Ground loops and equipment design
Ground loops are not only a problem in connections between equipments. When you design electronics circuits you should always avoid causing unnecessary ground loop, because they cause many annoying problems. A typical problem when the grounding is not done properly is that the electrical noise from the noisy parts of the circuit gets to the parts of the circuit which should be free of noise. Here is some guidelines for designing grounding systems used in equipments:
System part interconnections
When you have two circuits that are tied together electrically, but one of them
is high current then you should direct the ground and power paths to "feed"
them separately. You want the current of the driver to stay on the driver side
and the current of the logic to stay on it's own side. The thin trace inbetween
is still needed because this is not galvantic isolation.
|---------- | |
| | <<<< physical CURRENT POWER FOR
--REGULATOR LOGIC separation >>>> DRIVER DRIVERS
| | | |
| e-> | ground | <-e |
^ thick ^ thin ^
traces traces very thick
from reg to traces from
logic load drivers to supply
The common mistake is to "daisy chain" the ground by having the ground of
the high current item seek it's current path through the ground of the logic.
This causes ground spikes on the logic and thus logic errors due to bad
voltage levels at the logic chips.
Physical separation is to prevent electromagnetic coupling, of course. Even getting the grounds proper won't help if you couple the magnetic field back into the logic traces.
Always image traces to be resistors. Thick ones are small resistance and thin ones are large. The objective in laying out the board is to encourage the large currents to take the path back to their own source without getting onto the other grounds.
Separating current paths in this way can make a micro run right along side of a vicious current driver and not have logic problems in most cases. The cases in which it usually doesn't work is when the signal being sent to the driver is analog instead of digital. You're going to get some amount of ground differiental with the separate ground paths and so the analog signal will reflect this differance in the signal voltage relative to ground.
Current loop coupling of the signal to the driver could solve a really bad problem of ground differientals, but I have never used that technique. If your going to go to that extreme then you may as well isolate them altogether.
If your signal is digital you can clean it up abit by having a schmidt trigger on the driver side of the loop with it's ground relative to the high current load. This can provide a volt or more of tolerance in the ground differance.
If you get the currents going right you will see less problems with the logic side, but you might see more problems with the driver because it's signal from the logic is corrupted by lifting of the ground potential because of the high currents. When you have reduced this effect by minimizing the high current ground ohmage to the point where you cann't minimize it any more AND you have included schimdt buffering, then it's time to admit defeat and galvanically isolate the two circuits.
When you have power electronics and some microelectronics on the same circuit the layout of the current loops is critical. This also applies to situations where you have microelectronics and audio circuit on the same board.
What are the problem with treating both digital and analog grounds as the same ground source ?
Your digital circuit noise can get to your analogue signal path if you don't use separate grounding systems for digital and alalogue parts (those are interconnected only in one place). Digital grounds are invariably noisier than analog grounds because of the switching noise generated in digital chips when they change state. For large current transients, PCB trace inductances causes voltage drops between various ground points on the board (aka "ground bounce"). Ground bounce translates into varying voltage level bounce on signal lines. For digital lines this isn't a problem unless you cross a logic threshold. For analog it's just plain noise to be added to your signals.
How to eliminate this problem ? why the suggestion of keeping analog and digital grounds separate is made by having them separate on the IC, so use separate grounds. Two completely separate supplies allows switching currents to be speced and optimized for at the regulator, and to have analog supply optimized to be nice and quiet. Filtering and bypassing can help to short out some ground bounce currents. And if you connect the analogue and digital grounds together do it only in one single well-designed place.
If you are building an analogue to digital converter circuit, then you have to be very careful in keeping the analogue and digital grounds separare or otherwise you will not get the resolution you expect to get because of the noise. The AGND and DGND pins should connect to the analogue ground, preferably with it flooded under the whole IC package. If you don't do this, noise on the DGND pin fom the digital supply will get injected into the analogue section inside the IC. Decouplers should be placed close to the AVDD pins (analogue supply), which connect to the analogue supply plane. In order to stop noise being introduced to the analogue plane OR being injected into the analogue section of the chip from the DVDD pin, the DVDD pin should be connected to the analogue supply plane via a small filter, such as a ferrite bead, depending on the application and frequencies of interest. Make sure the decoupler for the DVDD is located close to the pin, with the bead not much further away.
The digital input lines should be close to each other and connect to the nearby digital IC. Underneath these lines, the analogue and digital ground planes should CONNECT. The reason for this is to keep the radiating loop area of the digital current path to a minimum. If you don't do this, every time a digital edge occurs, the pulse will either flow around rather a large path or get injected into the IC's analogue section, depending on your implementation. If your analogue side voltage needs to be very stable it needs its own regulator. Look for low-noise regulators, such as the ancient LM723. Other designs may find the digital supply regulation adequate and so just use an inductive filter to remove the necessary frequencies.
Why some circuits have separate analogue and digital grounds ?
Separating analogue and digital ground generally has to do with return currents and ground noise. Even though we think of GROUND as having zero impendence, it is a wire like anything else. Even ground planes have some impendence, and so currents flowing through these grounds cause voltage drops in the planes.
If we have a circuit that is operating on a senative analog voltage levels, or frequencies, we might not want these unexpected voltage drops occuring in the circuit. The are often injected into the circuit by filter caps, or bypass caps that were placed into the circuit to filter a signal or chip. So, the best way to keep this noise out is to create a separate analog ground.
Separate grounds does not always mean totally separated. Most analog ground and power splits are done by either just isolating the plane and not allowing the digital signals to flow over it. There is generally a single point connetion between the DIGITAL and ANALOG grounds. More than one point opens up the possibility of ground loops which cause the ground noise form the digital side to enter to your analogue electronics ground.
You should not run digital signals over or directly next to your Analog signals or planes. This will result in capacitive/inductive coupling to the "quiet" Analog planes, and/or result in horrible signal quality on the digital signal as its return current (which normally returns on the plane directly below the signal) is forced to travel "around" the isolated analog planes.
The need to have separate grounds in a printed circuit board or in a system is usually discussed and decided upon before the design begins. Often, when noise and other problems are perceived to be a possible problem, the designer will use separate analog and digital grounds in an effort to be conservative. Sometimes he's right and sometimes the grounds are separated needlessly.
The worst possible outcome is to design and build a printed circuit board with a common ground and then find that there is a substantial amount of digital noise getting into the analog circuitry. Then the entire board design must be redone. Sometimes separate grounds are needed and sometimes they're not.
Star ground or metal case ar ground plane approach for analogue signal systems ?
Star grounding and using equipment metal chassis as a common have both heir benefits and disadvanges when implementing then in equipments. Audio amplifiers are are fairly sensitive to grounding schemes in terms of hum, noise, and rf pickup. Star grounding inside is the best method for sighting against hummign and other noise problems, ecexpt RF noise. FOr preventing RF noise entering your system, it would be best to use metal case for your equipment and connect the connector and wiring shields (ground connector) directly to the case in the point where the connector enters the case.
Star grounding as it is s very goof methodf for wiring analogue equipments. But when considering high frequency RF circuits, it is absolutely the wrong way to go, because at radio frequencies wire is no longer just a wire, it is effectively an inductor, which causes that the star grounding scheme which works very well for lof frequencies does not work on RF.
The method of having RCA jacks insulated from the chassis and run the RCA grounds to the star ground is not recommended practice nowadays. Ground them to the chassis, this will maintain a continuous ground and supply the best immunity to RF interference. For the circuit design start with a continuous ground plane.
Unfortunately this ground plane solution is a good solution if only a relatively small dynamic range (20 dB for digital, 60 dB for video) is required. This means that ground plane approach is usable for many video circuits, digital circuits and not very demanding audio applications. However, high quality audio systems should be designed with 100 .. 120 dB dynamic range, thus with 1 V signal levels, any unwanted signals should be below 1 .. 10 uV. With various AC and DC currents flowing in the circuit, sooner or later they are going to end up into the ground plane. So there is a risk of several microvolt (AC) potential differences accross the ground plane. If the circuit topology is bad, this AC voltage (e.g. hum) is directly added to the signal. In a preamp, you also have to consider any leakage current flowing from the signal source, through the preamp ground to the power amp. Grounding toplogy does have very much effect how this currect affects circut operation.
One good idea for building a very high quality audio equipment is to keep the traditional star grounding topology inside the equipment and prevent any RF from entering the enclosure. By keeping the RF out of the main circuit board, the LF grounding can be more freely done, according to audio requirements.
To kepe the RF out, the equipment enclosure should be fully metallic and any removable panels should be tightened with screws (every 5 .. 10 cm so that they make good contact). For both input and output use use connectors isolated from the enclosure, but connect the connector ground terminal through a capacitor (1 .. 10 nF) with as short wire as possible. The capacitor will have a low impedance at RF, but 100 k .. 1Mohm impedance at hum frequencies and their harmonics, thus adding very little hum. This approach is recommended for all unbalanced connectors (RCA, 6.3 mm jack, 3.5 mm jack, BNC etc.). If you are using balanced connections with XLR connectors, wire the pin 1 ground directly to the case (do not connect it to the analogue signal star ground points).
For extra protection add ferrite beads to the wires inside equipment. Ferrite beads with a few turns of wire mounted close to the socket into both the signal and ground terminals of the connector will have an RF impedance of several hundred ohms but still practically no effect at audio frequencies. For adding effectively of this on input circuits, add a small 100pF-1nF chip capacitor after the ferrite beads from the signal wire to the ground (with 1000 ohm signal source impedance, this capacitor could be up to 1 nF, thus the corner frequency of 160 kHz). Do not use this kind of capacitor on the output circuits, since this can cause stability problems to some op-amps.
Tomi Engdahl <Tomi.Engdahl@iki.fi>