Hello Forum,

Hope all the mBed community is doing great!

I am actually writing to comment and ask for any advised in a issue I am facing, I am using Cool Components BaseBoard to do a Audio over Ethernet module in a school project...

Now up to the final stage of the project once I am trying to test if everything work nice and trying to wrap it up to make the final box and presentation I face a Noise coming from the ADC or DAC not sure which one exactly (actually I can hear it at the output of the DAC due to the speaker that its place there) I AM SURE that the noise came from the interface of the Ethernet Jack of the baseboard, still not sure if its coming through the power 3,3V or the GND of the Ethernet Communication Link and its not allowing me to listen correctly to the audio under it....

This is because the high frequency of transmission of the Eth Link is sharing either the power supply line or the GND line of my Input and output ANALOGUE filters and thats making like a square signal noise at the output of the DAC.

Anyone face this before or probably have a clue in how to solve it?

I am pretty sure that if I split the DIGITAL and the ANALOGUE GNDs and connect the the nearest point possible to the mBed pin #1 this noise might go down a lot, but I have no idea in how to do that with this base board.

I would really appreciate any help from you guys!

Here is the link for the baseboard main page: http://mbed.org/cookbook/Cool-Components-Workshop-Board

Thanks a lot

Fernando Ficoseco Argentina

Hi Fernando,

I had similar problems with my own designed PCB and yes, I had to cut tracks and get seperate analog and digital grounds right down to the mbed gnd to minimise the problem.

The other issue I had was power supply noise and I decoupled with ceramics as close as possible around the mbed supplies.

The final problem I had was USB lead noise when plugged into my laptop. This was minimised by using a Dell earthed power block (not the normal 2 pin type) and also I found a large clamp on ferite on the USB lead helped a lot.

So the clamp on ferite on the USB and ethernet leads and indeed power supply (DC feed) may be the easiest things to try first. Also keep all analogue wires short and screened and away from power and comms leads.

Hello Kevin! Thanks a lot for your answer! I have just tried out your advises and neither of them worked reducing the noise level (bad luck for me), modified again in my software the frecuency interval of packet sending and the noise frecuency changes

So that give me the clue that is only related to magjack conector of the ethernet rj45 conectore.... It is bad bacasue with for an audio intercomm develop it wont let the voice to be listen properly...

My last option is to rip off the mag jack conector from the cool components breadboard and wire it directly to the mbed pin outs... any advised with this idea? do you think it might help out?

Thanks again for your hep with this issue.

Sorry to hear that did not help.

Some of the key things with noise are:-

1. Impedance. Look to see if you can reduce any track impedance, remember a small inductance can have a big effect. So for your board (which I couldn't find the tracking layout), you could try soldering some wire from the 0V screen of the mag jack to 0V of the PSU. Treat the PSU 0V as the star point to route new 0V wires to.

2. Try a wire (in fact braid is better) from the mag jack screen/case to the 0V star point (created above). This provides a route for the noise that doesn't go through sensitive areas.

3. Add inductance to try to block the noise currents (as tried previously).

4. Ensure analogue signals are short and screened.

5. Ensure good HF decoupling is provided on supplies. Use good quality ceramic caps of several values in parallel (SMD parts are best) with thick short low inductance wires from the capacitor, i.e. no leaded capacitors with long leads.

I'll try to think of what else you can try. I actually also have a good earch screen on my pcb under the mbed and I know some people have fabricated a copper shield actually under the mbed itself and earthed at the 0V pin.

Maybe it isnt a possibility in your case, but what also works is making sure both the ADC and ethernet arent active simultaneously. So only do the AD conversions when ethernet is idle.

NXP has an application note on noise issues with adc/dac AN10974. It may be found here www.nxp.com/documents/application_note/AN10974.pdf

The Coolcomponents board may not be the best choice here. Seems to me that the ethernet wiring must be crossing you adc/dac wiring assuming you used the experimentation area to implement your lowpass output filters.

Are you sure the problem is due to hardware rather than software. Have you tried generating an onboard sinewave with/without running the ethernet code and unplugging the ethernet cable.

First of all thanks a lot for your advised I am reading and trying each one of them to see if any help in reducing the noise... not luck for me yet.

I did try in a single board conversion ADC-DAC and check if the noise came from the ADC or DAC but its not it work great in the same board, so that make me get sure that the issue is coming from the Ethernet link, specially from the way the mag jack is connected to the mBed board is my first guess.... The ADC/DAC works perfectly on their own, any way to reduce any other noise I declared as DigitalOut the not used pins from Analogic inputs....

I think that my two remaining options might be trying to develop a custom ground plane, but not so confident that this might help in something...

Or remove the magjack from the cool components base board, try with a different board. maybe, but my guess is the way they connect the mag jack and the mbed if that connection change the noise might reduce considerable.

The filtering and inductors in the power and gnd line were calculated and I placed them, they did not help in the ethernet from injecting the noise, but they did reduce some white noise that appears often.

I am not use the free pins that the red board brings with separates connection for 5/3,3/GND pins, I am using a different base board for the prototype of the filtering and analogue signal accommodation. So the cool compoents board remain free it that area

thanks again and any idea is usefull!!! Really I am kind of running out of options here.

I was giving a thought to this, and changing the magjack so the connection from it to the mBed change...

What if the noise came from inside the LPC1768 chip? or even the mBed 40 dip board can be introducing it....

I check out the inside electrical connection for the LPC1768 and it has several and different grounds for the ADC sub-module and separate grounds for the digital part. But for a size matter I think it was reduced to just one ground in the mBed board.

Any ideas that I can find useful about this?

Thanks a lot for the support!

Greets

Fernando

Fernando, from my own experiences there is noise on the ADC's and DAC's. My application needed a very low noise DAC, effectively providing a near DC low noise reference and I could not achieve this on the mbed. I studied the board, the chip and DAC's and what I believe to be the issues are:-

1. The mbed needs an analogue ground pin available near the ADC/DAC for the voltage to be measured across or referenced to.

2. DAC's and ADC's have a spec that is difficult to achieve in practice. The key issue which is often hard to find is that the main noise IMO is digital or clock feed through noise where the digital bus capacitively couples to the analogue signal, so noise spikes appear on the analogue signals at each digitital transition.

3. As above the NXP chip it self will generate noise as the DAC's and ADC's (and some other functions) are on an internal bus which is switching all the time.

For my application I added on an SPI bus DAC from Microchip which is powered as the only device on the SPI pins and hence clocks only when I change the output. It also has it's own linear voltage regulator and as the input signals of the DAC are high impedance, I treat the whole DAC, regulator etc., as an analogue component on the analogue ground line.

RESULTS: Very, very low noise on the DAC output (sub one bit).

The ADC I found similarly noisey, but kept the NXP part. As my requirements are very low frequency, I used a multipoint averaging technique, which was adequate for my needs.