Grinding Cams - Will this Work?

[mu38&Bg#]

Finally, found Achim's website, couldn't locate it in my bookmarks. http://metallmodellbau.de/

As for cam error using Dave's method. I actually drew the stuff in CAD and traced a path as described by the method using a 4" grinding wheel. There was no guessing involved. Yes, it will work fine in most cases, but don't pretend it's exactly as you'd expect it to be. If the discrepencies don't bother you, proceed.

 

[dsage]

Well I’m going to have to declare a TOTAL success in metal.
I prepared everything for using coolant but I didn’t have to use any - BONUS
There was so much air blowing out of the mini-grinder fan that it actually blew the sparks sideways and kept the metal cool too.
I made a little cardboard duct anyway to make sure all the air was directed at the grinding.
After each 5thou pass the metal was warmish to the touch but you could hold your finger on it ( so maybe 130deg F?)
By the time I got the next pass dialed in the metal was cool.

If you get the light just right you can see faint facet lines in the surface finish but you can’t feel them. With a flat lifer in place there is no sign of wavering in the readings as the cam turns. The surface finish is perfect even with the coarse wheel. I won’t change that. Not sure what the camera is picking up. It shows some radial lines that I don’t see on the real thing.

WOW.!! I never thought this was going to work out so well.

I measured the results:

The program assumes full contact for the lifter. I measured duration at the .002 lift points. The design of the exhaust lobe was 135deg duration, I got 122.5 I’m sure eliminating the .002 at each end would easily get back several of the missing degrees of duration. From experience probably all of it. About the same error for the intake cam I ground.
The lift is a few thou shy of the design at .071 instead of .074

I could go measure the duration at the zero points but that’s not how the engine will be operated. It would be easy enough to fudge the input numbers a bit to tune the result if I had to. I’m leaving it. I’m more concerned with overlap which I have total control over now because I just have to dial in the number of degrees and then zero the CNC reading and grind the next lobe.

So, bottom line for anyone with a CNC mill this is the cats meow for making cams.

I’m going to try to automate the machining passes a bit more. I have to do a bit of fiddling to get each pass set up. (about 5 seconds) as it is.

Now all I have to make is 16 lobes plus some spares.


Sage

 

[dsage]

Greg:

I'd be interested in what you found to be wrong with the profile when you drew it in CAD.
It measures better and was easier to made than any cam I've ever made.

Sage

 

[mu38&Bg#]

The greatest deviation from the desired profile occurs where the flank meets the nose radius. There is excess material left behind. This is more prominent if a small grinding wheel or a large flank radius is used.

Why do you think the lift is smaller than intended? There is no error using this method at the apex of the nose radius. The results do look good however.

 

[canadianhorsepower]

How many lines og Gcode did you need for this
great work

 

[dsage]

Greg and Luc:

I measured everything to do with the profile. The base diameter I made right on by tweeking the last path and the lift comes out only a couple of thou short at most. The duration is a bit short but I haven't measured it at zero lash. My engine needs about .002 of lash so I measured it there. It was about 10deg short at that lash. I suspect all of that will be made back if I measure it at zero lash as the program calculates for.

Luc:

The number of lines will vary I suspect based on the profile. I just took the output of CamCalc and converted the numbers to Gcode.
In this case there were 135 lines. There will basically one line for each degree of rotation so a maximum of 360. But when you get around the base circle a lot of degrees are zero cut so it becomes one line to move from one degree setting to the next with the same cut setting. No need for a new line until the depth of cut changes.

Having said that there are about 15 passes required of 5 thou each of the same code. If you were to make that all one program you'd have 15x135 lines. I've been setting the offset manually and running the same code over.


Sage

 

[dsage]

I've ground a couple of more cams and with careful attention to the base circle diameter in the final grinding pass the lift comes out only about 1 thou short. I also measured the duration at zero lash and it is right on for both the intake and exhaust cams I ground.
So I guess the bottom line is that this method works perfectly and produces the cam specified in the CaCalc program.$

I can't explain why except for perhaps one thing:
The lifter is already considered in the calculations. So perhaps the offsets in axis' and shifted point of grinding as you rotate are already figured into the calculations for depth of cut.
If you draw a 4.5" grinding wheel next to the cam lobe drawing, the grinding wheel is pretty much flat like a lifter.

Maybe this is why the cam turns out exactly as spec'd.

Sage

 

[dman]

I've ground a couple of more cams and with careful attention to the base circle diameter in the final grinding pass the lift comes out only about 1 thou short. I also measured the duration at zero lash and it is right on for both the intake and exhaust cams I ground.
So I guess the bottom line is that this method works perfectly and produces the cam specified in the CaCalc program.$

I can't explain why except for perhaps one thing:
The lifter is already considered in the calculations. So perhaps the offsets in axis' and shifted point of grinding as you rotate are already figured into the calculations for depth of cut.
If you draw a 4.5" grinding wheel next to the cam lobe drawing, the grinding wheel is pretty much flat like a lifter.

Maybe this is why the cam turns out exactly as spec'd.

Sage

been trying to say that.... the missing thou can be a number of factors but isn't the radius. most likely it is because the narrow radius on the nose of the cam gives less contact and the wheel cuts it more aggressively due to less pressure and deflection where there is more. you may also have a thou of backlash in the lead screw or if you have steppers they do lag behind because of lack of feedback. the driver can only output steps and micro steps that would give an ideal position under zero load. they do deviate from that under load.

 

[mu38&Bg#]

I can assure you the cams deviate from design where the flank meets the nose radius. Whether or not this is enough to cause problems with followers depends on the factors I've already mentions. They are not as designed though. Simple cams for display engines are not demanding of profile or accuracy so it won't matter 99% of the time. Real cams are much more than just duration and lift.

 

[dman]

I can assure you the cams deviate from design where the flank meets the nose radius. Whether or not this is enough to cause problems with followers depends on the factors I've already mentions. They are not as designed though. Simple cams for display engines are not demanding of profile or accuracy so it won't matter 99% of the time. Real cams are much more than just duration and lift.

again it is a little difficult to calculate the entire profile by hand but you can easily see that the geometry dictates it is impossible for it to deviate more than .005 at any point based on the distance of max lift from the axis of rotation. if you run it through the Pythagorean theorem you get a difference of .005 and that's in an impossible situation. the actual deviation of the contact point from the centerline is likely no more than a 1/3 of that and at that point the error is .001 or so.

i know exactly what the concern is, but i can estimate how much of a problem it is and the normal problems of deflection and such are a bigger concern.

 

[MuellerNick]

I measured everything to do with the profile.

I'm sorry, but you didn't. If the lift is as intended and your timing angle is off, the rest will be off even more.
Stating your measured everything, only shows you didn't understand the physics of a cam that well.
Are you sure, the acceleration/deceleration isn't completely off or are you just guessing?


Nick

 

[dsage]

You're right. I never pretended to understand cams to any great extent. That's why I titled this thread "Will this work".
I'm only interested in producing something easily. But that's not to say I'd be happy with something whittled out of a metal bar either.
I really appreciate all or your input (everyone).

Explain what you mean about timing angle.
If you mean is the lobe shifted around the shaft from where it is intended?
I doubt it (see last paragraph) but it might be. If so it makes no difference.
The lobes are ground in pairs (intake and exhaust) and assembled onto a shaft with bearing and spacers in between.
If I grind the pairs in one setup then both of them are shifted the same amount but relative to each other they will be exactly at the Lobe center angle I chose. It makes no difference, I just have to rotate the pair on the shaft at assembly time.
For each pair I intend to measure where the TDC point I want is and mark it. Then I will assemble it on the shaft which will be rotated to the cylinders TDC point and rotate the pair to the mark and pin it in place.

As for acceleration and deceleration being right or wrong:
I have no idea. I know that the base circle is right on diameter and (in case I didn't mention it) I used some radius gauges and best I can tell the tip radius is VERY close to expected. So I figure there are a limited number of ways to join those two circles with the flank. It could be the flank is flatter than desired (as eluded to by one of you) but it certainly isn't concave or bulging out. I can attempt to measure it to see if it has a radius of (can't remember off hand) someting like 0.56" as designed.

I don't want to disrespect anyone that grinds cams for real engines but I don't think whatever errors that are there are significant or will make any difference to the model engines operation. They are certainly leaps and bounds better than a hand filed cam and have abetter finish than the side milled cams I made previously.

I will go back to one thing. The CamCalc program that generated the data for the cam lists the acceleration G-forces etc. for each degree of rotation along with the cutting data. It is what it is, and I have to assume it is correct.
I remind all that I have simply added a slight modification on getting from one facet to the other to avoid the filing. The intermediate points (facets) are the same no matter the process used to join them.


Sage

 

[dman]

before somebody catches me i made an error on my last rough ass guestimation. i used the wheel diameter instead of the radius when trying to figure out chord height, well i did some more estimating, i can run precise numbers if i look up a few formula that i forgot or i can draft it in cad (i don't have it installed.) but by my new estimations the max offset of the contact point to the wheel is probably ~60-70degrees from the lobe being vertical and the error is .0025-.0035" again, just an estimation. if i play with cad i can give precise numbers but i really think the op has absolutely nothing to worry about.

 

[mu38&Bg#]

If one is happy with this method that's fine. If one understands what the downfalls are and is still happy that's even better. If one doesn't understand it's time to ask questions. I'm giving up on forums left and right because you talk talk as much as you want but people only listen to what they want to hear.

 

[canadianhorsepower]

I was under the impression that these forums were to exchange knowledge
that would allow any of us to build better engine

Right now this post looks like it's a fight that someone end out winning

I'm disapointed seeing so much knowledge bying over run by typical personality

 

[canadianhorsepower]

I was under the impression that these forums were to exchange knowledge
that would allow any of us to build better engine

Right now this post looks like it's a fight that someone end out winning

I'm disapointed seeing so much knowledge bying over run by typical personality

beeing over run

 

[dman]

yeah no more arguing from me.. dsage congrats on a beautiful part.

 

[dsage]

Thanks - to everyone.
I guess I've insulted some for not listening to reasons why these cams may not be "correct". But in the end, I have several in my hand and I can't see (or measure) enough wrong with them to be rejected.

BTW I attempted to measure the flank radius which is supposed to be 0.56" radius. The best I can say is that it is NOT flat because a straight edge rocks on it, and my 0.5" radius gauge shows a bit of light in the center of the arc so it must be slightly over 0.5". Not very precise I know but worth something.


Yup - may be crap theoretically but practically they're more than good enough for my purposes.


Thanks again

Sage

 

[MuellerNick]

I guess I've insulted some for not listening to reasons why these cams may not be "correct".

Accepted!

Yup - may be crap theoretically but practically they're more than good enough for my purposes.

I can't argue against that. When you are happy with them, you are happy with them!


Nick

 

[dsage]

I just thought I'd throw in an update on this cam grinding approach. Yes I'm still at it.

First off I have to apologize to a couple of you who said there would be something (not exactly defined) wrong with the shape of the cams. There does appear to be a problem but I previously had no way to prove it and the cams I've ground appear pretty good but then they are small and difficult at best to measure.
A friend of mine devised an excellent method of seeing exactly what the grinding wheel is producing. He suggested drawing a REALLY large cam on cardboard and then input that data into the Cam Calc program. Then install the cardboard template in the rotary table and set it up like I would a cam blank. Then I run the program a step (1 degree) at a time and trace the profile of the grinding wheel for each step to show the path of the grinding wheel to see how it deviates from the profile on the cardboard.
To be sure the method is correct I first used a straight edge in place of the grinding wheel to simulate the way the program was supposed to work which is to turn the blank, take a pass across it with a milling cutter, rotate the blank, take another pass, etc. etc.
Using that method the straight edge follows the profile perfectly so I know the Cam Calc program and the testing method is correct.

See attached picture of the path produced by the grinding wheel. Sorry, there is parallax in the picture I couldn't get the camera straight on so it looks like the wheel is way off the line.
The shaded area is the pencil lines from tracing the grinding wheel at each step so you can see where it actually grinds vs. the proper path CAD line on the cardboard. I have to admit that none of the cams I have actually ground look this far off. Of course the errors are magnified about 20 times (best I can figure).
Someone once mentioned that the problem is that for most of the profile you are not grinding where you would expect i.e on a line from the axis of the grinding wheel and that the point of grinding moves as you rotate the blank. This is true but using the straight edge test like the program expects it become obvious that the program needs for the grinding (or cutting) point to move. If you look at the straight edge setup (next post) the cutting takes place (on this enlarged mockup) as much as 2" off the centerline and this is exactly what is required to produce the correct profile. In this respect an infinitely large grinding wheel would be best. Anything smaller and some of the grinding is missed because of the curvature of the wheel.

Hmmm maybe I'll make a little belt sander instead of using the grinding wheel.

Another issue came into view as well. If you look carefully at the picture, down near the bottom of the tracing you'll see the grind path is separated from the proper path slightly whereas at the same point on the other side of the cam the grind path is right on the line. At first I thought his might be some eccentricity in the cardboard template but although there was some it was not this much. I've also noticed on the real ground metal cams that the lifter was not returning to the closed position at exactly the same zero point I had set at opening. It was about 2 thou off on the steel cams and is probably 20 times that in this test setup. So that gap appears to be real. I can't explain that. I guess it's a problem in the calculations.

So, I'm abandoning this method of making cams for the time being. The upshot of this exercise is that my friend who is a good programmer got interested and he seems to think he can fix this problem with a new program which will compensate for the off axis grinding and the diameter of the grinding wheel. I hope he's right.


Sage