A Simple Reliable Double Sideband Suppressed Carrier (DSB-SC) Demodulator

[KerimF]

Hi all,

In year 1979, I was told, at the university, that it is somehow impossible for a simple reliable DSB-SC demodulator to exist. So, I decided to prove, as an MS thesis and to myself in the least, that receiving more information (the two side bands of DSB-SC) should make recovering the modulating signal much easier than in the case of receiving less information (the one side band of SSB-SC).

To those who are interested in AM communications, I attached below LTspice files which introduce the topology of the simple reliable analogue AM demodulator. It works for all modulation indexes; from m=0 (no modulating signal) to m=infinity (no carrier).

It uses one PLL (not two PLLs in quadrature as in Costas Loop) and it doesn’t need, therefore, selective filters, passive or active (found in the Squaring topology). Naturally, it locks to the suppressed carrier (frequency and phase). So, it has a frequency lock range, much like in FM receivers. And being simple and reliable, it could be made as a low-cost analog integrated circuit. But this didn’t happen because it seems that I am the only person who really believes in its existence. I even took advantage of it, in the 80’s (being not known by the world), in my private short-range RF voice links (between home and workplace, 3 km), on MW band then FM band (more detail about this, on request).

On the attached example, the carrier 455 KHz (AM IF) is assumed suppressed (worst case). For simplicity, the audio signal is emulated by the model of a white noise generator with limited bandwidth.

Cheers,
Kerim

[radiolistener]

these days it's more effective to use some cheap STM32 MCU + audio codec for demodulation in digital domain.

[szoftveres]

How does it perform? I see some places in the simulation where there's an uncertainty in phase - not sure if that's acceptable in data transmission, and hence whether it's a true apples to apples comparison to a Costas-loop. Note that a simple BFO and a product detector can reproduce audio like this, without ever being phase coherent to the carrier.
Marked the screenshot with arrows.

Also, you don't need two PLLs to make a Costas loop - just run your VCO at 4f and generate the quadrature LO signals with two edge-triggered flip-flops.

[KerimF]

these days it's more effective to use some cheap STM32 MCU + audio codec for demodulation in digital domain.

Yes, demodulating DSB-SC signals could be done digitally in these days.

The purpose of my topic is 'simply' to show that, based on human logic, having more information makes things easier to do... not the inverse.

For example, even in these days and at all universities around the world, the undergraduate students are told that demodulating a DSB-SC signal (holding more information) happens to be much harder to do than demodulating a SSB-SC one (holding less information). This is not true... but who cares

[KerimF]

How does it perform? I see some places in the simulation where there's an uncertainty in phase - not sure if that's acceptable in data transmission, and hence whether it's a true apples to apples comparison to a Costas-loop. Note that a simple BFO and a product detector can reproduce audio like this, without ever being phase coherent to the carrier.
Marked the screenshot with arrows.

Sorry, I am not sure how we got different results from LTspice for the same schematic.
Please see the attached screenshot about the first 2.5ms (related to your first arrow).

Added:
My LTspice, unlike yours, shows clearly how the demodulator output follows the modulating signal always.
It is weird that, on your screenshot (post #2), the phase error (at VCOin) is steady!
On mine (post #4), and in a real demodulator, it cannot be so. It varies to keep the PLL locked while the duty cycle at the 'SIG' input varies, mainly at zero crossings of the DSB-SC input signal.

[KerimF]

Also, you don't need two PLLs to make a Costas loop - just run your VCO at 4f and generate the quadrature LO signals with two edge-triggered flip-flops.

I also wonder if this method is simpler than the one, I use.

[KerimF]

At the university lab and after I failed in all designs which I did during about 3 months (spring 1979), I tried the last one after which I had to leave the university and my high studies for good. I built it while I was almost sure that it will be another failure since, in theory, it shouldn’t work.

To my big surprise, it worked (much like it happened to Archimedes ).

First, I thought that the musical audio input (from a tape recorder) was connected, by mistake, to the output speaker. But when I varied, a little, the frequency of the ‘suppressed’ carrier, the sound went off.

Truth be said, on that day, I had no clue about how it was possible for this simple PLL to work (without needing two signals in quadrature as in Costas Loop). Only when I returned home, I had enough time to examine every node of it. I found out that it worked simply due to the natural imperfection of the high-gain comparator (as LM339) whose function was to convert the input analogue DSB-SC signal to squarewave. The duty cycle of the generated squarewave at its output was slightly different from 50%. And this small difference was enough to let the PLL lock though in a very narrow band.

Later, a further study let me know that the widest lock range occurs when the duty cycle at the output of the high-gain duty shaper is made to be 25% (or 75%). On the actual schematic (the attached one), there are two 25% duty shapers. But the DSB-SC demodulators which I used in the updated receivers of my private RF links, in the 80’s and for many years, worked fine with one 25% duty shaper only.

[KerimF]

Please note that this topic is academic now.

Since many years, FCC stopped giving licenses for the DSB-SC system for general use.

Knowing the trivial topology of this demodulator was likely useful to many radio users in the 80’s when their number was relatively small. And it was possible for some companies to produce low-cost ICs of it. At that time, many users would choose the DSB-SC system for being simple to implement at the transmitter and the receiver as well, besides its other advantages. Its only disadvantage is that it occupies on the radio spectrum twice the audio bandwidth and this cannot be tolerated anymore due to the big number of today's radio users.

But I still wonder which university in the world will dare to be the first one in correcting the universal scientific fallacy that says, 'demodulating a SSB-SC signal is easier than demodulating a DSB-SC one'.
For instance, how many years the world needed, after Galileo’s work, to correct the universal scientific fallacy that says, ‘Earth is flat’?

OT: Early 80’s, I presented my work to the local authorities (civil and military). I just got various negative reactions (each has its own story). The worst one was during an official meeting at the military headquarters. In this meeting, they insisted that it is impossible for someone who is born and lives in Aleppo city to know a scientific idea not known by any big company in the world working in its field. So, after about 20 minutes and by recalling the suffering of Galileo, I simply apologized for imagining unreal things and I was able to return home sane and safe
By the way, I used hearing that ‘Ignorance is bless’. I realized, from this experience, that ‘Being able to play ignorance is bless too’

[TheUnnamedNewbie]

I think nobody is looking at this because, to be blunt, nobody cares.
I've been working in the wireless world for a while now, and I have never, ever, heard anyone say anything like 'demodulating a SSB-SC signal is easier than demodulating a DSB-SC one'. At millimeter-wave we are reinvestigating SSB because it is easy to generate, but it is a bit of a pain to demodulate (as you need LO recovery and coherent demodulation, unlike DSB), so unless you really win a lot, you don't do it. To be clear, I'm talking about >50 gbit/second applications here.

And in any low-speed system, like audio or so on, you do all of that crap in the digital domain anyways.

[KerimF]

I think nobody is looking at this because, to be blunt, nobody cares.
I've been working in the wireless world for a while now, and I have never, ever, heard anyone say anything like 'demodulating a SSB-SC signal is easier than demodulating a DSB-SC one'.

I believe you.
Since about 3 decades ago there is no more real interest in DSB-SC system for many reasons. Therefore, since then, there was also no reason to compare the reception of SSB-SC and DSB-SC.

But if you are really serious in your remark, you may like asking an old professor in communications, at any university, the simple question:
"Which is simpler to demodulate, a SSB-SC signal or a DSB-SC one?"
He will reply without hesitation: "It is well known that SSB-SC is easier to demodulate".

Please notify me if you will get the opposite of this answer. Thank you.

[KerimF]

By the way, is it possible for two friends to get licenses for two adjacent SSB channels on a SW band?
Thank you.

Some of you will likely guess why I am asking this

[KerimF]

As I said earlier, I took advantage of this demodulator in the 80's to link home with workplace.
Although a conventional MW receiver couldn’t detect properly the voice signal transmitted as DSB-SC, the human brain of the listeners was able to differentiate between the sound of a random noise and a deformed speech. So, to hide completely the speech pattern, I had to also modulate the frequency of the suppressed carrier (if I remember well, by a 6 Hz sinewave with a frequency deviation about +/- 30 KHz, covering about 6 AM MW channels). This didn’t affect the reception of the DSB-SC signal, but the MW listeners used to have the impression that they were just receiving a noisy interference from a nearby machine.
After a few months, I decided to find a better solution than this very noisy one to scramble my two RF voice links (between home and workplace). I simply shifted to the FM band. The FM listeners just heard two quiet channels while a conversation was going on. The idea was trivial. I had a 32,768 Hz crystal which became the frequency of the suppressed carrier which, in turn, was modulated by the voice signal. I got this idea from the stereo FM protocol which uses 38 KHz to carry the difference of the two stereo channels. Obviously, I kept the baseband of my FM channel(s) empty, and I didn’t need to add a pilot (as the 19 KHz for the stereo FM).
I used this technique for many years till I got a Bell phone line at home

[KerimF]

Naturally, the gurus in communications around here realized already that this simple demodulator can also recover a second signal if it modulates the frequency of the suppressed carrier.

In this case, one may wonder what the effective bandwidth formula of such FM AM DSB-SC signals could be.

Although it seems that no one cared to mention it on any book or article, a good engineer (not a technician) in communications can deduce it rather easily.

[KerimF]

Perhaps there is one person among the billions in the world who may like to see an example of the FM-AM-DSBSC demodulator.
For him, I attached its LTspice files.

[KerimF]

I am very sorry; it took me too many days to realize that I had to start this topic in the forum ‘Projects, Designs, and Technical Stuff’.
I had to remember that, since many decades ago, the study of the advanced data communications doesn’t need any more to focus on the basics of electronics (at the level of basic discrete components) as it was the case on my days.

Although it is not noticed yet, the circuit, I presented, still has a weakness that needs to be solved in order to make it up-to-date. I didn’t do it because, on the one hand, its side effect is small in typical applications and, on the other hand, its solution seems not trivial. So, I hope when a moderator will read this post, he can help in correcting this situation by moving, if possible, this topic to ‘Projects, Designs, and Technical Stuff’.

Best Regards,
Kerim (an old-fashioned engineer)

[RFDx]

Hi all,
It uses one PLL (not two PLLs in quadrature as in Costas Loop) and if doesn’t need, therefore, selective filters, passive or active (found in the Squaring topology). Naturally, it locks to the suppressed carrier (frequency and phase). So, it has a frequency lock range, much like in FM receivers. And being simple and reliable, it could be made as a low-cost analog integrated circuit. But this didn’t happen because it seems that I am the only person who really believes in its existence.

Carrier recovery by squaring/frequency doubling is known since the end of the 50's, so technically the method is rather old. It has some drawbacks because the envelope of the input signal goes through zero and you can't square or frequency double something that is zero.

But I still wonder which university in the world will dare to be the first one in correcting the universal scientific fallacy that says, 'demodulating a SSB-SC signal is easier than demodulating a DSB-SC one'.

Why would any university in the world do that? Demodulating SSB is easier and simpler. That's a fact. How is your DSB-demodulator "easier" in comparison to a simple product detector made of a BFO and a mixer. No need for limiters, squarer/frequency doubler, VCO, PFD, PLL-filter or frequency divider.

[KerimF]

Hi all,
It uses one PLL (not two PLLs in quadrature as in Costas Loop) and if doesn’t need, therefore, selective filters, passive or active (found in the Squaring topology). Naturally, it locks to the suppressed carrier (frequency and phase). So, it has a frequency lock range, much like in FM receivers. And being simple and reliable, it could be made as a low-cost analog integrated circuit. But this didn’t happen because it seems that I am the only person who really believes in its existence.

Carrier recovery by squaring/frequency doubling is known since the end of the 50's, so technically the method is rather old. It has some drawbacks because the envelope of the input signal goes through zero and you can't square or frequency double something that is zero.

Please... Please... I wish you can give me, someday, one reference only which knows already the topology of the DSB-SC demodulator I presented here.
Sorry, I am almost sure that you didn't have time to read all posts of this topic attentively. For example, this demodulator has nothing to do with the Squaring Method that needs one selective filter in the least and cannot lock to the frequency of the suppressed carrier if varied, as my PLL demodulator does.

But I still wonder which university in the world will dare to be the first one in correcting the universal scientific fallacy that says, 'demodulating a SSB-SC signal is easier than demodulating a DSB-SC one'.

Why would any university in the world do that? Demodulating SSB is easier and simpler. That's a fact. How is your DSB-demodulator "easier" in comparison to a simple product detector made of a BFO and a mixer. No need for limiters, squarer/frequency doubler, VCO, PFD, PLL-filter or frequency divider.

It is not my fault that the world didn't gather, due to lack of knowledge, the 'limiters, squarer/frequency doubler, VCO, PFD, PLL-filter and frequency divider' in a low-cost IC, as it used to do for many other applications (one just needs to look how the 'digital' DSB-SC demodulator had to be integrated in relatively high-cost ICs).

Anyway, all universities have no choice but to teach 'Costas Loop' and 'Squaring Method' only for the DSB-SC system, as long they have no idea that a simpler reliable 'topology' with a frequency lock range does exist.

And because the world didn't know it, a pilot (19 KHz) had to be added in stereo FM broadcasting so that, the receivers can recover the suppressed 38 KHz.
Obviously, in the 80's, I didn't need to add such a pilot to demodulate my DSB-SC signal whose carrier frequency was 32768 Hz, in my private short-range voice links (on FM band).

[RFDx]

Please... Please... I wish you can give me, someday, one reference only which knows already the topology of the DSB-SC demodulator I presented here.

http://www.amalgamate2000.com/radio-hobbies/radio/dsbsc____demodulation_by_the_squ.htm

Besides carrier recovery with a squaring/doubling loop, you can find on the bottom of the website other patents dealing with the same topic.

Sorry, I am almost sure that you didn't have time to read all posts of this topic attentively. For example, this demodulator has nothing to do with the Squaring Method that needs one selective filter in the least and cannot lock to the frequency of the suppressed carrier if varied, as my PLL demodulator does.

Well, if you are (almost) sure it must be true. You are doubling the input frequency in some dubious "LTspice black boxes" and use the output as a reference for a PLL. Where is the innovation you are talking about? No need for a selective filter, the PLL does the job for you. A selective filter can be used in front of the PLL but you would later need to remove the phase shift introduced by the filter. 

Anyway, all universities have no choice but to teach 'Costas Loop' and 'Squaring Method' only for the DSB-SC system, as long they have no idea that a simpler reliable 'topology' with a frequency lock range does exist.

The universities teach the most common methods for carrier recovery. That is plenty. If a student wants to become acquainted with other methods he can visit a library.

[KerimF]

Please... Please... I wish you can give me, someday, one reference only which knows already the topology of the DSB-SC demodulator I presented here.

http://www.amalgamate2000.com/radio-hobbies/radio/dsbsc____demodulation_by_the_squ.htm

Besides carrier recovery with a squaring/doubling loop, you can find on the bottom of the website other patents dealing with the same topic.

Thank you for the reference.
I have no intention to hurt your feeling in any way.
I just try to talk professionally. So, if I say something that sounds wrong to you, I am all ears to learn from you.
I, like most engineers (in electronics), try to avoid as possible the use of coils. So, if a problem could be solved by two methods; one needs coils (even one coil only) and one doesn't need any coil, most engineers prefer the latter method. Am I exaggerating?
This is why I see my demodulator as being simple. And since no coils are needed, it could be integrated in one low-cost IC (instead of using many low-cost standard ICs).
Please note this demodulator is just an innovation, not an invention.
An invention could be, for example, of a dual-polarity gravity-wave generator so that it will be possible to travel in what is known as flying saucers.

Sorry, I am almost sure that you didn't have time to read all posts of this topic attentively. For example, this demodulator has nothing to do with the Squaring Method that needs one selective filter in the least and cannot lock to the frequency of the suppressed carrier if varied, as my PLL demodulator does.

Well, if you are (almost) sure it must be true. You are doubling the input frequency in some dubious "LTspice black boxes" and use the output as a reference for a PLL. Where is the innovation you are talking about? No need for a selective filter, the PLL does the job for you. A selective filter can be used in front of the PLL but you would later need to remove the phase shift introduced by the filter.

Sorry, if the word 'innovation' sounds really bad to your ears, I apologize.
Fortunately, it didn't sound so, to all other readers here.

For instance, among the designs I rejected while I was looking, in 1979, for a simple reliable DSB-SC demodulator, it was the topology which was patented for someone many years later (United States Patent 4430620). Its idea could be summarized:
“A DSB-SC signal could be demodulated without reinsertion of the carrier at the receiver. The received DSB-SC wave is amplified, demodulated by a full-wave rectifier. The rectifier output reversed in polarity at instants of successive zero crossings of the recovered signal.”

I rejected it because, at some zero crossings, the rectifier output doesn’t need to be reversed.  And to determine when the reversing is needed or not, mainly when the signal’s average amplitude become relatively very small (as in classical symphonies) cannot be achieved by a simple circuit.

Anyway, all universities have no choice but to teach 'Costas Loop' and 'Squaring Method' only for the DSB-SC system, as long they have no idea that a simpler reliable 'topology' with a frequency lock range does exist.

The universities teach the most common methods for carrier recovery. That is plenty. If a student wants to become acquainted with other methods he can visit a library.

Indeed, I wish this demodulator exists (or will exist) in certain libraries.
In fact, this is the purpose of this topic.

[KerimF]

To those who don't have the free LTspice simulator, I attached the circuit of the block 'DutyShaperIF' which is used on the main schematic.

[KerimF]

As I mentioned earlier, I used this demodulator in the 80’s for several years. But I used in its circuit one DutyShaper only instead of two. When I had the chance to download LTspice, I had the idea to add a second one, as seen on the attached schematics above, ‘DSB-SC_455K-v4’ and ‘FM-AM_DSB-SC_455K_v5’.

In ‘FM-AM-DSB-SC_455K_v6’, I removed the second DutyShaperIF in ‘FM-AM_DSB-SC_455K_v5’ (post #13) to see the difference in the demodulator’s response.

I guess a few readers only will notice the main difference of their responses by comparing ‘FM-AM_DSB-SC_455K_v5_plt.png’ and ‘FM-AM_DSB-SC_455K_v6_plt.png’.

[KerimF]

(Continued, previous post)

By a close look at the traces on ‘FM-AM_DSB-SC_455K_v5_plt.png’ (reply #13), the amplitude of the received AM modulating signal is not constant as it is supposed to be. It varies with the variation of the suppressed carrier frequency.

And by a close look at the amplitude of the received AM modulating signal on ‘FM-AM_DSB-SC_455K_v6_plt.png’ (reply #20), it is also not constant but, for the same variation of the suppressed carrier frequency, its variation is bigger.

For instance, I try not to bother the readers in this RF forum by adding detailed analyses (formulas, equations... etc) of this demodulator’s topology. I noticed lately that most of the today’s hobbyists and professionals in the RF communications have no real reasons to be interested in the basics of electronics while they construct/design their various modern complex devices in which most parts (blocks) are ready made. But I don’t mind discussing any part of it with those who may be interested.

[KerimF]

If there is an undergraduate student, among the guests of this forum, at a faculty of electronics, he has the opportunity to present the topology of this demodulator as a thesis project. He can be sure that none of his professors has an idea of this analogue DSB-SC demodulator (actually a versatile AM demodulator for all AM indexes) that doesn’t need any coil or selective filter (as in Squaring Method) or signals in quadrature (as in Costas Loop).

But I am afraid that this student needs to know how to find/calculate the values of its components (resistors and capacitors), besides choosing suitable modern ICs for it. By doing this, this thesis project will become really his.

For instance, the old version of the one presented here was made in 1979 and applied in the 80’s.
Then, it was published on the magazine ‘ELECTRONICS & WIRELESS WORLD, August 1987 (page 803).
Also, a version of it for 38 kHz (as in stereo FM) instead of 455 kHz (the MW/SW IF) was published on the same magazine, January 1987 (page 35).

Good luck.

[radiolistener]

But if you are really serious in your remark, you may like asking an old professor in communications, at any university, the simple question:
"Which is simpler to demodulate, a SSB-SC signal or a DSB-SC one?"
He will reply without hesitation: "It is well known that SSB-SC is easier to demodulate".

Please notify me if you will get the opposite of this answer. Thank you.

I'm not sure, why you are thinking that DSB is easier to demodulate? It can be easier if both side bands remains unchanged. But in real air, each side band is affected with different fading, different noise and different parasite carriers. As result DSB signal has two copies of the same signal but each copy is distorted in different way. And since we don't have third copy, we don't have enough information to make decision which sideband is more close to original signal. So, it doesn't helps much to demodulate it. The only advantage is that both sidebands can be summed in order to get better signal power and improve SNR. But if we put exactly the same power for SSB and for DSB, it seems that SSB looks better, because it has more high power spectrum density and as result SSB is less weak for background noise. Isn't it?

The problem here is that DSB signal uses twice more wide bandwidth, so it passes twice more noise power to demodulator than SSB signal with the same power. How can you eliminate that excessive noise in order to achieve better efficiency for DSB?

[KerimF]

I'm not sure, why you are thinking that DSB is easier to demodulate? It can be easier if both side bands remains unchanged. But in real air, each side band is affected with different fading, different noise and different parasite carriers. As result DSB signal has two copies of the same signal but each copy is distorted in different way. And since we don't have third copy, we don't have enough information to make decision which sideband is more close to original signal. So, it doesn't helps much to demodulate it. The only advantage is that both sidebands can be summed in order to get better signal power and improve SNR. But if we put exactly the same power for SSB and for DSB, it seems that SSB looks better, because it has more high power spectrum density and as result SSB is less weak for background noise. Isn't it?

The problem here is that DSB signal uses twice more wide bandwidth, so it passes twice more noise power to demodulator than SSB signal with the same power. How can you eliminate that excessive noise in order to achieve better efficiency for DSB?

I guess you also know that what you said applies on any known DSB-SC demodulator. The one presented here is just the simplest one.
It can be integrated in a relatively low-cost IC and doesn't need precise values for its components.   

What I know is that I used it in my private short-range RF links on MW in the 80's. The received voice was normal all the time while the transmitted power was relatively small since I didn't have at that time power RF transistors suitable for 1 Mhz.

About the noise/fading effect, I wish I still live in the golden years (before year 2011... when the world's Elite decided to also save the people among whom I was born and live) so that I would be able to build two transceivers of the same power (DSB-SC and SSB-SC) and compare their usefulness/responses in the same conditions.

For instance, there are applications that can take advantage of the FM detection of this demodulator. For example, in a multi-user walkie-talkie that uses one frequency only for the DSB-SC, a code could be transmitted, as FM and at a relatively low rate, to activate, for example, the ring at one user's side (besides adding, perhaps, a short message from the caller). But this system is practical among friends only since, as in a serious meeting room, there is, at any moment, one person only who talks while the others listen. And in this case, the advantage of this DSB-SC demodulator (having a lock range) is that the transceivers don't need to have exactly the same carrier frequency (in their transmitter and/or receiver).

If someday an analogue company will make ICs for it, many toys will likely use the DSB-SC system for being simple to implement, for transmission and reception as well. But since I am 74, I don't think I will live to see that day

[ftg]

Thank you for mentioning this project in the sinewave inverter thread, I thought it disappeared with the radio board.
I'll have to dig in my junkbox for some comparators and 4046's to test this out one of these days.

Direct DSB demodulation without costas or squaring loop would be quite nice and the lm339 duty cycle modifier looks like a neat trick for doing it.

[KerimF]

Thank you for mentioning this project in the sinewave inverter thread, I thought it disappeared with the radio board.
I'll have to dig in my junkbox for some comparators and 4046's to test this out one of these days.

Direct DSB demodulation without costas or squaring loop would be quite nice and the lm339 duty cycle modifier looks like a neat trick for doing it.

Lately, to show that this simple topology combines the two known techniques (Costas Loop and the Squaring Loop), I updated the duty shaper (50% to 25% or 75%) by adding to it a similar one but in anti-phase.
Their two open collector outputs are ORed and the result is a square wave signal whose frequency is 2*fc.

I called their block '455KHz_doubler' (its image is attached below).
(This block replaces the first/upper duty shaper on the main schematic which is driven by the DSB-SC signal. The second/lower duty shaper is no more needed, also the anti-phase DSB-SC signal that drives it).

Added:
To simplify the 455KHz_doubler, XOR gates (as of 74HC86) could be used to generate the 2*fc square wave signal.