FM demodulator using a 4046 PLL

[schmitt trigger]

I'm designing -as a hobby project- some IR headphones.

When I design anything, I divide the circuit into functional blocks, it allows me to test and debug each independently, before integrating the system.

The block I was testing is the classic demodulator circuit using a PLL, and a plain vainilla IC amp, as shown in the image. This schematic is essentially identical to many examples that appear on the web and app notes.

Since there is a significant source of IR noise at around 40 to 70 Khz, I decided to go for a higher center frequency of 120 Khz.

I assembled the circuit on a protoboard, and to test it I used an Agilent signal generator that can generate FM. The circuit works very well, and I can get a demodulated audio output of 100 mV RMS with 1% distortion to drive the LM386 audio amp, which then drives some 32 ohm headphones.

So far so good......the problem is that the circuit has a significant amount of white noise, enough to be very irritating.
Even if I feed an unmodulated carrier, the noise persists. I also figured out that the noise is being generated by the PLL, because if I inhibit it, the output from the LM386 is completely quiet. And before you ask, yes, the PLL is locked to the input signal.

I'm using a well regulated +6v supply, and the ceramic caps are all NPOs and the resistor are film types. As you can see from the schematic, I'm using proper decoupling too.

My question is: has anyone built a similar circuit? Have you experienced noise issues too?

 

[JimB]

I have never done anything quite like this, but have had head scratching experience of a different PLL in a frequency synthesiser which tured out to be a leaky varicap diode.
Yes I know you do not have a varicap, but the thought occurs that if there is something a bit off with the VCO resistors or capacitors, or there is a leakage path on the circuit board, that could cause a problem.

Looking at the TI application note for the 4046:
https://www.ti.com/lit/an/scha002a/scha002a.pdf
The tables of components for the VCO infer to me that you C is too low and your R is too high.
Try changing them.

The other thing of course, Agilent make good stuff, but are you sure that the noise is not coming from your signal source?

JimB

 

[schmitt trigger]

Jim: Thanks for your prompt reply.

1) Ooops, I did not update in the schematic that I'm using a 74HC4046. The timing values are different from that of a "normal" CD4046. I employed it because at lower voltages, it runs faster and with better VCO linearity. But I swapped it with a normal CD4046 and its associated resistor/capacitor changes, and the problem persists.

2) I suspected the signal source too. So I used another signal generator, plain vanilla, no modulation. The problem persists.

3) I'm also suspecting a leaky component. I've ordered some replacements.

 

[moffy]

The noise is coming from the loop filter from the mixer into the VCO. If you look at it with a scope you'll see lots of harmonics all modulating the VCO. You need to filter them out while maintaining your capture range. Not sure how as decreasing the filter loop BW reduces capture range. Could you increase the VCO frequency to say 500kHz? It will move modulation products further out of the audio range.

 

[schmitt trigger]

You are right, there is high frequency noise in the VCO's control voltage.
Cannot reduce the loop bandwidth too much, I'm attempting to demodulate audio.

On the higher VCO frequency, your idea appears to have merit and is worthwhile to explore further.
I've already moved the VCO frequency, but it was only about 25% higher. I'll double it up this evening, and hear the results.

 

[moffy]

Yeah, increasing the VCO center frequency seems the best. In such a case you can pre filter the higher components down without effecting your loop filter. Say prefilter with 5 to 10x the BW of the loop filter.

 

[schmitt trigger]

Thanks for everyone who replied...... I've solved the problem.

Turns out that the ground where I was returning the audio filter cap was noisy. Moving the cap position a pair of mm away diminished the noise level to a tolerable level.

I've been doing analog design for many years, and are very cognizant of proper grounding, but this still showed me a lesson. I had originally returned the cap's ground to the LM386 ground pin. When I returned it closer to the 4046's ground pin, the noise abated.
Upon a little reflection, I understand why: all the noise current should be returned to the source, which in this case is the 4046.

I hope this experience is helpful to other forum readers.

 

[moffy]

Glad to hear.

 

[ramuna]

Hi,
Its not just the 4046 which requires adequate bypassing of the power rails, but all CMOS logic ICs. They will oscillate at the slightest provocation, producing mysterious & often intermittent faults . Being stingy with bypass capacitors when laying out CMOS logic circuitry is often false economy, given the time & effort expended in tracking down and correcting such faults.

Edit: I just checked out your loop filter and this is what I found: I used a Frequency Lock range of 0 to 240 KHz with a centre frequency of 120 KHz, in the middle of the lock range.
From the 4046 datasheet, Tau = F_lock / (2 x pi x F_centre ^2) = 0.000002653 =R3 x C4.
Using your circuit's R3 & C4 I get 0.000033. Try changing R3 to 2k7 and see what you get.