I have wondered for a while whether or not I was doing this properly. Several years ago when I was working on my first mixed signal design, I searched for a reference or some good rules of thumb on how to separate ground planes. I remember there not being a lot of clear direction out there on this topic, or maybe I just didn't know what to look for. I can't remember how exactly I settled on the 0.1 Ω resistor, the 100 nF cap, and the 1 uF cap all in parallel. But, my goal was to minimize noise in the analog portion of the circuit, and for that purpose it seemed to work pretty well for me in the past. Admittedly though I didn't do a lot of experimentation with alternate configurations so I wouldn't be shocked in the least if this arrangement isn't ideal.
I can replace the 0.1 Ω resistor with a zero Ω one as you and Mike suggest. That's no problem. But that sort of leaves me with the question about what to do with the caps. Are they doing anything for me here? What would you think about simply removing them?
Also, what if I omitted the zero Ω jumper and just replaced it with a trace connecting the ground planes instead?
Hi VNE,
Grounding arrangements on nearly all circuits is always a critical area, even in digital systems, where you might think that because digital is only two-state, grounding would not matter. The reason for this, as I have said many times before, is that the nice solid line you see on schematics is an illusion- there is no such thing as a solid line in the real world. If you ever want to see this illustrated, just get a good low- noise differential input scope and connect one differential input to what you think is the best ground point in your system and then with the other differential input measure other 'ground' points on the circuit.
There are two broad ways to approach grounding:
(1) Isolate
(2) Solid
With the isolate approach, which you have used, you have islands of separate grounds, so you may have an analog ground and a digital ground to make sure that the noisy digital signals do not interfere with the sensitive analog circuits. This approach was widely used up to around 1985, and you can see this illustrated with audio power amplifier design especially.
Alternatively, the solid approach does not separate the grounds. Instead, it uses as solid a ground as possible (but you still separate sensitive circuits from digital circuits by physical layout). For example, you can specify a thick copper ground plane. In one case I stuck a sheet of copper on to the printed circuit board to get a solid ground. And in another case I used 5mm square copper bus bars to give a solid ground.
So where is all this leading: the problem with the isolate approach is that often it will catch you out, especially on a large system,because sooner or later the grounds have to be joined, so the modern trend is to connect all grounds together on a solid ground.
Of course, this is a broad philosophy and the are situations where you would completely isolate one ground from another, but this is where the two grounds can be separate. The classic situation is a processor controlling an industrial welder, for example. In this case you would most definitely separate the processor ground and welder ground, normally by a transformer or optocoupler. Another example could be a precision oscillator. Here you may need a separate and isolated ground for the oscillator to ensure low jitter.
So after all that, yes, I would advise joining the digital and analog grounds on your circuit by a solid conductor, but only in one place to avoid ground loops. Obviously, you can then ditch the capacitors that you mentioned.
spec