Using a spread spectrum oscillator with the PLL on the MCF52259

[drummer]

I am using MCF52259 in several projects and every time I bring the project to the emissions lab for FCC testing I get sneered at for 1) Not using multilayer PCB and 2) not using a spread spectrum oscillator.

The multi-layer PCB is a management issue and cannot be resolved.

The use of a spread spectrum oscillator is my only hope for a design that will have multiple frequencies and oscillators..

The SSO will change the frequency slightly (1 - 2.5%) over time so that a peak frequency will not stand out and thus pass emissions.

If I am using the PLL to increase the frequency (I require 80MHZ) will it not average out the oscillator frequency and ruin the effect of the SSO?

Has anyone out there used an oscillator?

What is the highest frequency I can use for my external oscillator?

[TomE]

> The multi-layer PCB is a management issue and cannot be resolved.

It will be resolved when you can never ship a working product and the company goes bankrupt and everybody loses their job. Easy.

Otherwise you need every square millimeter of both sides of the board covered with ground-plane and vias connecting all the ground areas on both sides together. Also filtering of all pins near the CPU.

How many dB are you failing by? If you're close (say 5dB) then you may be able to reduce the emissions by that much. If you're WAY over, then nothing will fix it except for good EM design, including multi-layer.

> The SSO will change the frequency slightly (1 - 2.5%) over time so that a peak frequency will not stand out and thus pass emissions.

Are you sure? We tried that on an MCF5329 and it didn't work as well as we thought.

On the quick "unofficial test" that looked for narrow emissions we could see the previous single spikes smeared out somewhat and turned into a "comb" of emissions.

The official emissions test doesn't look for very narrow peaks. It uses a bandpass filter. If the spread-spectrum "comb" is too narrow, all the energy can end up in the bandpass filter, resulting in the same result.

So you have to increase the spread spectrum modulation frequency, which is going to be limited by what the PLL can lock to.

If nobody else has done this, then you'll have to do all the tests yourself. Freescale don't give enough details of the PLL filter for you to model it, apart fomr the LOCK and UNLOCK frequency ranges. You'll be operating outside of the component specifications, so you'll have to reverse-engineer and then add your own safety margins.

> What is the highest frequency I can use for my external oscillator?

That information is in the Data Sheet.

Tom

[drummer]

Thanks Tom. You are always very helpful.

The ban on multilayer pcb's is the challenge I face every time I design a new product. My success is also my failure.

My current project worries me due to the complexities of multiple clock sources and multiple PCB's connected through flex ribbon cables.

I haven't brought it to the lab yet. I was just anticipating a plan B.

The highest oscillator frequency is 10MHZ so I guess the simple answer is no, its not possible to use spread spectrum oscillator on the MCF52259.

I have been successful in getting FCC approvals on every project submitted in the past 20 years but it isn't always easy.

Every project has its own characteristic noise problem that can't be predicted.

Once the plot is analyzed the source can be pinpointed and dealt with.

I discovered through experimentation that rf noise is sometimes injected into the ground plane and that its sometimes best to clear out ground areas underneath magnetic components (power supply and ethernet).

Power supply noise (switcher) can sometimes look like microprocessor noise due to ringing,

the frequencies can be coincidentally matching odd multiples of your clock frequency misleading you.

Unfortunately a spectrum analyzer and a Faraday isn't in the engineering budget so off I go to the lab with my prototypes.

[TomE]

> (I require 80MHZ)

If you can get away with 60MHz you can run directly with an external oscillator at that frequency.

An external 60MHz oscillator would come with its own EMC problems from the direct emissions from that signal.

> multiple PCB's connected through flex ribbon cables

Otherwise known as "Aerials". I hope you've only got low frequency signals on those cables, and with plenty of filtering on both ends.

> I guess the simple answer is no, its not possible to use spread spectrum oscillator on the MCF52259.

The simple answer is you don't know until you try it. The PLL will be able to keep up with a certain modulation frequency and deviation. The limiting factor would be expected to be the slew rate of the frequency change. That will either give you enough modulation to spread the spectrum enough to pass the test or it won't.

If your design is too many dB over the limit, then no single fix (like spread spectrum) will be enough.

We had excess emission from the TOP of a Freescale CPU. The RF seemed to be coming out of the top of the chip, related to the cache operation. We had to solder a big "lid" over the top of the CPU, RAM and Flash chips to fix this.

Tom