PLL question

The delay will cause phase shift. It's not a problem if the delay is constant: it just means the VCO edges won't coincide with reference edges - they'll be slightly offset.

A variable delay (e.g. due to interrupts, jitter on the CPU clock or concurrent processing) would cause phase modulation (and therefore frequency modulation) on the VCO.

If your microcontroller has a dedicated counter, which is clocked by the VCO, and which toggles its output without software intervention, then you should be fine.

I'm doing the same thing. I am using a 1 MHz reference oscillator (ECS-100AC), an MM74HC4046N, and a couple of counters in a Cypress microcontroller. I divide the 1 MHz down to 10 kHz. My 4046 VCO is running about 1.6 MHz and is divided by 160 for the 4046's phase comparator II. I only care about the frequency of the 1.6 MHz signal, not the phase.

The only problem I've had is phase noise (jitter). I have determined that the 1 MHz reference is rock steady. The 4046 VCO is very steady, but only if I disable the 1 MHz oscillator and the micro and drive the VCO frequency control with a fixed DC. When the loop is closed, the VCO control voltage is steady, indicating no loop oscillations. I conclude that there is 1 MHz noise or microcontroller noise getting into the 4046 somewhere (Vcc is bypassed at each chip).

It appears that I need to get the circuit off the breadboard (the type where one plugs the component into holes with spring contacts) and put it on a piece of copper ground plane and bypass the chips with short-leaded caps. I have been surprised to learn that even a half inch of component leads can cause problems even at this low frequency.

Anyway, be aware that you can have phase noise problems if you are not careful with your layout, bypassing, or filtering.

Good luck.

John

Hi John,

That's the ticket. I have never seen any RF stuff work well on these plug-in boards. Therefore, we do not have any of these boards in the lab. What I found to work really well are little islands of parts on copper clad. Wainwright strips are used a lot but cut up pieces of copper clad and some glue would also do. It doesn't look pretty and the more you touch the copper the uglier it becomes. But those kinds of prototypes can operate nicely into the GHz range.

Regards, Joerg

Hi, Joerg -

Yes, that's the way I normally do it when higher frequencies are involved. This time I was led astray by the 1.6 MHz maximum frequency. Stupid me. I wasn't thinking about the nanosecond rise and fall times of all three devices.

John