oscilator circuit design for use with precision ADC

Thread Starter

johnny411

Joined Mar 3, 2020
41
this is my "final" version of my pcb design. i am on a deadline to get this done so unless there are any major flaws in my design, which i assume there not to be, i will probably be ordering them today 4/29/20 at 7:00 CST. the digital section is quite cramped but the critical traces (power,usb,spi) are routed well enoughph (usb is in differential routing mode of course). also looking at the small section where the two grounds meet, is that how star grounds are supposed to b done?

i use easy eda so the full project (schematic, pcb) can be viewed here: https://easyeda.com/super7800/low-burden-urrent-sensor


pcb final 1..PNG

Edit: I ordered them. The design I went with is only slightly different from the image above.
 
Last edited:

kubeek

Joined Sep 20, 2005
5,794
I have two things, one is that lines crossing that cutout in the ground plane are not a good idea, because the return currents will all be messed together in the small area of the star point, but since you already ordered there is little that could be done.

Second thing, that you possibly could still change, is conencting the shield of the USB cable to your ground. I believe it should be kept on the isolated side, or diconnected alltogether. The way you have it all interference that gets coupled to the usb or measuring wires will be flowing from one side to the other and will be upsetting the ground plane potential, and most likely will introduce errors to your measurement.
 

Thread Starter

johnny411

Joined Mar 3, 2020
41
I have two things, one is that lines crossing that cutout in the ground plane are not a good idea, because the return currents will all be messed together in the small area of the star point, but since you already ordered there is little that could be done.

Second thing, that you possibly could still change, is conencting the shield of the USB cable to your ground. I believe it should be kept on the isolated side, or diconnected alltogether. The way you have it all interference that gets coupled to the usb or measuring wires will be flowing from one side to the other and will be upsetting the ground plane potential, and most likely will introduce errors to your measurement.

please excuse my ignorance but all the traces crossing into the analog side are digital (ignore the relay drivers, they have a trace leading right back to the dgnd with no vias to gnd in the agnd region), and are then fed into 74HC14 chips, which are supplied by a seperate analog +5v rail, so would'nt the return currents be small or nonissue?

second the usb is isolated using a usb isolator. its of course not laid out in a way that would allow taking advantage of its full effects, but it does keep the usb ground and system ground seperate (dont want the ground loops). since the most this device will ever see is maybe 15vdc, i wasnt too worried about isolation cutouts or anything like that.

oh. it is already connected. you are 100% correct. i should be able to carve those traces away. thanks. didnt see that. i will disconnect it alltogether. thanks.
 

kubeek

Joined Sep 20, 2005
5,794
please excuse my ignorance but all the traces crossing into the analog side are digital (ignore the relay drivers, they have a trace leading right back to the dgnd with no vias to gnd in the agnd region), and are then fed into 74HC14 chips, which are supplied by a seperate analog +5v rail, so would'nt the return currents be small or nonissue?
Yes, digital, and that implies rather steep rise and fall times, which implies non-negligible return currents due to trace capactiance and input capacitance. The return current from these has to pass through that starpoint and willl cause some voltage drop.
Now say the peak current on signal transition is just 5mA, and that it goes to U2. Lets assume the impedance of the ground plane is 1mohm, this could create a drop of 5uV in the ground plane underneath the reference U25, so the reference will have that error, and this error will propagate into the rest of the measurement.
 

kubeek

Joined Sep 20, 2005
5,794
I wouldn´t cut the ground plane, but if you really insist then make the connections to the digital side go above the star point and avoid any sensitive components to get undground those digital traces. What I din´t say in the last post, the area of the loop through the digital wire and back through the star point is very large and will act as an antenna, both receiving otside noise, and if it were higher frequency data also transmitting noise.

Also I would start with a pcb about twice the size, to be sure I have everything placed and connected as neatly and ideally as possible, and after that setup works according to expectations I would try optimizing it to smaller size to fit whatever case you are planning. Starting resonably big can also help with debugging and putting in zero ohm resistors so you can test out parts of the system, for example to bypass the PGA and see if some issues come from that or elswhere.
 
Top