Suggestion - Faraday's Cage

[jahonen]

The really strange thing I found was that they recommend to join the two split grounds at the ADC... Why do they take out of the package two separate pins for ground, if then you have to externally connect them just out the chip?
This is a great enigma to me.

They have two pins just because two grounding paths is better than just one. Bond wires inside IC package are extreme thin, and is a bit different situation than in PCB environment. Of course, much more ground pins would be even better but return benefit diminishes as number of ground pins increase, and larger packaging will cost more. If anything is high volume product, ICs certainly classify as one. So perhaps two ground pins is the optimum amount.

Complex digital chips like processors and FPGAs have very large number of ground pins too, to handle the massive return current produced by IO pin drivers. Too small number of ground pins would seriously degrade the SSN (simultaneous switching noise) performance. But usually there is a statement in the datasheet that all the ground pins should be kept at same potential for good performance (quite obvious).

Never ADCs seem to have more sensible recommendations in their datasheets, like in AD9650 (16 bit, upto 105 MSPS ADC) datasheet:

A single PCB ground plane should be sufficient when using the AD9650. With proper decoupling and smart partitioning of the PCB analog, digital, and clock sections, optimum performance is easily achieved.

I think that most of the classic statements found in ADC and DAC datasheets has been made in the era where multilayer boards were not so common. With only two layers, dedicating one side of just common ground plane with absolutely no slots in it, is difficult. Nowadays it makes more sense to just design a multilayer board.

AD5755 datasheet states that:

A microstrip technique is by far the best, but not always possible with a double-sided board. In this technique, the component side of the board is dedicated to ground plane, while signal traces are placed on the solder side.

Regards,
Janne

[scrat]

The really strange thing I found was that they recommend to join the two split grounds at the ADC... Why do they take out of the package two separate pins for ground, if then you have to externally connect them just out the chip?
This is a great enigma to me.

They have two pins just because two grounding paths is better than just one. Bond wires inside IC package are extreme thin, and is a bit different situation than in PCB environment. Of course, much more ground pins would be even better but return benefit diminishes as number of ground pins increase, and larger packaging will cost more. If anything is high volume product, ICs certainly classify as one. So perhaps two ground pins is the optimum amount.

Complex digital chips like processors and FPGAs have very large number of ground pins too, to handle the massive return current produced by IO pin drivers. Too small number of ground pins would seriously degrade the SSN (simultaneous switching noise) performance. But usually there is a statement in the datasheet that all the ground pins should be kept at same potential for good performance (quite obvious).

Thanks Janne! If it's only for a "bonding" issue, and they come from the same net, that's quite clear why they put more than one GND pin. However, that's pretty strange to call them "analog" and "digital" ground. If they come from different portions of the chip, it seems not a good idea to make a closed ring outside of the package to join them, it had much more sense to join them inside with a quite strong metal and then bond it to two pins to have a good conductor. Just wondering...

[Feanor]

I would also like to see a blog on Faraday cages. The theory might be well known to decide things like how hole size/ grid size relates to frequency, but practical solutions for the hobbyist can be tricky. For instance the screws you use to seal the Faraday cage can act like antennas radiating the noise nicely into your fully sealed box. Coaxial cables running into the box can act like waveguides doing the same thing.  If there are some tricks of the trade to get around these I'd like to see them. Also aluminium boxes are not cheap, is there a type of mesh out there that is any good?

I recon a blog could be done on this. How about a practical example? How about moding your fluke multimeter so that it is no longer susceptible to GSM interference? Now we are talking!

[migsantiago]

Yeah, lots of stuff can be said about noise and how to avoid it. Some expert users have proved it with their messages in this thread.

A basic tutorial on Faraday Cages would be enough. I have been studying the info on the links provided here and it's a whole world of things to consider... and even worse, most of them are based on the experience of the designer.

[.o:0|O|0:o.]

If properties of materials change in response to changes in frequency, how do you measure the capacitance between two circuit planes? Does a capacitance meter serve any purpose at all here or does it become a case of solving an equation and hoping for the best?

.o:0|O|0:o.

PS.
Interesting stuff about PCB design against noise and interference, http://ww1.microchip.com/downloads/en/AppNotes/00242a.pdf

[alm]

If properties of materials change in response to changes in frequency, how do you measure the capacitance between two circuit planes? Does a capacitance meter serve any purpose at all here or does it become a case of solving an equation and hoping for the best?

A Vector Network Analyzer (VNA) should be able to do this measurement, I believe. Don't ask about the price.

[migsantiago]

If properties of materials change in response to changes in frequency, how do you measure the capacitance between two circuit planes? Does a capacitance meter serve any purpose at all here or does it become a case of solving an equation and hoping for the best?

I'd say that the variation of the capacitance would not be important at that stage. Maybe if the circuit planes were really big, the capacitance would play a more important role (just as with coaxial cables and their inherent capacitance, for example).