Multilayer PCB - Shielding - Which supply on each layer?

Make a bluetooth device for each sensor that sends the digitized value to a receiver. Already having been quantified at the transducers removes all error. Bonus: No need for shields, and small circuit size,at least for the sensor senders. Also, you would incorporate a handshake in each that would keep another nearby unit from passing false data to your "data logger".

Have an inductive charging circuit for the battery device that powers each "sensor xmitter". They would all charge overnight and after each use while in the storage drawer of the receiver cart.

Sell tens of thousands of units.

Have a nice day...

I'd suggest...

Layer1 : signals + soldering pads for SMD parts. Layer2 : ground plane Layer3 : power pours Layer4 : signals

which is much easier to kluge, should you need to.

John

You don't make planes for power supplies that supply low currents. Making planes just increases the chance of capacitive pickup. Just use traces + a bunch of 1 cent caps. You might try putting power and tracks on layer 1, put a ground on layer 2. Now you need to work closely with your pcb fabricator to make sure you don't end up with an unbalanced stackup that might warp. What material ??

I'm with John Larkin here.

Ground/power plane underneath a track actually provides pretty good shielding; other signal tracks routed over a conductive plane radiate as if they were dipoles; the presence of the ground plane changes the radiation pattern to pretty much the one you'd get if there was an oppositely charged track as far below the ground plane as the real track is above it, and the electric field drops off a lot faster than square law.

Buried conductive planes within a board work a lot better than an aluminium plate bolted on to the back ("track side") of a board.

-- Bill Sloman, Nijmegen

Hi again,

Please read the last paragraph of the first article ("Input filter prevents...") in this pdf:

It says: "You should build the RFI filter using a pc board with a ground plane on both sides."

As I said, I don't have much experience with shielding uV, but that made some sense to me. Intuitively, I think I would prefer having more capacitive pickup, but it happenning to ground planes, where it doesn't do any harm, than having less pickup, but it happenning to signal tracks. Doesn't it make sense?

I now debugging will be more difficult (because the signal tracks will be buried), but reducing interference is a higher priority for me.

I also understand Bill's comment (about a track right next to a conductive plane radiating less at far distances --by image theory), but I'm not sure that the benefit from that[*] will be higher than the benefit from two ground planes at layers 1 and 4.

So... I don't know what to do. Are you guys convinced of what you say? Can you convince me, too? :-(

Thank you, Bill

Is the circuit going to live inside a metal box? If so, I wouldn't worry too much about shield layers both top and bottom. If not a metal box....well then things will be harder. What's the source imepdeance of the EEG sensor? You may not want shield layers near the sensor input lines till after the first amplifier stage. Capacitive effects might limit the bandwidth. Myabe not a problem at only

George Herold

You don't want that signal to stay uV levels for long after the sensor. First chance you get make it big! Then signal to noise problems get easier.

EDN will print anything.

Think about how many vias you'll need if the topside is a ground plane.

John

Yeah, you've got components living on the outside of the board. How much help are ground layers top and bottom going to do? (sheilding wise)

George H.

Really makes *NO* sense to have all of that radiation (bluetooth transmit & receive plus inductive radiation to the PCB) any where near those microvolt inputs and circuitry.....

Wouldn't having shielding planes surrounding sensitive signals be better?

Sorry, but I'll bet that examination rooms are already RF rich, and I'll also bet that there is no influence on low impedance transducers at all, particularly when the value gets converted right at the source. From then on, the data that gets sent is digital, and 100% error free. We can sense picoamp fluctuations in a PMT to see a missile launch 80 miles behind an aircraft without error introduction from local RF, I think we can handle this.

So, the transducer and ADC would be right there local to the sensor (ie no interference whatsoever), and the signal sent over the air is the quantisized values from sampling to sampling and are digital, so they cannot carry error either. Absoluetely a better solution. No more need for wired patch sensors. Far better, and it IS or WILL BE the future. Why not have it be him that goes wireless with body sensors?

Transparent Aluminum, anyone?

Why would one ever want sensitive signals taking long paths? Nowadays converting a signal to digital is cheap, even when 8 or 16 channels are involved.

Hell, a simple (or complex) data logger would be an excellent PCB layout to examine.

No one wants sensitive signals taking long paths, but sometimes you need to amplify before A-to-D conversion is possible. And sometimes, you need to amplify so much, that you need to break that gain into several stages, not to have output-to-input feedback. That's how you end up needing shielding.

It depends on the voltage levels and frequency bands it deals with. My application is not RF, but it deals with uV, so I'd rather learn from RF PCBs than from PCBs of data loggers.

I also think that is the future, and there are several companies that start offering wireless electrodes. I don't know how they do the strong filtering that is needed at the input to avoid saturation of the first stage, due to (self) HF interference.

The parts, pads, and connectors are already exposed to any local fields. The best thing to do is keep all the interconnects short. Doing the signal routing on an inner layer just adds a pincushion of vias and exposes signals to power-pour noise. And makes the board hard to work on.

If you're looking to process high-impedance nanovolt signals, put the board in a metal box so there are no noisy local electric fields.

It's not at all difficult to design pc boards that are limited by the physics: semiconductor noise, resistor Johnson noise, shot noise.

RF rectification can be another error source. Again, put it in a box.

John

If you really need 10 uV and 0.5 Hz, it has to be thermally shielded, not just ground-plane-against-capacitive-coupling shielded. Keep all wiring to the first amplifier/preamplifier short, and maybe random breezes causing temperature fluctuations won't dominate your signals.

Copper wiring, aluminum metallization on ICs, gold bond wires, Kovar leads on IC packages... you have lots of dissimilar metals, and there will be thermocouple voltages.

Hey, thanks for bringing this up. I never thought of it. I'll hope thermal changes won't be as fast as 0.5 Hz.

Not to mention ~1000-1500 uV/K for any metal vs silicon (very doping-dependent) and ~300 uV/K for tin oxide resistors. (Tantalum nitride and NiCr are much better, just a couple of uV/K.) Fortunately copper:Kovar doesn't come up much nowadays--there aren't a lot of ceramic and metal can packages left--and tubes are another issue entirely.

Cheers

Phil Hobbs