Two sync detectors

[w3jn]

Over the past couple weekends I've been homebrewing a couple of sync detectors.  One is loosely based upon Chuck W3FJJ's design here:  http://www.eecis.udel.edu/~hanavin/chuck/w3fjj/synch/synch.htm  I used a pair of SA612 mixers for the I and Q detectors, another SA612 as a VCO running at 4X the 455 carrier freq, a Johnson counter to generate the LO phases, a 74HC4046 phase detector, and a ULN2111 FM IF chip as a limiter and carrier regenerator.  The I/Q output is fed to a 4-stage allpass filter, with better than a degree accuracy from 100 hz-8KHz, using the design software that came with the latest ARRL handbook.  This affords very good opposite sideband rejection - theoretical is >50 dB.  I also rigged up a method to disable the phase detector and run the VCO with a pot to use the thing as a simple product detector (albeit one with sideband filtering).  There are three lock speeds, switch selectable, and a pushbutton to speed up lock, as the slowest lock takes ~5-10 seconds to lock if the VCO is way off.  Output is 2 separate channels:  both LSB, both USB, both AM, or ISB (USB = left channel, LSB= right channel).   With a narrow RX bandwidth, tuning the BFO to the opposite sideband results in absolute silence, indicating the allpass filter and combiner are doing their things.

The other one, based upon the 1994 Communications Quarterly article (partial copy here:  http://user.netonecom.net/~swordman/Radio/graphics/comm_qtr_f94_sync-det.djvu ) was based upon the MC13022 C-QUAM chip, which is now hard to find.  However, I noticed an eBay deal for 25 MC13122s in a 28 pin SOIP for ten bux.  The MC13122 is seemingly very similar in function to the MC13022 and it works in this circuit.  I built it pretty basic, no notch filter nor lock indicator as in the article.  The MC13122's VCO runs at 8X the carrier and presumably there's an internal Johnson counter.  It appears the phase detector in the MC13122 detects audio phase of the detected output rather than carrier phase.  Its lock is extremely fast - imperceptible as you tune across a signal except for a slight "pop".  The MC13122 does this all by itself - no other active devices, just a few resistors, caps, and coils.  I haven't built an allpass filter for it yet.

Comparing these two with what I remember of my Sherwood, neither of these does a good job keeping lock on rapidly fading skywave signals.   When the carrier fades on a signal, but the sideband(s) are still there (hence an extremely high modulation index), you can tell the SA612 based detector starts flipping its phase, particularly if you're listening to it in ISB and the phase detector tends to unlock momentarily causing a "growl".  The MC13122 just distorts heavily on these fades, like a conventional AM detector.   I never noticed the Sherwood getting cornfused on deep fades - it just rode them right through.

Although the MC13122 makes a really cheap and easy sync detector (about $5 worth of parts including the chip), I'm not too impressed with its sync performance.  Even trying to feed voltage directly into the loop filter, the VCO sounds really rough and is unsuitable to use as a straight SSB product detector.  I'm thinking that the C-QUAM circuitry might be causing this - it has an internal divider to generate the reference for the C-QUAM 25 hz pilot, and perhaps this is FMing the VCO when it's unlocked.  The MC13122 does a much better job of keeping phase lock on a really weak signal; the SA612 detector's Achille's heel is the stupid ULN2111 limiter.  As the signal gets weaker, its comparator triggers on the leading edge of the input signal, thus causing phase shift that then cornfuses the 74HC4046.  This is also the reason it unlocks on deep carrier fades - the comparator then triggers on the modulation.  

Still, although the SA612 detector doesn't approach the Sherwood, who is pretty much unfazed by any nasty signal you throw at it, it does pretty well with the added bonus of USB/LSB/ISB capability.  And you can't beat the MC13122's price and simplicity - on a decent signal it sounds really good on either the quasi-sync envelope detector or the full-sync.

I really should figure out a better limiter that that ULN2111 - but that's all I had in stock here that would even marginally fit the bill.  Gonna try some more messing around with the MC13122.  Perhaps build up another oscillator and directly drive it for SSB, and build another allpass filter.  If anyone has any experience using the MC13122, I'd like to hear your thoughts.   In any event, the Sherwood is still tops in my book, but I think the SA612 is a keeper, and I'm gonna do up a cabinet for it as soon as I figure out the engraving function on our shop mill.

[WQ9E]

John,

Did you ever try building the one described by Hershberger from Popular Electronics 1982?  I have often wondered how that one performs.

Rodger

[WA1GFZ]

John,
I played with that mot chip in my early SDR days. Tom VU bought a bunch of boards and I modified the loop filter to slow it down which effected the lock time but sounded better. I fed I/Q into a MOT DSP56303 Eval board. Pete SOV did the software and I hacked it to work very well. Then soft Rock happened and Sound cards were pressed into service making life a lot easier using Power SDR type programs.
One thing I wanted to try was to use phase detector #2 on the 74HC4046
(XOR) then drive a VCXO. The advantage of the XOR is the oscillator sits in the center of the tuning range when the signal drops into the noise. Typical phase detector would go into a search mode or slam to the low end of the tuning range.
I would actually like to add this to the AM detector in the Racal to limit the lock range and reduce noise.

[w3jn]

Roger - no I haven't.  That detector uses a quite unique and novel mixer - just a 4052 analog switch IC.  I took a look at it and figured there were better/easier ways to do it.  The allpass filter could be adapted to any I/Q sync detector though.

Frank - I'm using PD2 in the 4046.  Actually PD1 is the XOR in the 4046; PD2 is

This is a positive edge-triggered phase and frequency detector. When the PLL is using this comparator, the loop is controlled by positive signal transitions and the duty
factors of SINin and COMPin are not important. PC2 comprises two D-type flip-flops, control-gating and a 3-state output stage.

The nice thing about PD2 is its tristate output.  It's an open circuit when things are cool, connects to the Vcc rail or shorts to ground when a correction is needed.  So an active loop filter isn't needed.  With no signal the integrator does indeed float about midrange so lock range is good and it doesn't end up sinking or rising to the rails when idle.  However I did find that even with a passive loop filter the loop WILL oscillate - I neglected a damping resistor in series with the integrator capacitor at first.  A couple K resistor cured that.  The chip doesn't like to drive a lot of capacitance directly so I have 100K and 1 meg in series from the PD2 output.  I used some unknown varactor I had that has about 30 pF or so of capacitance at about 2.5V, and putting a bit of series capacitance with the oscillator tank results in a tuning range of +/- 4KC which is about ideal.  I've noticed that the very wide capture range and high loop gain of the MC13122 results in it getting cornfused with adjacent heterodynes, and it'll sometimes try to lock on an unwanted signal.  That's not the case with the Sherwood nor the SA612.

The bad thing about using PD2 with the ULN2111 is, of course, its leading edge phase detection.

I tried tacking 1000 uF in the loop filter of the MC13122 (orig was 47 uF) and it slowed it down some, but not enough.  Gonna try some series resistance and see what happens.  It is nice being able to tune around with no heterodynes,  I suspect the phase detector and loop design in the MC13122 is quite different from the MC13022.  Would be neat to try a 13022 and see if that's true.

I forgot to mention that I'm driving these with a WJ 9040 with the WJ8626A HF receiver in the frame.  The WJ's existing AM detector is very, very good, but it does some lowpass filtering of the audio and using a 16KC IF, you can definitely notice the difference with the sync detectors on a station with decent audio.  The Greek ERT outlet on 666 KHz is very well engineered and they have some outstanding music - it's a real pleasure to listen.  At night when stations start rolling in from Europe and the Middle East, you can still use the 16KHz BW and just reject upper or lower, as needed, and/or tune off center. to cut the interference.  The SA612 sync detector is an outstanding QRM fighter.

[John K5PRO]

One of the older Analog Devices log amps provides a very good limiter that works to 80 dB and to 50 Mhz, the AD606.

[WA1GFZ]

Yup John phase detector #1. An XOR will generate a 50% duty cycle without one input. Tristate phase detectors had a problem and fell out of favor because the loop input is unterminated during the tristate condition. This caused drift so it didn't really buy you anything.
I found it was better to live with a 1/2 second lock time rather than distortion. It would be cool to have the loop filter change time constant after lock plus some time delay. Yes you want a series resistor but don't go too high or the loop will go unstable when it tries to chase its tail. I think you would be better off limiting the frequency swing of the VCO which is why I thought a VCXO would be better and cleaner.

[w3jn]

I had thought of the time constant shifting upon lock, but the problem turns out to be keeping the larger capacitor at the same voltage as the smaller one without it loading down the loop.  If they're not the same voltage you'll get a "yoop" when the larger one switches in, then the loop unlocks, and it'll oscillate between states.  I'm sure there's a way to do this with an op amp, but I've really noticed little difference in distortion between the different time constants.

Another reason you might not want to do this is during a deep fade the loop approaches unlock.  My LD LED starts blinking a bit and that would then switch the loop time contants just when you need the longest.

I disabled the "blend" control on the MC13122 and that got rid of the growl and hum in the Q channel.  Increasing the AGC time constant (it has an onboard AGC-controlled IF amp) improved things some, but it still goes to distortion on deep fades.  I suspect that the carrier is actually phase shifting via propagation... but I never noticed this with the Sherwood.

Finally, I've reduced the VCO sensitivity on the SA612-based unit to the point where 0-5V results in 36 KHz of LO travel (which translates to 9 KHz once divided down by the Johnson counter).  The VCO is pretty quiet; there's very little residual FM or noise on it, and I think limiting the VCO sensitivity helped quite a bit in this regard.

[Steve - K4HX]

"It would be cool to have the loop filter change time constant after lock plus some time delay."

This is exactly what Hershberger did in his design. He also soft-switched from a non-sync detector to the sync detector upon lock. This way you get no heterodyne and 'yoop' as the PLL acquires lock.

[w3jn]

You are correct sir!  I did notice that he swtiched the audio from envelope to sync during lock, but didn't notice the dual time constant PLL filter.  However it's not a good idear to drive a 100 uF capacitor directly from the output of that phase detector!

[Steve - K4HX]

I don't recall the details of his circuit. I have the article here. Let me know if you need it.

[w3jn]

I have it, Rodger I believe posted it.  It's a cool and novel circuit, but a bit long in the tooth for nowadays.  Example, it uses a LED coupled to a CDS photoresistor in the AGC to reduce the gain 

[WA1GFZ]

John,
just put a high value resistor across the switch that adds the extra cap. Then delay the loop filter change long enough to allow the cap to charge to the correct voltage. Also have a bit of delay to go back to fast time to handle fades.  9khz still sounds a bit wide. I was thinking 5 should be plenty. Fades are tuff to avoid a woop because the phase detector is trucking without 1 input.

[w3jn]

You obviously don't have QSOs with Frank AHE and his drift-o-matic VFO    I want the range a bit wide so I can tune a fair amount away from center and still have the VCO lock.

[K3ZS]

In the late 1970's I  developed and constructed a detector from a pair of NE561's and a few opamps.  They are phase-lock-loop chips with a separate mixer built in.   My circuit was used as either an I and Q demodulator or a tracking synchronous detector.  This was for a type of sonar system, but the IC's were good to around 1MHz.  I don't think you can get NE561's anymore though.

[WA1GFZ]

NE561's were hard to get in the '80s
I bet analog devices makes something that duplicates the function.

[The Slab Bacon]

You obviously don't have QSOs with Frank AHE and his drift-o-matic VFO    I want the range a bit wide so I can tune a fair amount away from center and still have the VCO lock.

I resemble that remark ! !     

Leo Meyerson at his finest! !  (Globe VFO Deluxe) !!

[W4AMV]

I have a question. Let the signal applied to the synch. demodulator PD not be limited or controlled by AGC, but instead vary in carrier level as fading occurs. Utilizing the appropriate PD, the PD gain will change with the carrier level input. This would automatically alter the loop BW. Narrowing the loop as desired as the signal fades. This is the principle used in FM detectors permitting threshold extension. This is counter intuitive in FM as limiting was always the big plus. Now the question, is this applicable to the synch. demod? If so, care is required in choosing the PD. Thanks, Alan

[WA1GFZ]

Racal does that with the MC1357 in their AM detector. The problem is the carrier gets a lot of phase noise as it drops down. They use the carrier from the 1357 to feed a MC1496 used as an AM demodulator. I've considered putting a crystal filter between them to limit the oscillator noise and lock bandwidth.

[WA1GFZ]

I have some later Cubic detector modules with a detector like the K2CU that sounds nice. The K2CU with aa automatic sync option would be pretty cool.

[W4AMV]

Understood. Thanks! I started some time ago building a simple unit for a 9 MHz IF. Using balanced mixers as the PD. They have the right characteristic and added a frequency discriminator sweep circuit to aid acquisition. The input signal to the PD came from an IF strip that had RSSI and the RSSI gave me the input needed to further adapt the loop bandwidth. The item is still on the shelf and this thread may motivate me.   

[w3jn]

I discovered a few critical things about the MC13122.

First, I had pretty much ignored the "blend" pin, thinking it had more to do with C-QUAM demodulation than anything.  The Icom R-75 (whose sync detector is notably poor) uses the MC13022 and they ground that pin.  Not a good thing, as that pin is a quality-of-signal point, comparing I and Q, 25 hz stereo pilot, and RSSI and controls a stop-sense circuit (for scanning the broadcast band) AND the voltage at that point is used internally to narrow the loop bandwidth dramatically on strong signals.  One of the things the chip detects is % negative (ie, below carrier level) in the I channel.  As it turns out this is apparently only active in the presence of the C-QUAM pilot, but it's easy to bias that pin high externally a bit so it automagically selects fast or slow PLL bandwidth.

Another thing I discovered is the I and Q channel outputs do not like low impedance outputs, nor do they like any capacitive coupling between the two.  I had been capacitively coupling a computer amp/speaker directly from them and this was upsetting their balance.  Driving an opamp buffer/amp/active filter cured that.

Finally I had inadvertantly used a .001 uF cap to ground on the oscillator input pin, instead of the called-for 50 pF.  This reduced the oscillator level causing non-optimum detection.

It performs adequately on deep fades, but I still think my Sherwood is better.

[WD5JKO]

The 13022 Sync detector issue has been batted around for a long time on the Yahoo R75 group. I am attaching one modification used.

Interesting discussion,
Jim
JKO 

[w3jn]

Did a bit more experimentation with the blend pin.  Biasing the blend pin with a 15K resistor to Vcc will force the chip into sync mode.  The time constant on the PLL appears to be about 100 ms to lock in this mode, but once locked, if you tune the receiver 1 KHz it takes a a second or more for the PLL to re-lock.  With no bias, tuning around results in no beat notes but a bit of clicking as you tune across carriers, indicating that the PLL is locking very fast. 

With the 15K bias resistor and tuned to a very noisy CHU on 14.670, I could copy CHU enough to understand the voice announcements while the regular AM detector in the WJ9040 yielded no intelligible audio.  Removing the 15K bias resistor resulted in the detector repeatedly locking/unlocking on noise and fades.  I suspect at least a 3-6 dB improvement in S/N with the MC13122 on these weak signals, and it does pretty good on multipath-distorted fading signals also.

The R-75 group needs to take the blend pin into account, rather than just using it as a lock detector.  I noticed a transition between .7V on the blend pin and .75V with regard to the PLL lock time - that seems to be where the slow lock takes over.  Of course Icom just grounded the blend pin altogether, which is essentially crippling this very capable chip.

I built another copy that fits on a 2"X2" SMD proto board, with room left over for the 90 degree phase shifters.  I'm gonna hafta use regular DIPs for the op amps as I don't have any SMD chip pads left on the proto board. 

One last thing, this chip is REALLY sensitive to PS hum, which manifests itself as hum in the Q channel.   A 150 ohm resistor in series with the Vcc line, and 220 uF tantalum bypass at the chip cured that.  Bypassing the chip without the resistor does not result in any hum improvement; running it on a battery results in no hum at all, so it's definitely PS hum (or a ground loop between the PS and the radio).