To address repeated comments, the glass rod only needs to be positioned parallel to the axis by .050" end to end. Any angular variation under this .060" positioning is smaller than the .0001" resolution of the probe and is a moot point. We easily achieve parallelism of the rod by positioning the orange supports flush to the edge of the table. If an operator can't position it within .060" of flush, I'd have larger concerns for them. No, we don't want to make stops on the supports because it would limit use of the device to only the X axis on other machines. "The material Zerodur has better CTE characteristics." - The CTE of quartz glass is far less than the margin of accuracy of the probe so any better performing material is overkill. Also, it's impossible to buy a single piece of Zerodur for $10 and have it arrive the same day from a supplier.
would it help to probe the rod straight,just for the heck of it,just dont forget to canle the rotation afterwards or you will have wonky parts ,i know extreamly small numbers but you have it for free
forgot to mention that you can calibrate against a sphere that you use to calibrate cmm,first calibrate the probe to the sphere and then calibrate the machine with the newly calibrated probe.we have macroed our hermles so they grab a palete with a sphere mounted to it every 5 hours and calibrate the machine in all axes automatically
Make a rod from sphere, so both ends will share that sphere's radius. So any aligment up to couple degrees would be almost zero. But I am not sure if it can be manufactured.
For the rest of the World who use metric : the Glas pease is ca 305mm long , if you have it 1 mm wrong to the X Axis you have a length diferenz of 0,0016mm. a Normal CNC-Maschine have about 0,015mm on 1000mm accuracy. I dont know how accurat Haas maschines ar but i think the are not thits accuract (because the have no glas-scale)
RAMTIC parts-production concept was developed by Renishaw in the 1990's. They used a similar method but probed a master part, then applied scaling. I really like your approach Jay Pierson and plan to try your method in our tool and die shop.
I have added quartz encoder scales to my manual mill... I get the thermal accuracy of the quartz measurement system independent of the machine temperature errors. It also serves to provide direct positional feedback regardless of backlash because the scales are measuring the actual table position rather than motor shaft location.
@@TT-yu9uo yes they're an option on most machines. The problem is, when buying machines tools, glass scales are a factory build option only. Usually there is a long wait to get a machine from the factory, but the dealers bring a steady stream of machines in that they can offer immediately, all without glass scales lol
Yeah all things considered you're pretty much going to get the same level of precision, accuracy and repeatability with that methodology as with this dude and his HAAS. There are so many things this dude is ...just ugh posturing. He's like the opposite of Ben at Applied Science in terms of reporting the results of his scientific excursions and being all full disclosure about what he doesnt know. (I guess this sort of makes sense since this channel is about promo'ing his CNC business vs Ben who just does it because he likes science) Like oh wow, the CTE of alu is ~40x worse than quartz, sure. OK now what sort of cut is that quartz? There's a reason why AT vs SC vs BT all cost different amounts of money and are used in different oscillators based (hunt: quartz is not an isotropic mineral-- thats why a quartz watch crystal oscillator can cost 30 cents vs a rubidium standard that goes into a cell tower which is going to cost $300. See I can drop knowledge bombs about materials science too, granted I'm sure some PhD is going to jump in and school me good). If he's just using a rotary carbide bit or a surface grinder (or even if he lapped it with some diamond polish), unless he hit the right "grain of the glass", hes introducing systemic issues into his "transfer standard". (And arguably worse than that he's instilling false faith in his systemics). @Marco Reps is an EE who dabbled into building a CNC for..some reason (I think he's friends with Stefan G, those damn smart germans). He ended up getting pretty consistent repeatability within the nanometers. He did it with just closed loop feedback servos and the tried and true MIT design that Slocum came up with (gantry-based, as much mass as possible to dampen vibration out, 25mm ball screws with 2 aux. guides per axis). In retrospect, he would have done what you mentioned (LVDT scales, rather than shafts/rotary encoder absolute feedback). He also mentioned some weird damping aggregate, that I forget. If ya'll wanna see some real impressive work, go look at Huygen's Optics'. Or optics work in general. Zee germans and Japanese can charge 100k for a lens because their aspheric cine lenses are better than the Chinese clones on literally the nanometer scale (the entire visible light spectrum is only 300 nanometers.. so you can see chromatic abberation with a green looking "wrong" at 20 nanometers, especially on an IMAX screen)
@@wither8 Huygens Optics is Dutch, not German. But yeah, along with the Germans and the Swiss the Dutch are known for producing the most precise and insane tech on earth too (cough, ASML, cough).
I think I've done something relatively similar on the z axis just without a "known reference" so its all relative. We have some parts that required countersink diameters that were plus minus .01 but the surfaces were usually warped and we couldn't face them either. What we did first was split the countersink cycles up and go back and forth between the probe, drill and countersink which hurt our cycle times. What I was able to do was probe all the hole locations beforehand and store the Z deviation in macro variables for each location and compensate depths inside the countersink cycle all at once. Saved a lot of time!
. . . Wow. . . haha such a simple concept that could make a huge difference to those who can only afford, well, affordable machines! Learn to think outside the box a tad, and it can take you far! Sweet idea, simple to execute.
I worked for Ellenef in the 90s. We built wings and structure for f-14s. Used lead hammer, dial indicator, Allen key and a steel block with a known value, placed on the fixture sled--to set the cutter height and check for runout. Did not seem very high tech but it was good enough for them.
By big takeaway is your last line of code. Put something on the screen so the operator has a general idea if the machine is hot or cold. Without that, we're basically machining in the dark!
The rod can be at an angle as long as it is probing in the same spot then the length it becomes a constant. By the way Mr. Piercen you're a dam genius keep doing what your doing🤟
G103 P1 Limits look ahead to one line at a time. This prevents macro variables from being read before their correct value is calculated in a previous line.
I have seen it done an even cheaper way. The owner of a shop I was working a wanted an ice machine for cold drinks in the summer. He happened to see a couple at a motel on his way to work. The motel was under renovation and owner was happy to get rid of them for free. We would fill 5 gallon buckets up with ice and mix some coolant in. Then dump the cold coolant in to the tank. We did not monitor the temperature with thermometers, but the temperature inside of the machine enclosure would be much cooler after adding the cold coolant.
I like this idea. Would have helped me on some projects. An added suggestion on a machine like a Haas is to run the finish passes for whatever your critical features are in the shortest amount of time after probing as possible. If you have a long cycle some of these machines may be .001-.002 off 30mins after you probe your zero.
Great Idea - need to build one soon. I see its main use as verification to confirm your machine is operating within parameters rather than scaling. I believe if the temperature is far enough out that there is a significant scale factor the material itself (aluminum, steel, etc) will have its own thermal expansion issues. The part will be machined accurately but at the current temperature of the material. With an accurate temperature of the material this could be compensated. The temperature of the coolant may dominate with flood coolant and a long enough cycle time
Exactly! Finally someone totally gets it. If the "accuracy" macro displayed is pretty big (±.001") you should probably think about other thermal factors including material, coolant and ambient temps.
its probably easier to heat the colant,would it mess everything up if you mesured the highest temp in the machine and then add a few degrees with a heater? you would get a warm workpiece but stable temperature?
@@kristiankautto7616 heating/cooling coolant is what we do in my shop. 75° F is the temperature that works well for us year round. Coolant is heated/chilled after the pump and before entering the machine, we have parts in 6061, 7075, and 316 ss that need to hold 10ths or less and we do it with Haas machines. I have a 1 hour warmup routine that is implemented on startup days, say Monday morning, to allow the machines castings to stabilize then the coolant never stops circulating until the machine is shut down. Additional pumps have been added to flush the enclosures that run continuously which helps hold temperatures, and billets are warmed or cooled in a similar manner prior to machining as well as tools all calibrated at the same temperature. We used to have check routines, and we also used to use compensation and scalling values but the additional work was inevitably more costly compared to a few pumps and good chillers. Simple and reasonably priced modifications to the machines have removed any need for human involvement or error. What he is doing in this video is great for a job shop machine and is frankly genius but unnecessary for a production machine running continuously.
How do you keep the rod parallel to the axes? What about measuring the angle of the rod and correcting the result by the cosine? Another suggestion would be wrapping the rod in an insulator (styrofoam) and keeping it in a tempered box. If put in the machine for a short while it won't have enough time to change its dimensions due to the insulation.
See pinned comment at the top. The beauty of quartz is that you don't have to worry about any of the thermal mitigation you mentioned. An extreme temp swing of the rod (for a machining environment) of 25⁰ would still be under the resolution accuracy of the probe.
I think that's running Linux CNC under the bonnet? So you could do loads of stuff even more accurate than this. Add a temp sensor. Linear glass scales. Have it run a ball screw map every so often. Etc.
@@3dwezzy740 only thing I would buy from them again is a 1100MX the 8L lathe and plasma table have been a PITA with little support. Currently their support is refusing a simple request to transfer the 2 year warranty on a servo that is bad. They purchsed the servo from clearpath. we contacted clearpath and they refused because it was purchased by tormach and not us. Tormach refuses to understand a simple request to release the warranty to us so ClearPath will replace.
Seems like a great idea! Only issue I see is that when you place the device on the table, does it matter that it isn't perfectly parallel and perpendicular with the machine axes?
Edit: (I may be Wrong on this) Because the probe tip is a ball, as long as the ball is touching the opposite face (and this is why it was so important the faces were parallel) its self correcting.
@@NexusTrimean no its not self correcting. it may not be much, but see for example a parallelogram, when you probe the two angled sides, the measured value will increase. Yes i have validated it myself in a cad program :)
@@ipadize If taking single points yeah you could get some odd results if it's not perfectly straight to the machine axis. Taking a line on either side would help to correct. Ideally would get parallel to machine axis and probe X & Y separately. Not sure if the haas control has a single scaling function or allows individual axis adjustment.
The way its shown (single point) yeah those are super important. If you probe lines on either side you can correct for some angle, but best practice to isolate an axis as much as possible. I'm assuming they put a bit more work into this than the video shows, working it to be nice and straight when placed on the table. I know we 3d print some CMM fixturing and definitely have to do some minor post-work if we need things exactly the same height.
I would argue that you want your "master reference" made from the same material as the workpiece, not a material that has a different CTE. If you are only comping for ballscrew growth why would it matter if the length master length changes with temp? Said another way, if you care about safe guarding against all thermal growth (ballscrew, coolant, ambient air temp) then your reference standard needs to grow and shrink at the same rate as your workpiece. Use the following thought exercise: measure you standard at 21 degrees in the metrology room and record its length. Take it out to the shop floor where it is 25 degrees, and calibrate your machining process. At 25 degrees your standard and your workpiece now match. Take them back to the metrology lab and let them cool to 21 degrees. If you part is alum and your standard is glass, your workpiece will now be undersized because it shrank a lot more coming from 25 back to 21 than the glass grew going from 21 to 25. We use this process for a production job where we need to hold 2 tenths on 12" long aluminum part. The aluminum master reference is bolted to the side of the pallet and we use it to cal the WCS before a final cut. This is also why carbide gauge blocks exist: for tool and die makers who work with carbide punches. Having gague blocks of the same material as your workpiece makes you impervious to thermal growth. As long as they match, they match,
Lot of good stuff you mentioned, but careful to not mix a few subjects. The quartz rod also takes imperfections of the ballscrew into account. You definitely want to measure accuracy of the ballscrew with a known, thermally stable artifact like the quartz rod. If you wanted to measure the thermal status of a workpiece, then as you mentioned use an artifact of the same material that has a known measurement at a known temp. Carbide gage blocks under 1" have essentially no measurable growth under normal shop conditions (±15⁰) so it's most likely for wear resistance when working with carbide tools. Finally, what fresh-out-of-college engineer would design a part with ±.0002 over 12"? This is why engineers need hero machinists to set them straight.
I agree with Jonathan, your goal is the accuracy of the final product, and not the error of the machine ball screw. If you really care for accuracy, then a Renishaw ball bar test will give you the best results. As far I as know, Haas doesn’t use chillers because they use a software algorithm built in the control to compensate for thermal growth, and in my humble opinion it works good enough to hold half a thousand, I can come in the early morning and continue to work without any huge differences like in the old days. I’m sure you know you can see the machine temperatures in the control lol. Also, if you working with small tolerances, work in metric. 1 micron compensation = .00003937 inches.;)
@@PiersonWorkholding A master gauge of any material takes ballscrew growth in to account as well as environmental impacts of the workpiece (if matched with CTE). There isn't a mixing of subjects. Think of it this way: if only your ballscrew is growing, why does it matter if the master gauge is impervious to environmental changes?
Thanks ,… this is powerful …. Thanks for being so selfless and sharing so much of your knowledge it is shaving a decade at least off my learning curve as i prepaid to start a shop, thanks
I explored and experimented with these and similar concepts for thermal compensation over twenty five years ago. My employer said I was wasting my time and eventually fired me for "not being a team player". Oh, and I also perfected eight inch long valve spool bore finishing to within 0.00005" (that's five millionths of an inch) roundness, straightness and camber. With selective fitting of valve spools, we achieved less than 5 cc/min leakage at 3000 psi.
Hi i just wanted to add my pot to the fire :) There is a strong argument to not use quartz or zerodur. What i think is the material for temperature scaling should be out of the same material as your machined part. If your machine is hot say 30deg, and your part will be 30deg as well. Your calibraiton bar is going to be same temperature(or some other calibraiton artefact) if it is not from the same material once the parts cool, thy will contract differently. I am a beliver that the calibraiton artefacts should be designed and used and sett on the machine all the time
Don't confuse what you're measuring. This technique is only for measuring the machine accuracy. You're right about measuring an artifact of the same material, but that's measuring the workpiece at a specific temperature, not the machine.
With G103 P1 make sure you use it again following a renishaw shutdown command. Part of the shutdown routine re-enables look ahead. This can of course screw up any math you do.
Great tip! It's also possible to adjust servo scaling in the Haas control, but it requires a key to access on newer machines - have you tried this Jay?
Well knowing the thermal coefficient, calibrated length and measured length you can using simple math add an equation to calculate real displacement including thermal expansion of quartz rod. This is for super super accurate machining if not even grinding.
I really hope that quarts rod is indicated in straight before gauging the length. In the video just slapping it on the table with unprecise 3d printed stands witch throw x, y, and zed axis's off by id say more than a couple thousandths will have a big impact on precision. Well aerospace anyways, if where making farm parts id say its good to go, best dam farm parts in the land! If that rod is not indicated in within all three axis of the machine; I hope you have brushed up on some trig to throw all the deviations from the gauge setup into your C++ program. If all that was taken into consideration before this video id say yeah this is pretty banging video!
Hi i think your rod is not "perfectly" parallel to "X" axis. So measrung of lenght is wrong and will be diferrent every time you put rod in different angle to "X" axis. But maybe im wrong.
I don’t think it matters since they are probing both ends which were ground parallel. If you imagine the right side being angled and probing on the high side, the left side of the rod would be probed on the low side which would average to the correct length. I think this also applies to the potential error in height of the 3d printed parts. I’m happy to be told I’m wrong though.
As I mentioned earlier, the edges of the orange holders are positioned flush with the edge of the table. You would only start to see error once the misalignment is over .050”. And if someone can’t see or notice .050”, there’s no way they’ll keep working for me.
i think if you measure the rod on Coordinate-measuring machine you will find pretty accurate number (if u make coordinate system on rod). But on CNC machine rod is putted randomly on table not respecting coordinate system so you measure diffrent lenght... every time you put it in different angle you get different number. But maybe im wrong.
Happy Thanksgiving Pierson team and thank you Jay love the idea! Unfortunately (or fortunately for the machinist) my current shop usually plus minus 10 thou is enough for us but a beat trick to keep in mind none the less. Agree with you the getting it slightly out of angle is a moot point in terms of the machines accuracy.
great vid its detailed while being a reasonable length, the method is rather complicated and jank The CNC manufacturers are probably to blame. I'm used to taking the expected travel distance divided by actual travel distance and then multiply the current microsteps per mm value by the result. Then save the new value directly to the firmware. Essentially the same thing but simplifies gcode generation reducing errors. If this process was automated and part of homing the machine at every program start, that would awesome.
Wishful thinking on editing firmware. Machine builders have those pretty locked down. You still have to sample something to get your known and traveled distance and our method might be the solution.
@@PiersonWorkholding I probably could have emphasized it more but i meant to put the blame on the manufacturers. A machine shop could probably make loads of money if they start selling kits for industrial use not just hobbyist or garage operations that can be put together using open source software and boards. Marline already supports CNC, 3d printing and Laser cutting all one would need is to reach out to some board manufacturers like BIQU to support higher amperage drivers. 3d printers would be an other decade or two behind if it wasn't for open source and people printing parts for their printer then editing the firmware to enable new features.
This is a pretty brilliant idea, has me thinking about how I could mount a reference piece in my NTX to check the screws. Also, I could see with your system using it to adjust work offsets as the machine warms. Hell in fact, one could set G54 at machine zeros, and use program coordinate shift in the program in lieu of setting the work offset on the origin Then as you check the machine throughout the day and use scaling to apply a thermal comp the work origin would be adjusted automatically. You could finish the system out by mounting the rod on an actuator so that you can bring it into the probing envelope with an M code and periodically check your comps. Of course, it would be critical to make sure thermal scaling is applied accurately when you are finding the work offset, and you would want to write your work positions into variables. Might need to edit the post to call G54, G51 then immediately call G68 or whatever it is in Haas to shift the coordinate over
Do you have the supports placed near the Bessel points? Measuring something in the vertical and then laying it horizontal can affect the length. Have you performed the same measuring cycle in the Y axis as well to see if there's any discrepancy? I'd be curious if you got similar results.
@@PiersonWorkholding this process hinges on how accurately you can measure the rod. For the average electronic height gage linear error could be as much as .0002-.0003" over 12". FYI, Haas has a warm up function that applies a correction over a specified amount of time
@Pierson Workholding How about dealing with ball screw wear and doing all 3 axis' and them being different numbers. So instead of dealing with heat (or the lack of), dealing with wear?
Ballscrew wear would require a ballbar test which would cost you $250 minimum. This is an accurate, inexpensive first step to see if you even need a ballbar test to begin with.
Does the calibration factor tend to be the same on all axes? Seems like up/down is likely to be different because the ways are typically in a different place. Would be cool to have a 3 axis calibration system with a cube of glass
I can see that using the quartz rod will allow you to calibrate the machine. But what about the material being machined? Doesn't matter if the movements of the machine are dead nuts accurate if the material being machine has expanded in size due to heat and will shrink later as it returns to room temperature.
Careful to not confuse multiple factors. Material, machine, and ambient temps are all separate factors that should be measured with other devices, i.e. thermometers. The featured tool is only meant to measure and correct machine errors due to thermal expansion.
You are mistaken about how much a 5" piece of aluminum will change over 5°F. I think you meant to say it will change by three ten-thousandths (.0003") of an inch, not three tenths (0.300").
You could eliminate one step in your Gcode for scaling. Let's say the rod was 12.0000, and your measured length was 12.0005 Just divide 12.0000/12.0005 and that's your scaling factor. Multiple by this value everywhere as needed. You can get rid of the #903 variable (unless it's used somewhere else!)
You are correct and yes, we absolutely need the 903 variable. We explain it later in the video as one of the pieces of data to display on the control. It's a number that the operators can best relate to.
@@PiersonWorkholding, I'm probably in the minority here, but anyone truly calibrating a machine that doesn't indicate their standard on the machine wouldn't be in my shop long. That's just me though, I'm sure most would be fine with it.
Awesome Video! I love learning tips like this! Two questions. I know the probe is extremely sensitive but is it possible it could move the rod slightly if it is just free floating in the 3D printed stand? Is there a clamp or screw to keep the rod firmly in place within the 3D printed stand?
The glass isn’t free floating. It is pressed into the holder almost to the point where we though it would break. Fortunately quartz is ridiculously strong too.
with this method you are only probing on the x axis, are you just taking the x axis difference then applying it to x, y, and z by assuming all 3 ball screws are affected by the same temp?
How do you guarantee that the rod is precisely aligned with the X axis of the machine. If the bar has a tilt, either horizontal or vertical, it will appear to be longer than it actually is.
How much force does the probe put on a part? Honesty question because it seems like it could deflect the 3D printed plastic ever so slightly. I know the length assists but I would have made those pillars much shorter and allowed one end to float so as not to stress the rod even the slightest bit. Does that seem reasonable or am I being too precise? Thanks for the video though, it looks like a very inexpensive thermal compensation method.
I love the idea, but I am concerned by the error that could come from the installation angle of the rod. As it is not installed perfectly straight, the measured length is slightly bigger that the real rod length. I didn't run the numbers, so I am not sure if this error can be neglected or not, but that is definitely something to be cautious about.
Sure ? In my opinion you have to Line it with the Axis ( in x and z ) or i understand somthing wrong. ( Im from Germany and my english is Not the best,also i get confused with our inch- thing;-))
Looks like some Factory400 code in there, am I right? Or is that all Jay? Either way, it’s Great Stuff 👍👍💪💪😎😎⭐️⭐️⭐️⭐️⭐️. You probably already know by now, but if Block Lookahead were off the calculations could occur w/ ‘old’ data as it could be executing #___=__#____ before the desired data is assigned
I've never dealt with it, but I have seen a spot in Haas parameters for thermal expansion. Isn't this, exactly what those are for? I could be mistaken, but that's always been my understanding. Seems like it would be easier to update there, than run special programs from G51.
The thermal compensation in the Haas control is based on correction distance over time. It is meant to compensate for the machine warming up over the first part of the day. It doesn't actually take a temperature reading.
I'm not a machinist, but I have a question. In the video you probe the rod in X only to obtain the scale factor. How do you know that factor is the same as Y (or Z for that matter). Also , how do you know that you've aligned the rod perfectly in the X dimension? i.e. If the rod was rotated in Y ever so slightly then your probe's measurement would have a sine error in it, no? Seems to me that you'd want to set it up with a machinist's square with respect to the bed and probe, rotate it 90 degs and probe again. What'd I miss?
doesn't this machine have a glass scale? I have just ordered a set of glass scales last week for my CNC and I will set up the closed loop via this rather than the ballscrew pitch.
I suppose it would work easiest in a lathe with a probe. Lots of lathes don't have 12" of travel in X either so you'd have to use a shorter rod which reduces resolution.
At this level of precision ... Do you have to consider the bending of the 3D printed holding bases while probing the ends of the rod? I don't know the amount of force of the probing procedures.
Would there be some margin or error if the rod isn't parallel to the axis its being measured with? If those squared edges are canted, i think there would be some error introduced.
Zero dur glass has an even lower coefficient of thermal expansion than quartz. But in the end you will be limited by the accuracy of the measurement system you are using in most cases.
Ok, you calibrate the machine, of course likely at a significantly low or high shop temperature. Then you take some aluminum and machine it. Which temperature has this aluminum? The same as the room temperature? Obviously. Oh, what was the temperature coefficient of aluminum...? Means you are anyway trapped, with or without your quartz scaler you will not produce a good part at high or low shop temperatures. Your compensation is even working against a gain of accuracy, because it reduces the growth error of the machine and therefore increases the difference to the also grown material. You may of course do this compensation and run the machine at normal room temperatures. You will only do that compensation when there are significant deviations in the machining result, which then is a thermal problem of your machine even at normal temperatures (e.g. no core cooled ballscrews). For a very short machining time the compensation will still work, but with a long runner you will get more and other thermal problems which the machine obviously haas. 🙂 My conclusion: There is only one way to get precision. Work in a temperature stable environment and use thermally stable machines.
All true but you're oversimplifying the conclusion. Machines, parts and shops are always changing temperatures throughout the day. This method helps identify the current accuracy of the machine.
@@PiersonWorkholding I can‘t oversimplify it. It is only about a room with a precise air conditioning and highly accurate machines with an advanced temperature management. So that‘s just money. It is much more difficult to find skilled operators.
Sure, Z could be calibrated but would require another type of stand and some extra programming. Circular interpolation is more likely to be managed through firmware and backlash.
So if I get this right... This is the ever lasting way of making sure that years of wear and tear, will make sure.. you will be within tens... right? Just make a same scale for X....Y....and...Z ? I own a Okuma that has 4 digits after the dot(rest of the world)... I still want to have this under controll.... My head screams this is a yard stick... if your diviation of messured and reality is X.... maintance is needed!
Exactly! It can also be used to check ball screw wear and backlash. If you move the starting probe position to different parts of the table, you could check different sections of the ball screw for wear or damage.
"Micro inches"??? Just use metric properly alright? It's hilarious to watch people do anything to avoid metric and fail at it. Great idea no matter what units you use!
One critique would be that this measurement can't exclude machine backlash? (Which is likely of the same order of magnitude as you want to correct here, I think?) Depending on your milling operation this could be good or bad. Perhaps you could add an extra variable to subtract the backlash before calculating the scale factor? (Likely this is close enough constant)
@@PiersonWorkholding I think I would like to see you measure that? My experience is different. Put a 123 block at the end of the rod. Compare it's length measuring in the same direction at each end Vs the direction reversal when you just measure both ends. I would be surprised not to see quite a few microns difference?
@@PiersonWorkholding Trust but verify. A machine may not behave the same after several years of abuse as it did the day you received it. I think using the spindle probe is also a bit sketchy unless you're obsessively calibrating and checking the calibration on your machine. The temperature of the machine and gauge you use when calibrated could itself throw off all future measurements. Temperature differences do matter but this way of working just compounds errors in a manner that could make them painfully difficult to diagnose. Is a part out of spec? Was your gauge out of alignment? Is your probe out of calibration? Does your machine have some backlash ... The biggest problem with your solution is that it compounds these errors into a single complex variable that then affects the entire geometry of the part. Ideally, you'd probably just run critical parts and features on a machine that's been adequately warmed up in a climate-controlled environment ensuring reliable, repeatable performance with simple-to-diagnose issues. I personally think it's a solution searching for a problem. But the kind of problem it might find is the hair-pulling-inducing kind where everything is going wrong and you don't know where to start fixing it.
Would your measurement be off If that 3D printed part isn’t perfectly flat so taking it in and out of the machine get different results? Since it could be tilted slightly
2:31 - Wow I never knew it actually has the capabilities to warm the ball screws... interesting. Makes sense, needs to have dead on accurate temperature, regardless being chilled or warmed. 4:16 - Would be cool if you talked in unit of measure using microns instead of inches when speaking of high precision and accuracies. Precision fanatics only speaks in microns and sub-microns(nanometers) lol.
@@PiersonWorkholding "floor machinists" are not considered "real machinists" and have no clue how to design blueprints ordered from SpaceX, NASA, Boeing, and etc who needs micron tolerances... I guess this youtube channel is just another typical "floor machinists" content minded. Only a handicap would speak the language of defining small unit of measure using microinches (inches) and a professional machinist (not a floor machinists) using simple notation such as 1µm (1 micron = 0.000039"). The reason to use micron is because people like physicists/scientists/inventors in US and all over the world who sends CNC work to "professional machinists" always uses microns notation to create their stuff, they are taught in University to use microns and nanometers when talking about extreme small unit of measure, but in the "floor machinists" world, they still use inches... 🤦♂To put it simply why it literally sounds so stupid to use inches for small unit of measure specially in machining: Why not use "foot" unit of measure? 1 microinch is 0.00000008' foot. I'm sure you would work happily as a floor machinists using foot notation just like inches.
Doesn’t work would still affect the “length” or I guess x axis component of the length. Only thing that truly wouldn’t change would be a circle or a solid of constant width, as cool as a Reuleaux triangle gauge would be I think I that’s rather impractical
I see an issue with your G51. You are probing along the X axis, which determines the error of the X axis, however you are not probing Y or Z, yet you are applying the scaling factor derived from X to all axes. You are assuming that HAAS has thermal comp on all 3 axes and that the thermal error is the same for all 3 axes and that the HAAS control is compensating exactly for all 3 axes. Once upon a time they only had thermal comp on the Z axes.
@@PiersonWorkholding I get the premise, Quartz has a very low coefficient of thermal expansion and can be used as a standard, but you are stuck with applying the X-axis comp to all 3 axes at once because G51 doesn't have discrete scalars for each axis. I think this might be better as a test for a warm-up cycle to determine when it should end.
Hahaha, I thought you guys only used freedom fractions of the inch like twelfths, sixteenths... onemillionfortyeightthousandfivehundredseventysixths etc.😅 BTW great informative video.👍
Good idea, I mean I would just probe a large ring gauge and that would be good enough for me. I have 5-axis mills so the center of rotation I find is more important. Saying that, if you're going to these lengths to get your machines cutting this accurate then you would have to be making parts that are thermally controlled throughout their life too otherwise this is kind of pointless.
@@PiersonWorkholding well a large ring gauge is normally free if you have a large bore mic ;) I know it possibly could be nice to know this information but do you really have parts that need this sort of accuracy as all this would increase the accuracy on is true position. As far as diameters etc you would always change the offset? Like you would have to change your process to make the most of that accuracy, you couldn't drill and ream any holes as they would wander far more than the difference this would make between hot and cold.
@@fzriow Yes, the accuracy of our parts matter. If you're a job shop that does ±.005" all day long, skip this video and watch Titan wreck some endmills.
To address repeated comments, the glass rod only needs to be positioned parallel to the axis by .050" end to end. Any angular variation under this .060" positioning is smaller than the .0001" resolution of the probe and is a moot point. We easily achieve parallelism of the rod by positioning the orange supports flush to the edge of the table. If an operator can't position it within .060" of flush, I'd have larger concerns for them.
No, we don't want to make stops on the supports because it would limit use of the device to only the X axis on other machines.
"The material Zerodur has better CTE characteristics." - The CTE of quartz glass is far less than the margin of accuracy of the probe so any better performing material is overkill. Also, it's impossible to buy a single piece of Zerodur for $10 and have it arrive the same day from a supplier.
Exactly, Got to love those small angle approximations!
would it help to probe the rod straight,just for the heck of it,just dont forget to canle the rotation afterwards or you will have wonky parts ,i know extreamly small numbers but you have it for free
forgot to mention that you can calibrate against a sphere that you use to calibrate cmm,first calibrate the probe to the sphere and then calibrate the machine with the newly calibrated probe.we have macroed our hermles so they grab a palete with a sphere mounted to it every 5 hours and calibrate the machine in all axes automatically
Make a rod from sphere, so both ends will share that sphere's radius. So any aligment up to couple degrees would be almost zero. But I am not sure if it can be manufactured.
For the rest of the World who use metric : the Glas pease is ca 305mm long , if you have it 1 mm wrong to the X Axis you have a length diferenz of 0,0016mm. a Normal CNC-Maschine have about 0,015mm on 1000mm accuracy. I dont know how accurat Haas maschines ar but i think the are not thits accuract (because the have no glas-scale)
Simplicity is the key to perfection
RAMTIC parts-production concept was developed by Renishaw in the 1990's. They used a similar method but probed a master part, then applied scaling. I really like your approach Jay Pierson and plan to try your method in our tool and die shop.
I have added quartz encoder scales to my manual mill... I get the thermal accuracy of the quartz measurement system independent of the machine temperature errors. It also serves to provide direct positional feedback regardless of backlash because the scales are measuring the actual table position rather than motor shaft location.
This seems so much simpler in hindsight
Yeah I think Haas offers linear scales positional feedback as an option vs just having servo encoders. Other machines have them built in.
@@TT-yu9uo yes they're an option on most machines. The problem is, when buying machines tools, glass scales are a factory build option only. Usually there is a long wait to get a machine from the factory, but the dealers bring a steady stream of machines in that they can offer immediately, all without glass scales lol
Yeah all things considered you're pretty much going to get the same level of precision, accuracy and repeatability with that methodology as with this dude and his HAAS.
There are so many things this dude is ...just ugh posturing. He's like the opposite of Ben at Applied Science in terms of reporting the results of his scientific excursions and being all full disclosure about what he doesnt know. (I guess this sort of makes sense since this channel is about promo'ing his CNC business vs Ben who just does it because he likes science)
Like oh wow, the CTE of alu is ~40x worse than quartz, sure. OK now what sort of cut is that quartz? There's a reason why AT vs SC vs BT all cost different amounts of money and are used in different oscillators based (hunt: quartz is not an isotropic mineral-- thats why a quartz watch crystal oscillator can cost 30 cents vs a rubidium standard that goes into a cell tower which is going to cost $300. See I can drop knowledge bombs about materials science too, granted I'm sure some PhD is going to jump in and school me good). If he's just using a rotary carbide bit or a surface grinder (or even if he lapped it with some diamond polish), unless he hit the right "grain of the glass", hes introducing systemic issues into his "transfer standard". (And arguably worse than that he's instilling false faith in his systemics).
@Marco Reps is an EE who dabbled into building a CNC for..some reason (I think he's friends with Stefan G, those damn smart germans). He ended up getting pretty consistent repeatability within the nanometers. He did it with just closed loop feedback servos and the tried and true MIT design that Slocum came up with (gantry-based, as much mass as possible to dampen vibration out, 25mm ball screws with 2 aux. guides per axis). In retrospect, he would have done what you mentioned (LVDT scales, rather than shafts/rotary encoder absolute feedback). He also mentioned some weird damping aggregate, that I forget.
If ya'll wanna see some real impressive work, go look at Huygen's Optics'. Or optics work in general. Zee germans and Japanese can charge 100k for a lens because their aspheric cine lenses are better than the Chinese clones on literally the nanometer scale (the entire visible light spectrum is only 300 nanometers.. so you can see chromatic abberation with a green looking "wrong" at 20 nanometers, especially on an IMAX screen)
@@wither8 Huygens Optics is Dutch, not German. But yeah, along with the Germans and the Swiss the Dutch are known for producing the most precise and insane tech on earth too (cough, ASML, cough).
I love this, it may not be a $_$ calibration tool but it does not need to be a $_$ calibration tool either. Keep up the good work Jay!
I think I've done something relatively similar on the z axis just without a "known reference" so its all relative. We have some parts that required countersink diameters that were plus minus .01 but the surfaces were usually warped and we couldn't face them either. What we did first was split the countersink cycles up and go back and forth between the probe, drill and countersink which hurt our cycle times. What I was able to do was probe all the hole locations beforehand and store the Z deviation in macro variables for each location and compensate depths inside the countersink cycle all at once. Saved a lot of time!
. . . Wow. . . haha such a simple concept that could make a huge difference to those who can only afford, well, affordable machines! Learn to think outside the box a tad, and it can take you far! Sweet idea, simple to execute.
I worked for Ellenef in the 90s. We built wings and structure for f-14s. Used lead hammer, dial indicator, Allen key and a steel block with a known value, placed on the fixture sled--to set the cutter height and check for runout. Did not seem very high tech but it was good enough for them.
By big takeaway is your last line of code. Put something on the screen so the operator has a general idea if the machine is hot or cold. Without that, we're basically machining in the dark!
wow! That's pure genius!
That looks quite cool but wouldn't the tool not being perfectly parallel to the table on the machine introduce some variance?
It could but the edges of the orange holders are to be positioned flush to the edge of the table. This is what keeps it square.
The rod can be at an angle as long as it is probing in the same spot then the length it becomes a constant. By the way Mr. Piercen you're a dam genius keep doing what your doing🤟
@@PiersonWorkholding Assuming you guys also checked parallel on Z before putting the tool into use?
And see similar comments why it only needs to be square to the table by .050".
G103 P1 Limits look ahead to one line at a time. This prevents macro variables from being read before their correct value is calculated in a previous line.
We make some large (3'x2'+) aluminum aerospace frames with
I have seen it done an even cheaper way. The owner of a shop I was working a wanted an ice machine for cold drinks in the summer. He happened to see a couple at a motel on his way to work. The motel was under renovation and owner was happy to get rid of them for free. We would fill 5 gallon buckets up with ice and mix some coolant in. Then dump the cold coolant in to the tank. We did not monitor the temperature with thermometers, but the temperature inside of the machine enclosure would be much cooler after adding the cold coolant.
Coolant chillers are a thing
I like this idea. Would have helped me on some projects.
An added suggestion on a machine like a Haas is to run the finish passes for whatever your critical features are in the shortest amount of time after probing as possible. If you have a long cycle some of these machines may be .001-.002 off 30mins after you probe your zero.
Great Idea - need to build one soon. I see its main use as verification to confirm your machine is operating within parameters rather than scaling. I believe if the temperature is far enough out that there is a significant scale factor the material itself (aluminum, steel, etc) will have its own thermal expansion issues. The part will be machined accurately but at the current temperature of the material. With an accurate temperature of the material this could be compensated. The temperature of the coolant may dominate with flood coolant and a long enough cycle time
Exactly! Finally someone totally gets it. If the "accuracy" macro displayed is pretty big (±.001") you should probably think about other thermal factors including material, coolant and ambient temps.
its probably easier to heat the colant,would it mess everything up if you mesured the highest temp in the machine and then add a few degrees with a heater? you would get a warm workpiece but stable temperature?
@@kristiankautto7616 heating/cooling coolant is what we do in my shop. 75° F is the temperature that works well for us year round. Coolant is heated/chilled after the pump and before entering the machine, we have parts in 6061, 7075, and 316 ss that need to hold 10ths or less and we do it with Haas machines. I have a 1 hour warmup routine that is implemented on startup days, say Monday morning, to allow the machines castings to stabilize then the coolant never stops circulating until the machine is shut down. Additional pumps have been added to flush the enclosures that run continuously which helps hold temperatures, and billets are warmed or cooled in a similar manner prior to machining as well as tools all calibrated at the same temperature. We used to have check routines, and we also used to use compensation and scalling values but the additional work was inevitably more costly compared to a few pumps and good chillers. Simple and reasonably priced modifications to the machines have removed any need for human involvement or error. What he is doing in this video is great for a job shop machine and is frankly genius but unnecessary for a production machine running continuously.
How do you keep the rod parallel to the axes? What about measuring the angle of the rod and correcting the result by the cosine?
Another suggestion would be wrapping the rod in an insulator (styrofoam) and keeping it in a tempered box. If put in the machine for a short while it won't have enough time to change its dimensions due to the insulation.
See pinned comment at the top. The beauty of quartz is that you don't have to worry about any of the thermal mitigation you mentioned. An extreme temp swing of the rod (for a machining environment) of 25⁰ would still be under the resolution accuracy of the probe.
Who will be first to get this working on a Tormach mill? Asking for a friend! 🙂
$35 worth of parts is all it takes!
I think that's running Linux CNC under the bonnet? So you could do loads of stuff even more accurate than this. Add a temp sensor. Linear glass scales. Have it run a ball screw map every so often. Etc.
cant even afford a tormach😂
@@3dwezzy740 only thing I would buy from them again is a 1100MX the 8L lathe and plasma table have been a PITA with little support. Currently their support is refusing a simple request to transfer the 2 year warranty on a servo that is bad. They purchsed the servo from clearpath. we contacted clearpath and they refused because it was purchased by tormach and not us. Tormach refuses to understand a simple request to release the warranty to us so ClearPath will replace.
Excellent solution to thermal growth. Thankfully I don't make anything requiring that level of accuracy.
Seems like a great idea! Only issue I see is that when you place the device on the table, does it matter that it isn't perfectly parallel and perpendicular with the machine axes?
Edit: (I may be Wrong on this)
Because the probe tip is a ball, as long as the ball is touching the opposite face (and this is why it was so important the faces were parallel) its self correcting.
@@NexusTrimean no its not self correcting. it may not be much, but see for example a parallelogram, when you probe the two angled sides, the measured value will increase. Yes i have validated it myself in a cad program :)
@@ipadize If taking single points yeah you could get some odd results if it's not perfectly straight to the machine axis. Taking a line on either side would help to correct.
Ideally would get parallel to machine axis and probe X & Y separately. Not sure if the haas control has a single scaling function or allows individual axis adjustment.
I thought the same thing…
The way its shown (single point) yeah those are super important. If you probe lines on either side you can correct for some angle, but best practice to isolate an axis as much as possible.
I'm assuming they put a bit more work into this than the video shows, working it to be nice and straight when placed on the table. I know we 3d print some CMM fixturing and definitely have to do some minor post-work if we need things exactly the same height.
I would argue that you want your "master reference" made from the same material as the workpiece, not a material that has a different CTE. If you are only comping for ballscrew growth why would it matter if the length master length changes with temp? Said another way, if you care about safe guarding against all thermal growth (ballscrew, coolant, ambient air temp) then your reference standard needs to grow and shrink at the same rate as your workpiece. Use the following thought exercise: measure you standard at 21 degrees in the metrology room and record its length. Take it out to the shop floor where it is 25 degrees, and calibrate your machining process. At 25 degrees your standard and your workpiece now match. Take them back to the metrology lab and let them cool to 21 degrees. If you part is alum and your standard is glass, your workpiece will now be undersized because it shrank a lot more coming from 25 back to 21 than the glass grew going from 21 to 25.
We use this process for a production job where we need to hold 2 tenths on 12" long aluminum part. The aluminum master reference is bolted to the side of the pallet and we use it to cal the WCS before a final cut.
This is also why carbide gauge blocks exist: for tool and die makers who work with carbide punches. Having gague blocks of the same material as your workpiece makes you impervious to thermal growth. As long as they match, they match,
Lot of good stuff you mentioned, but careful to not mix a few subjects. The quartz rod also takes imperfections of the ballscrew into account. You definitely want to measure accuracy of the ballscrew with a known, thermally stable artifact like the quartz rod. If you wanted to measure the thermal status of a workpiece, then as you mentioned use an artifact of the same material that has a known measurement at a known temp.
Carbide gage blocks under 1" have essentially no measurable growth under normal shop conditions (±15⁰) so it's most likely for wear resistance when working with carbide tools.
Finally, what fresh-out-of-college engineer would design a part with ±.0002 over 12"? This is why engineers need hero machinists to set them straight.
I agree with Jonathan, your goal is the accuracy of the final product, and not the error of the machine ball screw. If you really care for accuracy, then a Renishaw ball bar test will give you the best results. As far I as know, Haas doesn’t use chillers because they use a software algorithm built in the control to compensate for thermal growth, and in my humble opinion it works good enough to hold half a thousand, I can come in the early morning and continue to work without any huge differences like in the old days.
I’m sure you know you can see the machine temperatures in the control lol. Also, if you working with small tolerances, work in metric. 1 micron compensation = .00003937 inches.;)
@@PiersonWorkholding A master gauge of any material takes ballscrew growth in to account as well as environmental impacts of the workpiece (if matched with CTE). There isn't a mixing of subjects. Think of it this way: if only your ballscrew is growing, why does it matter if the master gauge is impervious to environmental changes?
You had me there at 5:40 for ⅗ of a second. You were off by a factor on 1000.
Well your understanding of machinist nomenclature is off by 10,000.
freedom unit problems
Thanks ,… this is powerful …. Thanks for being so selfless and sharing so much of your knowledge it is shaving a decade at least off my learning curve as i prepaid to start a shop, thanks
Fantastic! Thanks for watching and best wishes for your future shop!
Great content as usual!
I explored and experimented with these and similar concepts for thermal compensation over twenty five years ago. My employer said I was wasting my time and eventually fired me for "not being a team player". Oh, and I also perfected eight inch long valve spool bore finishing to within 0.00005" (that's five millionths of an inch) roundness, straightness and camber. With selective fitting of valve spools, we achieved less than 5 cc/min leakage at 3000 psi.
Hi i just wanted to add my pot to the fire :)
There is a strong argument to not use quartz or zerodur.
What i think is the material for temperature scaling should be out of the same material as your machined part. If your machine is hot say 30deg, and your part will be 30deg as well. Your calibraiton bar is going to be same temperature(or some other calibraiton artefact) if it is not from the same material once the parts cool, thy will contract differently. I am a beliver that the calibraiton artefacts should be designed and used and sett on the machine all the time
Don't confuse what you're measuring. This technique is only for measuring the machine accuracy. You're right about measuring an artifact of the same material, but that's measuring the workpiece at a specific temperature, not the machine.
With G103 P1 make sure you use it again following a renishaw shutdown command. Part of the shutdown routine re-enables look ahead. This can of course screw up any math you do.
have not used a glass rod. however have used same material gauge blocks that have fill table of temp and size.
Great tip! It's also possible to adjust servo scaling in the Haas control, but it requires a key to access on newer machines - have you tried this Jay?
Well knowing the thermal coefficient, calibrated length and measured length you can using simple math add an equation to calculate real displacement including thermal expansion of quartz rod. This is for super super accurate machining if not even grinding.
In theory, yes but the real-life limiting factor is the four digits of resolution of the probe.
I really hope that quarts rod is indicated in straight before gauging the length. In the video just slapping it on the table with unprecise 3d printed stands witch throw x, y, and zed axis's off by id say more than a couple thousandths will have a big impact on precision. Well aerospace anyways, if where making farm parts id say its good to go, best dam farm parts in the land! If that rod is not indicated in within all three axis of the machine; I hope you have brushed up on some trig to throw all the deviations from the gauge setup into your C++ program. If all that was taken into consideration before this video id say yeah this is pretty banging video!
Gotta read that pinned comment at the top.
Did you ever try to probe a grid on a surface plate and use it to compensate for Z inaccuracy?
Hi i think your rod is not "perfectly" parallel to "X" axis. So measrung of lenght is wrong and will be diferrent every time you put rod in different angle to "X" axis. But maybe im wrong.
Think about a ball touching a plane. And moving over to an opposite plane that is parallel to the first. Any error will be canceled out.
I don’t think it matters since they are probing both ends which were ground parallel. If you imagine the right side being angled and probing on the high side, the left side of the rod would be probed on the low side which would average to the correct length. I think this also applies to the potential error in height of the 3d printed parts. I’m happy to be told I’m wrong though.
As I mentioned earlier, the edges of the orange holders are positioned flush with the edge of the table. You would only start to see error once the misalignment is over .050”. And if someone can’t see or notice .050”, there’s no way they’ll keep working for me.
i think if you measure the rod on Coordinate-measuring machine you will find pretty accurate number (if u make coordinate system on rod). But on CNC machine rod is putted randomly on table not respecting coordinate system so you measure diffrent lenght... every time you put it in different angle you get different number. But maybe im wrong.
@@PiersonWorkholding stick to hanging drywall is what you are saying basically.😂👍
If you got a bigger budget, Zerodur has an even lower thermal expansion coefficient than quartz
The better performance of Zerodur isn't needed. See pinned comment at the top.
Happy Thanksgiving Pierson team and thank you Jay love the idea! Unfortunately (or fortunately for the machinist) my current shop usually plus minus 10 thou is enough for us but a beat trick to keep in mind none the less. Agree with you the getting it slightly out of angle is a moot point in terms of the machines accuracy.
Thank you and happy Thanksgiving!
But what about the parallelity of the Rod to the X Axis? How do you account for that, when it's only '"loosely" placed on the Machinetable...
See the top pinned comment.
great vid its detailed while being a reasonable length, the method is rather complicated and jank The CNC manufacturers are probably to blame. I'm used to taking the expected travel distance divided by actual travel distance and then multiply the current microsteps per mm value by the result. Then save the new value directly to the firmware. Essentially the same thing but simplifies gcode generation reducing errors.
If this process was automated and part of homing the machine at every program start, that would awesome.
Wishful thinking on editing firmware. Machine builders have those pretty locked down. You still have to sample something to get your known and traveled distance and our method might be the solution.
@@PiersonWorkholding I probably could have emphasized it more but i meant to put the blame on the manufacturers.
A machine shop could probably make loads of money if they start selling kits for industrial use not just hobbyist or garage operations that can be put together using open source software and boards. Marline already supports CNC, 3d printing and Laser cutting all one would need is to reach out to some board manufacturers like BIQU to support higher amperage drivers.
3d printers would be an other decade or two behind if it wasn't for open source and people printing parts for their printer then editing the firmware to enable new features.
This is a pretty brilliant idea, has me thinking about how I could mount a reference piece in my NTX to check the screws.
Also, I could see with your system using it to adjust work offsets as the machine warms.
Hell in fact, one could set G54 at machine zeros, and use program coordinate shift in the program in lieu of setting the work offset on the origin Then as you check the machine throughout the day and use scaling to apply a thermal comp the work origin would be adjusted automatically. You could finish the system out by mounting the rod on an actuator so that you can bring it into the probing envelope with an M code and periodically check your comps.
Of course, it would be critical to make sure thermal scaling is applied accurately when you are finding the work offset, and you would want to write your work positions into variables. Might need to edit the post to call G54, G51 then immediately call G68 or whatever it is in Haas to shift the coordinate over
No squaring to the table of that rod ends?
See pinned comment.
@@PiersonWorkholding ay thanks / true
Awesome! Thanks for sharing!
Do you have the supports placed near the Bessel points? Measuring something in the vertical and then laying it horizontal can affect the length. Have you performed the same measuring cycle in the Y axis as well to see if there's any discrepancy? I'd be curious if you got similar results.
You're getting too theoretical talking Bessel points. The reality of four digits of resolution of the probe output is the limiting factor.
@@PiersonWorkholding this process hinges on how accurately you can measure the rod. For the average electronic height gage linear error could be as much as .0002-.0003" over 12". FYI, Haas has a warm up function that applies a correction over a specified amount of time
@Pierson Workholding How about dealing with ball screw wear and doing all 3 axis' and them being different numbers. So instead of dealing with heat (or the lack of), dealing with wear?
Ballscrew wear would require a ballbar test which would cost you $250 minimum. This is an accurate, inexpensive first step to see if you even need a ballbar test to begin with.
Does the calibration factor tend to be the same on all axes? Seems like up/down is likely to be different because the ways are typically in a different place. Would be cool to have a 3 axis calibration system with a cube of glass
i want a hass machine >.< My DIY machines, well, they work but with a bucket of headache attached to them. Envy every Hass owner out there 😅
I can see that using the quartz rod will allow you to calibrate the machine. But what about the material being machined? Doesn't matter if the movements of the machine are dead nuts accurate if the material being machine has expanded in size due to heat and will shrink later as it returns to room temperature.
Careful to not confuse multiple factors. Material, machine, and ambient temps are all separate factors that should be measured with other devices, i.e. thermometers. The featured tool is only meant to measure and correct machine errors due to thermal expansion.
All cool 😎but is your work shop temp control. Do you allow your workpiece to soak before machining. If not you are wasting your time. I believe 🤔
5:43 lol a 5 inch length of aluminum does not change in length 3/10 of an inch when the temperature changes 5 degrees.
Nothing in you brain made you stop and think what three tenths means to a machinist?
You are mistaken about how much a 5" piece of aluminum will change over 5°F. I think you meant to say it will change by three ten-thousandths (.0003") of an inch, not three tenths (0.300").
You gotta learn the lingo. No one says ten-thousandths and no one calls 0.1 a tenth.
5" X 5 Degrees = 25 X .000013 = .000325. is that how you correctly calculate the thermal expasion?
Yes. Your formula is right for aluminum. Multiply by .0000003 for quartz glass.
And how you line up this quartz glass rod along x axis?
See pinned comment at the top.
You could eliminate one step in your Gcode for scaling.
Let's say the rod was 12.0000, and your measured length was 12.0005
Just divide 12.0000/12.0005 and that's your scaling factor. Multiple by this value everywhere as needed. You can get rid of the #903 variable (unless it's used somewhere else!)
You are correct and yes, we absolutely need the 903 variable. We explain it later in the video as one of the pieces of data to display on the control. It's a number that the operators can best relate to.
@@PiersonWorkholding ah I see, didn't think of the front end for the operator, good to know, thanks!
Wait, did that quartz rod not need to be indicated along the X direction to make sure it was parallel to the axis before probing the ends?
Read the pinned comment at the top. ☝️
@@PiersonWorkholding, I'm probably in the minority here, but anyone truly calibrating a machine that doesn't indicate their standard on the machine wouldn't be in my shop long. That's just me though, I'm sure most would be fine with it.
Doesn't a machine of this class have quartz linear scales, though?
Awesome Video! I love learning tips like this! Two questions. I know the probe is extremely sensitive but is it possible it could move the rod slightly if it is just free floating in the 3D printed stand? Is there a clamp or screw to keep the rod firmly in place within the 3D printed stand?
The glass isn’t free floating. It is pressed into the holder almost to the point where we though it would break. Fortunately quartz is ridiculously strong too.
@@PiersonWorkholding What is securing the holder to the machine table?
@@conorg6273 magnets imbedded into the 3d printed bases.
@@conorg6273 Here's a link to this video in case you want to watch it: th-cam.com/video/Ds6ZQox2yPM/w-d-xo.html
@@PiersonWorkholding 😂😂
Take in consideration that some of us use older machines. ☹️
Works great with an edge finder too!
@@PiersonWorkholding but the edge finder is made from metal :P
with this method you are only probing on the x axis, are you just taking the x axis difference then applying it to x, y, and z by assuming all 3 ball screws are affected by the same temp?
We've repeated the tests in X and Y but the differences have been identical.
How do you guarantee that the rod is precisely aligned with the X axis of the machine. If the bar has a tilt, either horizontal or vertical, it will appear to be longer than it actually is.
See pinned comment
@@PiersonWorkholding There's no pinned comment
@@RavenRockMachine I see it. Pinned at the top of the comments.
@@CNC_Soup interesting on both computers weren’t showing it. Now it is. 🤷🏻♂️
How much force does the probe put on a part? Honesty question because it seems like it could deflect the 3D printed plastic ever so slightly. I know the length assists but I would have made those pillars much shorter and allowed one end to float so as not to stress the rod even the slightest bit. Does that seem reasonable or am I being too precise?
Thanks for the video though, it looks like a very inexpensive thermal compensation method.
Fair question. According to Renishaw, the force on the second probe touch is 1.1 oz. That's way under any force required to move the magnetic bases.
I love the idea, but I am concerned by the error that could come from the installation angle of the rod. As it is not installed perfectly straight, the measured length is slightly bigger that the real rod length. I didn't run the numbers, so I am not sure if this error can be neglected or not, but that is definitely something to be cautious about.
See pinned comment at the top.
Curious why you did not grind the rod to a nominal length? I understand it doesn't matter for function, but I would have done it just for fun.
Glass and ceramic swarf is deadly for linear rails. It's best to machine off the minimum required to clean up the ends.
How do you make shure that your glas rod is in line with x axis? If it would be slightly off you will get wrong length
The edges of the orange holders go flush with the edge of the table. They just need to be flush within .050" before angular error is a factor.
Sure ? In my opinion you have to Line it with the Axis ( in x and z ) or i understand somthing wrong. ( Im from Germany and my english is Not the best,also i get confused with our inch- thing;-))
@@woiskorn Thank you for watching from Germany! The rod holder needs to be flush to the edge of the table by 1.3mm.
Interesting low cost solution for teml comp!
Looks like some Factory400 code in there, am I right? Or is that all Jay? Either way, it’s Great Stuff 👍👍💪💪😎😎⭐️⭐️⭐️⭐️⭐️. You probably already know by now, but if Block Lookahead were off the calculations could occur w/ ‘old’ data as it could be executing #___=__#____ before the desired data is assigned
So the glass scale mount is now as accurate as the 3D printer you used?
Excellent
good video mr pierson
Could also buy a calibration sphere. Renishaw sells them.
Nope. Too small of a sample size to be useful.
How do you make sure the glas rod is perfectly parallell with the axis you are measuring?
Read pinned comment at the top.
I've never dealt with it, but I have seen a spot in Haas parameters for thermal expansion.
Isn't this, exactly what those are for? I could be mistaken, but that's always been my understanding.
Seems like it would be easier to update there, than run special programs from G51.
The thermal compensation in the Haas control is based on correction distance over time. It is meant to compensate for the machine warming up over the first part of the day. It doesn't actually take a temperature reading.
Where is the McMaster Car links?
Part number towards the bottom of the description.
Dead nuts accurate... Is that quantifiable? LOL
It's an industry term - th-cam.com/video/y8hcfTFVJ9k/w-d-xo.html
I'm not a machinist, but I have a question. In the video you probe the rod in X only to obtain the scale factor. How do you know that factor is the same as Y (or Z for that matter). Also , how do you know that you've aligned the rod perfectly in the X dimension? i.e. If the rod was rotated in Y ever so slightly then your probe's measurement would have a sine error in it, no? Seems to me that you'd want to set it up with a machinist's square with respect to the bed and probe, rotate it 90 degs and probe again. What'd I miss?
You missed the pinned comment at the top.
doesn't this machine have a glass scale? I have just ordered a set of glass scales last week for my CNC and I will set up the closed loop via this rather than the ballscrew pitch.
Would a method similar to this be possible/make sense on a lathe?
I am curious to
I suppose it would work easiest in a lathe with a probe. Lots of lathes don't have 12" of travel in X either so you'd have to use a shorter rod which reduces resolution.
The only way I could imagine that would work would be to mounting the rod in a tool holder? Idk lathes are deceptively simple yet complicated.
At this level of precision ... Do you have to consider the bending of the 3D printed holding bases while probing the ends of the rod? I don't know the amount of force of the probing procedures.
Kinda like 2/3 N in Renishaw probes.
Would there be some margin or error if the rod isn't parallel to the axis its being measured with? If those squared edges are canted, i think there would be some error introduced.
See pinned comment at the top.
Zero dur glass has an even lower coefficient of thermal expansion than quartz. But in the end you will be limited by the accuracy of the measurement system you are using in most cases.
See pinned comment at the top.
Can this be used on the DRO on manual machines.
Yes, works with any machine. Just use an edge finder instead of a probe.
Lots of potential for cosine error to become relevant here with the mounting method. Probe calibration also going to play a significant role.
Nice trick
Ok, you calibrate the machine, of course likely at a significantly low or high shop temperature. Then you take some aluminum and machine it. Which temperature has this aluminum? The same as the room temperature? Obviously. Oh, what was the temperature coefficient of aluminum...? Means you are anyway trapped, with or without your quartz scaler you will not produce a good part at high or low shop temperatures. Your compensation is even working against a gain of accuracy, because it reduces the growth error of the machine and therefore increases the difference to the also grown material. You may of course do this compensation and run the machine at normal room temperatures. You will only do that compensation when there are significant deviations in the machining result, which then is a thermal problem of your machine even at normal temperatures (e.g. no core cooled ballscrews). For a very short machining time the compensation will still work, but with a long runner you will get more and other thermal problems which the machine obviously haas. 🙂
My conclusion: There is only one way to get precision. Work in a temperature stable environment and use thermally stable machines.
All true but you're oversimplifying the conclusion. Machines, parts and shops are always changing temperatures throughout the day. This method helps identify the current accuracy of the machine.
@@PiersonWorkholding I can‘t oversimplify it. It is only about a room with a precise air conditioning and highly accurate machines with an advanced temperature management. So that‘s just money. It is much more difficult to find skilled operators.
Can Z get this type of correction? And can circular interpolation have corrections made?
Sure, Z could be calibrated but would require another type of stand and some extra programming. Circular interpolation is more likely to be managed through firmware and backlash.
In metrology, the term for an item such as this is an “artifact.”
So if I get this right... This is the ever lasting way of making sure that years of wear and tear, will make sure.. you will be within tens... right? Just make a same scale for X....Y....and...Z ?
I own a Okuma that has 4 digits after the dot(rest of the world)... I still want to have this under controll....
My head screams this is a yard stick... if your diviation of messured and reality is X.... maintance is needed!
Exactly! It can also be used to check ball screw wear and backlash. If you move the starting probe position to different parts of the table, you could check different sections of the ball screw for wear or damage.
"Micro inches"??? Just use metric properly alright? It's hilarious to watch people do anything to avoid metric and fail at it.
Great idea no matter what units you use!
Metric is better but Imperial built the world.
@@PiersonWorkholding True. But mixing them is silly!!
Since the X, Y, and Z ball screws are different lengths, do you apply the scaling factor to each axis? Or just a blanket?
You can if needed, but we generally use this method to know if the machine is running hot or cold.
I'm guessing that Look Ahead is the CNC preprocessing the next line of code to the one being executed? If so, is that a Haas thing?
Something like that. Look-ahead is next level nerd stuff that disinterests me.
One critique would be that this measurement can't exclude machine backlash? (Which is likely of the same order of magnitude as you want to correct here, I think?)
Depending on your milling operation this could be good or bad. Perhaps you could add an extra variable to subtract the backlash before calculating the scale factor? (Likely this is close enough constant)
Backlash is a whole different topic. Modern machines have backlash in the millionths so sloppy machines are beyond the scope of this tool.
@@PiersonWorkholding I think I would like to see you measure that? My experience is different. Put a 123 block at the end of the rod. Compare it's length measuring in the same direction at each end Vs the direction reversal when you just measure both ends. I would be surprised not to see quite a few microns difference?
@@PiersonWorkholding Trust but verify. A machine may not behave the same after several years of abuse as it did the day you received it. I think using the spindle probe is also a bit sketchy unless you're obsessively calibrating and checking the calibration on your machine. The temperature of the machine and gauge you use when calibrated could itself throw off all future measurements. Temperature differences do matter but this way of working just compounds errors in a manner that could make them painfully difficult to diagnose.
Is a part out of spec? Was your gauge out of alignment? Is your probe out of calibration? Does your machine have some backlash ...
The biggest problem with your solution is that it compounds these errors into a single complex variable that then affects the entire geometry of the part.
Ideally, you'd probably just run critical parts and features on a machine that's been adequately warmed up in a climate-controlled environment ensuring reliable, repeatable performance with simple-to-diagnose issues. I personally think it's a solution searching for a problem. But the kind of problem it might find is the hair-pulling-inducing kind where everything is going wrong and you don't know where to start fixing it.
Can't the software compensate for temoerature?
Would your measurement be off
If that 3D printed part isn’t perfectly flat so taking it in and out of the machine get different results? Since it could be tilted slightly
It needs to be flat within .050" to still give useful accuracy.
Is critical to put the rod aligned to X axis?
See the pinned comment at the top.
How do you like the VM table slots compared to the standard VF style slots?
I made a whole video about that. Check it out 14 mins in: th-cam.com/video/_lfYqrd-2tg/w-d-xo.html
well even with a 20 degree temp swing summer to winter im still in tolerance for my customers haha.
and my machine has glass scales so theres that
Sounds like you got some great high-end equipment.
2:31 - Wow I never knew it actually has the capabilities to warm the ball screws... interesting. Makes sense, needs to have dead on accurate temperature, regardless being chilled or warmed. 4:16 - Would be cool if you talked in unit of measure using microns instead of inches when speaking of high precision and accuracies. Precision fanatics only speaks in microns and sub-microns(nanometers) lol.
This is intended for shop floor machinists, not precision fanatics. One micro inch equals one millionth of an inch or 0.000001"
@@PiersonWorkholding "floor machinists" are not considered "real machinists" and have no clue how to design blueprints ordered from SpaceX, NASA, Boeing, and etc who needs micron tolerances... I guess this youtube channel is just another typical "floor machinists" content minded. Only a handicap would speak the language of defining small unit of measure using microinches (inches) and a professional machinist (not a floor machinists) using simple notation such as 1µm (1 micron = 0.000039"). The reason to use micron is because people like physicists/scientists/inventors in US and all over the world who sends CNC work to "professional machinists" always uses microns notation to create their stuff, they are taught in University to use microns and nanometers when talking about extreme small unit of measure, but in the "floor machinists" world, they still use inches... 🤦♂To put it simply why it literally sounds so stupid to use inches for small unit of measure specially in machining: Why not use "foot" unit of measure? 1 microinch is 0.00000008' foot. I'm sure you would work happily as a floor machinists using foot notation just like inches.
Lap a ball shape on the end of the rod as long as they are the exactly the same?
That would make it harder to align. We only need two parallel planes.
Doesn’t work would still affect the “length” or I guess x axis component of the length. Only thing that truly wouldn’t change would be a circle or a solid of constant width, as cool as a Reuleaux triangle gauge would be I think I that’s rather impractical
I see an issue with your G51. You are probing along the X axis, which determines the error of the X axis, however you are not probing Y or Z, yet you are applying the scaling factor derived from X to all axes. You are assuming that HAAS has thermal comp on all 3 axes and that the thermal error is the same for all 3 axes and that the HAAS control is compensating exactly for all 3 axes. Once upon a time they only had thermal comp on the Z axes.
So you’re assuming that I’m assuming… roger. Why didn’t you assume we have 3 programs for each axis?
@@PiersonWorkholding I get the premise, Quartz has a very low coefficient of thermal expansion and can be used as a standard, but you are stuck with applying the X-axis comp to all 3 axes at once because G51 doesn't have discrete scalars for each axis. I think this might be better as a test for a warm-up cycle to determine when it should end.
Hahaha, I thought you guys only used freedom fractions of the inch like twelfths, sixteenths... onemillionfortyeightthousandfivehundredseventysixths etc.😅 BTW great informative video.👍
Yuck! Fractions are for contractors and woodworkers. Decimals for the win.
Would the reading be different on a different axis?
Possibly. If you had a part with a lot more X movement than Y movement, the X ballscrew might be warmer. We haven't noticed anything like that though.
It's an industry term
Yes, taught on day 1 at machining school.
@@PiersonWorkholding th-cam.com/video/y8hcfTFVJ9k/w-d-xo.html
@@claushollen That's hilarious 😂
What Camera do you use??
Expensive ones that use quartz glass lenses.
Good idea, I mean I would just probe a large ring gauge and that would be good enough for me. I have 5-axis mills so the center of rotation I find is more important. Saying that, if you're going to these lengths to get your machines cutting this accurate then you would have to be making parts that are thermally controlled throughout their life too otherwise this is kind of pointless.
You missed the point. Ring gauges are susceptible to thermal expansion. And come back and tell us how much a “large” ring gauge costs.
@@PiersonWorkholding I understand better now after reading all the comments.
Indeed out of the box thinking.
@@PiersonWorkholding well a large ring gauge is normally free if you have a large bore mic ;) I know it possibly could be nice to know this information but do you really have parts that need this sort of accuracy as all this would increase the accuracy on is true position. As far as diameters etc you would always change the offset? Like you would have to change your process to make the most of that accuracy, you couldn't drill and ream any holes as they would wander far more than the difference this would make between hot and cold.
@@fzriow Yes, the accuracy of our parts matter. If you're a job shop that does ±.005" all day long, skip this video and watch Titan wreck some endmills.
@@PiersonWorkholding 😳😂