Hi,

I want to make a low noise 9V power supply for audio. I have an unused IRM-45-12ST. That's a noisy 12V power supply. Photos of the 12V output with a 1k load are attached. Triggering is difficult but the switching frequency is either ~300Hz of ~600Hz, not sure. There's ringing in the MHz range. Voltage noise is around 100mV to 150mV.

The plan is to use a pi-filter to get rid of the high frequency stuff and an L7809 to get rid of the low frequency stuff.

How would you go about sizing the pi-filter? Can I just make C1 and C2 equal? Apparently pi-filters can be used for impedance matching RF stuff. Should I care about the impedance for this application?

Thanks,
hrs

 

Hi,

I want to make a low noise 9V power supply for audio. I have an unused IRM-45-12ST. That's a noisy 12V power supply. Photos of the 12V output with a 1k load are attached. Triggering is difficult but the switching frequency is either ~300Hz of ~600Hz, not sure. There's ringing in the MHz range. Voltage noise is around 100mV to 150mV.

The plan is to use a pi-filter to get rid of the high frequency stuff and an L7809 to get rid of the low frequency stuff.

How would you go about sizing the pi-filter?

By intuition and guesswork.
If all the components were perfect, then the interference wouldn't have occurred in the first place.
Whilst it may be possible to calculate inductance and capacitance values, the self-inductance of the capacitors and intra-winding capacitance of the inductors will spoil it.
Start with some low impedance ceramic capacitors across the power supply, 100nF and 1nF in parallel (because the self-inductance of the 100nF will be too high).
Try a ferrite bead for L1.
Then some small ceramics for C2 to make up for the self-capacitance of C3.
Then get a better regulator! There are regulators with much better PSRR and noise than the L78 series.
Then when you measure it make sure that you use the little spring thing that comes with the scope probe for grounding right next to the tip, rather than using the usual clip lead, otherwise you might get misleading results.

 

Try a ferrite bead for L1.

There's ringing in the MHz range.

Many of the beads do not do much at 1mhz. They do well at 10mhz and above. I think you will need some more inductance than a bead can do. Pick a part for 1.5A, or at least 1A. The inductance drops as you approach the max current.

Here is an example of a bead. It adds little at 1mhz. It adds 600 ohms at 100mhz.

 

Thank you gentlemen, these are some interesting avenues I wouldn't have thought of.

The guy around the proverbial corner sells 1, 2 and 6 hole beads for pennies. I'll get some of those and experiment.