Great AM demodulator chip

[IN3IEX]

http://www.analog.com/static/imported-files/data_sheets/AD8361.pdf

Works perfectly at "ANY" IF and almost no distortion. Forget complex OPAMP/diode circuits or synchronous demodulators.
I suggest CFLTR = 10 nF and input and output coupling by 100nF capacitors. Read the data sheet. Input resistance is 225 ohm, see pag 15.

Now in my Drake R4B. Connected to T10 output through a 2.2K resistor and connected to the base of Q5 by a 10k resistor, remove C160 and C179.
Got a SOT-32 free sample from AD. Built manhattan style on top of six 1pF chip capacitors on 20x20 mm ground plane. 5V from a 78L05 with 12V power from the calibrator rectifier.
This thing is much better than the original circuit.
A theory paper is here: http://www.ing.unitn.it/~fontana/AmDetCompare.pdf
73

Giorgio

[W0BTU]

http://www.analog.com/static/imported-files/data_sheets/AD8361.pdf

Works perfectly at "ANY" IF and almost no distortion. Forget complex OPAMP/diode circuits or synchronous demodulators.

How could the performance of this IC be anywhere close to a good synchronous detector for AM?

[w3jn]

Under selective fading conditions, it won't.  It's essentially a square-law detector.

The problem with run-of-the-mill AM detectors is they fall apart at high modulation indexes and produce copious amounts of distortion.  This one is really nice and simple, and sounds like it works great.    I've tried the precision op-amp detectors and wasn't too impressed - an old infinite impedance detector does as well, or better.  If you can manage a SMD chip this one sounds like a winner.  Thanks for bringing it to our attention!

However the sync detector will be markedly superior under the afore-mentioned selective fading conditions, and if the sync detector takes advantage of an I/Q output with attendant allpass filters and matrix combiner, you have the decided advantage of being able to select sidebands, or have ISB outputs.  The MC13122 I'm currently messing around with works exceedingly well and you need only a dozen or so external parts for a functional sync detector (mostly .001, 20 uf and 47 uF capacitors).  A couple dozen more if you want to select sidebands.

[IN3IEX]

The MC13122 I'm currently messing around with works exceedingly well and you need only a dozen or so external parts for a functional sync detector (mostly .001, 20 uf and 47 uF capacitors).  A couple dozen more if you want to select sidebands.

Any schematic with the MC13122?

[w3jn]

Datasheet is here:  http://pdf1.alldatasheet.com/datasheet-pdf/view/162232/MOTOROLA/MC13122.html

Attached is a copy of the schematic.  To use the MC13122 as a 455 KHz sync detector, C4 should be ~47 uF (not critical, it sets the AGC time constant)

Don't connect anything to pins 7-15.

R26, C18 and C31 are deleted.  Connect a 10 uF or so capacitor to ground at pin 18.

Don't connect anything to pin 21.  R20 and D1 are deleted.

C23 should be ~47 uF.  Q2 and R11 are deleted.  Bias pin 23 to Vcc thru a 22K resistor, in series with a switch.  This will allow a fast/slow lock.  

Replace X1 with a 5-15 uH variable inductor, .1 uF capacitors in series with both leads.  The VCO range is wide enough that you shouldn't have to pad it with a capacitor, but if it won't tune to 3.64 MHz (455*8 ) try 20-100 pF in parallel with the coil.

The Q output amplitude is twice the I amplitude, so if you're going to connect this to allpass filter network you need to take that into account.  Don't connect anything < 10K impedance to the I or Q outputs or it'll mess with the PLL locking.  The L-R output is the Q output that's level changed coincident with the voltage on the blend pin.  Don't use that output, but it does need to be bypassed with the .001 cap.

Pin 1 is a quasi-synchronous output, and is <1% distortion on its own.  It doesn't do well with selective fading signals, though.  If you just want to use the pin 1 output, I don't think the chip needs anything connected to pins 15-21, pin 23, and pins 25-28.

L1 can be any old 1 mH choke.  It's not critical in the least, and I deleted C2.  R10 however sets the AGC response and shouldn't be deleted.

I found that the Q output had a considerable 50 hZ component riding on it.  I put a 1K resistor in series with Vcc and bypassed pin 25 right to ground with a 220 uF tantalum capacitor, and I increased the supply voltage to 10V to compensate for the voltage drop in the resistor.  The 50 hZ wasn't present when I ran it on a battery so that may or may not be necessary.  

Because the MC13122 has its own AGC it can take a very wide range of input voltage.  Adjustment is simple - just adjust the VCO coil for 3.64 MHz with no signal input.  If you're connecting it to a tube radio you might need to knock down the input signal a bit.

More discussion on this chip here:  http://amfone.net/Amforum/index.php?topic=28843.0

[IN3IEX]

I think that synchronous demodulators alone are not enough, in fact the fading affects the sidebands and the carrier. All of them affect the AGC. Maybe it is much better to isolate a sideband with a very selective filter that excludes the carrier, apply an AGC on the sideband and then demodulate synchronously. It is possible to emulate the various conditions with the R4B by using the notch filter to suppress the carrier in SSB mode. Suppressing the carrier improves the operation of the AGC. Maybe the Sherwood is capable to apply separate AGC to the sidebands and the carrier or a similar trick. (!!) I need to think about that.

[N4LTA]

Looks like a good device. I tried to sample a couple but the AD site is acting funny and I couldn't log in. I'll order some samples later and make a sm board.

[w3jn]

I think that synchronous demodulators alone are not enough, in fact the fading affects the sidebands and the carrier. All of them affect the AGC. Maybe it is much better to isolate a sideband with a very selective filter that excludes the carrier, apply an AGC on the sideband and then demodulate synchronously. It is possible to emulate the various conditions with the R4B by using the notch filter to suppress the carrier in SSB mode. Suppressing the carrier improves the operation of the AGC. Maybe the Sherwood is capable to apply separate AGC to the sidebands and the carrier or a similar trick. (!!) I need to think about that.

What is important to remember is that the VCO in the PLL provides the carrier, and if the time constant is long enough, it'll ride through any deep carrier fades or ionospheric-induced phase flips.  The carrier can be down quite a bit, I'd estimate 30-40 dB down from the sidebands and the MC13122 will stay in lock - if you bias the blend pin up as I've suggested.

This circuit can really dig an AM signal out of the noise.   I've connected it to an allpass network and the selectable sideband feature is very nice.  It doesn't work as well with the MC13122 as it did with my other sync detector, though.  I think the MC13122 has quite a bit more phase noise.  I coupled a 3.64 BFO to it so I could use it as a product detector for SSB, CW, etc., and tuning the opposite sideband still lets some noise through, whereas my other sync detector was completely silent (ie, tune the BFO for USB but select LSB with the allpass network).

Also, as you state, the AGC in the receiver and the AGC in the MC13122 act on both sidebands and the carrier - or any QRM that's present in the receiver passband.  If you're listening to a LSB signal and there's QRM in the USB portion of the passband, you won'[t hear it thru the allpass filter network but you can tell it's there due to the AGC acting on it.  No way to avoid it other than to narrow down the RX passband.  Still, it allows copy of absolutely unreadable signals - judicious tuning and usage of the receiver filters will allow the MC13122 to properly lock on a signal that has another carrier 400 hz away, and the allpass network almost completely eliminates that QRM.  Q5 copy when previously the signal was unreadable due to the close-in QRM.

[k4kyv]

It could also be used to measure true average (mean) power of a complex modulated signal.  A "wattmeter" like the Bird 43, with the simple diode rectifier, reads "average" power based on average rf voltage, which is a meaningless measurement for anything other that a steady unmodulated sine wave. Average, or mean power, by definition, is an Ohm's Law function of RMS voltage, not average voltage.

A Bird 43 in "average power" mode will indicate less than the true average power of a SSB or DSB suppressed carrier signal.  It will show no increase under modulation with a full carrier AM signal, even  though we all know that at 100% modulation with a sine wave, the average power of an AM signal increases by 50%. This is useful for measuring the carrier power output of an AM transmitter, since the reading is not muddled by the modulation.

One of the reasons the FeeCee gave for the p.e.p. bulls##t instead of an average power output standard was that at the time, there were no true average-reading rf power meters based on Vrms or Irms readily available, other than those using a thermocouple rf ammeter, and those are too sluggish for modes like slopbucket.

To-day Bird has the APM-16 that reads true average power, but it is pricey.  They are undoubtedly based on one of those chips or something similar, basically a square-law device. Shouldn't be much more difficult to homebrew an average-reading power meter using the chip, than a typical Hammy Hambone pseudo-wattmeter, that uses a simple diode rectifier.

Under selective fading conditions, it won't.  It's essentially a square-law detector.

To  clear up one possibility of confusion, this is not the same thing as the "square law detector" found in early AM receivers. Those used a grid-leak detector that produced an audio output waveform that is a function of the square of the rf envelope voltage, thus producing severe distortion at high percentages of modulation. Ironically, the square law detector in the old fashioned sense pre-distorts the signal in the exact opposite fashion as "AM"  with one sideband, so theoretically, an old-fashioned square-law detector should reproduce the detected audio from SSB + carrier without distortion, impossible with a regular envelope detector.  The square-law function of the chip under discussion allows the detector to produce instantaneous DC output as a function of the true RMS voltage of each rf cycle of the rf signal that generates the RF envelope, while the audio output waveform from the detector remains a linear function of the envelope of the modulated envelope, not its square. In other words, it works in the real world like a hypothetical ideal diode detector.

And one last word: THERE IS NO SUCH THING AS "RMS POWER".  Average power = rms voltage X rms current, NOT average voltage X average current.  The latter is a meaningless quantity.

[Steve - K4HX]

All that may be true but the current through the diode is not linear, it's exponential (squared or higher order).

The square-law function of the chip under discussion allows the detector to produce instantaneous DC output as a function of the true RMS voltage of each rf cycle of the rf signal that generates the RF envelope, while the audio output waveform from the detector remains a linear function of the envelope of the modulated envelope, not its square.

[N4LTA]

I got my samples and just finished a SMT board at lunch. It is a tiny chip with .025" pitch. I will use solder paste and a reheat gun  -  Looking forward to testing it as a demodulator.  My eyes are getting too old for this small stuff.

[John K5PRO]

I used to play with MC13022 and its predecessor the 13020, when I was at Broadcast Electronics and we were developing our own version of CQUAM encoders. Several of us in the Engineering Dept had modified radios to detect CQUAM with those chips. My best effort was an Alpine auto stereo that had been extensively modified to accept a 13022 with the filter network in thick film package. When I tossed the radio after the cassette deck died (it even had Dolby C), I saved the little hand made board I built. Someday gotta dig those out and try it for AM like you say.

[IN3IEX]

Look at this envelope demodulator:
http://www.rhichome.bnl.gov/AGS/Accel/Reports/Tech%20Notes/TN386.pdf
It seems that by accessing the FLTR pin of the AD8361, two chips can be combined for even better results.
If this is possible, an additional 90deg phase shifter at IF is what is necessary to build the ideal AM demodulator.
For the R4B a quadrature network similar to those used for phasing SSB could be built and capable of operating accurately between 45 and 55 kHz.
I think that a test could be made by connecting the FLTR pins of two AD8361 by a capacitor (1 uF ?), connect the inputs of the two chips to the phase network and combine the two outputs with two equal resistors (10k ?). We need the detailed design of the chip to know for sure.

The theory paper has been updated with the design of the phasing network: http://www.ing.unitn.it/~fontana/AmDetCompare.pdf

73 Giorgio

[Steve - K4HX]

Thanks for the info Giorgio. You simulations are quite revealing. The RMS detector appears to be quite superior to  a typical diode detector.

[WA1GFZ]

Yes, very interesting.
I wounder what is the limit of demodulated audio frequency range these RMS detectors can produce?

[IN3IEX]

Hi GFZ. The data sheet says:
"The AD8361’s internal 27 pF filter capacitor is connected in parallel with an internal resistance that varies with signal level from 2 kΩ for small signals to 500 Ω for large signals. The resulting low-pass corner frequency between 3 MHz and 12 MHz provides adequate filtering for all frequencies above 240 MHz (i.e., 10 times the frequency at the output of the squarer, which is twice the input (carrier) frequency)."

If we operate at 144 MHz and we demodulate directly the RF signal, the "audio" bandwidth can be as high as 3 MHz. A "video" bandwidth indeed.
In my prototype I put a 10nF capacitor in parallel to this 27pF capacitor, the resulting audio bandwidth is about 9 kHz. There is some 100 kHz ripple at the output because my IF is 50kHz. It is filtered by the audio chain of the R4B. I do not care too much about the ripple, it sounds so good.  

Many thanks to Analog Devices. They sell this kind of products almost only to mobile phone companies, BUT they have a couple of them that can go very low frequency. A ham radio guy in the design office?   All similar TI products can't go at IFs that are of our interest.

[IN3IEX]

By looking at the internal design of the AD8361, we see that it is possible to use it as an almost ideal rectifier.
This possibility is not documented, but should work.
The chip is a squarer -> filter -> square rooter. If we locate the filter at the output we have: squarer -> square rooter -> filter, that is mathematically equivalent to an "ideal rectifier" -> filter.
Just do not use any CFLTR, and filter the ripple at the output.
Because there is an internal CFLTR, I assume that the method will work only for IF less than about 1 MHz. Just fine for Drakes and Collins and many other rigs. This final  thought may also simplify the construction, that is good.

[IN3IEX]

Now I also have a receiver with a synchronous AM detector that includes sideband selection. I never use it because when conditions are worse it loses locking. Is is much better to switch to USB or LSB.