Filter type usage

[[Ja...]]

I understand how the different filter types work but I’ve never seen anything as far as guidelines on when you should use one filter over another – IE: Band pass vs low pass vs high pass. Is it strictly a judgement call – like – Well, this is a really smooth surface so I’ll use a low pass filter. Or, this is a really rough surface so I’ll use a high pass filter?

I’d like to know what you personally use for guidelines.

[[Ma...]]

If you know how those filters works, then you should know when to use them.
Each will give you different result. You either want coordinates/position or you want surface/profile.
Then you have filters used on turned parts.

But i am no expert on this.

[[Je...]]

.

Please sign in to view this quote.

.

Martin, I think what Jack was mentioning was that he has not seen documentation that clearly spells out guidelines on when to use each type of filter in a way that connects it to surface texture.

I do think that Calypso needs much improvement on making this type of information accessible and front-and-center to the user. One example of an excellent improvement towards in-software prompting is found in version 7.8 (2024) regarding pattern and loop syntax (different issue than this post is discussing but it makes my point):


.

.

Also, this forum exists for the very purpose of asking questions like this in a safe, constructive environment.

.

[[Ky...]]

Hello all,

Perhaps I'm missing something, but we do have documentation on what filters to use when measuring particular features for particular characteristics in the cookbook. It fully defines all settings, to my knowledge, to make measurements as repeatable between machines as possible.

Besides that, really when measuring something with a CMM like ours, you will almost always be using low-pass. The intent of these sorts of machines are to measure form and location mostly, which is going to be best served by low-pass filters. I describe it in a bit more detail in this thread: viewtopic.php?p=57868#p57868.

I hope I'm not barking up the wrong tree here, I might have misread something, but hopefully this is helpful.

[[Ja...]]

Thank-you to those who have responded. Kyle - If I look at, say, planes in the cookbook (pages 87 – 100) (my cookbook version is 2015, don’t know if there are newer versions) I’m not seeing anything that mentions filter types. From what I understand, and please correct me if I’m wrong, the low-pass, band-pass, and high-pass filters are for adjusting to different surface conditions. I was under the impression that if you have a smooth flat surface say like that of lapped, ground, etc. I’d want to use a low-pass filter because I’m not really looking for spikes or dips. Conversely, if I have a rough surface say like that of a band saw, I’d want to use a high-pass filter because I am looking for the high spots. That’s why I’ve been reluctant to use a low-pass filter for everything, and why I was asking the question of what criteria do I use to determine the different filter types. What determines the transition points from low-pass to band-pass, or from band-pass to high-pass, or vice-versa? I wasn’t able to find anything in the cookbook. Am I looking at filter types in the wrong way? Or did I miss something in the cookbook?

[[Ky...]]

Hello Jack,

Thank you for clarifying the question. I did double check the cookbook and it doesn't specifically mention that they should be low pass. It does give the cutoff wavelength, but not the type of filter, which is strange. I could have sworn it was some where.... Maybe the new ones have it or perhaps they just assume that we cover that in class.

Regardless, even though the cutoff wavelength is dependent on the expected roughness (and this is mostly due to the fact that if you have a very low roughness, you will need a high cutoff frequency/low cutoff wavelength to see much of anything), the type of filter is going to be determined by the type of callout.

For a form or location, which should be the vast majority of what you are doing on a CMM like ours, then you will want a low pass filter because it is the low frequency information that controls this. You may occasionally see a waviness callout, which can be done on something like our machines and that is where you want a band pass (so you only see the wavelengths you are interested in). It is for callouts like roughness, which I wouldn't do on a CMM unless I had a ROTOS head, that you would want a high pass filter because roughness is specifically looking for high frequency data.

I'm going to try to give a practical example and see if this helps:
I had to play around with this because it is highly dependent on the number of points and the size of the geometry, but I made a program off the CAD here, specifically I am looking at the flatness of the wavy side on the +Y side:

I measure it as a freeform (50000 points) and recall it into a plane. I then make two flatness characteristics, one if filtered with a .25 mm cutoff wavelength low-pass, the other with a .25mm cutoff wavelength high-pass.

Here are the results. The flatness on the high-pass is pretty good, considering the "dents" on that plane are something like 20 mm from peak to peak. The low-pass, though, reflects that this is a very bad plane (which we expect).


If you want me to upload the program let me know, I can do that (just keep in mind it is in Calypso 2023), but for the time being, I'm not sure what you mean by "high spots", but you might actually want to consider decreasing the cutoff wavelength to increase the number of higher frequency items you are catching. You may need to significantly increase the number of points though.

Hope this helps.

[[Ja...]]

Thanks Kyle. The examples helped a lot. A couple more questions. 1) So would a high-pass filter be good to use on something like a machined surface where the cutter is moving too fast so you have lots of little ridges to deal with? In that case I’m not looking for a surface finish specifically but it would affect the results in a similar manner. 2) For band-pass, you mentioned looking at only wavelengths I’m interested in. How am I to know what those wavelengths are? What are the guidelines?

[[Ky...]]

Hello Jack,

So, for the first question, my question in response is are we talking part of a callout (like flatness or position or something), or are we really only interested in like showing it to someone to show an issue or something like that? If it is the former, I would really not recommend using a high pass filter because it really gets rid of the form, which is mostly the larger wavelengths anyway. If it is the latter, I can see maybe using a high pass filter to get rid of form to see that. You will need to estimate what wavelength those would be and choose the cutoff wavelength appropriately (I don't know about the specific standard, but a Gaussian filter has something like %50 attenuation at the cut off. RC circuits have like 70%). Specifically, you would need to figure out how far the peaks of the particular features you are trying to find are (if it is from like a tool bit or something, you might estimate it by how far it travels along the piece in one rotation perhaps) and then make sure that is well inside of that cutoff wavelength.

As for the second question, usually when I would use a band-pass filter, it would be in like a waviness callout and I think most that I have seen, admittedly I don't see it often, list the frequencies it cares about in the call out, so that's how you would figure out when you need your cut off wavelengths to be. Besides that, you usually have to make educated guesses, as I mentioned above if you are looking for specific tool marks, you can base it off of the specifics of the cutter machine and just give a reasonable range around it, or sometimes you just need to ask the design what they are interested in if they are asking for it.

A couple things I should add, though:
1. Make sure you take enough points for the filter in question, otherwise Calypso will not apply it. I was having issues with this a bit yesterday when I was making that example program. I believe it is at least 7 points per wavelength you need.
2. Remember that the stylus effectively acts as a morphological filter. This knowledge base article is specifically for freeform features, but the idea is still the same. Since the probe has a finite size, it cannot fit into everything. https://portal.zeiss.com/knowledge-base?id=400534
3. Lastly, and I have alluded to this before, a bridge type CMM is mostly meant for measuring form and location callouts. I have done waviness, at least a little bit, but to really do like surface finish and the like, you would likely need something specialized, or at least a ROTOS on the machine. I seem to remember when I went through AUKOM I was shown a document that showed the different type of finishes and what they represented physically and we were discussing where CMMs were most useful for, and it started tapering out right around tool marks I believe. I can't find it in my notes, unfortunately, so I can't verify that, but just remember that the machine is really meant for the lower frequencies, like form and position.

I hope my answers don't come off as too vague. It is hard sometimes to figure out what a customer needs over a forum like this and even if it wasn't, there are always exceptions to the rules.

[[Ja...]]

Thanks for the clarifications Kyle. They’re definitely appreciated. Yes, I can verify the 7-point requirement. In fact, it needs to be 7 points or more after the filtering is done so one needs more than that to start with. Especially if it’s a surface with a lot of form error. I’m a strong believer in high point density so that’s not a problem. For the mechanical filtering aspect, I use the smallest tips I can. Most of the programmers I’ve run across don’t seem to care much about the mechanical filtering aspect of probe selection, and also aren’t aware of the ISO 12180-2 guidelines for the largest probes to use in a bore. To a point, I get it. Personally, I’ll take accuracy over programming/run time any day. And yes! I’d love to have a Rotos! That’s what I’ll be pushing for down the road. But for right now, we need an O-Inspect Duo a lot more. 😃

Thanks again.

[[Ke...]]

Please sign in to view this quote.

As long as we are here...
I have been asked about "Harminic Analysis" of precision bearings... bearings that are too large for our roundness machines. Is that something that is possible to evaluate properly on a ACCURA II, with a RT-AB Rotary Table? Digging around in Calypso, I was wondering if this is something that the Waviness/Fourrier Analysis would handle. As an added curveball, we often "benchmark" our competitors, and I have no specification to work to 🤣

[[Je...]]

Please sign in to view this quote.

I'm with you on the O-Inspect Duo. It looks like a great solution.

[[Ky...]]

In answer to your question, Keith, a "harmonic analysis" does make me think about waviness, specifically the Fourier analysis option in there. I'll show an example of what it looks like, but it will effectively give the amplitude (or peak-to-peak value depending on settings) of the harmonics. Do bear in mind, though, I think the tolerance and values are actually set in microns in the waviness characteristics, but still show up as mm in the report I believe.


Unfortunately I can't really give you an idea of the harmonics to look for. That would require application specific information that I really don't have, but hopefully some of your engineers might have an idea on.