Computing stuff tied to the physical world

Different op-amps

In Hardware on May 25, 2012 at 00:01

One last experiment I wanted to do after the recent sine wave circuits, was to compare a few different op-amps.

I’m including the original one here as well – the LM358, running at ±13.6V:

SCR64

Here’s the LT1413, running at ±14.4V:

SCR65

And here’s the NE5532ANG, running at ±15.3V:

SCR66

In each case, the supply voltage was adjusted until the output sine wave was ±10 V, with all other components identical. Note the slight difference in oscillation frequency.

What’s also interesting, is the mean output voltage: it should be 0V with an ideal circuit. Looks like the NE5532ANG performs best – within 1%. It’s described as being an “Internally Compensated Dual Low Noise Operational Amplifier”. Second harmonic is at -51 dBm, i.e. 0.28% harmonic distortion – an excellent signal!

As a quick test with that last op-amp, I reduced the supply voltage to ±2.5V – the effect was a slightly higher frequency of 522 Hz, a much lower output of 2.14 Vpp, i.e. ±1.07V, but relatively far off-center: 240 mV. Harmonic distortion rises to 3.5% in this case. But that’s not surprising: the NE5532ANG is only specified down to ±3V, and it’s not a “rail-to-rail” op-amp, which means it cannot generate an output voltage too close to its supply voltage (with a ±5V supply, distortion drops back to 1.25%).

Lots of op-amps. Lots of trade-offs.

Ok enough op-amp chit-chat for now, I’ll stop :)

  1. “-51 dBm, i.e. 0.28% harmonic distortion ” is there perhaps some trick I don’t know yet for quickly getting from -51dBm to that 0,28% ? dB’s are nice, i know that adding them is like multiplying, but how to “unlog” dB’s into percentages easily ?

  2. Ahem! When quoting a measured distortion figure, it’s not ‘-51dBm’, but ‘-51dB’. The Decibel, like the percentage, is a way of expressing a ratio. When it acquires a trailing letter (like m) this signifies a reference level, so that the ratio is of the measured quantity to the reference level. The letter ‘m’ is short for a reference level of 1 milliwatt (usually into 50 ohms).

    Decibels are power ratios, so a distortion of -51dB means the power in the harmonic is 51dB less than the power in the fundamental, which is understood to be the reference. The definition of a decibel is 10 times the logarithm to the base 10 of the ratio of the powers: 10*log10(P1/P2). So -51dB corresponds to a power ratio of 7.94*10^-6. When expressed as a percentage, the distortion is given by the ratio of the two voltages. Since power is proportional to voltage squared (assuming the same resistance), the voltage ratio is the square root of this power ratio: 2.82*10^-3, or 0.28%. The quick way is to take this into the dB calculation:

    (power) ratio in decibels = 2*10*log10(V1/V2) – where V1, V2 are the two voltages

    (power) ratio in decibels = 2*10*log10(P/100) – where P is the voltage ratio in %

    (voltage) ratio in % = 100*10^(B/20) – where B is the power ratio in decibels.

    Incidentally – notice the tiny little steps on the output of the LM358, that occur just below the zero voltage output. This is crossover distortion, which this family of devices are famous for. Its mostly high order, so won’t affect the component tester application, but it sounds horrible. You can get rid of it by adding a load resistor from the output pin to the negative supply rail (about 10k IIRC). This effect doesn’t show up well on an FFT, but is obvious when looking at the output of a distortion analyser.

    • Whoops – thanks for setting this straight. Also interesting info about the crossover distortion, I never knew what was meant by that.

      PS. Nothing like this sort of feedback to keeps pushing me forward – thank you!

  3. Your oscilloscope screen shots look stunning! What make/model ‘scope are you using?

Comments are closed.