Noise on Audio/psu

Started by Robert Leacroft, September 01, 2014, 09:43:49 pm

Previous topic - Next topic

Robert Leacroft

Hi all,

I am using some Limes (Debian) for high quality audio over IP transmition and reception and they are working very well with an "audio cape" I have build to access the audio pins etc. and provide balanced audio in and out.   With suitable filtering and playing around with the ground connections I can get quite clean audio output but I do have a problem of a repetitive noise pulse running at about 3Hz. I have traced this to pulsing on the 5V rail regardless of what psu I use (including a decent laboratory supply). These pulses on the 5V are only in the order of 30mV but that translates to audio pulses around -50dB that I would like to get rid of if at all possible.

Does anyone know what might be running in the background on the processor, such as a watchdog or similar, that would increase the  load (hence pulsing the voltage) repetitively every 300mS. Any thoughts would be much appreciated.



This coul be some background kernel process (300ms is too fast for a watchdog). Try looking into /proc/interrupts or try someting like LatencyTop to see what is waking the CPU.

PS: Have you tried adding a low-ESR capacitor to 5V rail?

Robert Leacroft


Thanks for your comments.  I did a watch on interrupts and the only things clocking up lots of interrupts are sunxi lcd0 (even though it is disabled in script.bin); sunxi scaler0 and aw_clock_event  but all of these seem to be running a lot faster than my 300mS repetition rate. So far I have not been able to find anything else running at that speed.

I have also tried a range of Cs across various point on the power rails which have some effect but not enough.

I have also been playing around with the PSU and earthing/common reference point.  Connecting my audio earth reference to different points on the Lime card makes  phenomenal differences on the overall noise levels and especially to these pulses.  If I connect my reference to the 0V point at the psu connector the noise levels are around -50dB. Connecting to the labelled GND pad (next to one of the rounded cutouts) is not bad but if I make a connection to one of the solder tabs holding the HDMI connector case in position then the audio noise drops below -80dB which is very good indeed.  There are obviously some impedances in the GND tracks within the card that add or subtract to the noise.  Running on two separate supplies (one for Lime and one for my audio board) with the only common pint being the HDMI case makes it all perfect. Not the best solution with a flying wire soldered on to the Lime but as my audio card is supplied by an isolated + - 12V power converter with luck I may be able to rejig my pcb and its 0V to suit and perhaps in the end reference to one of the GND pins on the GPIO headers. Interestingly the pin labelled Analogue GND  is just connected by a track directly to GND.

At present I am only working in mono and it is a great shame that you cannot invert one of the audio outputs so that I could get a differential output from the codec that would overcome all these problems. Reading the codec data sheet you can set up a differential input but for some reason known only to Allwinner not an output.

If I find an eventual reason for the pulses I will let you know.