This is a power supply for a 6 channel power amp I've put together. The transformer is a 1200VA toroid, rectified and filtered puts out about +/-55Vdc. As can be seen on the scope, I'm getting spikes on the DC output at 120Hz, and I'm hoping someone will help me get rid of them. Although you can't tell in the picture below, the height of the spikes is asymmetrical, every other one being 50mV, then 20mV, then 50, etc.
Attachments
I took the design from the National datasheet pertaining to amplifier power supplies, AN-1849. Here is the schematic. Notice I've drawn on it a little - there are two different 'grounds' - the two blue circles denote what is connected to the third pin of the power cord. The green circles denote everything else, which is referenced to ground via the preamps ground connection (not shown, but is explained in the app note). Notice, I had to draw on the ground symbol for the stuff after the rectifier!!! If you trace the layout in the app note, it is there. Hmmm...
Attachments
Here is the layout for my board. Pretty simple, and follows the schematic exactly. The reservoir caps are mounted off the board, and connect to the pads between the large resistors. Two other caps aren't shown - like C8 and C11 (in the layout), there are two .1uF 250V caps across the legs of the diode bridge not shown, for a total of four, one across each leg.
I'm guessing there is a grounding problem, but I figured the HF spikes should have been killed by the caps across the legs and the caps from the legs to ground right under the bridge. Is it possible that the bridge rectifier it self is at fault - aka, its putting out unusually large amounts of hash? It is a 1000V, 35A standard recovery bridge from digikey.
PS - the noise can be heard on the output of a speaker, like a quiet HF chopping.
I'm guessing there is a grounding problem, but I figured the HF spikes should have been killed by the caps across the legs and the caps from the legs to ground right under the bridge. Is it possible that the bridge rectifier it self is at fault - aka, its putting out unusually large amounts of hash? It is a 1000V, 35A standard recovery bridge from digikey.
PS - the noise can be heard on the output of a speaker, like a quiet HF chopping.
Attachments
This is a problem. The larger the distance to the capacitors, the higher the resistance/inductance in series with them. This will cause spikes as the high charging currents flow into them. Treat the common connection point between the two capacitors as a star ground. Similarly, treat the other ends of the capacitors as star connection points for the rails.cuibono said:
Mr. Evil, right you were!
I redid the power supply, tossing the board and re-doing it point to point between the reservoir 12 gauge 'rails', and indeed the HF spikes are basically gone. I might add that using 1oz copper for the board, and having the tracks only .15" wide was probably a big part of it too.
Thanks!
I redid the power supply, tossing the board and re-doing it point to point between the reservoir 12 gauge 'rails', and indeed the HF spikes are basically gone. I might add that using 1oz copper for the board, and having the tracks only .15" wide was probably a big part of it too.
Thanks!
Glad to see it sorted out. This is why you see many DIY PSUs where the reservoir capacitors are bolted together via massive plates of copper. That's overkill, but the principle is the same.
There are a lot of other unexpected surprises with PSUs, and indeed all circuits, that can be fixed by considering where the current flows, and by remembering that there are more components in a real circuit than appear on the schematic (resistance and inductance of wires, capacitance between PCB traces etc).
There are a lot of other unexpected surprises with PSUs, and indeed all circuits, that can be fixed by considering where the current flows, and by remembering that there are more components in a real circuit than appear on the schematic (resistance and inductance of wires, capacitance between PCB traces etc).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.