Penri wrote:Is there a load on the output of the filter?
It is connected to the Audio Amplifiers.
Penri wrote:Is the negative rail earthed? if not I suggest running a wire, the bigger the better, to earth on your domestic mains. The scope probe croc clip should be adequate although I've next checked it myself.
As installed, it is not earthed to anything because we aren't bringing mains into the case. We did have the grounds terminating at a central earth point on the case, but that was a nuisance from a maintainability viewpoint and appeared to make no difference to the noise levels (which were fairly low anyway back in April).
The scope does provide a good earth and I just tried grounding the 12 V DC -ve pin; unfortunately, this made the hum louder (but only slightly).
Penri wrote:From the look of the output trace the noise is spiky, high frequency stuff, unless there's a low frequency component that can't be seen at this timebase setting, and well outside what the human ear can detect.
Yes. It's in the kHz / MHz regions, although the spike repetition frequency is a bit lower.
Penri wrote:I'm trying to cast my mind back to what we heard from the bell tower speaker when in-situ, I think is was higher in pitch than 50Hz mains hum but still something I could hear and not too shrill, so sub 1kHz perhaps.
Can you see any noise, probably sinusoidal in nature, on the output of the amplifier?
I have an app on my phone called VisualAudio that uses the phone's microphone as an input device and gives you various display option, including a power spectrum graph, of what it "hears". Can you download such a thing onto your phone and see what the sound spectrum coming out of the speaker is? I've only recently downloaded if so wasn't available when we were listening to the noise. If you can get such a thing it may give us a clue of what the offending frequency(s) is/are.
I've installed an App from the Android Play Store called Spectroid (see Screenshot). This was taken with the Pi unpowered (no music), but it sounds exactly the same if the Pi is running with the music off. There's quite a lot mush from around 50 Hz up, but the peak is consistently a bit higher.
The O/P from the amplifier looks pretty identical to the input (and the rails).
On the trace, the red line appears to be the spectrum at start up; the yellow trace is the one that moves. The 1031 Hz peak is misleading, I've seen it as low as 700 Hz and as high as 3000 Hz; it just corresponds to the current peak. The one thing that I haven't been able to see anywhere is a larger spectral line that corresponds directly to the hum.
Penri wrote:By the way there's nothing wrong with the circuit, as far as I can see, my only comments are that the 220uF cap is probably overkill and properly earthing the negative rails is important.
OK. Unfortunately, the earthing seems to set up a hum loop, part of which I think is through the Pi.