Impatience is the enemy...or is it the mother of invention?

I’m starting to have some success with this CNC-gig. I’m getting good results. I’ve caught the bug. All good.

But I find that I’m always trying to push the feed rate and cut depth on my projects to cut down the amount of time they take to create. Invariably, I can cut a 1:20 job down to :45 if I just play with plunge rates and feed rates on each of the bits. I know I’m probably sacrificing quality by doing this, but honestly, I’m impatient - and I’m planning to babysit the CNC, so time matters.

So, I’ve read a lot (as I tend to do) and have tediously worked my way through all of these chip-clearance calculations, etc. etc. - but frankly, implementing these formulae proves to be aggravating - either the bits don’t come with the specs I need, the software doesn’t allow me to specify everything, or I lose patience because the formulae are for milling metals and such and nothing as straight forward as cutting wood. I was a math major in college, but even so, the guides seems way too “scientific” for what SHOULD be a simple ask: How fast and deep can I drive a #201 through hard maple? How about Walnut? Mahogany? What about a #112?

Then there’s the question of what I should be looking for in the result that indicates that I’m going too fast? I have the feeling that if a more experienced CNC-er were to look at my results, they would say, “You could go faster, you know”…or…“Slow that down and you’ll get cleaner results”.

On that last question, there is a perfection issue (Voltaire: Perfection is the enemy of good)…if a symptom of “going too fast” is a fuzzy result - and I can knock that fuzz off in about 3 minutes of hand sanding - is it worth an extra 15 minutes of CNC time to not have the fuzz? Is there wear and tear on the router that I’m not considering? Something else?

What I know for sure is that, when I’m pushing work through my router table, or cutting patterns by hand, I feed FAR deeper and faster than I’m doing on my CNC. There is the 1/4" vs. 1/2" shank, but even taking that into consideration, it seems to me that I should be able to “hog away” more wood than I’m advised according to the (albeit conservative) standards that are packaged in CC.

Does anyone know of a non-scientific guide to “depth and feedrate” by media type that’s simple, straight forward and non-mathematical? Guidelines for specific bits (or “bits like this one”) on specific wood species?

If not, by the way, who else would want such a thing…and would you be willing to help contribute to one?

Much appreciated.


The problem is, the CNC doesn’t have a sensor to detect a lost step, so feed and speed rates have to be conservative so as to preclude them.

For feed and speed rates, I like the testing technique from:


Have you seen my feeds & speeds page and if so did you turn away because of the math ?

The reason I am asking is because I wrote this after spending countless hours (and a fair number of reading and writing posts in this thread) myself to answer those same questions (or at least find something I would be comfortable using myself)

It looks bad, but it boils down to very, very little math:

And even that is taken care of by the spreadsheet

Basically: pick a tool diameter, pick a material, look-up target chipload, input RPM, boom, get feedrate and DOC

EDIT: but it’s still a coarse grain approach. For figuring out the optimum between two different species of hard wood, then experiment, as @WillAdams pointed out


Chances are @Julien and @WillAdams, I have already read the referenced docs…but I WILL go back and look to see…I’m smarter now than I was six months ago…so maybe I missed something valuable…

But what I’m looking for is something non-empirical (I know that will offend all of the engineers in the group, so sorry)…something along the lines of: "Hey guys…I’ve found that I can push a #201 through hard maple at 95 at a depth of .1 before I get into trouble. If enough people confirm the same result, then I know where to start and adjust.

I’m a math major, I’m not afraid of math. But I’m also an artist…and sometimes I feel like engineering can be an art form.


This is a very valid approach too, and I understand it’s pretty much what CC does (tested-recipe-based F&S) but CC plays it safe as you mention, because…very large user base. Your idea I guess would be to come up with a crowd-sourced “optimal MRR recipes” database.

Where I think you would maybe struggle, is that there are as many upper limits as there are Shapeokos

  • due to various mods (HDZ or not etc…)
  • due to how perfectly tuned the machine is
    so an aggressive recipe that would work perfectly for someone, would break things for someone else.

I went back to the docs…and yes…I did read through those before. They’re amazingly thorough and clear. They’re also encouraging everyone to repeat the same process for the same things.

How many people have used a 201 on hard maple? Does every one of them need to go through the calculations, select a chipload and RPM, plug it in and get the answer? Why not just say, For a 201, on maple, here’s your starting point. Period. Why make everyone take out the calculator / spreadsheet?

Seems to me Carbide could do that for all of their bits, once, for everyone…and post it. It might make me want to buy the Carbide bit vs. some cheaper bit I find on eBay…if all the work is done for me already.

This is pretty much what @wmoy does in his #MaterialMonday videos, but…he has to settle for “comfort zone” settings. A middle point could be to have the current CC database, enriched with many more sub-recipes (i.e. not just “hardwood”, but woodX, woodY, …). Interesting endeavour in any case, and I’ll be happy to help and contribute with tests of mine (but c.f. my point above…if an HDZ can cut X deep and Y wide at Z feedrate in hard maple with a 201…what happens with a stock Z ? There are just too many parameters in this F&S thing to come up with look-up tables for every possible case)


You do understand you’re presenting a challenge, right?

This is a perfect crowd-sourcing solution waiting to happen.


I really did not mean to sound pessimistic about such an approach, just sharing what I think could be the pitfalls, because I walked a similar path :slight_smile:

There is a crowd-sourced set of feeds and speeds on the Shapeoko wiki, but when consulting it I often felt like “…darn, this one important detail/parameter is missing from this otherwise great example” or “these two things are completely contradictory”, which then led me down the rabbit hole of the chipload thing. Anyway, I have said too much already, I’ll let others be more enthusiatic ! (and again I’m always one for testing & reporting, so don’t hesitate to count me in the “crowd” for this initiative)

1 Like

Ask 10 people for their settings on carving Maple, and you’ll get 17 different answers.


I tried crowd-sourcing numbers at:

let me know if you want an account to try to continue there.

Alternately, we could work up a collection of the CSV files which Carbide Create uses.

EDIT: or, I did have a tableau up which was going to accept input from a Google Form feeding into a Google Spreadsheet, something like:!/vizhome/Carbide3DCNCFeedsandSpeeds/Sheet1?publish=yes

if folks actually want to contribute I’ll put that back up, but it will have to be separate from the Carbide 3D page and folks will have to only input tested values.



Different people have different views on what finish quality they expect, how much post-machine finishing they are prepared to do etc.

There’s differences between the spindles, HDZ, Z plus, regular Z, how well tuned in and tight the V-Wheels and belts are, XXL, XL and regular size Shapeoko (belt stretch is worse on larger machines) and that’s before we talk about toolpath strategy.

So far I’ve;

  • Started from Winston Moy’s defaults for the cutters in “Wood stuff”
  • Learned to use roughing and then finishing paths in my stock to maximise removal rate and then come back for a clean finish
  • Learned what I think “happy” sounds like and pushed speeds up and down in the feed override in Carbide Motion when the machine sounded like it could go faster or was screaming horribly

I’m not a woodworker, I’m just an engineer who’s learning about this inconsistent material; wood.

Even cutting in relatively homogenous and consistent materials though, like baltic birch ply or MDF, the speed and depth you can run an adaptive clearing is way higher than a countour, and the feed rate at which it starts to chatter and scream depends on the direction of the cut too.

If what you’re after is faster cuts, a mixed strategy of roughing out fast with a bigger cutter followed by shallow finishing passes can buy a lot of time without compromising the end product.

All of that said, if it was feasible, a crowdy DB of how hard people push their machines would be interesting, the question is how to categorise the data to make it readable, there’s so many parameters to a cut and I’m not sure that just MRR is particularly useful.


Oh, seems to still be up at:!/vizhome/CNCFeedsandSpeeds/Sheet1

but I’ll have to find the Google Forms or spreadsheet link


Yup…I am a woodworker and not an engineer — and I have learned a LONG time ago, that wood is alive and fickle. There is a science, but it boils down to art. Everything gets tried on a scrap test piece and adjusted. But having a solid starting point is the time-saver.

Yes! And that’s the power of crowd sourcing. Get enough people voicing their opinions and you develop a trend of combined knowledge. It’s a statistical analysis. You accept that statistics are a guide and not a gospel, you use them as a starting point, guideline, and sanity check. If we can generate enough participation, the results will provide an amazingly accurate starting point.

@WillAdams: This looks like a wealth of info…I’ll have to go through it.


That is rather handy Will, looks like an excellent start.

It hadn’t occurred to me to try and machine Limestone either :wink:

1 Like

There’s also a lot of subjectivity. For example if you’re using double sided tape, your holding force will be a lot weaker than with clamps. Or if you’re cutting around tabs, you might want to have a gentler cut that doesn’t rip your part out.

There are way too many cases and caveats for us to be able to say “hey, here’s some kick-ass speeds and feeds… but use your own discretion.” It’s far easier to have an advanced user opt to go faster, than to have CNC newcomers know when to opt to go slower.


there’s also the current lack of “Dynamic F&S”, CC uses the same feedrate for slotting as it does for all of pocketing etc…

you can post-process the gcode file to correct for this:

but without correction, the F&S will be for the worst case, not typical case. Once you can correct for it you can push the common case harder


It’s important to remember that, even in statistics, we need to be careful with categorising and labelling our data, otherwise what we have is just noise that we can’t draw any useful inference from. Just having lots of data doesn’t improve accuracy or relevance, having good labelling and categorisation along with knowledge of the data collection mechanism is a precursor to that. Without this the output is as likely to mislead as to inform.

If the data doesn’t have sufficient context and grouping such as roughing or finishing pass, machine type, Z axis type then it’s not really going to speed up my homing in on what I can run on a particular job or give me much context to say “Hey, howcome lots of other folks can do xxxx at yyyy than me?” and go look at my setup.


Why are artists always like that? They can poke a paintbrush at a 1/4" square of canvas for an hour, but can’t wait for the toast to pop up out of the toaster! :smiley: :smiley: