Isolating Disturbing Loads

Source Voltage drop cannot be minimised if they are fed from the same source. But, at least the bus voltage drop can be minmised, if they are fed from different busses. Moreover, the impedance between the two busses will reduce the interference from the non-sensitive loads affecting the sensitive loads.

Both the disturbing load and the sensitive load are ultimately powered from the same source. However, unless your power source is directly connecting to these devices at the local distribution voltage (like a portable generator) then all of the grid transformers are doing a very subtle thing that most people do not grasp. On every cycle of your power fluctuation the magnetic core of each transformer can only store and deliver a fixed maximum amount of energy. The grid upstream of this transformer can provide much more energy so the voltage sag upstream of the briefly overdriven transformer is minimal. By having the sensitive load connected to a transformer that is not overdriven (isolated) the sensitive load gets power with little to no fluctuation.

Connect the disturbing load to its own distribution transformer. The disturbance will not be eliminated - but greatly reduced.

Adding to post #4. An important effect is distance.

1. Suppose welder and sensitive equipment [SE] are both 20m from the power origin. Each is connected by separate cable,kept apart as suggested in post #4.
2. The cable run between welder and SE is 40m.
3. It is generally found that interfering switching devices are within 10m of the affected equipment, so 40m separation is good.
4. If the SE is powered from the same cable as the welder, there may be just a few metres of cable between the two.
5. N.B. ordinary power wiring has enough attenuation at high frequency to suppress the really sharp "spike" edges which affect digital equipment within the 10m mentioned. The PVC insulation commonly used has high loss at high frequency compared to e.g. polyethylene, used for high frequency signal cables.
   

If the SE is powered from the same cable as the welder, there may be just a few metres of cable between the two. what do you mean with this?

1. Suppose welder and SE are both plugged into sockets of one supply circuit.
2. Welder and SE will include, typically, 1.5 m of cable each (to its plug).
3. That is 2 x 1.5 = 3 metres cable, which interference must pass through between welder and SE. There could also be cable between sockets.
4. For small numbers, one can say "few", especially when the number is not known or variable or is much less than usual.

Thank you for your explanations. But I wonder what is the golden rule of this interference situ. Is there any rule you can cite?

One rule! Difficult!

1. Some interference problems will not be solved until you have done everything the right way - leaving out just one "best practice" measure or restoring one thing "as was" leaves the problem in "full health". But they are usually the cases where no-one thought about interference in making or installing.
2. The best Project Manager for whom I worked [his contracts made money - even when it looked like a loss when he was handed it] said often "Doing work on-site costs ten times as much as in the factory and fixing design problems costs even more". He was very supportive when a problem was identified. And very stubborn when what customers wanted had been identified in our tender as "not provided" but was in the customer's specification and unusual. I particularly remember a customer who wanted all control panels to be made of 3 mm minimum thickness steel - he got them - at extra cost!
3. In the same way, fixing an interference problem - finding out what it is and then curing it - is far more expensive once equipment is in operation than during the design and manufacture. If something is new to you, it is better to bear the cost of investigation and tests than have to fix them in operation with an unhappy customer.
4. If there is a rule it is "Always think IF an equipment you want to use is likely to be susceptible to interference or be a generator of interference - and specify some installation and equipment standards when there could be a problem".
5. Ask suppliers what EMC (ElectroMagnetic Compatibility - emissions and susceptibility) standards with which their equipment conforms. AND what EMC tests they have have done. Ask to see their test results! It is possible to declare conformity without doing tests! I well remember a foreign company which got a sub-contract for a governor on their low price. It turned out they had only done radiated susceptibility radio frequency tests, with the equipment in an "RF proof" cabinet and all the cabling under the metal floor of the test room. Since they were required to meet military standards, which include susceptibility and emissions at the cables and with radiated fields their submission was an example of what my first boss called "weasel wording". Meaning words carefully chosen and intended to "worm" out of a potential contractual obligation, like the long, thin agile predator called a weasel!
6. The European EMC standards are strict. I suggest you look at them. But they assume you use good practice for use of segregation of cabling and shielded cable/housings, when needed, as specified by the maker. But there is a "loophole" for the "weasels" in the susceptibillty criteria - it is up to the maker to specify what constitutes an acceptable performance during test, so if it were a meter they could have specified 10% error even though the brochure says 1%!
7. Note, the RF field strength specified for heavy industrial equipment to EN generic standards is 10V/m - but military goes up to 200 V/m on the outside of ships and aircraft. Motor industry RF standards [with so much electronics used] are now very good - if you park a car on a public road next to an airport 200 V/m is possible.