My 5V rail supply, a home build job using a 5V step-down regulator which is being fed of one of my 12V rails is starting to hit its limit at about 40W. If I start up an additional Raspberry Pi the thermal protection activates and the regulator shuts down. Bugger.
Time for a new supply. I have an old (>25 years) AT switch mode power supply (SMP) which had at one stage been put into service as a bench supply to supply +-5V and +-12V for projects. Since I have upgraded my bench supply this unit was sitting in a box ready to be ....
Anyway cutting a long story short I decided to use this SMP as the unit to supply the shack with 5V. Now this unit is not the best when it comes to EMI and so I rummaged through my store and found a few items to make the supply as noise free as possible.
Below is the design of the whole supply. This AT-SMP is a 200W unit. With a rating of 20A for the 5V rail (100W), about 60W more then the current supply. I'm now able to run a few more Raspberry Pi churning away in the shack.
All parts fitted nicely into the original case. Here are a couple of photos of the modification.
Here we see the filtering for the fan, a 220µF and a 0.1µF capacitor and a 10Ω resistor to slow the fan down a bit.
And the 5V output stage with the 135µH choke, the 2200µF capacitor and the 0.1µF capacitor to frame/chassis ground.
This is the little Schaffner FN610, a 10A single stage line filter. 10A is a bit overkill but is was the best size to fit into the case. Since the SMP already had a single stage input filter I hope that an additional single stage would be enough.
Next I started to setup the test bench to see if I managed to quieten the unit down.
This is a picture of the noise the SpecAn is picking up with only the mag-probe attached. Note the big spike in the middle at 148.7MHz is one of the local pagers. This pager is causing me endless grief. Working on a solution currently, though.
Note, the Power supply is switched off.
The above picture shows the probe attached to the power cord of the SMP.
The next picture is with the power supply switched on and the probe attached to the 5V rail.
Note: No load attached to the 5V rail.
This is the noise voltage as measure on the 5V rail with the mag-probe.
The next picture is with the probe attached to the 5V rail, SMP switched on and a load (LED globe) attached.
We can see that we have a bit of noise from about 13 MHz on wards. However, it is still less than 85 dB down. But we also can see that below 1 MHz noise has increased, so let's zoom in, to have a better look.
So from about 200 kHz we see a steady increase of noise to about -18dBm at 10kHz. So the VLF spectrum will be still a bit noisy. However, from 200 kHz to 1 kHz the noise is less than -80 dBm and from 1 kHz the noise is below -90 dBm. So I'm happy with the result.
And the noise voltage between the negative rail and the frame ground measure with a 10x probe.
Conclusion
The VLF/ELF noise might be reduced further if a FN610-3 would be use as this unit has 4.5 time more induction than the FN610-10. Reducing the FAN speed has quieten the wind noises so the power supply is also acoustically quiet. The next few days will show if it is as good as the old 5V regulator (or better).
And yes, "Hot Glue" keeps everything in place.
Appendix
No comments:
Post a Comment