Output signal distortion from op-amp

Question

I am currently using a TLE2144, and the input signal of 600 kHz, 4.5 Vpp is generated from a function generator.

Using a voltage follower with TLE2144, I was able to get an output of 4.5 Vpp with an input from the function generator, but after it reaches above 510 kHz, the signal starts to distort on the negative cycle.

The gain bandwidth of TLE2144 is around 5 MHz, and the slew rate is 27 V/μs. Therefore, I should be getting a good signal on the output, yet the negative cycle of the signal looks slightly distorted. Not sure what the issue is. I do not know what I need to look for.

Schematic:

Dual Supply of +/-12 V

Image shown below (input: cyan signal, output: yellow signal):

I don't care about the phase difference.

I may have found the issue; even though it mentions that for rhe TLE2144IDW, the gain-bandwidth is 5.9 Mhz, they have provided a chart where the signal starts attenuating after 500 kHz. I will have to look into this more.

Answer 1

If the 330µH inductor has a high enough self-resonance frequency and does not begin to behave as a capacitor at 600kHz, then its impedance will be pretty high and with 4.2Vpp at the input it should limit output current to a very reasonable +/- 1.7mA. So it's not an output current problem.

The opamp is used at a relatively high frequency compared to its gain bandwidth product: it has 6MHz GBW, so at 600kHz, only an open loop gain of 10 (20dB) remains to correct nonlinearities. From the internal schematic, internal compensation doesn't seem to wrap around the output stage, so only global feedback can reduce distortion in the output stage. However, the amount of feedback that counts towards reducing a certain distortion harmonic does not depend on the open loop gain at the fundamental frequency, but at the frequency of the harmonic in question. So, for example, to correct harmonic 2 at 1.2MHz, it loses 6dB of feedback, leaving only 5 (14dB). And for harmonic 4, only 8dB of feedback remains.

As frequency increases, crossover distortion will increase too, and feedback available to correct it is reduced. If the output stage is not included in the compensation, this causes a sharp rise in THD at high frequency, as shown by the datasheet:

This is in agreement with your measurement: due to the inductor, current is shifted by almost 90°, so the output stage crossover will occur on the peaks of the output voltage waveform, right where a kink appears on your output voltage. The output stage is not symmetrical, and both sides will switch at different speeds, so it is not surprising the distortion is not symmetrical and only occurs on one polarity of the signal.

Distortion should rise sharply with output current, so you should be able to adjust the input voltage and find the point where it pops up.

Basically, if you want low distortion on a signal, the opamp needs to have enough feedback remaining at the harmonic frequencies to be suppressed, which means it needs GBW much higher than the input frequency. A symmetrical linear output stage that generates little distortion to begin with is also a plus. Most "oldschool" rail to rail topologies will not work for this, but recent rail to rail outputs which wrap the compensation around the output stage can have excellent performance.

Answer 2

Before researching into harmonic distortion data and other fine matters as possible causes of your observation, make sure these do not fall victim of measurement artefacts. Doing measurements, when there are other sources of significant current pulses synchronous with your function generator signals, isolate your voltage follower circuit from any other circuits of which it may be a constituent part. If pushed to the brink of despair in vain search for truth, I would not hesitate to make a one-off PCB with only this circuit. Why? Your inductor may be a receiver of EMI. Then, the interference can leak through supply voltage: at 600 KHz the supply voltage rejection ratio decreases down to mere 60 dB. Also, it may be helpful to read TLE2141 and TLE2141-Q1 EMI Immunity Performance.

Interference and supply leakage having been excluded as a result of consideration into the matter or the DUT arrangement reshaping following my recommendation, take care of the phase difference issue and the timing data: if the standard deviations of 15 for a phase shift and 14.4 MEG for the frequency shown on your scope display have a legitimate explanation, attach these to the question, if you are still interested in the quality answers from the community. If there is no legitimate explanation, take care of this matter! It may give a clue capable to solve your problem.

Do not discard the simulation data from consideration: however sound, speaking in general, is disbelieve concerning simulation predictions, I've examined the TI simulation model for TLE2141 and not found any implementation detail (at least those I am aware of) that the developers had missed to include. For the record: in my opinion, the model covering of harmonic distortion is quite trustworthy.

No visible distortion at all. You may become interested in output current:

Finally, the .ac analysis simulation run

gives an \$A_{VD}\$-vs-frequency plot, exactly representing the graph of Figure 15 of the datasheet. The inductor in the opamp feedback is only a technical trick used for the .ac analysis, permits to compute the DC operating point, helps avoid the opamp being sticked to the rail voltage, and has no meaning for the real device operation.

Answer 3

Using a voltage follower with TLE2144, I was able to get an output of 4.5 Vpp with an input from the function generator, but after it reaches above 510 kHz, the signal starts to distort on the negative cycle.

Unless a "wiring" problem on test board ...

Don't know if this can help, but my simulation does not show distortion until 700 kHz is reached with a peak-to-peak voltage output of 20 V (as OP third graph shows).
Don't know where the "bug" is?

For ~ 5 Vpp, frequency should be 2.5 MHz (?).

Answer 4

For an op amp to be accurate, its open loop gain at a given frequency should be “infinitely high”. The op amp you’re using has open loop DC gain in the tens of thousands. But at 0.5MHz, the open loop gain is down by more than 3 orders of magnitude. A good follower for a 600kHz signal would need, say, open loop gain of 1000 at 600kHz. The GBW needed is 600kHz×1000=600MHz, not 27MHz!

The load you have is reactive, and some op-amps may not like it even if it is not capacitive. It depends on how the inductor is constructed perhaps.