Inverse clipper circuit discussion

Hi, A typical clipper circuit, clips all signals above a certain level.

I am looking for a circuit that does exactly the opposite. That is ignore all lower level signals and pass through only the higher level ones.

Any ideas of how such a weird circuit can be made?

Hi,

comparator and analog switch.

Klaus

The name of the circuit is gated AGC, which use a particular treshold to open and close a gate.
The gate is open for strong signals and closed for low level signals.
This approach was used by the old tape noise reduction circuits.

vfone said:

The name of the circuit is gated AGC, which use a particular treshold to open and close a gate.
The gate is open for strong signals and closed for low level signals.
This approach was used by the old tape noise reduction circuits.

Click to expand...

Ok that reminds me of the squelch circuit as used on HF radios.
The thing with squelch is that under the presence of high level signals the "gate" will also pass the lower level ones.
I am looking for a way (if it exists) to pass only the high level signals above a threshold and completely attenuate/ignore the low level ones.
It migh sound a bit weird, or might be totally impossible, I do not know.

Which kind of signals do you want to manipulate, which signal processing methods do you consider? I understand "typical clipper circuit" mentioned in the initial post as a strongly nonlinear signal processing, e.g. cutting the instantaneous signal with limiting diodes. Respectively it causes harmonics and intermodulation, it would be e.g. inappropriate for audio signals, even plain voice. Similarly you can cut low level signal components by a pair of diodes in series circuit, but also involving strong signal distortions. The mentioned noise gate or squelch circuits are in contrast manipulating the signal envelope and not causing strong distortions. On the other hand they are keeping the level relation of simultaneously appearing sound components.

To process the level of sound components separately, you need to decompose the sound e.g. by filter banks. There's no simple way. In the context of your question, that's probably the same as "totally impossible".

I'm confused as to what you want exactly.
What type of signal is this?
Could you draw a picture of the signal for various amplitudes and tell us what output you want?

crutschow said:

I'm confused as to what you want exactly.
What type of signal is this?
Could you draw a picture of the signal for various amplitudes and tell us what output you want?

Click to expand...

I am trying to find a way to extract very high level noise spikes (ignition noise) from usable signal on HF. Then use these spikes with phasing circuits to actually remove them from the initial signal.
The problem I face is that under the presence of high level signals, also the lower ones are passing through the circuits I am trying, so they are also phased out (simple amplifiers just process voltage levels no matter what these are). I am trying to find a circuit/technique that somehow differentiates these signals, strong and weak, based on their amplitudes.

To remove just spikes you need a Noise Blanker circuit...search the net for schematics. It is named "noise" blanker but actually they remove spikes (from ignition, from fluorescent lamps, etc.)
Generally they use also a gated circuit driven by a feedback detected signal. A real Noise Blanker circuit would have adjustments not only for the amplitude of the spikes, but also for their depth and width.

vfone said:

To remove just spikes you need a Noise Blanker circuit...search the net for schematics. It is named "noise" blanker but actually they remove spikes (from ignition, from fluorescent lamps, etc.)
Generally they use also a gated circuit driven by a feedback detected signal. A real Noise Blanker circuit would have adjustments not only for the amplitude of the spikes, but also for their depth and width.

Click to expand...

Yes I am aware of the noise blankers thank you. They work in the IF, before the narrow filtering. They convert the detected spikes into DC then use this to cut off the IF at some point. If I am not mistaken, the settings of the width and the depth are done in the DC part, not the RF. That is after the DC has been created, triggered from the noise pulse.
Noise blankers work only in relatively narrow channels. Imagine the noise from 1MHz below to affect your 1MHz above listenning frequency. However, I am aware at least of a noise blanker product from the 60s, the one used in the Collins KWM-2, which sampled 40MHz with a separate noise antenna, for car ignition spikes, so as to cut off the IF stage of the HF rig even at 20MHz away. This was not an X-phaser, it was a pure noise blanker with a separate antenna. Sampling the noise from so far away in frequency, is a faulty assumption. It could work only in the case that the noise presents in the whole spectrum up to 40MHz and was consistent (timing duration of the noise spikes). In fact they mention this in the manual.

Noise canceller circuits (using a separate noise antenna) don't work well for spikes and fluorescent lamp noise, as the noise blankers works fine.
Noise canceller circuits are good for reducing industrial noise (hum few MHz wide) and also for reducing the ionospheric noise.
But, if you do a combination of both you will get good results.

two parallel schottky diodes wired tail to head, then put the pair in series, will do that. rf loss will be high until the signal amplitudes are big enough to "turn on" the diodes. this works really well at RF frequencies, but due to semiconductor transit time effects, works less well at microwave frequencies.

One can certainly think of using an algorithm to do this. If at audio types of
frequencies. And it has advantage that when shutting off analog stream
one holds the last value, to avoid step transients a simple comparator and
gate would exhibit. And of course control the turn on transient as well. One
could even envision regenerating the pulse with known parameters, if that
would enable any additional criteria. Or use correlation if that enables more
discrimination.

Of course the algorithm would introduce some latency, if thats an issue, to be
concerned with.

Regards, Dana.