ER20 collet nut tightening torque / tooling?


So far I have been tightening my collet nuts “monkey tight, not gorilla tight”, and it has worked out nicely. Now that I use ER collets with super low runout, I am wondering if I should be a bit more accurate than this, and have some specific tooling to make sure I apply a controlled and repeatable torque on the collet nut.

I read somewhere that the recommanded tightening torque for ER20 collets is 59 ft/lbs.

What’s a cheap option/tool to do this ? (bear with me, I’m not a machinist or even a mechanically-inclined person, the answer may be as simple as “get a wrench of just the right length, you doofus”)

EDIT: oh and by cheap I meant much less expensive than this


If you already have a torque wrench maybe you could cut a “head” that fits the socket drive end similar to that expensive wrench you posted?

1 Like

How about using something like this crowfoot wrench attached to a regular torque wrench

1 Like

I have never used any special tooling to tighten my ER20 collets. Nor have I had a problem…


Hmm, 14 ft. lbs. for ER11:


Yeah probably overkill, I agree. But it’s always good to own one of those gismos that makes you feel/pretend like you know what you are doing :slight_smile:


@WillAdams thanks, the ER20 model is still 66$, much better but still above my arbitrary psychological price threshold.
@Winters636 and @luc.onthego this makes sense, I don’t own one but I looked and I could buy one in the low 30$ range, and then “just” change the head indeed.


If you do use a crows foot or anything that extends the arm of a torque wrench, you will need to do a mathematical formula to ensure you are getting the correct torque. I used to know them, but Im sure if you search google for torque wrench offset it will pull some examples up


Is 59 ft-lbs appropriate for Carbide 3d’s HDM Spindle (ER20) collets?

IMO, monkey tight is more than fine for the machine and spindle capabilities you have. You are running much less radial and axial depths of cuts than industrial machines that require a bit more detail to the tool assembly. An ER collet is a compression collet and holds the tool shank with much less torque than your standard router collet.

Add: Two wrench system is more than adequate.


We used a tool very similar to this when working with fuel and refrigeration connections. Maybe a local tool vendor could offer some suggestions.

1 Like

I’m actually more concerned about over tightening (with those huge wrenches) than under tightening. But not if 59 ft-lbs is acceptable.

Nice kit - perfect if the wrenches are the right size for Carbide 3d’s ER11 and ER20 spindle collets! :slight_smile:

Usual preface, I’m with PreciseBits so while I try to only post general information take everything I say with the understanding that I have a bias.

Short answer is it depends on the collet bore size you’re using.

Most times you will just see the max torque that the standard (ER20, ER16, etc) is rated for. Personally I go back to REGO-FIX for this as they are the inventors of ER collets. Their spec for all the collets is based on the nut type and the diameter of the bore of the collet (you can find it listed in their catalog (Link), currently on page 298).

For a standard ER20 collet with a standard nut it breaks down to:

  • 12 ft-lbs for up to 1mm (0.0394")
  • 25 ft-lbs for 1.5-6.5mm (0.0591"-0.256")
  • 60 ft-lbs for 7-13mm (0.276"-0.512")

That being said I have a lot of experience with mounting and testing collets in our measuring systems. For a good collet these are way past the point you need to reduce TIR but you may need more torque to reduce vibration in cutting. It does wear the collets and especially the nuts faster the higher the torque you use. When testing we could get around 400 cycles out of an ER20M nut at 20 ft-lbs before it started showing increases in runout (testing with ER20 1/4" collets). Reducing it to 12.7 ft-lbs would over double this with collet TIR within margin of error of the 20 ft-lbs measurements.

All of the above will depend on the fit, finish, materials, hardness, coatings, etc of the collets and nuts. The point I’m making is that like for like there’s more wear for the higher torque used in a meaningful way. Haven’t got any good data on cutting with it, and if the lower torque isn’t reducing vibration enough that’s going to wear the collet more in cutting.

Hope that’s helpful. Let me know if there’s something I can help with or expand on.


Thanks! Are speeds and feeds recommendations available for PreciseBits products?

I won’t get into too much brand specifics here for fear of crossing the marking line.

It’s something that we always fight with. In the real world, especially when you are talking about application specific cutters, there’s no useful easy answer to give. Your best cut is going to be limited by either the bit, the system, or the material.

To give an example let’s make up some numbers and materials. Let’s say that I have a 2 flute cutter that’s been designed to cut well at a 0.004" chipload in X wood/plastic/metal. It’s been tested, it gives a beautiful cut, has long tool life, etc. So now I put that up as the recommend chipload.

Now we sell that cutter to a customer with a system with 0.001" TIR and now my “recommend” number is giving poor results. Instead lets say that have good runout but their system can’t take the deflection, or a more brittle but just as hard similar material that then fails running that chipload, a wood that’s close to what we tested but has silica in it, or the same wood with twice the moisture etc, etc.

One solution to this is to add a margin to the chipload numbers. That typically results in giving someone with no issues worse finish, cycle time, and tool life. Another is to add a caveat that no one reads etc. What we decided on was to instead give guidance to test and a process to find the right feeds and speeds for YOUR material on YOUR machine (which is what I recommend you do with ANY tool no matter the brand). We obviously have internal data and numbers and we will give more specific numbers when we know the system, material, and setup used. We’ve also been working on some new ideas for the existing method especially as we started making bigger tools. But that’s why you won’t really find “recommended” chiploads or feeds and speeds on our site (some exceptions).

One last thing to take away from this in general. EVERYONE has this issue. Some manufactures will just go and put what the tool can take without a horrible cut on a perfect industrial machine. Some will give a very conservative number that will leave a lot on the table for cycle, finish, and tool life. Some will give a range, use a formula, etc. There’s problems with all of those, and frankly also with our “solution”. I personally don’t know what the right answer is, I just know that I hate all my available options. (Gets even worse after you throw marketing into it).


For their standard ER11 Collets/Clamping Nuts, Rego-Fix also recommends a maximum torque of

  • 6 ft-lbs for 1.0-2.9mm (0.039"-0.098")
  • 24 ft-lbs for 3.0-7.0mm (0.118"-0.256)

Maybe half those values would be appropriate for Carbide3d HF Spindle collets/nuts?

I wouldn’t say it’s that simple. The problem is that it really depends on the finish and how well your collet and spindle match. If you have a good finish and well matched and conforming tapers then you can get away with a lot less torque.

For all of the below, don’t hold me to this, this isn’t an official recommendation, etc. Basically this is all at your own risk and your evaluation.

If I was going to try and find a method that didn’t require new equipment or tools it would be this. With everything freshly cleaned, take ideally a blank (or a tool but don’t hurt yourself doing this) and insert it into the collet without the nut. Make sure that the blank is far enough in that it’s making contact with the entire length of the collet. Then put that in the spindle (again, without the nut) and put a decent amount of hand pressure pushing the collet into the spindle. While doing this grab the end of the blank and try to move it in multiple directions (at least one per leaf). You want to do this with a decent amount of force.

What you will find when you have a good match is that their no spots where you can feel movement. In cases where there’s not a good match there will be “soft” areas where it can move. It’s not really going to be enough to see but you’ll be able to feel it. You should also be able to do this by a bit more than finger tightening the nut but it’s going to be a lot harder to tell as the taper in the nut also starts to come into play.

Just to cover it now, if you do have some wiggle in the collet it doesn’t mean that the collet is bad. It also doesn’t mean that it’s a good collet if you don’t. e.g. you could have a massively off center bored collet that would feel good because the taper matches and there’s a good bore size for the blank. Also to cover it now, a worn nut is going to throw any of this right out the window. Make sure you don’t have any grooves or smeared metal in the nut taper.

If you have any of those “soft” areas I’d say just go to the torque limit. Either the tapers or the bore don’t well match and your going to need the extra torque to compress the collet. If you don’t then you can probably scale back the torque. On my one set of data I wouldn’t say you can scale by half though. For that ER20m test I referenced above, we saw changes in the TIR going from 20 ft-lbs to 10 ft-lbs, hence the 12.7 ft-lbs I referenced. It didn’t change on all of them so it might have just been on the ones that weren’t a perfect match. Regardless, for some margin I’d say go 65-75% of max. Again this is not an official recommendation just my best stab at it with what I know of collets and the loads those here are usually hitting.


Shouldn’t axial loads be considered?

Not sure what you mean? In general yes, to the test above it isn’t a test for load, it’s a test to see how much the collet matches the spindle. If there’s not slop in the match between the collet, spindle, and blank then you will get good clamping barring either a bad nut (wear or incorrectly made) or a spindle that has been reground one too many times or incorrectly machined. I guess you could also have a matching taper collet with too short a length before the groove for the extractor ring. No way you’re getting load or really any empirical numbers without test equipment though (might be able to use a phone with a vibration app but I don’t really trust those that much).

Only other thing I can think of here would be to see if you could pull the blank out of the collet (don’t try that with a cutting tool).

Again though, it’s not my official recommendation. I’m just proposing something that at least has a check before wholesale lowering the torque. To really test this you would need to run at both torque numbers under your maximum cutting load and measure the vibration. That would be on top of checking the TIR at both the torque numbers.