I am using the CN0349 as a fully "isolated" conductivity sensor for seawater. Interfacing with the part is fine and it's data output is fine.
However, it causes huge amount of interference (RF?) with other devices on my master PCB board. It can successfully jam my GPS and cause a large reduction in signal quality for my LTE modem. It is apparent because removing it from the header shows a spike in signal quality. Are there any suggestions on how to shield this part from interfering with the rest of the board without turning it off when my gps or modem needs to be used?
The most likely culprit here for your RF interference is the ADuM isolator parts. There are techniques on how to handle this, and thankfully someone wrote them up in a nice little app note. Have a read through this and try implementing these techniques and see if you're able to cut down on the noise.
I assume the situation is depicted in Figure 4. Dipole Radiation Between Input and Output. Is there any way to mitigate interference without remaking the eval board?
Are the ADuM parts necessary for the circuitry?
Well I think its a mix of both, I am prototyping a PCB to interface with the CN0349. If the AD5934 is operating at high frequency wouldn't that produce considerable noise on the power and ground planes,without isolation? I could just redesign the pcb board to include the ADUM5020 or without isolation but my supervisor was wanting the product to not have to go through another pcb design/wait period. Would it be worth trying a low ESL cap instead?
Re: AD5934 producing ground noise - not necessarily - if the supply pins are well bypassed, any high-frequency activity inside the part is confined to a small area. And you don't have the dipole situation that you have with the ADUM5000 straddling between the two ground planes. I'd try it without isolation and see what you get - in fact, you could start by just shorting grounds together and leave the ADUM5000 in place; this would be a useful debugging exercise anyway.
Hm, this is good information.
Would I also short both the iso3.3V and 3.3V rails together?
If the ADUM5000 is still in place, then don't tie them together. If this test determines that isolation is not necessary, that is, no degradation to your conductivity measurement as a result of shorting grounds, then the next step would be to remove the ADUM5000, shorting the supplies (such that the formerly isolated side of the board is now powered directly.)
Should the same be done with the ADUM1250?
If you're defeating the isolation (tying grounds together), then it can be left in place for initial testing, then removed and the I2C signals connected directly (wire SDA1 to SDA2, SCL1 to SDA2).
I removed both isolation chips which did the trick. However for a more permanent design are there isolation chips out there that avoid having any RF interference?
All isolated supplies will have some level of interference, but the ADUM5020 mentioned above would be a good step in the right direction. There is quite a bit of reference material on the product landing page:
What about optically isolating, is that an option?
For the I2C signals, there are ways of opto-isolating but they're not very convenient due to the open-drain, bidirectional nature of the signals. But it's really the power that is more troublesome - there are esoteric (including optical) ways of isolating power without transformers but they won't be nearly as convenient or compact as a transformer-based isolated supply (like the ADUM2050.)
And since it's questionable that you need isolation at all, if there's a way to make a non-isolated solution work in your application, that would be the simplest approach.