# How does this demodulator exactly work?

I have the following circuit which is an AM demodulator (I think). RF input is between the diodes left:

Plotting the bode plots for the negative and positive input of the comparator, I get a lousy allpass for the positive input (with a small notch at 10kHz) and a multi-order lowpass for the negative input.

• What is R1+C6 for? If V+ should have the unfiltered input, why not directly connect RF?

• If V- should be lowpass filtered, why not choose the components such that we get actual 2nd order rolloff rather than having the poles so widely spaced (or the zero in between)

• At the input of the comparator is a high frequency signal which would toggle the bits at a high rate. Now I get it that the comparator is probably too slow for that but I feel it's odd relying on parasitic elements rather than actually creating the desired behavior.

• Does the circuit work for for low modulation index, i.e. when I encode bits such that 1 represents a full carrier wave and 0 represents 20% attenuated carrier wave?

I also simulate this system in MATLAB, creating a 1GHz RF signal with 10kHz baseband data, demodulating the RF signal with "rf*sign(rf)" and filtering with the two calculated filters and comparing the values. I cannot get meaningful results.

• Plotting the bode plots for the negative and positive input of the comparator Bode plots are only for linear circuits, a comparator isn't linear so making a Bode plot is meaningless. – Bimpelrekkie Nov 12 '19 at 7:45
• R2 C7 R3 do quite a bit more than JRE says. C7 charges to the signal mean DC value and set a ref point for the comparator to toggle around. R3 sets V_C7 means at about 90% of the mean DC value giving a positive bias to the data out when 0/1 ratio is constant. – Russell McMahon Nov 12 '19 at 7:48
• @Bimpelrekkie No it is not meaningless because the ideal comparator is the function "Vdiff = Vp-Vm > 0" with Vp=Hp(Vrf) and Vm=Hm(Vrf). – divB Nov 12 '19 at 14:36
• @RussellMcMahon You did not describe what R2 is good for. Can you elaborate how these values are picked? If I tell you for example the bitrate is 100kbit/s (each bit is 10us long), how would I pick R2, R3, C7? – divB Jul 16 '20 at 5:17
• @divB R2 & R3 combined divide Vin so that pinn 4 is about 90% of Vin and filtered to provide the DC component. The time constant of R2//R3 & C7 is longer than the average bit time - increasing this time makes pin 4 vary less with individual bits and short runs but makes the cct less responsive to changes in the mean DC level and slower to lockon initially. R3//R2 ~= 20k. Tc 20k x C7 ~= 20 uS. 100 kb/s = 5 us/bit - pin4 will track Vin reasonably rapidly - so if anything C7 could be slightly larger. – Russell McMahon Jul 17 '20 at 6:33

What you have there is a classic, simple data slicer.

D2 and $$\C_{p2}\$$ form an AM demodulator. What goes to C6 and R1 is no longer RF.

C6 and R1 form a sort of high pass filter/differentiator.

R2, C7, and R3 are there to slightly delay the edges of the detected low frequency signal.

U3 is a comparator. It compares the signal (V+) with the delayed signal (V-.)

When the signal is higher than the delayed signal (that's the rising edge,) the comparator's output snaps to high.

When the signal is lower than the delayed signal (that's the falling edge,) the comparator's output snaps to low.

The output is a series of pulses that follow the originally transmitted on-off keying from the transmitter.

This simple data slicer will output random "data" when there's no received signal. The link I posted goes into some detail on how that is avoided with better designs. It also includes a section on the simple data slicer, though.

• Do you have an explanation for why we need a differentiator (R1+C6)? It seems to me the circuit works much nicer and easier without. – divB Jul 16 '20 at 5:19
• The link you posted is very nice but those circuit seem to be much more simple & intuitive. Do you have any reference for this circuit? Can I find it in a textbook or so? – divB Jul 16 '20 at 5:20
• @See my comment from November re role of R2 R3 C7 of setting a decision DClevel slightly below the mean signal DC level. – Russell McMahon Jul 17 '20 at 6:34