads are apart I will make sure they stay apart on the next build.

Strip boards do not work above about 100Khz, PERIOD! The capacitance between parallel strips is about 10 to 20pF. This causes havoc with high-frequency operation, just as you have already experienced.

 

[BrownOut]

Are you sure your sig-gen output doesn't change when you change frequencies? How are you measuring the output? O-scope? DMM? You either have a measurement error or a spurous reactance from your construction method. Gain won't drift.

 

[MRCecil]

You might want to place a resistor from the non-inverting input to ground equal to the value of your feedback resistor, R2. This will let your sig. gen. work into a resistive load rather than the reactance of the input alone.

 

[crutschow]

What you are observing is not unusual. Any stray capacitance at the inverting node will cause a peak in the response. You need to add a small compensation capacitor (likely 10 to 100pF) across the feedback resistor (R2) to eliminate this peak. It's value can be determined experimentally. A square-wave is a good way to do this. The output square-wave will have no overshoot when properly compensated.

Of course this will reduce the high frequency response below what the gain-bandwidth value would indicate. Thus for maximum bandwidth you want to minimize the stray capacitance at the inverting node (short leads), which reduces the value of the required compensation capacitor. Lower resistor values will also reduce the effect of the stray capacitance.

 

[antknee]

I'm sure it is stray capacitance from the stripboard and leads. I will tidy up, reduce lead length, add a compensation capacitor and reduce the R1 and R2 resistor values.

Thanks for your comments.

 

[unclejed613]

having a peak in the response is also an indication of instability, and the circuit could become an oscillator under certain conditions (such as stray capacitance across the output). the causes of the strange behavior here are probably stray capacitance from the inverting input to ground and the unterminated noninverting input being coupled to the inverting input by stray capacitance, which would create a positive feedback mechanism as the frequency rises. since the stray signal is out of phase with the inverting input (because it's capacitively coupled), it's effect isn't cancelled by the inverting input.

 

[Mikebits]

You should also make sure that you do have some sort of load on the output, and I also agree with brownout that you should verify your sig gen output is not shifting with frequency.