1

I have been wondering this for a while. When we look at the electrical characteristics of a device such as an opamp there is a gain versus frequency plot.

I have couple of questions regarding measurement of frequency response in practice:

1-) How is an opamp's frequency response obtained in reality/on bench? Do they sweep or measure the output voltage for each frequency input and obtain the plot? Or do they apply an impulse or step input to the input, obtain the response and do the math and plot the frequency response?

2-) If they do the sweeping freq. method but not the impulse or step input method; where is the second one used in practice? I mean is the step input or impulse input used in any case in the engineering world to obtain the frequency response? Or are these only used on paper(mathematical modelling or simulations)?

Transistor
  • 168,990
  • 12
  • 186
  • 385
user16307
  • 11,802
  • 51
  • 173
  • 312
  • *"Do they sweep or measure"* who are the "they" you refer to because your question seems to hinge on this. – Andy aka Sep 15 '18 at 13:18
  • 1
    I meant the manufacturers when they generate the datasheets. – user16307 Sep 15 '18 at 13:20
  • 1
    Lets say you designed an opamp at a semiconductor company and want to plot its freq. gain response before putting on the datasheet. I thought it is done on bench. – user16307 Sep 15 '18 at 13:21
  • I expect they begin the design with a goal and that goal is both the stability of the device and its maximum operating frequency. This in turn leads to a predicted bode plot (or transfer function) and the aim would be to achieve that transfer function so, the typical graphs shown in the data sheets are based on goals and should those goals not be met then someone is for the high jump. – Andy aka Sep 15 '18 at 13:23
  • What the manufacturer does is a frequency sweep AND and step response AND an impulse response AND modelling, on many batches, for several process variations. And if they all agree, to within a spread, then it gets put on the data sheet. If you have a system, it might be more amenable to step, or impulse, or sweep tests. You use the one you can, or what's most convenient. – Neil_UK Sep 15 '18 at 13:24
  • 1
    @Neil_UK Is what you wrote based on opinion or fact? Just curious. – user16307 Sep 15 '18 at 13:27
  • @user16307 based on my experience in industry. If you can test the same thing several different ways, you do. If they don't agree, you know you have a measurement problem. You only put in the data sheet what you can prove, and you know you can repeat. – Neil_UK Sep 15 '18 at 13:35
  • Test time is expensive if it is a bottleneck to production, so some parameters are guaranteed by test and others by design. But for BW, the rise time and unity Gain breakpoint and thus BW can be measured otherways simply at 1 frequency for min, nom and max. The scope captures are representations of nominal. Then statistical Process Controls are used on results. (SPC) – Tony Stewart EE75 Sep 15 '18 at 14:10
  • BTW if you study datasheets carefully, learn how to design by using a table or list of specs, before you design **anything** including a description of inputs and output responses. – Tony Stewart EE75 Sep 15 '18 at 14:15
  • Often the UGBW is set by interaction of an onchip current and an onchip capacitor. Either or both can be trimmed to tweak the Bode response. – analogsystemsrf Sep 16 '18 at 02:32

1 Answers1

2

The test is done with a circuit like this (and this is how analog devices does their AC gain or open loop tests). You could build one of these circuits and make an open loop gain test (with your device in the spot labeled DUT) There are more ways to make this measurement, but this is one of the best. enter image description here

Figure 5, the ac signal is applied to the DUT input via a 10,000:1 attenuator. This large value is needed for low-frequency measurements, where open-loop gains may be near the dc value. (For example, at a frequency where the gain is 1,000,000, a 1-V rms signal would apply 100 μV at the amplifier input, which would saturate the amplifier as it seeks to deliver 100-V rms output). So ac measurements are normally made at frequencies from a few hundred Hz to the frequency at which the open-loop gain has dropped to unity—or very carefully with lower input amplitudes if low-frequency gain data is needed. The simple attenuator shown will only work at frequencies up to 100 kHz or so, even if great care is taken with stray capacitance; at higher frequencies a more complex circuit would be needed.

Source: http://www.analog.com/en/analog-dialogue/articles/simple-op-amp-measurements.html

Voltage Spike
  • 75,799
  • 36
  • 80
  • 208
  • See also this: https://electronics.stackexchange.com/questions/339950/practical-ways-to-obtain-bode-plots-for-an-unknown-circuit/396730#396730 Some use network analyzer for the purpose? – user16307 Sep 18 '18 at 22:26
  • @user16307 that method can be susceptible to output or input impedance loads\parasitics. If your estimating and you don't need to be exact, that method would be easier than building this circuit at the potential cost of accuracy – Voltage Spike Sep 18 '18 at 22:29