Envelope detector and distortions

Question

I am attempting to design an envelope detector, however, I'm having problems with regards to diagonal clipping and negative peak clipping of a certain input.

While trying to follow this video's simulation file, without modifying its input, I noticed that there isn't a lot of hassle when picking for the RC values since the modulation index was okay for the input wave.

Input: (modulation index: 0.5)

Output:

However, I am unsure of what RC values to choose when my input is: $$f_m = 10kHz\quad f_c=600kHz \quad \text{amplitude}=1V \quad \text{modulation index}=0.5$$

I first tried getting the geometric mean of my messenger frequency and my carrier frequency but the plot shows that these values large

Input:

Output:

I then tried lowering my RC value in an attempt to fix the diagonal clipping, but as seen on the plot, the output still isn't a clean sine wave (compared to the very first output. Decreasing the RC values further would then result to negative clipping and ripples on the upper portion of the output.

Some questions I want to ask are:

1. How do I get the RC values that would still maintain the shape of the envelope given the input? Is it possible to do so without adding components to the envelope detector?
2. If it is not possible to get the optimal RC values, is it still possible to retrieve the message in later stages? (i.e. inputting the output to a high pass filter/low pass filter or amplifier)

Answer 1

If your diode is of silicon, voltage drops around 0.7 V and peak of output is only 1 - 0.7 = 0.3 V at maximum. That's why blue plot looks apart from green plot however you change RC constant.

Answer 2

Sure you have got distortion on the sim. You will get it in real life. Try making the input volts 10 volts to reduce diode effects. If this helps then place a DC voltage source in series with your AC generator. Experiment with a start value of say half a volt. Most AM detectors are pre-biased to remove some diode nonlinearity. Then reduce the RC time constant accepting more ripple. Now you can post-filter to get the ripple down to an acceptable level. Often an RC section is adequate. I have used full-wave bridge to get low 2F ripple that is easier to post-filter. I have benched a germanium pre-biased full-wave bridge that performs well. If you have spice models for these then you could try. Now things can get worse when modulation % is increased like 90%. When you do this on the sim you will have to do more optimization.