DSBSC demodulation by the squaring loop technique

There are a lot of small simple DSB transmitter projects on the air , these are usually low power transmitters of 1 to 10 watts ,they are easy to make and also efficient on power as no wasteful RF carrier is sent along with the information sidebands.
The usual way that  most radio amateurs demodulate double sideband suppressed carrier signals  is by means of a single sideband receiver , thus  using upper or lower side band mode then  zero beat the DSB signal for best intelligibility;


The the other way to  truly demodulate dsbsc that I know of and have seen circuitry and theory  is by  two techniques, 1,  the squaring loop  or 2.  Costas loop . either technique to synthesise the bfo carrier from the upper and lower sidebands  and enable efficient demodulation  using both upper and lower sideband simultaneously in a product detector or mixer. .
This is an experiment to test this first method and to examine the function of the components  of the squaring loop . My experiment was started at 455 Khz  and the main reason I picked this common frequency was its universal use in radio receivers  and my  signal source ,  the HP8904A  will only go to 600 Khz ! it is a versatile synthesised source that can perform many types and mixes of modulation function ,AM,FM,PM,Noise,and Double sideband suppressed carrier  to name a few .   so there is my modulated DSB generator on hand  . To demodulate DSB you must reinsert the carrier synchronously with the two sidebands to get the speech to sound natural . This fact is the saving grace that enables the demodulation of the  dual sideband suppressed carrier signal and enables one to derive the correct carrier  ( b.f.o) relationship to the DSB signal .   THE CARRIER frequency IS ALWAYS MIDWAY BETWEEN THE UPPER AND LOWER SIDEBANDS ,  Just imagine  a 1 Khz modulation tone , the transmitted signals  are  carrier frequency ( remember its missing) plus 1 Khz and minus 1 Khz  these are the frequencies of the two transmitted signals above and below the suppressed carrier ,  so all we have to do is  some electronic  manipulation  of these two sideband signals  and we can derive the original suppressed carrier frequency. midway between them .

    MC1496 configured as a doubling circuit  I found an increase in the wanted output if the value of the coupling capacitor feeding pin 10 was 1/5 of the input cap value
Enter the Squaring circuit, There are several configurations to achieve this  I decided I'd use a MC1496 mixer as a suitable squaring circuit ,its got lots of gain ,they work well, also hoping the carrier suppression features of the IC will minimise the feed through of the 455KHz DSB  to the output of the squaring circuit and we will only have 910 KHz to contend with .
For any  DBM to act as a doubler or squaring circuit you feed the same signal into both input ports ie the carrier port and the signal port  and the product will appear at the output .I set up the signal generator at 100mV,  feeding the two input ports and spectrum analyser to look at the output of the mc1496  wired in the doubling  mode as per the IC  data sheet ' I also ran mine on a one sided 12 volt rail ( no dual volt source)  I used the recommended component values in the MC1496 applications Note,  downloaded from the web .

 I'm not going to get into the math of this process , suffice to say there is plenty of that on the net if you look around  ( but no circuits of how you do it practically) hence the project , There are a couple of excellent articles  on DSB demodulation out of the now extinct  HAM RADIO  magazine . and my experiment is based on the one using the squaring loop technique , In the Article by ..H Priebe  Auto product detection of DSB,........HAM RADIO March 1980   he uses a Fet transistor biased into square-law ( non linear) to multiply the 455 KHz  DSB signal  by itself and derive the wanted  product s.    It turns out if you take a 455 KHz DSB signal and square it ( multiply it by itself ) you get some useful signals out ..  By doubling the 454Khz and 456Khz sidebands, you will get 908 KHz  and 912 KHz out of your squaring circuit , that  is as we mentally visualize the  product of the "times itself"  operation will produce  BUT, there is a third product of interest,  the single  product 910Khz!!  divide this by two and the result is 455 KHz ! midway between the upper and lower sideband .  feed this into our product detector as the BFO and voila  out pops 1 KHz  tone  exactly what we  want . So there it is , the DSBSC squaring loop in a nutshell !  easy huh !  but do you think I can find additional practical circuit on the net  to build one  ....no !   a Good place to start when you are researching projects  is to do a patent search on the topic , all sorts of stuff turns up that will assist in your understanding of the idea  but don't get distracted when doing patent searches,  there is so much curious  and interesting stuff out there ,its easy to get waylaid .

I was somewhat dismayed at the low conversion efficiency of theMC1496 circuit as a doubler ,at the input  powers I was using  , there was quite an amount of loss in the squaring process.   however when I tried an SBL-1 as a squaring circuit it just would not work at the low levels of power one would expect from a radio receiver I.F so I guess the 1496 is not so bad after  all .                      ( today found another simple doubling  circuit to try  using just 3 transistors)
I fed  912.4 KHz into the squaring board input and looked at the level of 912.4 KHz  at the output , it was -55 dBm . a fundamental suppression of 35 dB  . note this signal is going into both  carrier and Rf ports so I dont expect to get the usual 50-65 db rejection of  signals one would get with the MC1496 configured as a mixer .

                

       100 mV  DSB at 455 KHz 1 KHz Mod    into MC1496     -22 dBm                           Output of MC1496   note 910 Khz carrier generated  sidebands at 2 KHz !     

       
                    F

The wanted product at 910 KHz is about 6 dB above the other products ,  this 910Khz  frequency will have to be selected or filtered to isolate it from the other sideband ( note they are twice the original sideband frequency from the new carrier  ie 2 KHz ! .  a single series quartz crystal at 910 Khz is what the HAM RADIO article used to pass the 910 Khz . this is where the project gets very  expensive , custom made crystals are not cheap ,I will try and obtain one to duplicate the original article . (  subsequently gave up on this idea they are too expensive , so  I changed IF frequency )

(The only reason I was keen on 910 Khz meant I could easily use the 455KHz  output of my Icom r75 receiver with fiddling or additional mixing  )

 There is of course no reason why one couldn't in practical terms  use   for example a  9 Mhz  i.f  and then for a filter , an 18 Mhz  microprocessor crystal,  readily obtainable inexpensively from Mouser., Radio Spares , Element 14 , or many other  component suppliers   . or use a completely different IF that suits the filter crystal you have ???
 I did some tests ramping the 455 KHz DSB signal up and down from 1 mV input to 200 mV and saw the 910KHz wanted product was always at least 6 db above all the sidebands  ..Im looking at the 910 KHz output spectrum and thinking "that now looks like an AM carrier signal ?", the wanted product is 6-7 db above all the rest of the carriers  so what if I pass it through a limiter like an FM radio does and hopefully the capture effect will hold the 910 Khz and strip the unwanted carriers  .  I tried the FM IF limiting  and there was some increase in carrier  level above the existing sidebands but i figured it was  much too complicated  for the amount of extra carrier level derived  plus the generated intermod from limiting  so I gave it away ,                                                         
   Feed through  of 455Khz at doubler output  -66.5 dBm                                                                     

What i did notice was the output conversion of the squaring circuit using the mc1496  is not good  ie  a -24 dBm input at 910 Khz gave a product   1.843 MHz @ -36 dBm   also  if there is a 10     db change in input level there will be a corresponding 20 db change in output level  ie drop the input to the doubler by 10dB and the wanted "doubled carrier product"  falls by 20 dB   !!  following the square law !!

FILTERS  :
Back to filtering techniques to prise the wanted synthesised carrier  out of the squaring circuit.  I have been playing around with simple single quartz   crystals .as band-pass filters , I experimented with phasing crystal filters  with the moveable passband rejection notch but came to the conclusion that the rejection notch is not required in this application . the single phasing crystal did work but the circuit was a little fiddly  ,so I looked at other filter ideas .
Just a note : When sweeping very narrow high Q filter circuits you must sweep at very low rates, I was getting a strange ringing on the high side of the pass-band ,these looked like ripples on the trace , I  fiddled for ages with components trying to determine what was causing it only to realise that it was due to the excessive  sweep rate of the Agilent E4407  spectrum analyser in tracking mode . Just because these newfangled machines have an auto function that mixes and matches sweep speed to resolution bandwidth and sweep width  it can still get it wrong ! . beware and be prepared to override auto settings if things just don't look right ......... you can never sweep  too slowly but you can sweep too fast !! ..I experimented with a number of HC49  crystals , 10 Mhz, 4 MHz and 1.8430 MHz , sweeping them to see which one gave the narrowest response and as you would expect the 1.8 Mhz was the winner !  typically 250 Hz  but often less , this gives me  -3dB  selectivity 125 Hz away from the wanted carrier and since voice comms tends to start above 300 Hz this amount of roll-off from the simple single crystal should give a good usable  wanted carrier to unwanted sidebands   Ratio ;   Amplify this  filtered carrier up to a level where it will trigger a divide by two counter and then I have my regenerated carrier at the correct frequency for injection  into the product detector to recover the audio from the received  DSB .  well that is what I  imagined anyway!!

             1.84 MHz crystal filter swept response                                                                             1.8 MHz microprocessor xtals    see  LA8AK  website for more details

To test the concept  I had to put together a simple DSB generator at 1.843 Mhz  .The 1000  Hz audio from the HP8904A into an SBL-1 DBM  , 921.523 Khz ( half of 1.84304 MHz)  from the from the Agilent  N9310A  Signal  generator   and the resulting  DSB output into the MC1496 squaring circuit with the 1.84304MHz  crystal as a series  filter on the "product " output port  . I did several audio modulation frequencies to see how the sidebands  showed in relation to the carrier , and the wanted synthesised carrier was always much higher than the sidebands .I originally tried the  classical   simple single phasing crystal filter circuit ,  a la 1930 radio techniques  but the poor shape factor allowed a lot of  unwanted sideband energy either side of the wanted  derived carrier , even though the wanted carrier was 15-20 dB above the crud I felt we could do better ! when I get things operational i can fiddle to see how much minimal carrier we can get away with and the demodulator still function ,  Extensive searching on the web brought me to  LA8AK web pages " Improved stopband of crystal ladder filters " I replicated one of his circuits with some changes in component values, scaling with frequency  , the response of this simple filter is exceptional ,   look at the  swept response in the Dual Window above ,  note  the 10 Khz  span and the 1 Khz span . This LA8AK website filter has stopbands up to 80-90 dB below the minimum bandwidth .

     

 100 Hz modulation     at     921.523 Khz   +/- 1 Khz   (  squaring  output )                             Filtered  carrier derived  from MC1496 squaring circuit  (300-3K  modulation)

The squared output  signal was fed through the LA8AK  website  dual crystal filter  and viewed on the analyser   see filtered signal above, The Fixed audio frequency tests where very satisfying but how would it go with  real comms quality speech?? so I then decided to use a local Govnt agency comms channel as my speech source (  Automated continuous speech from local airport  on weather conditions)  as an audio feed  into the DSB modulator.  I was interested to see the processed  RF spectrum from the squaring circuit after it had passed through the new series twin crystal filter . I took the audio output of my trusty old Yeasu VX-5 and fed it into the audio port of the sbl-1 DBM .  , At no time was the wanted carrier ever approached in significant  levels  by the sidebands around it , see filtered carrier picture
You will notice the losses of the very narrow  dual crystal filter is quite high , around 17 dB , these loss appears to be consistent with these extremely narrow crystal filters (although maybe my poor matching is the source of the great loss??) . A much wider filter,  3KHz or so will have much less loss , maybe a couple of dB or so . But that's not a problem as the DSBSC demodulator will sit at the end of the  receiver IF chain,  mollycoddled by the AGC and be supplied with  plenty of RF  level  and a  good signal to noise  ratio to work with . 

                 

                          Modified novel cascade if amp    off  Harry 'SM0VPO  web  page on IF  amps , NOTE:   the ZL1BTB MODIFICATION  replacing one of the decoupling
                          capacitors with a crystal on the frequency I wish to enhance 

There is still a large amount of amplification to bring the derived carrier up to a suitable useful level , I tried a CA3089 FM  IF   strip but was not happy with the broadband noise coming from the limiter ,the output never looked clean on the scope  ,  So I looked through the web  ( as one does) looking for suitable circuit when I came across this one in Harry  SM0VPO website  ,it is a cascode  almost totem pole amplifier configuration where the first three transistors are in series  . I built up the circuit on double sided PC board . I swept it on the spectrum Analyser in tracking mode and man has this circuit got some low frequency gain!  80 dB  plus , see the screen shot , The tracking generator out (amptd was -60 dB)  with 30 dB in series with that !! ( I had to reduce the tracking generator RF output to almost 90dBm with external pads to stop the cascode amplifier  from limiting at the output ) , it has your typical ultra high gain broad band roll off .  BUT it is a useful circuit,   To cut down on the broadband noise ,  I replaced one of the  1nf emitter decoupling capacitors with a microprocessor crystal at 1.8430  MHz  and obtained a very useful response ,  I have marked the 10dB bandwidth points as I wish to see how it performs filtering out noise on the multiplied sidebands ,note this simple single crystal  filter has poor stopband attenuation ,but I only want my wanted signal 15-20 db above the crud to use it , I connected the oscilloscope probe to the series amp output and . the scope pattern looked a clean sine wave ,  I suspect we can get some good clean logic level triggering of the following  divide by two  with this output !

      
 Harry's  cascode IF  strip   wide open .....terminated in 1K at each end                              Emitter bypass cap now  replaced with 1.837 MHz crystal

        
   This filter has much less loss as the twin xtal  filter.  I  suspect this is  all I require                                           block diagram of the tail end !

I have four 1.843 Mhz crystals  hence the different centre frequencies  on display 


The level of filtered carrier is quite low and around the 20-25 uV level. It  will need a good deal of amplification to bring it up to a level where is can  reliably trigger a "divide by two"  to generate the BFO injection frequency . ,

I made up the above SM0VPO   cascode amplifier  with the ZL1BTB crystal  modification  to narrow the bandwidth and remove excessive  broadband noise . I built this on double sided PC board to minimise potential interaction and to maximise screening , the output of the squaring circuit was connected to the high gain narrow band amplifier , I connected the Oscilloscope to the output of this  amplifier  and feed varying  levels at 1.83 Mhz   into the SM0VPO amp  from the signal generator and noted the high gain output (terminated in 1 K )  amplitude .  -60dBm  input  gave 12v pk to pk . -63dBm  11 v , -66dBm  gave   6,5v , -70dBm  gave 2.8v    and -80 dBm gave 300 mV, however ,  a noisy  display  .

I suspect with the average  IF level of  most receivers being up near the -20 dbm level at the end of the I.F  so  there will be plenty of signal and lee way for the circuit to operate over a moderate range of I.F  outputs,

I then built up the second  half of the H Priebe  Auto product detection of DSB,........HAM RADIO March 1980  circuitry once again on double side PC board , This contains a 1.84 MHz class A  broadband amplifier, logic level  pulse shaper , transistor cross coupled multivibrator as a divide by two  , a class A 920 KHz  IF amp  and a mc1496 product detector with audio  output .  I have the most trouble with the divide by 2 circuit and if I was going down this road again I would use a dedicated CD4013 cmos logic I.C. , I had to change the component values quite markedly from the original to get the circuit to play ball.....   so frustrating  for such a straight forward  circuit . !

                                                                

                                                                         -100 dBm into the series Amp board and digital divider board

just a note there is an error in the circuit of the divide by 2 module ,  the two emitters connect together  and  to earth ,  but not as shown in the diagram,  sorry  my mistake 

 

To test thes sensitivity of just the "flip flop" board I connected the Agilent  N9310A  Signal  generator  to the input at the "doubled IF Preamp"  and the oscilloscope probe on the  pin 10 of the MC1496  . I set the signal generator  to 1.8248 Mhz and ramped the output amplitude up and down , the minimum level I could get reliable toggling was          -- 33dBm .

I connected the two pc boards together .  fed the 1.84MHz from the signal generator into the series Amp , placed the scope probes one on pin 10 of the MC1496 product detector  and the second probe on the  collector of the double if preamp transistor .I ramped up the amplitude of the signal generator from -127 dBm and watched the scope  traces,   At  weaker than -100 dBm I could easily get reliable solid half frequency square waves on pin 10 !  .  Any way after getting it all operational,  I decide to test it . The 1 Khz  DSB  out of the SRA-1  DBM ( I suspect my original SBL-1 was dodgy)   was resistively split in two,  Half feed to the I.F input of the product detector and the other half was fed to the input to the squaring circuit ,     it was all connected up and the Rf  level from the signal generator set to show good clean square waves  coming from the divide by two ,on the scope  and I listened to the audio output of the product  detector   and there was the 1 Khz audio tone (as generated by the  HP8904A feeding the I.F port of the SRA-1 DSB generator )  so the concept works   though its somewhat "complicated" .. The next real test is to pull some real  dsb off the air  on HF  and feed that into the demodulator and see what it sounds like !!

                                                                      


When running the DSB demod board with communication audio coming from the vx5 , it was amazing to watch the square waves from the divider on the oscillioscope  appear in time  and solidly locked , along with the audible  speech coming from the amplified computer speakers , I was able to simulate varying the I.F  signal levels  with the switched attenuator  feeding the squaring input .. when the signal level dipped  below trigger threshold the audio was stilted  or  muted .
...
What I did note that was somewhat amusing ,  I  could twiddle the signal generator frequency backwards and forwards from 921.4857 KHz  to 921.4257  KHz rapidly and the audio didnt change in pitch at all !!  so stability is not so stringent as  SSB  ( figures really when you think how the circuit actually works ) .. beyond that range of frequencies  there was some distortion appearing but Im sure that was due to the divider stage being starved of signal as the passband of the crystal filter reduced the level  of synthesised carrier .
 

Things to try       yet to try MC1590 as squaring circuit ,                          

                                             

                                                   and newly found 3 transistor doubler out of RSGB  "Amateur radio techniques"
I made up the "Andren "  broadband doubler up on some double sided PCB  but I modified the circuit slightly , a tuned output with link coupling for low impedance out and a small modification to adjust the bias  and see what it did to performance , I used BFY90 transistor because  I have a couple of  dozen or so ..
, I noticed the original circuit had a fixed 5 K bias resistor , I found this was too much for the BFY90's,   they all turned on too hard and got hot and bothered  so I changed to an adjustable bias which I set to just below for maximum  signal out  (0,813 V on the base of the  BFY90's )  These are  the results of the experiment  as a doubling  circuit ,I used 4.17 Mhz input and looked at 8.34  output  ( only because the hand 120 Pf cap I found on the bench  and the toroid I had in the output circuit resonated at that frequency ).

input  -25 dBm            out -13 dBm
           -35                            -28
           -45                            -41
           -55                            -53
           -61                            -59
           -63                             -61
           -66                             -65

Well this little circuit performs very well  for what I require!  The  MC1496 was -24 dBm in and -36 dbm out ,( -12 db conversion )   this little circuit is -25 dBm in and -13 dBm out !!! (+12 dB  conversion ) The next test will try a link coupling to the input to see if matching improves the performance  now I have to try it  with DSB and see how it goes  producing the frequency  product I require ..  
  This is the circuit Ill try on the front of my filter/divider board combination . with all these little improvements here and there I may be able to reduce the complexity of the demodulator  and the overall required gain hence improve the stability of the whole circuit ...
I may even try just the divider board on its own connected to this Andren doubler circuit, as the divider board  will trigger reliably at -33 dBm  but I will use a 1.83  MHz crystal to replace the emitter bypass capacitor in the "doubled IF Preamp"  part of the divider  circuit ...watch this space...

 

 

more to come

Once it is all operational I will try to measure the minimum signal required at each stage to make the circuit work,( done this)    Im sure the whole circuit can be greatly simplified , ie a mc1590  in mixer configuration , could replace the MC1496 squaring circuit and thus have extra gain  before passing into a more simple high gain IF amp of maybe just two transistors with  emitter bypass crystal , the "divide by two"  would be a CD013 and the product detector could be a simple dual gate mosfet configuration ??

perhaps one could incorporate a dedicated NBFM receiver chip mMC3362??? etc   as they have most of the circuitry we require .

  you can see the simple ssb product detector  using only one sideband is attractive BUT the tuning has to be spot on and  no drift of the receiver  VFO  .However with this circuit , drift is not so much a problem ,so long as you can get signal through the "doubled IF"  crystal filter at an amplitude sufficient to enable the  divider to work ,then drift wont cause audio distortion .. .

Double sideband demodulator  mk3  squaring loop

 DSSC    demodulation by the squaring means     diode doubler technique

.watch this space more to come  .

Double Sideband   suppressed carrier Demodulation   Patents

These are all the References I can find on Double sideband suppressed carrier demodulation , there are bound to be more ,  they are all US patent Office registration No..  The E  ones are European Patent office ,
Have a look on the application of File dates and the Granting dates and some people had to wait a long time before their patents  were granted .

2924706 ,2992326, 2999155, 3047659,  3060383, 3088069,  3108158, 3141066, 3151217, 3252094, 3286183, 3345571, 3457513, 3537017, 3593149,  4253066

 4821322, 4475218, 4642573,  4862098, 2979611,2397961, 4130839, 4430620, 4485487, 4510467, EP0291826B1, EP0293828B1, 2193801, 2276863, 2784311,

3375453 , 3430151,

3719903, a  novel dsb modulator ,

 

back to home page                                                          Hit Counter