HOW ENVELOPE DETECTION (SUPPOSEDLY?) WORKS
Most of us grew up with the above diagram of how a receiver detects (demodulates) an AM signal. Here is how they say it works:
-- Because of the way the sidebands and the carrier in the transmitted signal interact, we end up with a signal whose "envelope" matches the frequency of modulation. And we just need one side of the envelope.
-- We used a simple diode to rectify the incoming signal.
-- A simple filter gets rid of the RF.
-- We pass the resulting signal through a capacitor and we get audio, which we listen to.
REASONS FOR SCEPTICISM
But recently, a member of my local radio club has questioned this explanation of AM detection. He maintained that "envelope detection" is not real, and that was actually happening was "square law" mixing. I guess there are reasons for skepticism about the envelope detection explanation: The envelope detection explanation does seem very (perhaps overly) simple. This does sound a bit like the kind of "dumbed down" explanation that is sometimes used to explain complex topics (like mixing). Envelope detection does seem consistent with the incorrect insistence from early AMers that "sidebands don't exist." (Of course, they do exist.) All the other detectors we use are really just mixers. We mix a local oscillator the incoming signal to produce audio. Envelope detection (as described in the diagram above) seems oddly different.
Denial of envelope detection can even be found in the ARRL handbook: On page 15.9 of the 2002 edition we find this: "That a diode demodulates an AM signal by allowing its carrier to multiply with its sidebands may jar those long accustomed to seeing diode detection ascribed merely to 'rectification.' But a diode is certainly non-linear. It passes current only in one direction and its output is (within limits) proportional to the square of its input voltage. These non-linearities allow it to multiply."
ISN'T THIS REALLY JUST MIXING, WITH THE CARRIER AS THE LO?
It is, I think, tempting to say -- as the ARRL and my fellow club member do -- that what really happens is that the AM signal's carrier becomes the substitute for the VFO signal in other mixers. Using the non-linearity of the square law portion of the diode's characteristic curve, the sidebands mix with the carrier and -- voila! -- get audio. In this view there is no need for the rectification-based explanation provided above.
But I don't think this "diode as a mixer, not a rectifier" explanation works:
In all of the mixers we work with, the LO (or VFO or PTO) does one of two things:
-- In non-switching mixers it moves the amplifier up and down along the non-linear characteristic curve of the device. This means the operating point of the device is changing as the LO moves through its cycle. A much weaker RF signal then moves through the device, facing a shifting operating point whose shift is set by the LO. This produces the complex repeating periodic wave that contains the sum and difference frequencies.
-- In a switching mixer, the device that passes the RF is turned on and off. This is extreme non-linearity. But here is the key: The device is being turned on and off AT THE FREQUENCY OF THE LO. The LO is turning it on and off. The RF is being chopped up at the rate of the LO. This is what produces the complex repeating wave that contains the sum and difference frequencies.
Neither of these things happen in the diode we are discussing. If you try to look at the diode as a non-switching mixer, well, the operating point would be set not by the carrier serving as the LO but by the envelope consisting of the carrier and the sidebands. And if you try to look at is as a switching mixer you see that the switching is being controlled not by the LO but by the envelope formed by the carrier and the sidebands.
Also, this "diode as a mixer" explanation would require the diode to be non-linear. That is the key requirement for mixing. I suppose you could make a good case for the non-linearity of solid state diodes, but the old vacuum tube diodes were quite linear. The rectifying diode mixer model goes back to vacuum tube days. The "diode as rectifier" model worked then. With tubes operating on the linear portion of the curve, the diodes were not -- could not -- have been working as mixers. We have just substituted solid state diodes for the tubes. The increased non-linearity of the solid state diodes does introduce more distortion, but the "detection by rectification" explanation remains valid.
Even in the "square law" region (see diagram below) an AM signal would not really be mixed in the same way as signals are mixed in a product detector. Even in the square law region, the diode would be responding to the envelope. Indeed, the Amateur Radio Encyclopedia defines "Square Law Detector" as "a form of envelope detector." And even in the square law region, the incoming signal would be rectified. It would be moving above and below zero, and only one side of this waveform would be making it through the diode. Indeed the crystal radio experts discuss "rectification in the square law region" (http://www.crystal-radio.eu/endiodes.htm ) So even in the square law region, this diode is a rectifying envelope detector.
RF Cafe has some good graphs showing the linear and "square law" portions of the crystal diode's curve (see above): http://www.rfcafe.com/references/electrical/ew-radar-handbook/detectors.htm
The crystal radio guys have a good take on square law detection (note, they just see it as rectification, but on a lower, more parabolic portion of the curve): http://www.crystal-radio.eu/endiodes.htm
Here is a good booklet from 1955 on AM Detectors: https://worldradiohistory.com/BOOKSHELF-ARH/Rider-Books/A-M%20Detectors%20-%20Alexander%20Schure.pdf
Posts
Nicely done Bill. I don't recognize the simulator- do you?
ReplyDeleteDale W4OP
Thanks Dale: I don't recognize the simulator, but it definitely does the needed work! I stick with LTSpice because I have already climbed the learning curve on that one, and usually know which buttons to push. 73 Bill
ReplyDeleteDale: I just built the same thing in LTSpice. Same results, of course. 73 Bill
ReplyDeleteA thorough analysis of AM detectors, including square law detection and many well known receiver examples (Collins, Hammarlund, ...) may be found in http://www.tonnesoftware.com/appnotes/demodulator/EnvelopeDemodulators.pdf
ReplyDeleteI have Hewlett Packard's Genesys- and while it does many things extremely well- like filter synthesis and Monte Carlo analysis, this simulator is fascinating.
ReplyDeleteWow, thanks for the Tonne document (are you W4ENE himself?). I am going to study it in more detail tomorrow. 73 Bill
ReplyDeleteYes, the simulator is fascinating. The guy in the video has a very good one, but I found that I could do all of what he did using LTSpice. This simulator is a great way to test the validity of one's understanding of the circuitry. 73 Bill
ReplyDeleteEven today there are people who.believe the carrier goes up and down in amplitude with modulation.
ReplyDeleteThe concept of SSB dates to 1915, but even after SSB started becoming a ham thing, for a long time AM was in one Handbook chapter, and SSB in another, as if two different things. They need to be explained together.
Listen to SSB with an AM receiver. That's envelope detection. You get the amplitude variations, but things are missing. It needs a carrier.
Then listen to an AM signal with fading. It goes from readable to sounding like SSB, since the carrier is lost.
The carrier is needed, either from the transmitter or reinserted at the receiver.
This is a very intriguing discussion. One thing I'd be interested in seeing (if it exists) is some analysis outside the amateur community. That's not to say professional sources (engineers and physicists) are free from superstition and folk belief, but it's likely to be attenuated, and it will certainly be of a different character since it resonates in another echo chamber.
ReplyDeleteMost fascinating of all is that the detection effect was discovered and employed (using naturally-occurring semiconductors such as galena) without knowing how or why it worked. Another cargo cult? I don't think so. In the long history of human tool-making, it's never been necessary to know *why* something worked in order to know that it *did* work and worked predictably. The knowing of *why* has often helped only at the margins of development.
For instance, very useful steel alloys--so important that the industrial and technological world we know today would have been impossible--were created and used long before the crystalline chemistry of metals was understood. Such understanding has made for *improvements* in steels for highly-specialized purposes, but nothing revolutionary. The same is true of early vacuum-tube development. DeForest himself admitted he didn't know why an Audion worked, only that it did. Likewise, the later use of alkaline coatings on tube cathodes to increase emission was discovered only experimentally and not by deep understanding.
These intuitive and practical technologies do not amount to application of cultic lore. In the absence of
supernatural belief, they're only good-old seat-of-the-pants tool making which, even more so than the scientific method, is one of the few human claims to cosmic fame.
Todd: Yea, it is amazing that people were using crystal diodes before we knew how they worked. But at least we knew that we didn't know! That makes it much different than cargo-cultism. I found this very cool site about diode envelope detectors from Analog Devices. It comes complete with math and simulators and oscilloscope traces. And it is from ANALOG DEVICES, not from some ham (like me!). Check out the part about having to bias the diode when the incoming signal is small. Note that they bias it just to the point at which the diode is fully on. So even with the bias, they are using the diode as a rectifier. You can also use the biasing to test the assertion that what is really happening is just mixing, not rectification: increase the bias to the point where the diode is on all throughout the cycle of the input signal. Some argue that the non-linear curve of the diode will allow the carrier to mix with the sidebands to produce audio. But no. I tried it. It doesn't work. The diode detector needs rectification as part of the process of recovering the audio modulation. Here is the Analog Devices site:
ReplyDeletehttps://wiki.analog.com/university/courses/alm1k/circuits1/alm-cir-envelope-detector
73 Bill
Bill--Hey, that AD page is excellent! I see now you had already referenced it in your original post. Biasing the detector diode is one of those "Oh, sure" ideas one doesn't necessarily think of oneself but instantly recognize as obvious. This characterizes most of my mental activity; i.e., one step behind the rest of the world. I find it's not as crowded there. --73, K7TFC
ReplyDeleteOne other thing . . . I'd like to get to the bottom of this "non-linear" thing I've heard much about but I neither understand nor fully accept the idea. It seems to be used (at best) as a shibboleth or (at worst) an incantation. I accept, of course, that a switching mixing is "non-linear" (i.e., you don't just get a bigger version out of what you put in *and* it involves cosine terms), but its linearity seems to me entirely beside the point. It's issues such as this that make me wish I had a better command of mathematics.
ReplyDeleteTodd: I posted the AD link in response to your question. But hey, non-linearity is key. Without it, you end up with just two signals, one on top of the other. Addition, not mixing. Two types of mixers: switching and non-switching. In the non-switching mixer the LO rides up and down along the non-linear curve of the device. Along comes the modulated signal. Depending where the LO is at the moment, it goes though the amp at that operating point. A moment later the operating point changes. This produces the complex repeating wave that has within it the sum and diff frequencies. Switching mixers are a bit different, but equally valid. 73 Bill
ReplyDeleteThe "LO rid[ing] up and down along the non-linear curve of the device" is a good, almost visual explanation. At first I thought the LO level would have to be very precise, but I guess as long as some portion of the LO waveform runs in the "square-law" region of the diode curve, it doesn't much matter that most of it is in the linear region. It did occur to me, though, that the LO waveform would be in the square-law region *twice* per cycle--once on the rising edge and again on the falling edge.
ReplyDeleteI can't quite picture how that non-linearity in the toe of the IV curve would apply to a switching mixer, even those that use diodes. In the case of a Tayloe mixer, for instance, the commutation is pure switching by square wave (as Mini-Circuits recommends for its diode rings and as other authorities say is best as well--more efficient and lower IMD). In the broadest sense of the term, such mixers *are* non-linear--so much so they're actually *digital*. But their mixing occurs as a result of the switching and not because of the non-linearity of the IV curve of any of the devices they contain.
I'm probably at the edge of mere semantics here, but now that I have an idea of what some mean by "non-linear" mixing I think they should call it "square-law" mixing or something like that. Moreover, if my intuition is correct that switching mixers don't use non-linearity as the multiplication principle, then that significantly reduces the times when it can be claimed that "mixing is the result of non-linearity." Does a Gilbert cell rely on IV-curve non-linearity, or does the differential flow of currents in them mix by some other principle? Are they commutators of a different sort? Something to chew on, but don't feel obliged to reply. Tempus fugit.
Todd: It is important to understand this. I'm finding out that very few do. So it is time for you to join the Mixer Illuminati. The scales will fall from your eyes: First this:
ReplyDeletehttps://soldersmoke.blogspot.com/2022/10/adding-diode-ring-to-direct-conversion.html
Then this: https://soldersmoke.blogspot.com/2022/10/how-diode-ring-multiplies-by-1-and-1.html
And here is the Gilbert Cell: https://soldersmoke.blogspot.com/2021/11/how-to-understand-ne-602-and-gilbert.html
Once you tell us that you "get the particular go of it" we will schedule your induction ceremony.
73 Bill
Mixer Illuminati? Very few? Challenge accepted! You certainly know how to motivate a guy. Will I get a decoder ring? --73, Todd
ReplyDelete