Sign up Calendar Latest Topics Donate
 
 
 


Reply
  Author   Comment  
Shurafa

Sr. Member / Supporter
Registered:
Posts: 310
Reply with quote  #1 
Anti-aliasing Filter Details

Hello.

I wonder if crrently there are common commercial vibration analyzers that operate without anti-aliasing filters.

Im asking this to understand the practical value to teach this concept and its details in the signal processing of machinery vibration courses/books. The concept is quite important and can be generalized to sampling data other than the time based waveforms. Nevertheless, Im not in favor of the elaborated details usually considered as a part of the classic training.

My background and exposure are limited to certain kind of machinery and data collection. So Id like hear from other people specifically who set up their own systems and carry out non-routine vibration data collection/tests.

What level of details (of anti- aliasing) do you really need to use in your jobs?

Regards- Ali M. Al-Shurafa
OLi

Sr. Member
Registered:
Posts: 1,917
Reply with quote  #2 
It is for me to be regarded as classic, legacy information or should be, all modern system should have aa filtering. If they claim to be sigmadelta A/D they are builtin the A/D not saying that avoid every problem. However lately in regard to mems sensors the problem did emerge again in respect to a major analog device producer and post processing, downsampling etc Nyquist rule still rule and the devices alised like he..... In another case where the manufacturer didn't realize the severity it also was a problem. So for old manufacturer that passed that decades ago it is no problem, for new systems, startups where they don't have a clue and worse the the seasoned sensor manufacturer where you no longer can trust their datasheet giving the sensors to newbies w/o a clue....... it is a severe problem and may dent the credibility for the technology. It cost us 2 man year work because I am stupid and trust datasheets as far as I know the fault still seriously remain but the not true diagram is removed from the datasheet, that is a modern problem fix. So yes and no, it should still be warned of and known but unless you test a new system or new technology it should really not be a problem but do not always trust the datasheet regarding mems. Only my 1SEK. Olov
__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
Curran919

Avatar / Picture

Sr. Member / Supporter
Registered:
Posts: 599
Reply with quote  #3 
It may not be exactly what you are looking for if you are teaching some really hands on guys, but I would consider using downsampling to show the effect. If you have a greatly oversampled signal, you can downsample (as opposed to decimating) and still get all the aliasing effects of the frequency content you've cut out reflecting in your new bandwidth.
Shurafa

Sr. Member / Supporter
Registered:
Posts: 310
Reply with quote  #4 
OLi,

What do you mean by "mems"?

Regards- Ali M. Al-Shurafa
OLi

Sr. Member
Registered:
Posts: 1,917
Reply with quote  #5 
MicroMechanic sensor aka the same tech as you have 5-10 in your car to trigger the crash cushions etc. 
__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
MarkL

Avatar / Picture

Sr. Member
Registered:
Posts: 1,121
Reply with quote  #6 
Isnt that what WWT is using? or am I wrong, I have seen Analog devices on about advances they have made to them recently for use in predictive maintenance.
electricpete

Sr. Member
Registered:
Posts: 647
Reply with quote  #7 

I certainly don't have any experience designing any custom equipment.

 

I would venture to say that every major vibration data collector has analogue anti-aliasing filter built in ahead of initial a/d sampling.   Therefore it is not something most practitioners need to be concerned with.  Certainly not me.

 

If you elect something different trying to piece together a cheaper custom solution, than you would be responsible to review / evaluate the effects.  Primarily what kinds of frequencies might appear above half your selected sampling rate and what would be the consequences of allowing these higher frequency components to show up as lower frequency components below half sampling rate (would you be able to recognize them easily, and how much would they disrupt your ability to rapidly and accurately report and interpret the data).   And weigh that against the cost to include a simple (and possibly standard / integrated) anti-alias filtering.  fwiw from my small vantagepoint it seems that most often the cost to include anti-aliasing would be small and easily quantifiable. The potential for disruption might be low but certainly much less quantifiable (than the cost) and it seems like there could be a pretty big downside if the equipment or the human interpretter gets fooled due to lack of something that that the Monday morning quarterbacks would undoubtedly consider "standard".  Also I think analysts job is already difficult enough (especially when trying to identify unexpected/unknown frequency) without adding more options for him to consider (aliasing of a higher frequency).  Just my opinion, again not much direct experience.   Sorry if I missed the mark.

OLi

Sr. Member
Registered:
Posts: 1,917
Reply with quote  #8 
Yes it was a Analog Devices device specific made for this use that was not as the datasheet promised as the diagram was a excel sheet describing what they aimed to make not a verified documentation of the product they still sell, they admitted this but made nothing to fix it... It was by testing as you suggested it was discovered to be serious aliasing. Since you can't access the raw signal any additional filtering is not possible, the output from the sensor is already digital. You are locked by the software in the sensor in this case. So you need to verify the system with sensors before using them or it may bite your rear seriously. It is not a general problem with the technology but facing it with a what I believed a serious supplier is really bad and if it get widely used with this problem it will be really bad. There are other variations that is not bad in this way but have some limitations in hi frequency range in my view but who knows what hi frequency you need for bearing condition if you even care to bother with such, that is not always the case. Olov
__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
Ron Brook

Member
Registered:
Posts: 35
Reply with quote  #9 
I learned many things about digital signal processing working for Nicolet Scientific.  When asked to visit the engineers at Hydro Quebec to see if they would want to buy a 660 dual channel fft, they showed me their system. They had 'done it themselves' with a Mass Comp computer.  They showed me there data and asked why their plots had many more peaks than what my 660 screen showed.  I asked if they had anti-aliasing filters on their front end and they smiled ear to ear.  They had just ordered them that morning realizing their error!
OLi

Sr. Member
Registered:
Posts: 1,917
Reply with quote  #10 
I also run a Mass Comp system for vibration analysis on a turbine or actually endwindings a long time ago, 3-4 full height cabinets with a teletype console, that was mobility. I met my wife at that keyboard, many things you can thank a Mass Comp for. Olov
__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
OLi

Sr. Member
Registered:
Posts: 1,917
Reply with quote  #11 
Or actually I think it was one of these
http://bickleywest.com/modcomp.htm

You used the big buttons to enter the hex boot sector address and pressed load and lampen blinken poppen fusen and mostly it did boot.
Olov

__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.