Tracing and Fixing Noise

[skrasms]

I am debugging a mixed A/D circuit that has a problem with noise. I have re-built it several times, and each time the noise floor has changed unexpectedly. I am hoping to re-design such that the analog noise floor is below the dynamic range of the A/D converters (12-bit).

My first issue is how to accurately measure what is generating noise. If I hook my oscilloscope up to almost anything it shows a pattern of voltage spikes at a rate around 10 kHz. That same pattern shows up on circuits with different power supplies. Would that be related to the oscilloscope itself? It is an old Tektronix 2235.

Right now I am looking at the analog side by itself, with the digital side turned off and disconnected from the circuit. According to my oscilloscope even my voltage reference with bypass capacitors is giving out more than 40 mV pk-pk of spikes at 10 kHz. It is an Analog Devices ADR02 rated to have only microvolts of noise. Looking at the same signal with a Fluke 8060A set to rms voltage it reads 0 mV, -74dB.

How can I approach this? What tools can I rely on?

 

[nickt]

There's a very good online book (pdf) called "opamps for everyone" (think it's by Walter Jung - had a quick look but can't find it in the meta)

Yeah yeah opamps etc BUT it has an excellent section on board layout and specifically talks about mixed analog/digital boards and issues with ADCs.

If you can't find it PM me and I'll email my copy.

cheers
Nick

 

[clintrubber]

That pdf might be by Mancini (Ron, not Henry)

http://www-s.ti.com/sc/psheets/slod006b/slod006b.pdf

http://www.tangentsoft.com/elec/reviews/free.html

 

[nickt]

[quote author="clintrubber"]That pdf might be by Mancini (Ron, not Henry)[/quote]
Yep that's the one, wonder where I got Jung from? :roll:

 

[clintrubber]

[quote author="nickt"][quote author="clintrubber"]That pdf might be by Mancini (Ron, not Henry)[/quote]
Yep that's the one, wonder where I got Jung from? :roll:[/quote]
Walt Jung has also written a thing or two (to say the least), so it's not strange. And as it happens he also appears on the html-link above.

Bye,

Peter

 

[skrasms]

Thank you for those articles, they're really helping me understand the PCB layout side of things better.

Right now I am just working on a breadboard trying to figure out where my major noise sources are. How can I measure the noise I am getting on the power lines, for example?

 

[dale116dot7]

A few things about A/D and D/A's that I've noticed...

For audio, clocking issues - noise on the clock line, unstable PLL performance, incorrect shielding on AES or SPDIF cables can create noise that sounds just like white noise.

For normal microcontroller A-D systems, check out the sampling capacitance - understand how the sample and hold works. Charge coupling will introduce noise into the readings. The voltage references to the A-D need to be clean, but in most cases, a microcontroller (or pretty much any unbuffered successive approximation A-D converter, prefers a capacitive input - not an op-amp directly. I usually make an R-C filter and that works well.

Generally, grounding is important. On a 12-bit system, you can probably get by without doing a lot of special grounding or power supply magic if you have the luxury of a 4-layer PC board, if you use adequate decoupling capacitors, and if you pay some attention to ground currents of any power devices, if you have any. I do this sort of thing on my engine control modules (my day job), and 11 or 12 bit without noise is very possible.

Also, check for op-amps driving capacitive loads, or oscillating.

Generally, breadboards are a bit of a pain in the arse, and noise performance will almost always be better on a PC board.

-Dale