Hello,

I've got a question regarding OP -AMPs. I'm using an OPA548 in the basic non-inverting Op-Amp config. My input signal is a 100 kHz square wave. Unfortunately, the output doesn't look anything like the input, more like a sine wave. According to the data sheet the Op amp should be fine at that frequency.

Has someone got an idea where I'm going wrong or suggest a different OP-Amp (it needs to be able to handle ±20V and up to 10W)

Thanks,

Nikoh

 

As a thumb rule I once learned. That to amplify a square without to much distortion you need a bandwidth 10x wave frequency. So what is bandwidth of your IC? Also remember the GBW rule if you have any gain in your circuit.

 

Well, the gain bandwidth product is 1MHz.

However, a square wave is the sum of all of the odd harmonics of the fundamental frequency, so you really need a lot more bandwidth to pass all of those odd harmonics.

If you had the opamp wired for unity gain and were buffering a sine wave, you would be just fine. However, you're only getting the fundamental and maybe the first couple of odd harmonics, which makes for a lousy-looking square wave.

 

Are you sure the signal you see is a distorted signal, and not any oscillations? A power opamp is more likely to oscillate than a normal type opamp.

 

Did you use the recommended bypass caps on the supply pins?

What are your values for R1 and R2?

 

At 100kHz the gain of the OPA548 power opamp is only 10. It has a lot of distortion above 20kHz so I think 100kHz will be very distorted.

 

Thanks for your help. I did use all the caps and the resistors were 1k3 and 130R. Which OP-Amp could I use for my purposes? The signal is definitely not caused by oscillation though.

 

How much gain do you need? That is part of it.

There is also another spec in datasheets they throw out with op amps called slew rate. It is usually measured in V/µs. This is the maximum rate the op amp can change the output. A bad slew rate usually shows up as a linear ramp, almost like integration.

 

A gain of 10 would be nice.The output didn't look like a ramp, just like a distorted sine wave.

 

Slew rate and freq response are different things, though they do interact. Slew rate shows up on large signals, on square waves the edge is changing very fast, this is where you will see it.

With freq response you need the op amp to go higher than the fundamental. As other folks explained a square wave actually has a lot more than is obvious, for a 100Khz signal you'll need something that will go 1Mhz at the same gain. This sounds like you'll need a 10Mhz op amp (at a gain of 10). Not cheap.

Something else to look for is a differential amp. They are similar to op amps, with dual outputs. Some of them can go very high freq indeed. Many cases this is what they were designed for.

It may be easier to go with a simple transistor amp. As a class they go to much higher frequencies.

As you are finding out, op amps don't do high freq well. This could change over time, but right now the devices that can are expensive.

 

What is it that you're trying to do with your circuit? What is your application?

I can only guess that you're using that opamp because you have some sort of heavy load that you're trying to drive with higher-than-typical rail voltages. The OPA548 is capable of sourcing/sinking up to 3A continuously with the rails up to 60v apart.

With that amp, like I suggested before, you might get a halfway-decent looking sine wave output if you were operating at unity gain (Av=1; output connected to noninverting input) - however, you are expecting a gain of 10 with a square wave, which just isn't going to happen.

If you have a heavy load, you might consider using a half-H bridge instead. There is no gain; it simply pulls the output from one rail to the other.

As an aside, you are using very low values for R1/R2. While this is a power opamp and at lower frequencies it could handle the extra load, you are generally better off using higher values for gain resistors. Otherwise, you're dissipating power in the feedback loop for no apparent reason. Somewhere in the range of 10k-100k is usually pretty good. There isn't a hard-and-fast rule for it; it depends on the application.