I am interested in ways to reduce/eliminate electrical noise in UT measurements systems. We frequently find ourselves inspecting highly attenuative materials at high gain and electrical noise is a problem, particularly from DC servo motors on our scanning systems. It is also a frequent problem with portable flaw detectors in electrically noisy environments. Any suggestions or comments would be welcome.
02:14 May-08-1998 Rolf Diederichs Director, Editor, Publisher, Internet, PHP MySQL NDT.net, Germany, Joined Nov 1998 602
Re: Noise Reduction in UT Instrumentation : I am interested in ways to reduce/eliminate electrical noise in UT measurements systems. We frequently find ourselves inspecting highly attenuative materials at high gain and electrical noise is a problem, particularly from DC servo motors on our scanning systems. It is also a frequent problem with portable flaw detectors in electrically noisy environments. Any suggestions or comments would be welcome. -------------
Electromagnetic noise can be transmitted to your UT measurements system through air or cable from the noise source, maybe a DC-motor.
You may work with shielding on the noise source to prevent electromagnetic waves in air. Sometimes it is better to put efforts on the source of the noise than doing a lot of work on the receiver side. We assume that the UT equipment is manufactured in a shielded unit. The probe cable and the transducer must be also shielded.
So far it seems that the shielding is simple. However, what electrical potential should lay on the shield. Basically ground, but what kind of ground? The center power supply provides the best ground, from there it is spread to different units (scanner, UT equipment, fane, cooling unit, pump, crane, ....) Usually at those units you can still use the ground for analog and digital devices together. However, behind that you must be carefully. For instance if the DC motor frame provides a voltage supply and you use it for the UT equipment as well, that can course a problem. The theory says that all grounds should be provided separately from one center point.
But often you cannot prevent that analog ground contact happens elsewhere, e.g. if your transducer is immersed in the water of a C-scan tank. That makes it often necessarily to connect the analog ground of the transducer cable with the 'dirty' digital ground at the tank together. In practice people trying several options, connecting thick ground cables from one place to another until the best result is found. UT systems designer commonly providing pulser receiver units as much as possible close to the transducers. For special applications pulser transmitters may build also inside the transducer, as so called active probes.
The main problem exist at the transducer side. The transducer needs usually a relatively high impedance matching. That means the cable shield can be inducted with a much higher noise than it would occurs with a low resistor matched cable. If you read 75 Ohm on your cable that does not helps, important is the load at both sides of the cable. It is obviously that this problem increase direct proportional with the cable length.
During the work it may help the use of a frequency analyzer which will show you the noise frequency spectrum. By switching on/off different devises you can locate the parasite of a specific noise carrier frequency and watch the spectrum until you find the optimum grounding.
Other sources of noise can be coupled into the UT unit by a wire connections, maybe by the gate alarm output to a SPS/PLC. An opto-coupled interface can help. A connection via a low pass filter (R-C or L-C) can be a more simple solution.
Finally you must be aware about that noise exist statistically, that means you never can prevent a false alarm coursed by noise, it may come just once a day. That lead to noise reduction methods which are provided by the UT equipment. Simple amplifier filter or other digital algorithm are in practice applied. Another forum message may look more close to this subject.
05:52 May-08-1998 Linas Svilainis R & D, Kaunas University of Technology, Lithuania, Joined Nov 1998 67
Re: Noise Reduction in UT Instrumentation
Mr.Garner from Alliant Techsystems has posted very interesting question which is complies with nowadays equipment level. I believe such organisation as Mr.Garner is representing has already been through various conventional techniques to struggle the noise. Rolf has already mentioned all the possible ways for solution search. But I believe you are interested in more rough methods of noise reduction. J.R.Barnes in his book "Electronic system design: interference and noise control techniques" says - "It's too late to care about the EMC problem once system has already been designed" I fully agree with him. It's the same as boiling the soup - before trying to get some sufficient results, some sufficient attempts have to be put here. First-some "conventional" techniques details. The noise band and source should be examined first. Once the noise is induced in receiving transducer cable, double shielding or twisted pair inside dense shielding plus differential preamplifier should help. If the proper cable shielding has been applied, then one could expect that usefull signal return current will be flowing inside the cable ground and noise will be flowing outside the cable shield. If proper shielding/ grounding of receiving amplifier has been applied, those currents can be separated and noise influence reduced. Sometimes ferrite bead with sufficient permeability placed near receiving amplifier outside the cable can solve the problem. You've mentioned the problem with DC servo motors. The spectra of such noise is concentrated in low frequency domain and looks more or less as "pink" noise. Proper damping of low frequencies might reduce the noise. Also, good choice could be the application of DC motors with rotating magnetic field instead of brush motors. Since there is less sparkling, it should echibit lower noise level. Better results can be obtained if time selection is applied-i.e. shanner head moves when acquisition channel is not active. Of course, in such case inspection time should be sacrified.
Now-more complicated methods: 1. Averaging, temporal and spatial will help to improve the SNR. Especially effective are various signal processing methods based on this aproach. The best known is SAFT - what is attractive in this method, that the spatial resolution is improved and the electrical AND STRUCTURAL noise are reduced pertaining the conventional samples number, so the inspection speed. Other representative here is the Split-Spectrum and Spread-Spectrum processing. In first case conventional equipment can be used, in second case more complicated equipment has to be used. Both are based on aproach, that structural noise is casued by reflections smaller that the defect size one is looking for. Tadeush Stepinski from Uppsala University has developed the adaptive procedure for parameters adjustment, so it can be easily implemented into equipments data processing package. Unfortunately, because of velocity dispersion with frequency, this procedure will work on limited depths. The frequency selective loss compensation technique, we've developed with coleagues from University College London, allows for conventional way of attenuation vs frequency compensation. Phase compensation is under development. 2.The time has come to forget the convetional way of ultrasonic imaging, when the reflected signal is directly presented on oscilloscope or computer display. When spike or step exciting signal is applied to excite the transducer, the data presented as A-scan is very convenient to analyse manually, so proffesionals in this field can tell almost everything about defect by looking at reflected signal shape. When it comes to highly attenuating materials, especially with high structural noise, manual inspection becomes complicated or impossible at all(as manual I understand the eye interpretation of A- or B-scan image). By applying the compex signals for transducer excitation, one can achieve high energy feed into material and better detectability of inhomogeneities. But the reflected signal direct representation on-screen is very confusing. Here is a lot of space for NDT equipment manufacturers - with real time generation of ecxiting signal and received data pre-processing very good results can be achieved. Presentation of such results on-screen might be confusing for old-fashioned operators, but possible further benefits are evident.
Hope my notes have been of some help for you
Goog luck in fighting this dragon
P.S. Optional image presents the SAFT'ed and temporal domain deconvolved image of point reflector.