And that is why it works so well with SSB. NOT. With suppressed carrier the envelope is there but there is nothing to mix with (no reference) so the result is shear distortion. Life is so much simpler with AI !! Don, ND6T
Hello Don: Yea, we are very fortunate that the envelope in an AM signal corresponds to the AF modulating signal. This lets us envelope detect: We just rectify it, run it through a Low Pass filter then a cap and -viola! -- AF. As you note, this does not work with SSB. Here the envelope does not correspond to the modulating signal. So we have to use product detectors. 73 Bill
I don't think you can use AI as support in a discussion like this. As I understand it what you get is just a reflection of the majority written views on the subject. Science should not work on consensus because if it did all progress would stop because we could convince ourselves we already know everything. Personally I am becoming suspicious of anything that AI plays a part in. AI is already well on the way to ruining Youtube and similar services. The amount of BS you can see on Youtube now is just amazing. Often its got great graphics or pictorial content but no one seems to do any quality checking because often its pure nonsence.
Well, I chose to use AI! I wrote, "I know that AI sometimes gets it wrong, but I think that on this one, it has it right." Just like the people who wrote the underlying web sites, sometimes AI gets it wrong. Would you be happier if I went to the underlying web sites and dug info up there?
To move forward on this issue I think we need experimental evidence rather than reference to written sources unless those sources refer to experimental evidence. We (society) seem to have reached the point that Issac Asimov forecast in his great series of Foundation books written so long ago where research relies on reference to written text (rather than experiments). AI can't provide evidence unless we carefully frame the question in such a way that we tease out new knowlege hidden in existing knowlege. Perhaps if we asked AI for proof that detection works from rectification rather than mixing it might provide some help. I would do this but due to my distrust of AI would not know how to start.
Bill I think I may have thought of a way to prove this. I have tried to add a sketch but the system keeps rejecting it so I will just describe the idea instead. Imagine you have a balanced modulator and your inputs are 1KHz and 3600KHz. In a well adjusted BM your outputs should be 3599 and 3601KHz plus 1 KHz and hopefully a very attenuated 3600KHz. Feed this output through a high pass filter to reject the 1KHz. Now feed this output to a diode detector similar to what you showed in your earlier sketches. Assuming that you are using levels that don't overload the detector what do you get out of the detector? If its 2 KHz then you have proven mixing detection. By the way I hope I didn't offend you with my earlier comments about AI. I view AI as being similar to the introduction of calculators. Everyone uses calculators these days but only those that predate the introduction of calculators seem to be able to do a quick mental calculation to confirm the calculator's results. I have seen some real crazy calculation results where the person doing the calculations just accepts the results. I fear we are heading that way with AI. I have only just realised that my posts have been Anonymous. I guess this is because I don't post often. I am Mark VK6WV.
Mark VK6WV wrote: "Imagine you have a balanced modulator and your inputs are 1KHz and 3600KHz. In a well adjusted BM your outputs should be 3599 and 3601KHz plus 1 KHz and hopefully a very attenuated 3600KHz. Feed this output through a high pass filter to reject the 1KHz. Now feed this output to a diode detector similar to what you showed in your earlier sketches. Assuming that you are using levels that don't overload the detector what do you get out of the detector? If its 2 KHz then you have proven mixing detection."
Is that really proof, Mark? All that is left after suppressing the main RF carrier to insignificance and the rest of the 1kHz audio was removed, is a 2-tone signal, 2 kHz apart, which has a very distinctive looking envelope. Guess what (fundamental) frequency the envelope has? A diode would just follow the envelope and after filtering the RF part out, you get the 2 kHz fundamental + lots of harmonics.
Envelope detection can of course also be applied to a SSB signal. Because the envelope of a SSB signal doesn't correspond to the baseband signal, you can't get the original audio back by simple envelope detection and have to resort to actual frequency mixing to do that.
Bill I am not in a position to build and test the circuit I described in an earlier post at present but I have run an LTSpice simulation. I know it’s just a simulation and to be really sure we would need to construct the circuit and run tests, but all the same, It clearly shows a 2KHz signal output and hence mixing detection. I just wish I knew how to attach a diagram to this comment. 73 Mark VK6WV
Hello Mark! No offense taken OM. I too am often skeptical about AI results. I have seen a lot of hallucinations! But I have also seen a lot of wrong stuff in published books and on web sites. At least the AI seems to try to look at a number of different sources. Sometimes it is useful, other times, not so much. Now, on to the simulations: I too am playing with LTSpice. I set up a very small AM signal (like the one you might see in a crystal radio) going into a silicon diode (no germanium in Spice). I keep it in the diode's "square law region." You can see in LTSpice how the there is square law detection by the diode: becasue of the curvature of the curve (!) you see that the positive peaks of the input signal are going through at higher voltage than the lower portions of the signal. This is square law detection. And sure enough, at the output you can see the AF tone. But look what happens when I move the signal up above the square law region -- up into the LINEAR region of the characteristic curve. I do this by adding a DC voltage. This keeps the diode on. The AF tone disappears. I think this supports the idea that without ordinary rectification and without square law action, you just won't get detection from a diode. Kind of cool to see this in LTSpice. Listening to VK6IA on 20 meters now (early in the morning here). 73 Bill N2CQR
Mark: I presented to Google Gemini the "prove it" challenge that you posed. Here is what I got (in seconds): Conversation with Gemini
Prove that a diode detects AM signals by way of rectification.
The diode in an AM detector circuit proves its ability to detect AM signals by performing half-wave rectification, which is the first step in a process called envelope detection.
An Amplitude Modulated (AM) signal is composed of a high-frequency carrier wave whose amplitude varies in proportion to the low-frequency modulating signal (the audio information). The original audio signal is represented by the "envelope" of the AM wave.
Here is the step-by-step process demonstrating detection by rectification:
Rectification by the Diode (Non-linear Device):
The AM signal is an alternating current (AC) signal with both positive and negative voltage cycles.
The diode, a non-linear device, acts as a half-wave rectifier. It allows significant current flow in only one direction (usually the positive half-cycle, but can be configured for negative).
This action chops off one-half of the AC waveform, resulting in a pulsating DC signal. Crucially, this rectified signal is no longer symmetrical around zero, and its average value follows the peaks of the original AM envelope.
Envelope Following (Peak Detection):
The output of the diode is passed through a simple low-pass filter (typically a parallel resistor-capacitor (RC) network).
The capacitor quickly charges up to the peak voltage of the pulsating DC signal (the peaks of the rectified AM wave) during the positive half-cycles.
When the input voltage drops, the diode cuts off. The capacitor then slowly discharges through the resistor.
The RC time constant of this filter is chosen to be long compared to the high-frequency carrier wave but short compared to the low-frequency modulating signal. This forces the capacitor voltage to follow the overall shape of the AM signal's envelope (the slower audio variation) while smoothing out the rapid, high-frequency carrier pulses.
Recovering the Modulating Signal:
The output voltage across the capacitor is a DC voltage with the original modulating audio signal riding on top of it.
A final DC-blocking capacitor (a high-pass filter) is often used to remove the remaining DC offset, leaving only the original alternating audio signal, which is then sent to an amplifier and speaker.
Thus, the diode's rectification transforms the AM AC waveform into a pulsating DC signal whose average value tracks the original audio information, allowing the low-pass filter to extract the modulating signal. This entire circuit is often referred to as a diode envelope detector and is the simplest form of AM demodulator. ============= Pretty good for a few seconds work, don't you think? 73 Bill N2CQR
Bill I agree that its a pretty good description of detection by rectification but I don't see it as proof. I have been researching the topic online today. It seems that references seem to fall into one camp rectification or the other mixing. No where did I come across a source that looked at both answers. I did find something interesting. Apparently when the mathematics of modulation were first described a lot of people did not accept the concept of sidebands and made drawings on graph paper of modulated waves to "prove" that only one frequency (the carrier) existed. I read about an experiment that used one of the first crystal filters that was narrow enough to only pass the carrier. Signals passing through this filter removed the sidebands restoring a plain carrier proving the existence of sidebands. Not sure where we go from here. I ran my LTSpice model through a range of input amplitudes, every case showed mixing occuring. I used a Schottky diode in place of a germanium device. I know how to add a model of say a 1N34 but wanted to keep the simulation so it only used included parts so anyone could reproduce it. 73 Mark VK6WV.
Mark: I think it is best to go back to James Clerk Maxwell's question: "What's the go of it?" In other words, how does it work? I think the envelope detector is explained very well by rectification. I haven't heard any other real explanations. Some just pull out the trig and say, look, there is your explanation. For your simulations, start by keeping the signal level very low, low enough to stay completely within the square law region. Then move the whole thing up into the linear range, looking at the output signal in FFT. I see the mixing product disappearing. I too was mindful of the AMers who denied sidebands. I mention this on one of my posts. But I don't think this is the same thing. This may be a disagreement between those who seek an intuitive understanding, and those who think that the math is all you need. 73 Bill N2CQR
Actually, both views might be true... If you consider the time domain, it is obviously that the diode detector does do rectification, After low pass filtering you then have your recovered audio signal. You can also look at it in the frequency domain. If you take the AM signal and multiply it with a square wave signal with the same frequency and phase and with the amplitude varying between "1" and "0". You have the same signal as with the rectified AM signal. ! In other words it is much like a direct conversion receiver for SSB, only the carrier is being delivered together with the modulation, so you don't need to have a local oscillator.
Anecdotical evidence 1) : Peter Parker VK3YE designed a 80m DC receiver with a single diode. It is very much like a straight AM receiver. Only because there is no carrier he adds it using a Local Oscillator.
Anecdotical evidence 2) : look at what AI claims is typical for the working of a product detector: - It uses a nonlinear element, that sounds like a diode - it creates new frequencies including the sum and difference. That is the same for both views as well - The desired audio is obtained by a low-pass filter, Same for both, again!
- Only the last claim (it can be used for SSB) is of course not true for there is no carrier to do the switching.
Cor: I disagree. Can you point to a circuit that simply takes an AM signal, runs it through a non-linear device (without rectifying!) and manages to mix the carrier from the signal (no coherent or synchronized LO please) and the sidebands to produce audio? I have never seen one! You will see BJT and FET circuits that purport to do this, but if you look carefully, you will see that they are biased close to cutoff, and are, essentially rectifying. You will also hear about what happens in the square law region. But even square law detection is just a form of rectification. I think the argument that "both sides of this argument are right -- one in the time domain, the other in the frequency domain" is bogus. The frequency domain argument is just based on the math, the trig. There is no physical description of "what's the particular go of it." We have a good, intuitive, physical description, for example, of what happens in a diode ring switching detector. We should have something similar for the diode envelope detector. And we do: rectification. Envelope detection. Terman was right.
And that is why it works so well with SSB. NOT. With suppressed carrier the envelope is there but there is nothing to mix with (no reference) so the result is shear distortion. Life is so much simpler with AI !!
ReplyDeleteDon, ND6T
Hello Don: Yea, we are very fortunate that the envelope in an AM signal corresponds to the AF modulating signal. This lets us envelope detect: We just rectify it, run it through a Low Pass filter then a cap and -viola! -- AF. As you note, this does not work with SSB. Here the envelope does not correspond to the modulating signal. So we have to use product detectors. 73 Bill
DeleteI don't think you can use AI as support in a discussion like this. As I understand it what you get is just a reflection of the majority written views on the subject.
ReplyDeleteScience should not work on consensus because if it did all progress would stop because we could convince ourselves we already know everything.
Personally I am becoming suspicious of anything that AI plays a part in. AI is already well on the way to ruining Youtube and similar services. The amount of BS you can see on Youtube now is just amazing. Often its got great graphics or pictorial content but no one seems to do any quality checking because often its pure nonsence.
Well, I chose to use AI! I wrote, "I know that AI sometimes gets it wrong, but I think that on this one, it has it right." Just like the people who wrote the underlying web sites, sometimes AI gets it wrong. Would you be happier if I went to the underlying web sites and dug info up there?
DeleteTo move forward on this issue I think we need experimental evidence rather than reference to written sources unless those sources refer to experimental evidence.
DeleteWe (society) seem to have reached the point that Issac Asimov forecast in his great series of Foundation books written so long ago where research relies on reference to written text (rather than experiments). AI can't provide evidence unless we carefully frame the question in such a way that we tease out new knowlege hidden in existing knowlege. Perhaps if we asked AI for proof that detection works from rectification rather than mixing it might provide some help. I would do this but due to my distrust of AI would not know how to start.
Bill I think I may have thought of a way to prove this. I have tried to add a sketch but the system keeps rejecting it so I will just describe the idea instead.
DeleteImagine you have a balanced modulator and your inputs are 1KHz and 3600KHz. In a well adjusted BM your outputs should be 3599 and 3601KHz plus 1 KHz and hopefully a very attenuated 3600KHz. Feed this output through a high pass filter to reject the 1KHz.
Now feed this output to a diode detector similar to what you showed in your earlier sketches. Assuming that you are using levels that don't overload the detector what do you get out of the detector?
If its 2 KHz then you have proven mixing detection.
By the way I hope I didn't offend you with my earlier comments about AI. I view AI as being similar to the introduction of calculators. Everyone uses calculators these days but only those that predate the introduction of calculators seem to be able to do a quick mental calculation to confirm the calculator's results. I have seen some real crazy calculation results where the person doing the calculations just accepts the results. I fear we are heading that way with AI.
I have only just realised that my posts have been Anonymous. I guess this is because I don't post often. I am Mark VK6WV.
Mark VK6WV wrote: "Imagine you have a balanced modulator and your inputs are 1KHz and 3600KHz. In a well adjusted BM your outputs should be 3599 and 3601KHz plus 1 KHz and hopefully a very attenuated 3600KHz. Feed this output through a high pass filter to reject the 1KHz.
DeleteNow feed this output to a diode detector similar to what you showed in your earlier sketches. Assuming that you are using levels that don't overload the detector what do you get out of the detector?
If its 2 KHz then you have proven mixing detection."
Is that really proof, Mark? All that is left after suppressing the main RF carrier to insignificance and the rest of the 1kHz audio was removed, is a 2-tone signal, 2 kHz apart, which has a very distinctive looking envelope. Guess what (fundamental) frequency the envelope has? A diode would just follow the envelope and after filtering the RF part out, you get the 2 kHz fundamental + lots of harmonics.
Envelope detection can of course also be applied to a SSB signal. Because the envelope of a SSB signal doesn't correspond to the baseband signal, you can't get the original audio back by simple envelope detection and have to resort to actual frequency mixing to do that.
Bill
ReplyDeleteI am not in a position to build and test the circuit I described in an earlier post at present but I have run an LTSpice simulation. I know it’s just a simulation and to be really sure we would need to construct the circuit and run tests, but all the same, It clearly shows a 2KHz signal output and hence mixing detection.
I just wish I knew how to attach a diagram to this comment.
73 Mark VK6WV
Hello Mark! No offense taken OM. I too am often skeptical about AI results. I have seen a lot of hallucinations! But I have also seen a lot of wrong stuff in published books and on web sites. At least the AI seems to try to look at a number of different sources. Sometimes it is useful, other times, not so much. Now, on to the simulations:
ReplyDeleteI too am playing with LTSpice. I set up a very small AM signal (like the one you might see in a crystal radio) going into a silicon diode (no germanium in Spice). I keep it in the diode's "square law region." You can see in LTSpice how the there is square law detection by the diode: becasue of the curvature of the curve (!) you see that the positive peaks of the input signal are going through at higher voltage than the lower portions of the signal. This is square law detection. And sure enough, at the output you can see the AF tone. But look what happens when I move the signal up above the square law region -- up into the LINEAR region of the characteristic curve. I do this by adding a DC voltage. This keeps the diode on. The AF tone disappears. I think this supports the idea that without ordinary rectification and without square law action, you just won't get detection from a diode. Kind of cool to see this in LTSpice.
Listening to VK6IA on 20 meters now (early in the morning here). 73 Bill N2CQR
Mark: I presented to Google Gemini the "prove it" challenge that you posed. Here is what I got (in seconds):
ReplyDeleteConversation with Gemini
Prove that a diode detects AM signals by way of rectification.
The diode in an AM detector circuit proves its ability to detect AM signals by performing half-wave rectification, which is the first step in a process called envelope detection.
An Amplitude Modulated (AM) signal is composed of a high-frequency carrier wave whose amplitude varies in proportion to the low-frequency modulating signal (the audio information). The original audio signal is represented by the "envelope" of the AM wave.
Here is the step-by-step process demonstrating detection by rectification:
Rectification by the Diode (Non-linear Device):
The AM signal is an alternating current (AC) signal with both positive and negative voltage cycles.
The diode, a non-linear device, acts as a half-wave rectifier. It allows significant current flow in only one direction (usually the positive half-cycle, but can be configured for negative).
This action chops off one-half of the AC waveform, resulting in a pulsating DC signal. Crucially, this rectified signal is no longer symmetrical around zero, and its average value follows the peaks of the original AM envelope.
Envelope Following (Peak Detection):
The output of the diode is passed through a simple low-pass filter (typically a parallel resistor-capacitor (RC) network).
The capacitor quickly charges up to the peak voltage of the pulsating DC signal (the peaks of the rectified AM wave) during the positive half-cycles.
When the input voltage drops, the diode cuts off. The capacitor then slowly discharges through the resistor.
The RC time constant of this filter is chosen to be long compared to the high-frequency carrier wave but short compared to the low-frequency modulating signal. This forces the capacitor voltage to follow the overall shape of the AM signal's envelope (the slower audio variation) while smoothing out the rapid, high-frequency carrier pulses.
Recovering the Modulating Signal:
The output voltage across the capacitor is a DC voltage with the original modulating audio signal riding on top of it.
A final DC-blocking capacitor (a high-pass filter) is often used to remove the remaining DC offset, leaving only the original alternating audio signal, which is then sent to an amplifier and speaker.
Thus, the diode's rectification transforms the AM AC waveform into a pulsating DC signal whose average value tracks the original audio information, allowing the low-pass filter to extract the modulating signal. This entire circuit is often referred to as a diode envelope detector and is the simplest form of AM demodulator.
=============
Pretty good for a few seconds work, don't you think? 73 Bill N2CQR
Bill I agree that its a pretty good description of detection by rectification but I don't see it as proof.
ReplyDeleteI have been researching the topic online today. It seems that references seem to fall into one camp rectification or the other mixing. No where did I come across a source that looked at both answers.
I did find something interesting. Apparently when the mathematics of modulation were first described a lot of people did not accept the concept of sidebands and made drawings on graph paper of modulated waves to "prove" that only one frequency (the carrier) existed. I read about an experiment that used one of the first crystal filters that was narrow enough to only pass the carrier. Signals passing through this filter removed the sidebands restoring a plain carrier proving the existence of sidebands.
Not sure where we go from here. I ran my LTSpice model through a range of input amplitudes, every case showed mixing occuring. I used a Schottky diode in place of a germanium device. I know how to add a model of say a 1N34 but wanted to keep the simulation so it only used included parts so anyone could reproduce it.
73 Mark VK6WV.
Mark: I think it is best to go back to James Clerk Maxwell's question: "What's the go of it?" In other words, how does it work? I think the envelope detector is explained very well by rectification. I haven't heard any other real explanations. Some just pull out the trig and say, look, there is your explanation.
ReplyDeleteFor your simulations, start by keeping the signal level very low, low enough to stay completely within the square law region. Then move the whole thing up into the linear range, looking at the output signal in FFT. I see the mixing product disappearing.
I too was mindful of the AMers who denied sidebands. I mention this on one of my posts. But I don't think this is the same thing. This may be a disagreement between those who seek an intuitive understanding, and those who think that the math is all you need.
73 Bill N2CQR
Actually, both views might be true...
ReplyDeleteIf you consider the time domain, it is obviously that the diode detector does do rectification, After low pass filtering you then have your recovered audio signal.
You can also look at it in the frequency domain. If you take the AM signal and multiply it with a square wave signal with the same frequency and phase and with the amplitude varying between "1" and "0". You have the same signal as with the rectified AM signal. !
In other words it is much like a direct conversion receiver for SSB, only the carrier is being delivered together with the modulation, so you don't need to have a local oscillator.
Anecdotical evidence 1) : Peter Parker VK3YE designed a 80m DC receiver with a single diode. It is very much like a straight AM receiver. Only because there is no carrier he adds it using a Local Oscillator.
Anecdotical evidence 2) : look at what AI claims is typical for the working of a product detector:
- It uses a nonlinear element, that sounds like a diode
- it creates new frequencies including the sum and difference. That is the same for both views as well
- The desired audio is obtained by a low-pass filter, Same for both, again!
- Only the last claim (it can be used for SSB) is of course not true for there is no carrier to do the switching.
73 cor PA3COR
www.pa3cor.nl
The link for Peter Parker VK3YE 80m DC receiver : https://vk3ye.com/projects/proj80dir.htm
ReplyDeleteCor: I disagree. Can you point to a circuit that simply takes an AM signal, runs it through a non-linear device (without rectifying!) and manages to mix the carrier from the signal (no coherent or synchronized LO please) and the sidebands to produce audio? I have never seen one! You will see BJT and FET circuits that purport to do this, but if you look carefully, you will see that they are biased close to cutoff, and are, essentially rectifying. You will also hear about what happens in the square law region. But even square law detection is just a form of rectification. I think the argument that "both sides of this argument are right -- one in the time domain, the other in the frequency domain" is bogus. The frequency domain argument is just based on the math, the trig. There is no physical description of "what's the particular go of it." We have a good, intuitive, physical description, for example, of what happens in a diode ring switching detector. We should have something similar for the diode envelope detector. And we do: rectification. Envelope detection. Terman was right.
ReplyDelete