Talk:Single-sideband modulation

The article could do with more information on vestigal sideband. 213.40.114.230 (talk) 16:25, 16 October 2009 (UTC)

25 years ago I added the following sentence to the Vestigial Sideband paragraph: "The Milgo 4400/48 modem (around 1967) used vestigial sideband and phase-shift keying to provide 4800 bit/s transmission over a 1600 Hz channel." This sentence was written based on personal memory of the the Milgo 4400 rather than any reliable source. I have deleted the sentence as this reference shows that my sentence is wrong.^[1] I don't know how many years it will take the ten sites which have copied this sentence to make a correction.Rdmoore6 (talk) 13:48, 17 September 2018 (UTC)

The use of the Mixing Equations will help explain the mathematics using just cos and sin. The math in its current state is needlessly complicated for an encyclopedic article. --Crcwiki (talk) 18:56, 10 July 2015 (UTC)

The maths is very poorly explained. For example, it is not clear what "j" refers to. All symbols must be explained clearly. —Preceding unsigned comment added by 202.124.67.78 (talk) 16:08, August 24, 2007 (UTC)

I've suppressed some of the maths and the rest of it also looks dubious. This explanation needs a diagram. A Hilbert transform is not the obvious way to explain SSB to our target audience. --Wtshymanski 19:41, 16 November 2005 (UTC)

- Multiple levels of explanation are possible and need not be mutually exclusive. Layering is part of the power of a hyperlinked encyclopedia. It can appeal to a broader audience. Rather than delete the mathematics, you could find or create a more appropriate place for it. Bury it a little deeper. Those who don't want to see it will stop digging when they get beyond their comfort level. --Bob K 03:14, 19 November 2005 (UTC)

- Ah ha! I have discovered subpages. But there seems to be a controversy surrounding their usage. I don't understand what all the fuss is about, but I guess I will find out. So I jumped right in and created a subpage for the suppressed mathematics and linked to it from the article. I named the subpage "Proofs", according to the suggestion at Wikipedia_talk:Subpages. Now I'm going to go catch up on my real job. Keep up the good work, have fun, and good luck. --Bob K 18:10, 19 November 2005 (UTC)

I've also eliminated the word "subhetrodyne" which I've never seen before (neither has Google). --Wtshymanski 19:41, 16 November 2005 (UTC)

subhetrodyne [sic] is not the right spelling, which presumably accounts for the 0 hits. But I agree that it is unusual, and I understand your concern, particularly since super historically derives from supersonic. Or maybe in this century it is time to start putting that archaic modifier to more practical use. Just a thought.

--Bob K 23:23, 17 November 2005 (UTC)

Thanks for the spelling correction, but even typing in "subheterodyne" gives only 6 Google hits, one of which is our very own Wikipedia. The usage is non-standard. It's not Wikipedia's place to create new terms or redefine language. --Wtshymanski 03:16, 18 November 2005 (UTC)

You're welcome, and thanks for all that you do. Also, I believe one of those 6 hits has yet another interpretation. It uses superheterodyne and subheterodyne to respectively describe the sum and difference frequencies produced by the mixer, regardless of whether the injection is high or low. So people look for true meaning in the prefix. It doesn't naturally occur to them that it is just a useless vestige of the naive, early days. Encyclopedias and dictionaries do not redefine or even define language. They just try to keep up with it. --Bob K 16:40, 18 November 2005 (UTC)

"High side" and "low side" works for me, and I believe is commonly accepted ( I'll have to check my ARRL handbook!). And yes, we should keep up with the current usage - but not try to lead it! --Wtshymanski 17:46, 18 November 2005 (UTC)

- My favorite dictionaries are the ones that give all the usages they know about and then explain which one is the "preferred" usage (and hopefully why). In this case especially, I think that would help deter proliferation of the incorrect ones, because people would know that they are incorrect, whereas now they all appear more plausible and more useful than Armstrong's original concept. --Bob K 20:44, 18 November 2005 (UTC)

A good example can be found at http://www.patentstorm.us/patents/5461426.html
In particular the 11^th paragraph of the DETAILED DESCRIPTION section.  The desired output signal is at 10.74 MHz (see 9^th paragraph). The 11^th paragraph describes two techniques:

"... using a lowpass FIR digital filter and a 10.84 MHz oscillator to carry out a superheterodyne ... rather than the bandpass FIR digital filter and the 10.64 MHz oscillator to carry out the subheterodyne...".

Another example is http://www.engr.usask.ca/classes/EE/352/2005/Ch/Ch3-AM-04m4.pdf (excerpt below)

3.22  A 4.02 GHz satellite television signal enters an Earth station receiver with IF frequency 70 MHz. What is the LO frequency and the image frequency for the cases of:

(a) High-side injection (superheterodyne)?

(b) Low-side injection (“subheterodyne”)?

  --Bob K 23:23, 17 November 2005 (UTC)

We don't use subpages for this kind of thing, period. The maths needs its language cleaning up, but it should remain firmly in the article. When I have a little mroe time, I will copy it back, and add a cleanup tag. -Splash^talk 03:52, 20 November 2005 (UTC) Bold text

Why is there a separate page for Single-sideband suppressed-carrier transmission when "single-sideband" almost always refers to that, and is the main modulation technique discussed here? 72.75.98.88 (talk) 06:48, 4 June 2009 (UTC)

It now seems to have been changed to a redirect, which is fine. David Spector (talk) 21:19, 24 March 2015 (UTC)

I think the author of this new section has good intent, but...

It's a false claim that "This approach avoids the use of the complex representation of baseband signals ...", because the proof offered at http://i.imgur.com/xj0moI9.png relies on arguing that:

${\displaystyle {\mathcal {F}}^{-1}\{M(f)u(f)\}={\frac {1}{2}}[m(t)+j\cdot {\hat {m}}(t)],}$

which is a complex baseband representation. By moving the proof to the RF domain, it simply becomes much more difficult and is not customary. This section is a front for the user's blog, which makes another false implication that it contains fewer "links" than Wikipedia's treatment here. And finally, the links it provides include several seconds of promotional material for something called adf.ly.

--Bob K (talk) 18:51, 24 January 2013 (UTC)

In the section "History", added "and channel status signalling at or in the vicinity of 3,825 Hz." Akld guy (talk) 01:05, 24 August 2014 (UTC)

I took it out. Bring us a source that explains what it means and maybe we can put it back. Dicklyon (talk) 02:29, 24 August 2014 (UTC)

Means this. When either of the parties hangs up a call or a data transfer is terminated, there needs to be a way to signal to the opposite end that the call is terminated. The presence or absence of the 3.825 KHz signal does that without the need to introduce inband (300Hz-3.4KHz) signals that would be apparent (and LOUD) to the non-hungup party. I thought that those who had already contributed to the page would see instantly what I meant. Perhaps you don't have any actual experience with FDM (aka broadband) systems. I'll try to find a source, but frankly I'm not going to waste a lot of time. If there aren't people here who don't know that OBS (Out of Band Signalling) was used, I'm not going to fight them. My experience consists of quite a few years in the telecoms field, but my terminology may not be appropriate to North American experts. It's what we used in my country, which of course met international standards. I used "at or in the vicinity of 3,825" because while pretty sure that was the frequency, it's been several decades and my recollection may be wrong. Akld guy (talk) 03:41, 24 August 2014 (UTC)

"in order to support the fidelity required and desired for relative high fidelity, full range clean and articulate vocal audio."

I'm pretty sure the person doing the talking is the one responsible for the audio being articulate or not, not the hardware. Perhaps a different word would better articulate the thought that was intended to be conveyed.

2601:5CE:4101:8553:D0C1:F6:D0C0:3F4C (talk) 03:58, 11 July 2017 (UTC)

There is something similar, developed by Bell Labs, called the articulation index. Probably what the editor meant. WhaleFarm (talk) 17:10, 25 January 2026 (UTC)

In response to the change by Bob_K ( https://en.wikipedia.org/wiki/User:Bob_K ) at https://en.wikipedia.org/w/index.php?title=Single-sideband_modulation&diff=next&oldid=854177708 , which undoes part of my recent changes, let me explain why I made the changes I made.

Consider an audio signal consisting of a pure tone at 300 Hz with a carrier radio frequency of 10 MHz. The audio signal has a bandwidth of 0 Hz. The resulting AM signal will have components at 9.9997 MHz, 10 MHz, and 10.0003 Mhz, with a bandwidth of 600 Hz, which is not equal to 2 times the bandwidth of the audio signal, but is equal to 2 times the maximum frequency of the audio signal.

Is there something wrong with that example?

If so, we should undo all my changes and clarify matters to avoid my example.

By the way, I don't understand the remark accompanying Bob_K's edit, which was "compare apples (bandwidth) to apples". All the quantities above are measured in Hz, so there is no comparison involving mismatched units.

Daniel R. Grayson (talk) 18:20, 10 August 2018 (UTC)

Your statement "is equal to 2 times the maximum frequency of the audio signal." is correct. I believe we should consistently say it that way. The case of a narrow audio signal (tone, for example) works with this phrasing, but not when we talk about the bandwidth of the audio. Most audio has a BW nearly as wide as the maximum frequency, but certain types of data and MCW do not. I am not totally in favor of saying "audio" instead of the "baseband" or "modulating" frequency since not all modulation is audio, but since most is, I am not against using that term which everyone will understand. JNRSTANLEY (talk) 22:06, 10 August 2018 (UTC)

Thank you for the positive response. I'm not totally in favor of using the term "baseband", because I see problems with the definition at https://en.wikipedia.org/wiki/Baseband -- what does "close to 0" mean? Is an audio signal with frequencies all above 75Hz a baseband signal? I have no way to tell. Daniel R. Grayson (talk) 01:34, 11 August 2018 (UTC)

I see what you mean about the definition of baseband at https://en.wikipedia.org/wiki/Baseband, although near the end of the article under "Modulation" it is correctly explained using audio as an example, and thus seems to call it baseband even without the low audio frequencies. Well how about we stay with "audio" and go back to your last entries on the bandwidth. that is, "equal to 2 times the maximum frequency of the audio signal". Unless Bob K can defend his last edit, I say go back to yours. JNRSTANLEY (talk) 09:56, 11 August 2018 (UTC)

Okay, thanks, let's wait a day or two for Bob_K to chime in. Daniel R. Grayson (talk) 12:00, 11 August 2018 (UTC)

By the way, at https://en.wikipedia.org/wiki/Baseband#Modulation , it is not clearly stated that the lower limit of the audio signal is artificially declared to be 0Hz for the purpose of defining its bandwidth there. Maybe a clarification should be added. Daniel R. Grayson (talk)

Explanation of my comment "compare apples (bandwidth) to apples": My interpretation of Dan's edit was that the difference between "bandwidth" (apples) and "maximum frequency" (oranges) was due to the fact that bandwidth has several possible definitions (even for full-spectrum audio), as does "maximum frequency" for a finite-duration signal (6 dB rolloff?, 12 dB rolloff?, 90 dB rolloff?). I was not considering the case of a baseband tone. I guess if that's what we're going to talk about, it should be more obvious/explicit. Sorry for the misunderstanding.

--Bob K (talk) 15:17, 11 August 2018 (UTC)

Bob_K, thanks!

By the way, in my example above, I was thinking of the audio signal as eternal.

Daniel R. Grayson (talk) 18:49, 12 August 2018 (UTC)

Reading the discussion for the first time I have the following comments:

This Occupied Bandwith (OcBw) is only true if we had an ideal transmitter without any distortion.

In a real transmitter the distortions and frequency response/rolloff filtering for the modulating signal, the IF and RF filtering define the radiated bandwidth. It depends what minimum signal level you specify to measure the bandwidth -20 dB, -40 dB, -60 dB or an other value.

- for Double Side Band with Carrier CDSB, Double Side Band with a reduced Carrier DSB(RC) and Double Side Band with suppressed Carrier DSB(SC) the 2x the max. modulating frequency defines the OcBw.

-Vestigal Sideband which is from my understanding no a AM single sideband, since the complemetary second sideband is also radiated the OcBw depends on the filter rolloff for the second sideband and the attenuation that has to be achieved.

- for SSB Suppressed Carrier or SSB(SC), by the highest minus the lowest modulating Signal define the Occupied Bandwith (OcBw), e.g. for min/max 300 Hz/3000 Hz, for a SSB signal with suppressed carrier is 2700 Hz.

- for Carrier Single Sideband or CSSB, by the carrier and the highest modulation frequency defines the OcBw, e.g. for a 3000Hz tone, the OcBw to 3000 Hz

For on-/off-keying of a tone it is the max. modulating frequency and the increase in bandwidth which increases with keying speed. Snn47 (talk) 17:12, 3 November 2025 (UTC)

I think the discussion of the bandwidth of a 300 HZ tone being zero adds little to the description of SSB. Technically, it can only be 0 if it never started or stopped, which gets into a deeper rabit hole. Seems unnecessary for the task at hand. I went for a shorter version of the basics, with some citations. WhaleFarm (talk) 16:57, 25 January 2026 (UTC)

If you look at it this way, the same would apply to the bandwidth of a single carrier (ITU CW continious wave), which by your view cannot have a Bn = 0Hz, because the carrier has started and ended at some time. However that is not the way ITU defined the bandwidth.

However if you consider the switch-on and switch-off, independently of the duration that a carrier or tone is transmitted, you have to apply the bandwidth of the RF-spectrum of Carrier On-/Off-Keying and for SSB tone On-/Off-Keyingwhere. The bandwidth increases mainly with the keying speed, the rise-, decay- and duration of the pulse(s-shape), which is viewing a CW signal as pulse modulation. If you only switch the signal once on not off during the duration of 1 s the spectrum and the bandwidth will not change. Only for the time that the carrier is switched on and off will the spectrum and the bandwidth.

For a continious carrier and for a SSB tranmitter with a single tone modulating, in absence of a carrier and AF and RF distortions, the bandwidth of single tone is identical to an unmodulated carrier or 0 if it exists >1s the time normally applied. The bandwidth is

bandwidth = highest modulating AF-frequency - lowest modulating AF-frequency

With only one tone there is only one AF-frequency therefore no bandwidth, if the tone is not keyed and the AF-tone and the RF-transmitter does not introduce any distortions.

For any SSB signal with two or more frequencies, the frequency difference between the highest and the lowest AF-frequency will define the bandwidth bandwidth = highest mines lowest frequency), with only one tone therefore also 0 Hz Snn47 (talk) 07:58, 26 January 2026 (UTC)

How would you like to work this into the SSB article? If we take speech band as 300-3000 Hz (telephone grade), we would potentially save by not using the 0-300 section, or 10%. If we said natural speech was 50-4000 Hz, we could save 12.5%. But in the real world, brick wall filters don't exist because of the impulse duration in time. Even if we key CW on once during a 1 second period, there is an increase in BW above 0.

If you think something differently should be presented about the bandwidth issue, I'd love to discuss it. WhaleFarm (talk) 16:39, 26 January 2026 (UTC)

You are correct analogue filter don’t cut of higher audio frequencies right away, but are steep enough to meet the required bandwidth definition.

The max. audio frequency of Aeronautical Radio Navigation Systems (ARNS), like (D)VOR or ILS, VHF-COM AM-CDSB ground transmitter for ATC En-Route control are limited to 3000 Hz. With some VHF-COM transmitter that were limited to 2400 Hz for technical reasons. No pilot complained in the many decades they were operated, since they were high quality and produced only minor distortions which often is mistaken for a transmitter having a larger bandwidth.

The same applies for the min. audio frequency in ARNS, like (D)VOR or ILS, VHF-COM AM-CDSB voice communication. The min of 300 Hz is the min. value used . ARNS Systems, like (D)VOR or ILS, transmitt below 300 Hz additional information e.g. 30 Hz, 90 Hz or 150 Hz.

The bandwidth varies also with how RF bandwidth is defined, e.g. teh -3 dB-, -20 dB-, -40 dB-, or -60 dB-points. For which purpose, is the bandwidth defined and applied e.g. to achieve Electro Magnetic Compatibility (EMC) between transmitter/systems using the same analogue-/digital-modulation, or transmitter/systems using different analogue/digital modulations.

Often you have to take not just the necessary/occupied bandwidth into account but e.g. also spurious/out-of-band (e.g. oscillator leakage) or the noise floor may be interfering (e.g. when several AM-CDSB voice-channels are generated simultaneously in a digital transmitter the noise floor may be just 40 db to 60 dB below the peak signal).

The bandwidth will increase with the keying speed, but the bandwidth depends to a large degree how the keying is done. The bandwidth for on/off keying of a carrier depends like a pulsed system on the duration of the rise and decay time, the form of the the leading and trailing edge and the duriation of the on time, which is why hard keying generated often bad key klicks that were heard many kHz away. For SSB using a tone to generate a carrier there is a much lower impact due to the slow rise of the audio tones.

How the bandwidth increases depends also on the definition how the bandwith is generated.

The point I was trying to make was that contionious tone into a SSB transmitter like a continious carrier has once the transmitter is stabil has 0 Hz bandwidth, since the tone is not referenced to 0 Hz. Snn47 (talk) 10:51, 27 January 2026 (UTC)

Thanks, I agree with your points. Bandwidth could be managed to pack more signals in. I don't know of an analog system that dynamically does that, managing fixed allocations is enough. Is there somthing you think we should add to this section? What you're saying sort of presages the digital world.

All the modulation techniques for SSB are really about filter skirts and bandwidths. If you have a good reference, it would be great to work it in. WhaleFarm (talk) 13:35, 27 January 2026 (UTC)

For the spectrum attenuation in relation to the carrier level in dBc only, I can think only of the CDSB-AM VHF-COM MOPS from EUROCAE, ED-23C 3.2.7.2 Transmitter Occupied Spectrum for 8.33 kHz Mode, FIGURE 3-1: SPECTRAL MASK, which specifies Attenuation from the carrier level (dBc) 2.5 kHz 0dB 3,2 kHz -40 dB, 5 kHz -60 dB. This mask applies to both 25 kHz and 8.33 kHz channel spacing.

I remember an ARRL article from the 1980's, where the speech spectrum was reviewed with the result that found around 1800 Hz few hundred Hz without much content and proposed to shift a few hundred Hz of the higher frequencies there to reduce the bandwidth, while maintaining understandibility. While I still may have this ARRL in the attic, I could not find it when I searched the internet. Snn47 (talk) 16:16, 30 January 2026 (UTC)

I found an old qst that points here: http://www.bitsavers.org/magazines/EDN/EDN_V18_N10_19730520.pdf

could that be it. April 1979 QST - just a quick note references EDN. I tried every other QST search, not much to find WhaleFarm (talk) 17:09, 30 January 2026 (UTC)

That is not what I remember from the QST article I read. I tried searching with narrowband voice communication and think I found some papers on narrow bandwidth voice communication. Since I could not find a link to the QST December 1977, (Dr R W Harris and J C Gorski) I cannot confirm that this was the article I remembered.

Others did similar work, e.g. June 1979 https://apps.dtic.mil/sti/tr/pdf/ADA074445.pdf Raymond Roy/Hitchcock and later others VK5BR, who has some additional referenceshttps://www.qsl.net/vk5br/TransmissionSystems/NarBandVoice.htm

vy 73 Snn47 (talk) 20:40, 1 February 2026 (UTC)

I looked through my older versions of the ARRL-HB and in the 1981 version I found the voice spectrum measurement diagram from 80 to 8000 Hz that I remembered, NARROW·BAND VOICE MODULATION 14-33. The HB referenced the 1977-12 QST, but this issue does not contain this diagramm. Snn47 (talk) 21:08, 1 February 2026 (UTC)

Thanks, Snn47 — I appreciate you digging up the references and adding the practical and regulatory context. I agree that real-world bandwidth depends on filtering, keying, and how it’s defined and measured, and that idealized cases only go so far.

I’ve added a short, sourced NBVM section to the article as a historical example of analog bandwidth-reduction techniques within SSB. Unless there’s a specific, well-sourced change you’d like to suggest beyond that, I think this discussion is in a good place. Thanks again for the constructive input. WhaleFarm (talk) 22:38, 1 February 2026 (UTC)

An editor claims that "The carrier recovery doesn't solve the frequency shift. It gives better S/N ratio on the detector output." To my experience, I don't know why this would be the case. It's also not clear what a "BFO oscillator shift" is, or where this "agreement" comes from; 1.7kHz from where to where? In any event, it requires a citation to support the claim. algocu (talk) 20:29, 24 June 2020 (UTC)

The use of "inverting stage" is confusing. I think most engineers would thing voltage inverting. Spectrum inverting happens if the LO is above the signal when being mixed. If that's what was meant, it whould be described, or removed. WhaleFarm (talk) 17:07, 25 January 2026 (UTC)

I read this again, I think there could be two separate parts.

first, that you need an accurate bfo. If off by 30 hz, Donald Duck appears (with a reference).

Second, the USB/LSB issue. The hetrodyning signal needs to be in the same place it was before removal, or else the spectrum would be inverted. Should be easy to find a reference on that, many ham transcievers used an IF at 9 MHz, causing the usb/lsb to flip for some ham bands.

The USB/LSB discussion only marginally belongs in Demod. WhaleFarm (talk) 16:00, 27 January 2026 (UTC)

I rewrote this section to clarify the signal flow in the Weaver modulator and explain the purpose of the first frequency translation (to create spectral separation that allows simple low-pass filtering). The earlier text mixed concepts and implementation details and didn’t make this clear. This edit should make the section in line with Weaver's origonal paper.

The 300 Hz and ~2 kHz values are given as speech examples, not requirements. The description is aligned with Weaver’s 1956 Proc. IRE , while keeping the explanation non-mathematical and encyclopedic. WhaleFarm (talk) 15:45, 27 January 2026 (UTC)

I love this book, but it's beyond many people interested in SSB. It's very dense. Is it appropriate to have so many references to it? WhaleFarm (talk) 16:06, 27 January 2026 (UTC)