ADC Noise - please check my PCB!

[sleeper1987]

Hi, I'm having a lot of noise from my external ADC, and I suspect it's the fault of the board.

The ADC is 10MHz, and 12-bit, and I'm getting around 4-bits of noise, equivalent to 7mV on it. Lower clock frequencies have no effect on it.

Can someone please take a quick look at my PCB / schematic and see if there's anything obviously wrong with it?
View attachment DAQ pcb..pdf
View attachment DAQ schematic..pdf
View attachment Eagle project..zip

Any help would be greatly appreciated!

 

[crutschow]

It's difficult to tell from a cursory view of your design, but do you have separate analog and digital grounds? If not, then it likely will be difficult to get the noise any lower than what you see. It's probably the digital noise on the ground you are seeing. Keeping digital noise out of an analog circuit on the same pcb generally requires two grounds and careful layout to keep the digital noise from contaminating the analog signal. Usually the application notes for the A/D will tell you how to do that.

 

[sleeper1987]

It's difficult to tell from a cursory view of your design, but do you have separate analog and digital grounds? If not, then it likely will be difficult to get the noise any lower than what you see. It's probably the digital noise on the ground you are seeing. Keeping digital noise out of an analog circuit on the same pcb generally requires two grounds and careful layout to keep the digital noise from contaminating the analog signal. Usually the application notes for the A/D will tell you how to do that.

Hi. Thanks for the reply. As I understand it, I shouldn't actually be using *separate* analogue and digital grounds, but instead just distinct areas of the ground plane. As it is, there's little high frequency switching on the right side of the board, (turning it off altogether has no effect), and there are lots of local vias to the ground plane to minimise ground loop currents.

Is there anything else it could be? - I notice the circuit's pretty capacitive as I've not included any inductors - could that be having a detrtimental effect on the ADC?

 

[Gayan Soyza]

When I get ADC noise stuff I'd more look on those things.

*Input sensors must powered from a regulated supply.
*Must use precision voltage references.
*Over sample should be done if you have enough time between samples.
*Hysteresis must be included.

 

[crutschow]

Hi. Thanks for the reply. As I understand it, I shouldn't actually be using *separate* analogue and digital grounds, but instead just distinct areas of the ground plane. As it is, there's little high frequency switching on the right side of the board, (turning it off altogether has no effect), and there are lots of local vias to the ground plane to minimise ground loop currents.

Is there anything else it could be? - I notice the circuit's pretty capacitive as I've not included any inductors - could that be having a detrtimental effect on the ADC?

The grounds should be separate or physically split except for being tied together at one point (and that location of that point can be important. It's often near the A/D converter). Since an A/D converter obviously has both analog and digital portions, it's important to keep the two sides as separate as possible, thus the ground plane split is usually down the middle of the converter chip. For this purpose many A/D's have separate analog and digital grounds on the chip. If they are both tied to the same plane than it's easy for the digital signals to contaminate the analog.

Are the analog and digital power supply also separate? That can be another source of contamination between the two. If they are the same supply then, at a minimum, there should be a ferrite bead filter inductor type connection between the analog and digital connections with both sides of the power decoupled to the ground plane very close to the A/D power pins with ceramic caps.

It may be difficult to reduce the noise without redesigning the board. Can you amplify the signal to raise it above the noise? Also as Gayan mentioned, can you over-sample and do a digital average to reduce the apparent noise?

Here is an article that discusses the layout of an A/D converter board in more detail.
View attachment A-D HS Converter &#8.doc

 

[sleeper1987]

Hi again, thanks for getting back to me.
I've done a fair bit of reading on the ground plane subject. The article below asserts that actually splitting the plane isn't necessary, as ground currents tend to flow directly underneath their traces, (provided the path is uninterrupted). There's an interesting table about the percentage of ground current a given distance away from the trace - 87% is contained within 5 trace widths!
Presumably, then, splitting the board is only worthwhile if you actually need to redirect the return path to avoid a nearby trace?

https://www.electro-tech-online.com/custompdfs/2010/04/june2001pcd_mixedsignal.pdf

Regardless, this doesn't seem to apply to my board. I've got separate and distant analogue/digital traces, and followed the ADC manufacturer's recommended trace layout for the chip itself. Also, the power supplies are separate and chosen specifically to be low noise. The voltage references are from precision sources, and buffered through a low noise op-amp. I feel like I've done all I could!

Additionally, there should be very little high frequency noise anyway - the uC I'm using is the only source, and is doing very little, far away. I've tried setting it in low power mode - which shuts off clocks and peripherals - to take measurements, so I'm pretty confident it isn't the source of the noise.

Short of getting hold of an oscilloscope to identify the source, can you suggest anything else?

Thanks,
Dan.

Edit: Sorry, I realise I haven't answered your question about supersampling - and no, it's not possible really as I need to capture a lot of samples quickly.