I have my mbed on a pcb layout with good analog and digital segregation and layout techniques. The pcb has a digital ground plane under the mbed and it is powered by 4 x NiMH cells giving good low noise stable supplies.

The noise is about the same as other posts at 100mV, but doing and FFT shows the noise to be 400kHz. I'm not using I2C (uses 400kHz)or indeed any other logic outputs, so does this give a clue to the noise source?

My Analogue out is 600mV in this test and I have 100pF across the output as recommended in other posts.

So any comments, feedback, links appreciated...

I have found issues with the power supply noise on the mbed. I have reason to belive this comes from the conections between the pcb and mbed, made worse by sensors periodicaly switching on and off. check what your code is doing on a 400kHz cycle, if it draws a lump of current to do this, you have found your noise source.

The fix I applied was to treat Vin, Vout and GND as a L78xx series voltage regulator. decoupling capacitors of 100nF between Vout and GND and 330nF between Vin and GND. This cleared a fairly repetitive noise signal on the ADC in my project. I will drag out the files and find out what the frequency was, but it *was* related to power drops caused by sensors switching on.

Christopher,

Many thanks for these details. Yesterday I was investigating the power supply noise and found the 3.3V had relatively large narrow spikes present, but Vin had significant dips in the supply volts even though I was using 4 x NiMH cells. I shall follow your advice and see what improvements I can achieve.

Some more info..... It would appear that my noise is some kind of internal mbed counter, not I believe anything to do with my code, but open to correction!

It seems to be a noise pulse every 1mS and a longer burst every 10mS, hence counting?

My code at this point is a simple single output of the DAC voltage and then an infinite loop (no checks).

{.....

signal = 0.68; Voltage step

lcd.locate(0,0); lcd.printf("Finished");

while (1) { } }

End of test routine

Any further thoughts what might be generating the spikes? I'll investigate an improved 0V to 0v side of 100pF.....

Hi All,

Update. DAC output noise at 1mS interval appears to be proportional to the value used to set the DAC output voltage. i.e. get zero amplitude spikes when s=0 and max amplitude spikes at s=1.0.

Looks like the a reference voltage plus a noise spike are multiplied by 's' to get the output voltage?

The 10mS noise seems to be present continuously and at constant amplitude.

I'm off to try ferrite beads on the output and more hf 3V3 supply decoupling, but interested to hear from anyone.....

Progress!

Need to do more testing, but appear to have removed most if not all of the noise! Cure was to use a 4S2 grade ferrite bead Farnell 245-500 on the output of the DAC and increase the Cap across the output from previous 100pF to 100nF (just one I had available). I need to check the DAC response time, but it seems to work very well. (bead and 100pF did not help).

Strangely this removes the 30mV input supply noise that was occuring in synchronism with the DAC noise, so can only assume it is some kind of instability which is stabilised by the load capacitor? Any thoughts? I didn't try without the bead, i.e. just the capacitor as the pair should form a nice LPF.