Uneven Ground and References

Thread Starter

himmels

Joined Mar 25, 2018
13
I have a circuit with about 24 op amps in it, they are all powered by +-15V rails. In the same circuit I have a bank of analog multiplexers and transistors acting as switches, they are running off +-5V.

The +-5V was being supplied by a 7805 and 7905, with input/output caps. Unfortunately they were causing the ground to rise about a millivolt, which was upsetting the virtual ground on my op amps. After protecting against backflow with diodes and reverse polarity etc. I was still having the same problem. I decided to try two zener diodes and I got the same issue!

This issue is occurring with or without the load connected, so it is not to do with the load. If I disconnect these power sources ground returns to zero. I assume it probably has something to to with uneven distribution of ground (it's on a breadboard). But even when I connected grounds accross with more wires it happens. My last guess is that there's oscillations, even with stabilising caps. Does anyone have any ideas?

TLDR: Power sources causing local ground to rise.

Cheers, Himmel
 

danadak

Joined Mar 10, 2018
4,057
Maybe a pic of breadboard layout ?

Grounds chip to chip chained or all come to single point ?

How is the virtual ground established ? Reliance on rail regulation and matching ? Is
PSRR from rails referred to input a possible cause ?

Schematic.

Regards, Dana.
 

dendad

Joined Feb 20, 2016
3,569
The power ground and signal ground should only be connected at one point.
It sounds like you have current flowing in the signal ground. Good analog layout is important. And not real easy to get right on a breadboard.
 

Thread Starter

himmels

Joined Mar 25, 2018
13
Maybe a pic of breadboard layout ?

Grounds chip to chip chained or all come to single point ?

How is the virtual ground established ? Reliance on rail regulation and matching ? Is
PSRR from rails referred to input a possible cause ?

Schematic.

Regards, Dana.
Cheers Dana, sorry I can't post a picture,

I've got the chips connected to the rails, there are 4 distribution strips. I've got 0.1uF caps at all Vin. Virtual grounds are in the context of transimpedance amplifiers, and buffers maintaining 0V output when not selected. PSRR is probably an issue, I think in total theres some reactive elements at play that are going to happen on a breadboard implementation. I have found that the addition of a 330uF capacitor on the output of the zener reference instead of 0.1uF halves the effect, no more, no less though.
 

Thread Starter

himmels

Joined Mar 25, 2018
13
The power ground and signal ground should only be connected at one point.
It sounds like you have current flowing in the signal ground. Good analog layout is important. And not real easy to get right on a breadboard.
Cheers Dendad,

Sounds like youve probably hit the nail on the head here. Both the LMXXX regulator and zener reference setups always sink current to ground and its probably uneven and there's loops etc. I was hoping there was an easier way to fix the issue, perhaps with different capacitance or inductor or a diode to try and direct flow.
 

Thread Starter

himmels

Joined Mar 25, 2018
13
Ok, so I 'fixed' the issue by running a dedicated ground line from the LMXXXs to the power supply and it cleaned up the ground. This solution is ok for my situation, but I think I'm going to look deeper into analog design with different references on board, as I will need to do this with the PCB implementation.

Could you offer me any good references on the subject?
 

MisterBill2

Joined Jan 23, 2018
6,350
The power ground and signal ground should only be connected at one point.
It sounds like you have current flowing in the signal ground. Good analog layout is important. And not real easy to get right on a breadboard.
Separating digital power ground from the analog power ground is not only important, it is often VITAL!! AND, for the analog section, if you can connect both ends of the ground trace back to the analog supply common that may bring a slight benefit. Likewise with the positive and negative. Adding additional capacitors across the analog power occasionally may also help. AND, the exact arrangement of the analog-digital star point may also need a bit of experimenting.
 

MaxHeadRoom

Joined Jul 18, 2013
20,920
I have always subscribed to the opposite, I connect all grounds together where possible and in most case where noisy environments exist, connect to earth ground at a star point.
This is often done in most industrial situations.
The motion cards I use sit in a PC slot where the P.C. common is at earth ground together with the analogue servo out and digital I/O.
In industrial equipment environments it is also known as equi-potential bonding.
Max.
 

MisterBill2

Joined Jan 23, 2018
6,350
OK on the cards in a PC slot. NONE of the stuff I did went into PCs, we purchased those boards from places like Data Translation. Our PCBs lived i custom card racks inside electrical cabinets. Mostly those systems were not motion control but rather instrumentation. So things were quite different.
 
Top