Anti-logarithmic RF amplifier for reducing harmonics without filtering

[neazoi]

Hello,
Considering a wideband RF oscillator followed by an amplifier.
This system usually produces harmonics can be filtered with a LPF.
In multi-band operation more than one LPFs are switched accordingly.


My idea is to use an ANTI-logarithmic amplifier (if such thing exists) to reduce these harmonics without any LPF on multi-band operation.


It works like this:
The anti-logarithmic amplifier will amplify the larger signals (which is the carrier) more, but it will amplify the smaller signals (which are the harmonics) less.
Using such an amplifier and combined with an output attenuator possibly, the harmonics levels compared to the carrier, will be reduced.


1. Does such an anti-logarithmic amplifier exist?
2. Will this idea work?

 

[FvM]

1. yes, an exponential circuit. Maybe extended to symmetrical form, like y = sign(x)*exp(abs(x))
2. not at all

The "anti-log" circuit distorts the sine wave, creating new harmonics.

- - - Updated - - -

I think, the problem is that you are confusing the intended signal manipulation in frequency domain with manipulation in time domain.

 

[neazoi]

The "anti-log" circuit distorts the sine wave, creating new harmonics.

Yes I understand, but even the LOG amplifier is also non-linear and yet used in measurement equipment.
Moreover the exponential amplifier may create harmonics, but these won't be amplified as much as the carrier. So more harmonics possibly but at lower levels compared to the carrier.

 

[volker@muehlhaus]

Yes I understand, but even the LOG amplifier is also non-linear and yet used in measurement equipment.

... but not in the RF path as you suggest.

Moreover the exponential amplifier may create harmonics, but these won't be amplified as much as the carrier. So more harmonics possibly but at lower levels compared to the carrier.

As FvM says, you confuse frequency domain amplitude and time domain. The amplifier would process a voltage that is the sum of sine wave + harmonics. It doesn't process the harmonics level independently from the sine carrier.

And even an ideal sine wave processed with an exponential amplifier will generate harmonics. Nonlinear amplifier -> distortion from the sine-> harmonics

 

[neazoi]

As FvM says, you confuse frequency domain amplitude and time domain. The amplifier would process a voltage that is the sum of sine wave + harmonics. It doesn't process the harmonics level independently from the sine carrier.

I am sorry but I struggle to understand this.
Won't an amplifier amplify whatever signals come into the input?
A distorted input signal isn't the sum of the carrier and the harmonics?
I am talking about a "hypothetical" exponential amplifier not an ordinary amplifier.
Please let me know.

 

[volker@muehlhaus]

I am sorry but I struggle to understand this.
Won't an amplifier amplify whatever signals come into the input?

Yes, but your idea is based on frequency domain with harmonics signal = small level = less amplification. You treat signal separately, as if the amplifier would see carrier only (large amplification), then harmonics only (small amplification).

A distorted input signal isn't the sum of the carrier and the harmonics?

Exactly, it is the sum! So with your exponential amplifier, the amplifier gain will then depend on the time varying instantaneous voltage of carrier + harmonics. Example for a heavily distorted sine wave: make it look more like a square wave, by adding a lot of odd harmics. How can any theoretical amplifier make that a sine again?

(If you are not familiar with the theoretical concepts of time domain <> frequency domain, look for "fourier transform" examples)

I am talking about a "hypothetical" exponential amplifier not an ordinary amplifier.

Your amplifier would need to split signals into spectral components first, and amplify all of them separately. FvM and I understand that your idea is inspire by the spectrum analyzer display, where that frequency separation happens. But an RF amplifier sees all the spectral components added, and amplifies at any time the input voltage which is the sum of all these spectral components. Then there is no "small harmonics voltage", there is only one sum voltage that is carrier + distortion.

 

[neazoi]

Yes, but your idea is based on frequency domain with harmonics signal = small level = less amplification. You treat signal separately, as if the amplifier would see carrier only (large amplification), then harmonics only (small amplification).



Exactly, it is the sum! So with your exponential amplifier, the amplifier gain will then depend on the time varying instantaneous voltage of carrier + harmonics. Example for a heavily distorted sine wave: make it look more like a square wave, by adding a lot of odd harmics. How can any theoretical amplifier make that a sine again?

(If you are not familiar with the theoretical concepts of time domain <> frequency domain, look for "fourier transform" examples)



Your amplifier would need to split signals into spectral components first, and amplify all of them separately. FvM and I understand that your idea is inspire by the spectrum analyzer display, where that frequency separation happens. But an RF amplifier sees all the spectral components added, and amplifies at any time the input voltage which is the sum of all these spectral components. Then there is no "small harmonics voltage", there is only one sum voltage that is carrier + distortion.

All right, I think I understand now,
So a signal waveform in the time domain, can be analyzed, or is the RESULT of, the carrier and the harmonics. And if one adds the same harmonics as separate signals to a clean carrier then he will end up with the result distorted waveform.
However, the opposite is not true, it does not mean that the amplifier seed this distorted waveform as separate signals (like displayed on the spectrum analyzer), although the signals are there because you can detect them in a near by receiver.
Am I correct on this?

This makes me think of how they do the exponential display of the signals in spectrum analyzers....? Is it just a part of the detection process...

 

[volker@muehlhaus]

Yes, it is really just two views (time domain: distortion, frequency domain: multiple frequency components) of the same signal.

This makes me think of how they do the exponential display of the signals in spectrum analyzers....? Is it just a part of the detection process...

I think the DC signal at the detector output is going to that log amplifier for display.