Spread Spectrum: FM Modulation or Dithering to Reduce Harmonics?

[dnyberg2]

So, can one achieve similar SS effects by simply FM modulating the offending source or does it have to be some sort of complex dithering scenario? The problem is harmonic emissions that so far do not respond to any other treatment such as shielding or absorption. The design is generating the stuff by the buckets and spewing it out the spigot. To root of the evil is a nasty fast rising square wave... required for high efficiency.

Thanks!

 

[the_emi_guy]

So, can one achieve similar SS effects by simply FM modulating the offending source or does it have to be some sort of complex dithering scenario? The problem is harmonic emissions that so far do not respond to any other treatment such as shielding or absorption. The design is generating the stuff by the buckets and spewing it out the spigot. To root of the evil is a nasty fast rising square wave... required for high efficiency.

Thanks!

The problem is harmonic emissions

At what frequency?

The design is generating the stuff by the buckets and spewing it out the spigot.

How far over your limit line? Class A or class B?

To root of the evil is a nasty fast rising square wave... required for high efficiency.

What do you mean by high efficiency and why do you need fast edges to achieve it?
Can you describe the transmission channel of this square wave? For example, is it going out onto a cable?

 

[dnyberg2]

Well, without divulging any secrets, the fundamental is about 50 MHz but this thing has issues past 2GHz. The ASIC that produces the energy is first then a BPF then a coax Conn to the next stage that could be up to a yard away. Even the shortest coax cable at 6" causes the thing to radiate like a banshee. There is a common mode choke that does a heck of a job without costing too much battery life but it has its limits due to everything is so small SMT. No coax - test below limits. Plug in coax, over class B by as much as 20dBuV/m! Help! I did try a ferrite cable clamp the other day and that helped a lot. Shield cans cause more problems than help...

 

[dnyberg2]

By the way, square waves for FET switching efficiency.

 

[the_emi_guy]

Well, without divulging any secrets, the fundamental is about 50 MHz but this thing has issues past 2GHz.

Is it failing at 50MHz? 2GHz? Both? How far past 2GHz?

There is a common mode choke that does a heck of a job without costing too much battery life

Why would common mode choke effect battery life?

over class B by as much as 20dBuV/m! Help!

I'll assume you mean over by 20dB.

I did try a ferrite cable clamp the other day and that helped a lot.

How much did it help? How far over are you with the cable clamp ferrite, and at what freqs?

Try modifying the receiving end as follows: disconnect whatever this clock is driving (FET?) and replace with 0402 50 ohm termination to gnd (good low inductance connection, I'm assuming you have gnd planes). See how this compares. The idea here is to determine if the clock transport across the cable is your problem, or if there is also a problem with the FET circuit.

Are you sure you actually need ghz harmonics to get this FET to work efficiently? Consider bandlimiting the clock before launching it onto cable. If you really do need to transport this clock with ghz harmonics, then make sure the coaxial connectors at each end have good transfer impedance specs at ghz. This means at least SMA type (forget BNCs). Can you bandlimit the clock to, say, 4ns rise/fall times before launch, then use a buffer on far end assembly to square it back up before driving FET?

 

[dnyberg2]

The beast only fails from about 300 MHz to 700 MHz. The common mode choke is tiny hence you can't pump a bunch of power through it before it gets hot and begins to sat. Past 1 GHz it doesn't matter because of the limits... The clamp on choke made the lowest power setting almost pass. Only several dB over 37dBuV/m @ the offending range. The output energy from FETS in ASIC go through a BPF and it does as good as it can but too much bad harmonics come out the coax con. There is no amp downstream just the target load. And yes, a dummy load passes all limits by many dB...

 

[Antiphon]

Your problem is you aren't filtering at all. Your common mode choke is no good by 700mhz. You need a completely different set of filtering components meant for sub-ghz rf. And who did your PCB layout? It sounds like it stinks.

 

[dnyberg2]

Well... I thought the same thing but ADS swears the filter is working and the filter comps all by themselves on a similar PCB stack and routing work just great on a VNA. And, the harmonics are down by what the filter should be doing, but I suspect you're right, I suspect conducted is getting by the filter...

 

[the_emi_guy]

The common mode choke is tiny hence you can't pump a bunch of power through it before it gets hot and begins to sat.

Common mode choke will only saturate with large common mode current. The only common mode current will be your unwanted EMI, which is probably in the uA range and should not saturate it. If you are experiencing it literally heating up than it is probably not arranged as common mode, how many pins does it have?

Past 1 GHz it doesn't matter because of the limits.

Are you saying this because you are only seeking CE mark and not FCC? Please be as specific with your inputs as possible, you're getting free consulting here.

The output energy from FETS in ASIC go through a BPF and it does as good as it can but too much bad harmonics come out the coax con.

You are saying that you would like to filter these harmonics before the connector? (Your previous post seemed to suggest that you *needed* these harmonics to go across the cable). If this is the case then just enhance your filter. It is very easy to filter out 300-700MHz with 50MHz fundamental. This is 6-14th harmonics. 2nd order Bessel filter will create nice symmetric rising and falling edges and give you lots of dB attenuation of your unwanted harmonics.

 

[dnyberg2]

the_emi_guy, please check your private msg's... Thanks.