Power supply noise, how can I get rid of it? 2

[xj25]

Hi guys,

I have a 300kHz RF amplifier working (see thread248-366918 for info), but my small TRACOPower +15/-15Vdc SMPS is coupling a terrible 2 MHz noise up to 3-4Vpeak all over the circuit nodes (from PS, input and output).
The schematic is something like: 220AC - 24Vdc SMPS Carlo Gavazzi - +15/-15dc SMPS TracoPower - RF amplifier

The noise dissapears when I use a "good" lab source. If I supply with the lab source to the TRACOPower +15/-15 source it cames back again, so I guess the main problem is the Traco, not the Carlo Gavazzi 24Vdc source.

I thought of being switching induced glitches and prepared a LC filter for both PS outputs (4.7mH 0,1uF) that has quite high attenuation at that noise freq.

My surprise is that with the filter at the output the noise remains excatly equal, as if there were any filter there.
Even sensing just with the GND clamp of the probe at the output I see this noise. This seems to me to be a emmited EMI problem.

Any idea or help will be welcomed cause don´t know how to follow on, and would want to be able to use my "portable" source (that for the power given was not cheap!).
Regards

 

[LiteYear]

Radiation rather than conduction sounds likely. Try a grounded metal shield to physically contain the PS.

 

[Comcokid]

When you add a filter to get rid of noise, and it doesn't work, then you need to suspect common-mode noise. Take your output (+15, -15, and ground - all three wires), and put them through a large ferrite bead. The large ferrite bead you sometimes find on a computer video or other computer cable might do. If that helps, then try two or three turns through the bead.

 

[xj25]

Ferrite bead worked quite well, it limits the noise to about 0,2-0,3 volts in the pulses.
I will try also a common mode inductor to see if I can get a even cleaner DC rail.

is this common practice to follow with COTS switched sources?
it surprises me a little that I bought a supply from a known supplier to avoid problems, and finish reparing its problems...

btw, some radiation also, seen the difference if I put the source 1m away, but nothing comparing the common mode glitches.

Thanks!

 

[Comcokid]

Common mode noise is a frequent issue with switch-mode power supplies. It generally orignates from the parasitc capacitance between the power switching devices (MOSFETS) and their heatsink. As the MOSFETS switch ON/OFF they inject current through this parasitic capacitance to the heatsink, which is usually connected to the power supply chassis. These noise currents want to 'close the loop' back to the source and will circulate to find a path - in your case probably back into the suppply through the +/-15V and ground lines.

You may need to experiment with adding a common mode filter to the 230VAC input of your supply to see if that also helps. If you keep the output ferrite bead then another thing to try may be to put a small capacitor on each of the +15, -15, and gnd wires to the chassis before the ferrite bead/choke you put on the output. Try values of 100pf to 2200pf. This will provide a easier path for the noise currents to circulate without traveling into your RF amplifer to find a path.

 

[hacksaw]

of course the ferrite is a good solution. a common trick at the higher frequencies is to take a non-inductive resistor(50-100 ohms)and wrap wire around it as an inductor. Basically accomplishes what a ferrite bead does in a smaller volume.
small rf by pass caps work too.

If you attempt to useL/C filters there is no attenuation, as you discovered.

 

[xj25]

With 2nf bypass caps to source case + 7 ferrite turns it is quite stable without the need of any LC output filter. It was somewhat difficult to solder the caps to the case, as tin doesn´t want to be adhered to the aluminiun case and no earth connection was included.

Other issues appearing now, may comeback soon
Thanks!