Andrew, Hats off for the effort. To echo a few, concentricity of the blank and the hob will have a big effect on the outcome. I work with hobbing machines daily. Hobbing machines are geared, and the indexing is done with a large worm gear turning the blank. There is significant rotating mass and inertia behind both hob and blank to insure the proper relative speeds throughout the cut. The hob and blank are also both commonly on shafting and are very rigid. When anything is loose or eccentric on a hobbing machine it greatly affects the outcome, but these are relatively easy to spot and fix. Seems like you have the code and counts correct, hurdles are the belt, rigidity, and concentricity. Technically you have the setup correct down to the Helix Angle (lead angle). Clearly you have reverse engineered the process. Again, incredible effort and persistence. Nice job. The other, other Andrew from across the pond.
Hi Andrew - thanks for the info. I will investigate the belt, rigidity, and concentricity. I'm beginning to wonder if gear hobbing in the home model engineering environment is too big an ask. All the best. Andrew
Excellent video Andrew. I would definitely try a different hob and see if you get the same or similar results. I can appreciate your feeling of not wanting to be defeated, so to speak. Again, excellent video and work. Looking forward to the next video.
Yeah - the hob is one of things that I would like to try to eliminate as the cause. Unfortunately they are quite expensive and I don't want to purchase one without having a need for it in a future project. I've looked at second hand ones on eBay, but not knowing their history it's a bit of a gamble. As I understand it, if a hob has been incorrectly sharpened it can produce all kinds of problems. Some people make their own hobs, but for me that would just complicate matters even further. Cheers Andrew
Hi Andrew, FWIW, there are two mechanical things that I think are worth investigating before delving any further into the electronics: 1) As others have pointed out, your arbour isn't running true - look at about 16:00 to 16:10, you can see both the chuck and the arbour running out against the background. Also at ~20:34 you can see it again, but it looks a bit more than a 'once per rev' eccentricity - maybe there are problems with the bearings in the work spindle? I think it would be worth sticking a clock on it for a while to eliminate this possibility. (Or at least, clock up your arbour to make sure it's running true before cutting.) 2) Even with perfect timing pulleys and belt, if there is any eccentricity in the pulley PCD (not necessarily the OD) after mounting, it will transfer to a cyclic error in position. If the belt has too little or too much tension (or is just a poor fit to the pulleys) it won't settle smoothly into the pulley teeth, and there will be some 'cogging' as they rotate (i.e. some variation about the 'ideal' position as each tooth engages and disengages). Given that you now have the closed-loop stepper, I think it would be worth trying to drive the work spindle with it directly (through a torsionally rigid coupling) and eliminate the belt drive completely. Good luck! I admire your perseverance! 🙂
Thanks for the suggestions - yeah there is definately runout. I will replace the collet chuck with a 4 jaw and try again. Yeah - I have wondered about the accuracy of the pulleys. They are supposed to be timing pulleys with teeth. Driving the spindle directly seems like a great idea. I will certainly give that some thought! Cheers Andrew
Hi Rich - yeah that would be interesting, but I have no idea how to interrogate the encoder built into the stepper motor. Food for thought! Cheers Andrew
@@learningturningmetal Hi Andrew, I'm thinking that you are losing pulses from the mill encoder. do you think that swapping the outputs would help? I would be thinking of replacing the mill encoder with another type Rich
Thanks Matt. I will certainly check out tool rigidity. I have a crude check in the Arduino code to try to identify if the CPU is struggling - and it doesn't kick in. I have since been testing with AndysMachines code (written in assembler) and that produces identical results. A bit of a brain teeser!. Cheers Andrew
One thing I might suggest if I may. You might get a better tooth profile if you don't back track on the moving cutter. It seems to me that you'll get a better profile when you cut through it once foward & shut the cutter off when you back track. Interesting project. Thanks for sharing.
@@learningturningmetal Thanks for commenting Andrew. I didn't think that was your total problem. I just noticed with the little bit of experimenting I did with making gears, when I backed off on the moving cutter, my teeth were distorted compared to others that I didn't.
Hello Andrew, I admire your perseverance getting this issue sorted out. As someone who really has no idea about your electronics, I would like to present my observations on strictly mechanical terms. Is there an accurate method to measure the teeth on the hob? I am somewhat in agreement with wino4340 that the issue might be with the hob itself. Also, could there possibly be some "tool flex" in the hob shaft as I observed a little more material being taken off the gear as it was returned to the "start" position? Perhaps stopping the cutting operation with the gear to the right of the hob after being cut and examining it then. I am confident you will get it sorted. Good luck !!!. Earl
Hi Earl - thanks for your thoughts on this. It would be good if I could eliminate the hob as being the problem - but they don't come cheep. I will look into rigidity - more testing needed! All the best. Andrew
One thing that you might try is only cutting in one direction. Each time I see you do a cut you bring the hob back through the work for a second time in the reverse direction. Could this introduce a small difference? Excellent investigatory work though. A great video.
I am wondering if it is the small eccentricity you see when the blank is on the mandrel. Have you tried indicating the blank with the headstock turning? Thanks for the video, I find this topic very interesting!
Hi Andrew. Thinking about this, where you are moving the gear blank sideways to the cutter the spiral of the cutter is what is causing the wider gaps between the teeth if the blank is at 90 degrees to the mill spindle, try having the cutter directly behind the gear blank and using the X axis to cut instead of the Y axis, it will be harder to get the right depth but that can be set as a zero on your DRO that you drive the axis into. Other than that you may have to set the blank rotation axis to match the spiral axis of the hob cutter.
Hi Robin - thanks for the suggestion. My understanding is that the hob design requires the hob to be cut along the x-axis, in order for it to produce to correct tooth profile. Plunge cutting the hob will not produce the required result, as the cutters on the hob don't have the same profile as a basic gear cutter. Cheers Andrew
@@learningturningmetal The hob to work setting is usually not at 90 degrees but corrected for the lead angle of the hob, The direction of travel is along the axis of the gear which you are doing. You could accomplish this by rotating the head of the mill slightly in the correct direction. The lead angle (hob setting angle) is usually marked on the hob. If you are not cutting in the "hobs normal plane", you will be producing a wider space, as you are.
Hi Ralph. Thanks for the info. The helix angle of the hob is marked 1deg 20min and I set the angle of the mill spindle to that by moving the spindle down by 1" and getting the offset to be sine of that angle. I checked this several times and I think it was spot on. I guess I could experiment by adjusting the angle and taking more test cuts. Cheers Andrew
When the video was sped up, you can see run out on the shaft holding the gear blank. That would definitely make that gears teeth smaller than expected. Can you setup a tailstock center for the gear blank shaft?
I don’t think your problem is with the electronics. I think your hobbing machine headstock doesn’t have enough rigidity, also your not using a tailstock to support the mandrel.the vibrations may be causing the teeth to cut oversize. If you have a tailstock for your dividing head or rotary table try setting that up to offer some support to the mandrel.
Your gear blanks appeared to have an ever so slight wobble while they were turning. Perhaps it’s my old eyes playing tricks on me. Either way It might be something worth looking in to.
Hi Andrew, the closed loop stepper motors have really come on, a lot cheaper than a servo & I would say perfect for this application. I would try another hob, cheap ones are NOT all equal! Also sort out the concentricity issue on your indexer. Check on the gear if there is a sort of phase shift as you go round from tooth to tooth. Hope this helps.
Hi Dave - thanks for the advice. It would be great to be able to eliminate the hob as being the cause. Problem is new ones are quite expensive and I would need to buy one suitable for a future project. I've looked at the second-hand market but I think that is a gamble, not knowing the history of the item. More testing needed! Cheers Andrew
Hi Ken - could be. Both programs produce the same outcome - so I doubt it's that. It could be and encoder/stepper issue - but hard to spot. Cheers Andrew
Andrew, This is a tough one for sure. During the high speed view it looked as if the gear shaft was wobbling but that could just be a video artifact. The counts are really close. It does appear that the gear is being over hobbed. Could the depth be a bit too much? Say trying a reduced depth of cut? Or breaking it up into 2 passes as the end of the gear shaft isn't supported. Granted rotationally you are beyond close. The question now may just lie in the effectiveness of the actual cutting operation. 32 teeth may not be enough for you to learn much from. You could increase the diameter of the gear to say 64t. see what sort of flex deflection you get at the end of the gear's shaft and the end of the hob's shaft. I think the issue at this point is more mechanical. I may not be in the right ball park. mostly just thoughts.
Hi Andrew, are you tilting the head of the mill in order to compensate for the helix angle of the hob? This is clearly visible in AndysMachines' video (probably also mentioned in the audio). The cut should be parallel to the rotational axes of the blank but not at an angle, right?
You can see that the involute-cutter gear teeth have a slightly different profile (they're thicker). So I'd be suspicious of the hobbing cutter profile. Also, the stick-out on the spindle is fairly significant, given its diameter...and I suspect you're getting some cutter deflection that might account for some of the difference.
Hello Andrew, I am wondering if you problem is interrupts .Unless you are running in assembler you really don't know what is happening when a position change comes in from the motor encoders. I don't know much about Adreno, I use to program Intel 8085 and 8051 microprocessors to do speed and position control. When a position change comes in it must take the highest priority , and everything else like LCD refresh must wait while the position is processed and the control of the stepper is processed and implemented . This is just my thoughts on the matter.
When you are cutting the gear at around 16mins there looks to be some runout on the chuck holding the gear blank. More noticeable at the higher 4x video speed. Could that explain the wider gaps?
Hi Jeffery. There does't appear to be any backlash in the belt system. The pulleys have teeth and the belt is sold as a timing belt. Mind you, belt drives are totally new to me. Maybe I should explore connecting the stepper motor directly to the spindle in order to rule that one out. Cheers Andrew
maybe there's a rounding error in the calculation so that there is a slight drift over time, have you tried calculating the ratio between pulses? does it end up as an exact whole number?
The hob is likely not causing the issue. I assume it works fine on the other hobbing machine. (Andy). I had massive problems with a really cheap parallelport interface, it was missing encoder steps. (optocouplers) just too slow to "catch all the signals", missing some. Try an encoder with less lines, slow down the spindle or use only A or B signal? (not both) Then there might be and issue that your signal to the stepper driver is not triggering a step every time. I have one driver that refuses to work on an Arduino, but it works on another controller. (seems to have too less power to drive the input reliably) To check this I would change the software so that the signal to the stepper driver and the signal coming from the encoder have the same frequency. 1 Encoder pulse, causing 1 step signal. Then hook up the oscilloscope and see if the two signals are matching. Does this make sense? If you have a speed issue in the signals, everything will look fine when moving slowly and if you reach a certain frequency it goes all downhill.
@@learningturningmetal Thank you for letting us participate. Do not give up. But ... computers do not deserve to cause frustration 🙂. Stay undefeated do not let the thing harass you. Cheers Timo
Hi Andy, very cool video you did. There might be an performance issue in your code. The digitalWrite function takes some time (If I remember correctly, it’s around 190 cycles per call which is ~10 uS for the 16MHz Arduino Uno). You are doing it two times in the for loop (20 uS). So you are able to handle 50000 Pulses per second. Depending on the pulses your encoder produces (and ignoring the fact, that the main loop does other things as well) this might be an issue (RPM/60 * PulsePerEncoderRev). Long story short: instead of using digitalWrite try to write the output registers directly. This will increase the speed drastically. Ideally you won‘t have not more than one iteration of the for loop per iteration of the main loop. Don‘t know if the encoder lib does support interrupts but this could be a better way to track each single pulse of the encoder and to ensure perfect synchronization (Depending on the hob speed you will overcome the capability of the arduino and you will need a faster cpu like the Esp32)
Hi David - thanks for the suggestion. I've since been testing using AndysMachines code which is written in assembler. So I guess his code would be far more efficient than my code. I tried to make my code as tight as possible, but included a crude check to try to identify if the Arduino was running out of steam. I will however investigate to see how I might be able to write the the output registers directly - not something I know how to do at the moment. Cheers Andrew
@@learningturningmetal I also think it's an issue with the timing. When you showed the traces in the previous video, there was very clear jitter in the timing between pulses. That would result in slight angular variations back and forth, which is consistent with the results you're getting. I believe you're using an AVR board, so one option would be to switch to an ARM board. Those clock speeds on the order of 10 times faster than an AVR. So would help to mask any timing issues with code.
Hi John - after further investigation I think the problem is hob related. I'm just in the process of uploading another video which might help explain. Cheers Andrew
Really enjoy the video! Could I put forward that it might be as much a synchronization issue but a synchro phasing issue. If there was Sensor on the mill shaft, and one on the closed loop motor shaft they could be compared. Say for example, both sensors are set so that a position on the cutter is “0”, and one on the motor shaft is also set to a “0” position. When the cutter makes contact, it slows a bit by friction as it cuts, so now the sensor reads “-1”. The sensor on the stepper motor might read “-3”, from being introduced into the rotational speed of the cutter. If the mills pulses need increased, the program counter will see the incorrect RPM count, and speed up accordingly. Independent of the mill, the stepper motor will sense a low RPM, and the program will speed it up to a null state. ( not +/-). Between the independent speed adjustments, the drift would be reduced I think. What say you?
Hi William - I see where you are coming from. I think I would need to get hold of another encoder and attach that to the motor shaft somehow. I will have a ponder! Cheers Andrew
Fascinating detective work. You seem very close to an acceptable working gear, but I know what you mean, it is annoying when it is not quite right. I can offer no help regarding the electronics but I do agree with other comments that a bit more rigidity may help. Can the hob shaft be shortened by holding it further into the collet? A tailstock might help if you can set it up so that it doesnt get in the way. Certainly cutting in one direction should give a cleaner finish, but if the synchronisation is correct you should be able to come back out of the cut without removing more material. A little more sleuthing required yet Mr Holmes 🧐
The sync issue is likely worse under load, I.e. while cutting. It sounds to me like your arduino is missing an encoder step. I have seen this problem in a similar application I am working on.
It has been decades since I have had anything to do with servo loops but I was thinking that even with a closed loop stepper motor it would still lose a step. The motor encoder would detect it but in this configuration it cannot correct and keep sync. Is there a way to read the output of the stepper motor encoder to see if a) it detects and issue or b) at least see if the two encoders jive? If they are out of sync try cutting with less depth of cuts or try sobbing in plastic or some softer material. Just a thought. But, as they would say in your country....Keep Calm and Continue Hobbing!.....OK I admit bad joke.
Your work. As I assumed earlier, no steps are lost. A properly designed stepper system doesn't lose steps unless some evil guy throws an anvil into the running setup. As a disaster detection an encoder can be used. After watching Your previous video I came across some hobbing videos launched the last 2 or 3 years. They used a much less advanced approach. Use an indexing plate for positioning the gear blank. Then move the table in the X direction and cut some teeth completely. Advance the indexing wheel a tooth or so, and make a complete cut of a tooth or two. Of course it ought to be possible to automate the index movements.... Maybe I'm wrong but I think You use an Arduino UNO as the controller. It's a handy and is a good friend but not very fast. Clough42 used much more powerful things in his digitally controlled lead screw project. But, of course You get a lot of experience from Your work. Looking forward to Your next attach on the issue.
Hi Stefan - thanks for the suggestions. I think there is some kind of Arduino clone which runs at twice the speed. I will look into what Clough42 uses. Cheers Andrew
Hi Stefan. Clough42 has a lot of info on TH-cam regarding his electronic lead screw. Also links to products he uses (eg TI launchpad) and source code on github. For his project he is processing a lot more encoder pulses, but I will investigate! Cheers Andrew
@@learningturningmetal I did look at several of the electronic lead screw videos. The lack of a complete list of parts as well as things have to be collected all over the market made me drop the idea. I do have an electric rigging for controlling the lead screw during turning. Automatic reverse of the lead screw, just to feed the cross slide... No assembled to tested yet.... Motor control, reverse proximity switches, code is ready. Just the final testing remains.
You are not accounting for the "phase shift" when you are feeding from the from face to the back face of the gear. When you feeding in, you are changing the relative surface speed of the spindle vs the workpiece, by adding a small amount of speed on top of it. In a mechanical gear hobber this problem solves itself, because gears and leadscrews are in mesh, and feeding in automatically speeds up or slows down the spindle in relation to the workpiece. You either need a precise and constant feed speed and account for it in the code, or, you need to electronicaly measure the distance covered by feeding in and account for that.
Since the hob angle of rotation is electronically locked to the gear blank angle of rotation, it is hard to see how the X-axis position or speed would have any effect when cutting a straight-cut spur gear. However, cutting the gear by moving the X-axis very slowly and steadily only in one direction should remove any such error.
Try cutting a gear with an odd number of teeth 21, 23 etc and see if it changes also maybe change the drive ratio to an even number like2:1or 4:1. Could be a mathematical problem . You have encoders running even numbers , micro steps of even numbers a gear of even numbers but a final drive that divides by 3 . Another thing I was thinking is how consistent is the speed of the mill spindle ? If the rpm is surging up and down maybe the rest of the system isn’t reacting fast enough to compensate ? A Dual channel oscilloscope would set to tread the pulses of both encoders might show this or maybe something as simple as a strobe light ( timing light ) shone at some marks on the gear blank and one cutter tooth would as well .
Hi Ian - thanks for the suggestions. I think that math is good at the moment. Effectively I've got a 1:1 ratio in relation to incoming encoder counts and outgoing stepper counts. So to make a 13 tooth gear, just read 13 and output 1. It's a real brain teeser - far too many variables that could affect the outcome. More testing needed. Cheers Andrew
Gday Andrew, you have eliminated a lot of possible factors and it really can be only a couple left, 1 the job is not correct and 2 it maybe be arbor flex on the hobb, there was a bit of runout on the gear arbor to, I know this won’t bet you and you get it sorted, Cheers
Hi Matty - yeah there is runout on the arbor - I will replace the collet chuck with a 4-jaw to eliminate that. I too am wondering if the hob might be root cause. Problem is that hobs are expensive, and I don't want to purchase one without a future project in mind. The second hand hob market looks risky too, without knowing the history of the hob. Apparently incorrect sharpening of the hob can introduce all sorts of issues. Cheers Andrew
i love you sir you are amazing work i want to make Gear shaper chan you help me can you provide me code for arduino i am waiting your kindness and very thankful to you for i make gear hobber thanks a lot of sharing file of gear hobber
Great job ANDREW
I like your perseverance.
👍👍👍👍👍
Ha ha - I'm beginning t olose the will to live. It's starting to drive me nuts!
All the best.
Andrew
Andrew, Hats off for the effort. To echo a few, concentricity of the blank and the hob will have a big effect on the outcome. I work with hobbing machines daily. Hobbing machines are geared, and the indexing is done with a large worm gear turning the blank. There is significant rotating mass and inertia behind both hob and blank to insure the proper relative speeds throughout the cut. The hob and blank are also both commonly on shafting and are very rigid. When anything is loose or eccentric on a hobbing machine it greatly affects the outcome, but these are relatively easy to spot and fix. Seems like you have the code and counts correct, hurdles are the belt, rigidity, and concentricity. Technically you have the setup correct down to the Helix Angle (lead angle). Clearly you have reverse engineered the process. Again, incredible effort and persistence. Nice job. The other, other Andrew from across the pond.
Hi Andrew - thanks for the info. I will investigate the belt, rigidity, and concentricity. I'm beginning to wonder if gear hobbing in the home model engineering environment is too big an ask.
All the best.
Andrew
Excellent video Andrew. I would definitely try a different hob and see if you get the same or similar results. I can appreciate your feeling of not wanting to be defeated, so to speak. Again, excellent video and work. Looking forward to the next video.
Yeah - the hob is one of things that I would like to try to eliminate as the cause. Unfortunately they are quite expensive and I don't want to purchase one without having a need for it in a future project. I've looked at second hand ones on eBay, but not knowing their history it's a bit of a gamble. As I understand it, if a hob has been incorrectly sharpened it can produce all kinds of problems. Some people make their own hobs, but for me that would just complicate matters even further.
Cheers
Andrew
Brill series Andrew, your persistance is a real inspiration. I've finally watched all your vlogs now, and really enjoyed them
Thanks mate!
Hi Andrew, FWIW, there are two mechanical things that I think are worth investigating before delving any further into the electronics:
1) As others have pointed out, your arbour isn't running true - look at about 16:00 to 16:10, you can see both the chuck and the arbour running out against the background. Also at ~20:34 you can see it again, but it looks a bit more than a 'once per rev' eccentricity - maybe there are problems with the bearings in the work spindle? I think it would be worth sticking a clock on it for a while to eliminate this possibility. (Or at least, clock up your arbour to make sure it's running true before cutting.)
2) Even with perfect timing pulleys and belt, if there is any eccentricity in the pulley PCD (not necessarily the OD) after mounting, it will transfer to a cyclic error in position. If the belt has too little or too much tension (or is just a poor fit to the pulleys) it won't settle smoothly into the pulley teeth, and there will be some 'cogging' as they rotate (i.e. some variation about the 'ideal' position as each tooth engages and disengages). Given that you now have the closed-loop stepper, I think it would be worth trying to drive the work spindle with it directly (through a torsionally rigid coupling) and eliminate the belt drive completely.
Good luck! I admire your perseverance! 🙂
Thanks for the suggestions - yeah there is definately runout. I will replace the collet chuck with a 4 jaw and try again. Yeah - I have wondered about the accuracy of the pulleys. They are supposed to be timing pulleys with teeth. Driving the spindle directly seems like a great idea. I will certainly give that some thought!
Cheers
Andrew
Hi Andrew, it would be interesting to see on the oscilloscope a comparison of both encoders while the machine is running.
Rich
Hi Rich - yeah that would be interesting, but I have no idea how to interrogate the encoder built into the stepper motor. Food for thought!
Cheers
Andrew
@@learningturningmetal Hi Andrew, I'm thinking that you are losing pulses from the mill encoder. do you think that swapping the outputs would help? I would be thinking of replacing the mill encoder with another type
Rich
Hi Rich - I think I might have found the problem. See the next video - uploading now!
Cheers
Andrew
Great series Andrew. Both my thoughts have been suggested below.. tool rigiidity.. and arduino clock speed/cost of digitalwrite function.
Thanks Matt. I will certainly check out tool rigidity. I have a crude check in the Arduino code to try to identify if the CPU is struggling - and it doesn't kick in. I have since been testing with AndysMachines code (written in assembler) and that produces identical results. A bit of a brain teeser!.
Cheers
Andrew
One thing I might suggest if I may. You might get a better tooth profile if you don't
back track on the moving cutter. It seems to me that you'll get a better profile when
you cut through it once foward & shut the cutter off when you back track.
Interesting project. Thanks for sharing.
Hi Todd - I have since tried one cut - but the same result. Thanks for the suggestion though.
Cheers
Andrew
@@learningturningmetal Thanks for commenting Andrew. I didn't think that was your total problem.
I just noticed with the little bit of experimenting I did with making gears, when I backed off on the moving cutter,
my teeth were distorted compared to others that I didn't.
Hello Andrew, I admire your perseverance getting this issue sorted out. As someone who really has no idea about your electronics, I would like to present my observations on strictly mechanical terms. Is there an accurate method to measure the teeth on the hob? I am somewhat in agreement with wino4340 that the issue might be with the hob itself. Also, could there possibly be some "tool flex" in the hob shaft as I observed a little more material being taken off the gear as it was returned to the "start" position? Perhaps stopping the cutting operation with the gear to the right of the hob after being cut and examining it then. I am confident you will get it sorted. Good luck !!!. Earl
Hi Earl - thanks for your thoughts on this. It would be good if I could eliminate the hob as being the problem - but they don't come cheep. I will look into rigidity - more testing needed!
All the best.
Andrew
One thing that you might try is only cutting in one direction. Each time I see you do a cut you bring the hob back through the work for a second time in the reverse direction. Could this introduce a small difference? Excellent investigatory work though. A great video.
Hi Terry - thanks for the advice. I've since tried completing a gear in just one pass - but it doesn't appear to make any diffterence. 😭
Cheers
Andrew
I am wondering if it is the small eccentricity you see when the blank is on the mandrel. Have you tried indicating the blank with the headstock turning? Thanks for the video, I find this topic very interesting!
Hi Andrew.
Thinking about this, where you are moving the gear blank sideways to the cutter the spiral of the cutter is what is causing the wider gaps between the teeth if the blank is at 90 degrees to the mill spindle, try having the cutter directly behind the gear blank and using the X axis to cut instead of the Y axis, it will be harder to get the right depth but that can be set as a zero on your DRO that you drive the axis into. Other than that you may have to set the blank rotation axis to match the spiral axis of the hob cutter.
Hi Robin - thanks for the suggestion. My understanding is that the hob design requires the hob to be cut along the x-axis, in order for it to produce to correct tooth profile. Plunge cutting the hob will not produce the required result, as the cutters on the hob don't have the same profile as a basic gear cutter.
Cheers
Andrew
@@learningturningmetal The hob to work setting is usually not at 90 degrees but corrected for the lead angle of the hob, The direction of travel is along the axis of the gear which you are doing. You could accomplish this by rotating the head of the mill slightly in the correct direction. The lead angle (hob setting angle) is usually marked on the hob. If you are not cutting in the "hobs normal plane", you will be producing a wider space, as you are.
Hi Ralph. Thanks for the info. The helix angle of the hob is marked 1deg 20min and I set the angle of the mill spindle to that by moving the spindle down by 1" and getting the offset to be sine of that angle. I checked this several times and I think it was spot on. I guess I could experiment by adjusting the angle and taking more test cuts.
Cheers
Andrew
Thanks for sharing 👍
When the video was sped up, you can see run out on the shaft holding the gear blank. That would definitely make that gears teeth smaller than expected. Can you setup a tailstock center for the gear blank shaft?
Hi Don - thanks for the suggestion - I will certainly set up a tailstock centre and try again.
All the best.
Andrew
I don’t think your problem is with the electronics. I think your hobbing machine headstock doesn’t have enough rigidity, also your not using a tailstock to support the mandrel.the vibrations may be causing the teeth to cut oversize. If you have a tailstock for your dividing head or rotary table try setting that up to offer some support to the mandrel.
Thanks for the advice - I will check regidity out.
Cheers
Andrew
Perhaps cut a few plastic (Delrin) gears to see if it improves? Is the tooth width on the Hob OK?
@@kenshelhamer9816 thanks for the suggestion. See my latest video for an update 😉
Cheers
Andrew
Your gear blanks appeared to have an ever so slight wobble while they were turning. Perhaps it’s my old eyes playing tricks on me. Either way It might be something worth looking in to.
Hi mate - no, your eyes are not playing tricks. It's an issue I need to address.
All the best.
Andrew
Hi Andrew, the closed loop stepper motors have really come on, a lot cheaper than a servo & I would say perfect for this application. I would try another hob, cheap ones are NOT all equal! Also sort out the concentricity issue on your indexer. Check on the gear if there is a sort of phase shift as you go round from tooth to tooth. Hope this helps.
Hi Dave - thanks for the advice. It would be great to be able to eliminate the hob as being the cause. Problem is new ones are quite expensive and I would need to buy one suitable for a future project. I've looked at the second-hand market but I think that is a gamble, not knowing the history of the item.
More testing needed!
Cheers
Andrew
Perhaps the encoders/steppers/program allow a small amount of angular variation, slightly enlarging the tooth gap?
Hi Ken - could be. Both programs produce the same outcome - so I doubt it's that. It could be and encoder/stepper issue - but hard to spot.
Cheers
Andrew
thats soooo much stickout on that arbor
Andrew, This is a tough one for sure. During the high speed view it looked as if the gear shaft was wobbling but that could just be a video artifact. The counts are really close. It does appear that the gear is being over hobbed. Could the depth be a bit too much? Say trying a reduced depth of cut? Or breaking it up into 2 passes as the end of the gear shaft isn't supported. Granted rotationally you are beyond close. The question now may just lie in the effectiveness of the actual cutting operation. 32 teeth may not be enough for you to learn much from. You could increase the diameter of the gear to say 64t. see what sort of flex deflection you get at the end of the gear's shaft and the end of the hob's shaft. I think the issue at this point is more mechanical. I may not be in the right ball park. mostly just thoughts.
Thanks for the suggestions. I will do some more investigative work.
Cheers
Andrew
It looks to me like you have a good bit of runout on your spindle, that would make it cut wider than it should be.
Hi Timothy - well spotted. I'll replace the collet chuck with a 4-jaw and get the runout eliminated.
Cheers
Andrew
Hi Andrew, are you tilting the head of the mill in order to compensate for the helix angle of the hob? This is clearly visible in AndysMachines' video (probably also mentioned in the audio). The cut should be parallel to the rotational axes of the blank but not at an angle, right?
Or in fact is it tilted the right way?
You can see that the involute-cutter gear teeth have a slightly different profile (they're thicker). So I'd be suspicious of the hobbing cutter profile. Also, the stick-out on the spindle is fairly significant, given its diameter...and I suspect you're getting some cutter deflection that might account for some of the difference.
Hi Tim. You are on the right track👍. See my latest video for an update.
Cheers
Andrew
is the shaft containing the blank running out causing excess removal and wider gaps?
Hi Daithi - yes there is some runout. I will look to correct this.
Cheers
Andrew
Hello Andrew, I am wondering if you problem is interrupts .Unless you are running in assembler you really don't know what is happening when a position change comes in from the motor encoders. I don't know much about Adreno, I use to program Intel 8085 and 8051 microprocessors to do speed and position control. When a position change comes in it must take the highest priority , and everything else like LCD refresh must wait while the position is processed and the control of the stepper is processed and implemented . This is just my thoughts on the matter.
Hi Brian. I see where you are coming from but I think I've managed to get to the bottom of it. See my latest video for an update.
Cheers
Andrew
When you are cutting the gear at around 16mins there looks to be some runout on the chuck holding the gear blank. More noticeable at the higher 4x video speed. Could that explain the wider gaps?
Hi John - well spotted. A think the runout is just a few thou, but certainly not perferct. I will test with a 4-jaw independent chuck.
Cheers
Andrew
I think that a closed loop stepper driving your mill spindle would eliminate any "Lost steps" in that drive.
Hi Malcolm - yeah a bit of a brain teeser!
All the best.
Andrew
How much backlash is there in that belt drive?
Hi Jeffery. There does't appear to be any backlash in the belt system. The pulleys have teeth and the belt is sold as a timing belt. Mind you, belt drives are totally new to me. Maybe I should explore connecting the stepper motor directly to the spindle in order to rule that one out.
Cheers
Andrew
maybe there's a rounding error in the calculation so that there is a slight drift over time, have you tried calculating the ratio between pulses? does it end up as an exact whole number?
Could it be jitter between when you detect an encoder step to when you tell the stepper to step?
Thanks for the suggestion Kieran.
Cheers
Andrew
The hob is likely not causing the issue. I assume it works fine on the other hobbing machine. (Andy).
I had massive problems with a really cheap parallelport interface, it was missing encoder steps. (optocouplers) just too slow to "catch all the signals", missing some.
Try an encoder with less lines, slow down the spindle or use only A or B signal? (not both)
Then there might be and issue that your signal to the stepper driver is not triggering a step every time.
I have one driver that refuses to work on an Arduino, but it works on another controller. (seems to have too less power to drive the input reliably)
To check this I would change the software so that the signal to the stepper driver and the signal coming from the encoder have the same frequency. 1 Encoder pulse, causing 1 step signal. Then hook up the oscilloscope and see if the two signals are matching. Does this make sense?
If you have a speed issue in the signals, everything will look fine when moving slowly and if you reach a certain frequency it goes all downhill.
Hi Timo - thanks for the suggestions. More testing needed!
Cheers
Andrew
@@learningturningmetal Thank you for letting us participate. Do not give up. But ... computers do not deserve to cause frustration 🙂. Stay undefeated do not let the thing harass you.
Cheers Timo
@@timogross8191 I think I might have found the problem. Hopefully the next video will explain.
Cheers
Andrew
Hi Andy, very cool video you did. There might be an performance issue in your code. The digitalWrite function takes some time (If I remember correctly, it’s around 190 cycles per call which is ~10 uS for the 16MHz Arduino Uno). You are doing it two times in the for loop (20 uS). So you are able to handle 50000 Pulses per second. Depending on the pulses your encoder produces (and ignoring the fact, that the main loop does other things as well) this might be an issue (RPM/60 * PulsePerEncoderRev). Long story short: instead of using digitalWrite try to write the output registers directly. This will increase the speed drastically. Ideally you won‘t have not more than one iteration of the for loop per iteration of the main loop. Don‘t know if the encoder lib does support interrupts but this could be a better way to track each single pulse of the encoder and to ensure perfect synchronization (Depending on the hob speed you will overcome the capability of the arduino and you will need a faster cpu like the Esp32)
Hi David - thanks for the suggestion. I've since been testing using AndysMachines code which is written in assembler. So I guess his code would be far more efficient than my code. I tried to make my code as tight as possible, but included a crude check to try to identify if the Arduino was running out of steam. I will however investigate to see how I might be able to write the the output registers directly - not something I know how to do at the moment.
Cheers
Andrew
@@learningturningmetal I also think it's an issue with the timing. When you showed the traces in the previous video, there was very clear jitter in the timing between pulses. That would result in slight angular variations back and forth, which is consistent with the results you're getting. I believe you're using an AVR board, so one option would be to switch to an ARM board. Those clock speeds on the order of 10 times faster than an AVR. So would help to mask any timing issues with code.
Hi John - after further investigation I think the problem is hob related. I'm just in the process of uploading another video which might help explain.
Cheers
Andrew
Really enjoy the video! Could I put forward that it might be as much a synchronization issue but a synchro phasing issue. If there was Sensor on the mill shaft, and one on the closed loop motor shaft they could be compared. Say for example, both sensors are set so that a position on the cutter is “0”, and one on the motor shaft is also set to a “0” position. When the cutter makes contact, it slows a bit by friction as it cuts, so now the sensor reads “-1”. The sensor on the stepper motor might read “-3”, from being introduced into the rotational speed of the cutter. If the mills pulses need increased, the program counter will see the incorrect RPM count, and speed up accordingly. Independent of the mill, the stepper motor will sense a low RPM, and the program will speed it up to a null state. ( not +/-). Between the independent speed adjustments, the drift would be reduced I think. What say you?
Hi William - I see where you are coming from. I think I would need to get hold of another encoder and attach that to the motor shaft somehow. I will have a ponder!
Cheers
Andrew
Fascinating detective work. You seem very close to an acceptable working gear, but I know what you mean, it is annoying when it is not quite right. I can offer no help regarding the electronics but I do agree with other comments that a bit more rigidity may help. Can the hob shaft be shortened by holding it further into the collet? A tailstock might help if you can set it up so that it doesnt get in the way. Certainly cutting in one direction should give a cleaner finish, but if the synchronisation is correct you should be able to come back out of the cut without removing more material. A little more sleuthing required yet Mr Holmes 🧐
Hi Tim - thanks for the suggestion. I will explore regidity.
Cheers
Andrew
The sync issue is likely worse under load, I.e. while cutting. It sounds to me like your arduino is missing an encoder step. I have seen this problem in a similar application I am working on.
Thanks for the suggestion Fred. More testing needed I think.
Cheers
Andrew
it looks like a bit of deflection while cutting it I'd try it with a tailstock at least you'd know it's solid.
Good point - I will introduce a tailstock and check regidity.
Cheers
Andrew
It has been decades since I have had anything to do with servo loops but I was thinking that even with a closed loop stepper motor it would still lose a step. The motor encoder would detect it but in this configuration it cannot correct and keep sync. Is there a way to read the output of the stepper motor encoder to see if a) it detects and issue or b) at least see if the two encoders jive? If they are out of sync try cutting with less depth of cuts or try sobbing in plastic or some softer material. Just a thought. But, as they would say in your country....Keep Calm and Continue Hobbing!.....OK I admit bad joke.
Ha ha- more like Keep Calm and do something else! More testing needed!
Cheers
Andrew
Your work. As I assumed earlier, no steps are lost. A properly designed stepper system doesn't lose steps unless some evil guy throws an anvil into the running setup. As a disaster detection an encoder can be used.
After watching Your previous video I came across some hobbing videos launched the last 2 or 3 years. They used a much less advanced approach. Use an indexing plate for positioning the gear blank. Then move the table in the X direction and cut some teeth completely. Advance the indexing wheel a tooth or so, and make a complete cut of a tooth or two. Of course it ought to be possible to automate the index movements....
Maybe I'm wrong but I think You use an Arduino UNO as the controller. It's a handy and is a good friend but not very fast. Clough42 used much more powerful things in his digitally controlled lead screw project.
But, of course You get a lot of experience from Your work. Looking forward to Your next attach on the issue.
Hi Stefan - thanks for the suggestions. I think there is some kind of Arduino clone which runs at twice the speed. I will look into what Clough42 uses.
Cheers
Andrew
@@learningturningmetal It's not an Arduino clone. It's a high performance device, maybe a Texas processor of some kind. But yes, taka a look.
@@learningturningmetal Why not post Clough42 and hear what he thinks?
Hi Stefan. Clough42 has a lot of info on TH-cam regarding his electronic lead screw. Also links to products he uses (eg TI launchpad) and source code on github. For his project he is processing a lot more encoder pulses, but I will investigate!
Cheers
Andrew
@@learningturningmetal I did look at several of the electronic lead screw videos. The lack of a complete list of parts as well as things have to be collected all over the market made me drop the idea.
I do have an electric rigging for controlling the lead screw during turning. Automatic reverse of the lead screw, just to feed the cross slide... No assembled to tested yet.... Motor control, reverse proximity switches, code is ready. Just the final testing remains.
You are not accounting for the "phase shift" when you are feeding from the from face to the back face of the gear.
When you feeding in, you are changing the relative surface speed of the spindle vs the workpiece, by adding a small amount of speed on top of it.
In a mechanical gear hobber this problem solves itself, because gears and leadscrews are in mesh, and feeding in automatically speeds up or slows down the spindle in relation to the workpiece.
You either need a precise and constant feed speed and account for it in the code, or, you need to electronicaly measure the distance covered by feeding in and account for that.
Since the hob angle of rotation is electronically locked to the gear blank angle of rotation, it is hard to see how the X-axis position or speed would have any effect when cutting a straight-cut spur gear. However, cutting the gear by moving the X-axis very slowly and steadily only in one direction should remove any such error.
I think that effect is real, but would be very small considering the rotational speed of the cutter vs the feed rate.
Try cutting a gear with an odd number of teeth 21, 23 etc and see if it changes also maybe change the drive ratio to an even number like2:1or 4:1. Could be a mathematical problem . You have encoders running even numbers , micro steps of even numbers a gear of even numbers but a final drive that divides by 3 .
Another thing I was thinking is how consistent is the speed of the mill spindle ? If the rpm is surging up and down maybe the rest of the system isn’t reacting fast enough to compensate ? A Dual channel oscilloscope would set to tread the pulses of both encoders might show this or maybe something as simple as a strobe light ( timing light ) shone at some marks on the gear blank and one cutter tooth would as well .
Hi Ian - thanks for the suggestions. I think that math is good at the moment. Effectively I've got a 1:1 ratio in relation to incoming encoder counts and outgoing stepper counts. So to make a 13 tooth gear, just read 13 and output 1. It's a real brain teeser - far too many variables that could affect the outcome. More testing needed.
Cheers
Andrew
Gday Andrew, you have eliminated a lot of possible factors and it really can be only a couple left, 1 the job is not correct and 2 it maybe be arbor flex on the hobb, there was a bit of runout on the gear arbor to, I know this won’t bet you and you get it sorted, Cheers
Hi Matty - yeah there is runout on the arbor - I will replace the collet chuck with a 4-jaw to eliminate that. I too am wondering if the hob might be root cause. Problem is that hobs are expensive, and I don't want to purchase one without a future project in mind. The second hand hob market looks risky too, without knowing the history of the hob. Apparently incorrect sharpening of the hob can introduce all sorts of issues.
Cheers
Andrew
i love you sir you are amazing work i want to make Gear shaper chan you help me can you provide me code for arduino i am waiting your kindness and very thankful to you for i make gear hobber thanks a lot of sharing file of gear hobber
Thanks Sikander. Unfortunately I don't have any spare time.
All the best.
Andrew
Interesting problem. Keep chewing on it. Thanks for the look!
Thanks John - my teeth are now totally ground down and dentist appointment booked.🤣
All the best.
Andrew