[RCaffin]

I have personally exploded a GT3-127T-12 GRN pulley spinning at 12,000RPM and stopping it
You wouldn't have a photo of the remians by any chance. I would LOVE to see that.

the length from the X-axis motor to X-axis driver is 12M but Y-axis cable length is only 3.5M and Z-axis cable length is 6M so using differential encoders is prefered
12 m? Have to agree there. Some nice RS422 driver chips such as 26LV31E maybe? That's what I have used in the past.

Cheers
Roger

[websrvr]

I have personally exploded a GT3-127T-12 GRN pulley spinning at 12,000RPM and stopping it
You wouldn't have a photo of the remians by any chance. I would LOVE to see that.

the length from the X-axis motor to X-axis driver is 12M but Y-axis cable length is only 3.5M and Z-axis cable length is 6M so using differential encoders is prefered
12 m? Have to agree there. Some nice RS422 driver chips such as 26LV31E maybe? That's what I have used in the past.

Cheers
Roger

I never thought to take pictures of the exploded pulley and there didn't seem to be any point in keeping it around so it was tossed at the time of replacement.

The 26LV31E is a line driver with four single ended inputs and four differential outputs so I think you meant to reference 26LV32D or 26LV32NS which is a line receiver with four differential inputs and four single ended outputs.

Yes in a pinch a quad RS422 driver chip would work but again, this requires adding an external board to provide encoder support, (board also needs a +3.3V REG to drop the +5V) which in my opinion should be present and I'd rather do it on the board where it's more appropriate and I can take advantage of the available encoder error detection provided by the MAX3097ECSE/MAX3098ECSE (tri-channel encoder line receiver) which will also accept being driven from NPN and PNP transistor driven encoder output signals.

[RCaffin]

so I think you meant to reference 26LV32D or 26LV32NS
Too much shorthand. Sorry.
I was thinking of how to make a common low-cost single-ended encoder drive 12 m of wire reliably. The use of a 26LV31E as a driver 'sort of' implies the use of a matching receiver, with the 26LV32 being the obvious choice. I carry both in stock.
The Avago etc 500 line (2000 ppr) encoders are very common, and I have been using them for .. um ... 30 years now? (Starting with their original incarnation as a HEDS unit from HP.) I find them very accurate and reliable. Plenty of similar units suitable for the LOW-BUDGET home hobbiest.

Extra PCB - true. But I make my own PCBs on the mill anyhow, and I don't mind knocking out a few for my own use. Could work out more expensive if you have to buy them.

Cheers
Roger

[websrvr]

I have personally exploded a GT3-127T-12 GRN pulley spinning at 12,000RPM
Um. Checking the Gates Powergrip GT3 Drive Design Manual I find that the 3 HP rating is no problem at all, but they don't spec beyond 10k RPM. Even lower RPM for the bigger tooth spacings. Could be outside the official specs?

Cheers
Roger

When I was told the HP was in spec and it was suggested the RPM was acceptable even though it exceeded it by 20%, stopping abruptly with this pulley was more than it could handle.
it's since been changed from the large pulleys to a multi-stage pulley design where the combined sets of pulleys are significantly smaller yet achieve the same ratio and only marginally noisier.

so I think you meant to reference 26LV32D or 26LV32NS
Too much shorthand. Sorry.
I was thinking of how to make a common low-cost single-ended encoder drive 12 m of wire reliably. The use of a 26LV31E as a driver 'sort of' implies the use of a matching receiver, with the 26LV32 being the obvious choice. I carry both in stock.
The Avago etc 500 line (2000 ppr) encoders are very common, and I have been using them for .. um ... 30 years now? (Starting with their original incarnation as a HEDS unit from HP.) I find them very accurate and reliable. Plenty of similar units suitable for the LOW-BUDGET home hobbiest.

Extra PCB - true. But I make my own PCBs on the mill anyhow, and I don't mind knocking out a few for my own use. Could work out more expensive if you have to buy them.

Cheers
Roger

Yes admirable however, not everyone has the ability to create PCB's for a powered cable extender or a differential encoder conversion board.

I like 10,000 pulses per rev from my encoders so 2500ppr quads are my preference but agree, 2000 pulses is more than acceptable for any application requiring some precision in it's movement.

One of the unmentioned modifications that was discussed by my friend and I tend to agree is I/we see no reason why this driver should be restricted to 24VDC and a 144W 24V BLDC motor due to voltage/current (24VDC/6A) limitations or can't be used with a 60VDC power source to drive a 300W 60V (10A peak) BLDC motor by making a few component changes and adding a buck to drop it to 12VDC for the rest of the driver. but this is unlikely to be done to the public released version so if you want to use some power, follow the thread for reference and modify and produce the design as you need remembering to credit macdowse for his work and efforts.

There is a large selection of SOIC-8 FETS that work under higher voltages with relatively low IRDs making heat a non issue, buck converters and switching PSU chips are inexpensive and require minimal parts to implement such as the LM2596HVS-12 (60VDC to 12VDC) so 60V/10A shouldn't be an issue to achieve in an already fantastic design.

[RCaffin]

I have personally exploded a GT3-127T-12 GRN pulley spinning at 12,000RPM
Um. Checking the Gates Powergrip GT3 Drive Design Manual I find that the 3 HP rating is no problem at all, but they don't spec beyond 10k RPM. Even lower RPM for the bigger tooth spacings. Could be outside the official specs?

Cheers
Roger

[RCaffin]

I like 10,000 pulses per rev from my encoders so 2500ppr quads are my preference but agree, 2000 pulses is more than acceptable for any application requiring some precision in it's movement.

Just for reference: I use a 500 line (2,000 ppr) encoder, a 3:1 GT2 2 mm toothed belt reduction and a 5 mm ball screw. That gives me 0.8 microns resolution. which is more than enough for me. Typically I set the zeros to within 5 microns. I suspect that below that I may be running into some backlash in the ball nuts ( a few microns?) and the linearity of the ball screws would be suspect. Also, the TIR on the spindle is at least 5 microns, so ...

Yes, modern FETs have become very good compared to the early FETs we were using in the 90s. Most of them now have integrated flyback diodes, which were NOT there at the start!

Cheers

[websrvr]

I like 10,000 pulses per rev from my encoders so 2500ppr quads are my preference but agree, 2000 pulses is more than acceptable for any application requiring some precision in it's movement.

Just for reference: I use a 500 line (2,000 ppr) encoder, a 3:1 GT2 2 mm toothed belt reduction and a 5 mm ball screw. That gives me 0.8 microns resolution. which is more than enough for me. Typically I set the zeros to within 5 microns. I suspect that below that I may be running into some backlash in the ball nuts ( a few microns?) and the linearity of the ball screws would be suspect. Also, the TIR on the spindle is at least 5 microns, so ...

Yes, modern FETs have become very good compared to the early FETs we were using in the 90s. Most of them now have integrated flyback diodes, which were NOT there at the start!

Cheers

Precision and tolerance in the hobby market is a byproduct of time, effort and finances but you can achieve more than acceptable tolerance, resolution and repeatability just by component selection without going completely broke so don't discount inexpensive components exclusively by it's cost.

Seems showing this project to a few others from work has inspired them to jump on the band-wagon and queried regarding the existence of for example, wiring the board in a lathe configuration to be controlled by Mach3 and a quick search of the thread yielded nothing obvious so macdowse, do you have some generic application wiring examples or is this too early in the game?

While I no longer use Mach3 or LinuxCNC in new builds, these do offer easy to use GUI type controls and a lot of people still use them so I think general or generic application wiring diagrams should be available so the general DIY hobbyist will be able to easily implrement the drive in his application.

[jfong]

The free version of eagle now is useless for larger projects. 80cm^2 board, two signal layers and two schematic sheets only. Standard for $100 year only does 2 signal layers. You need premium at $500 year if you want more than 2 signal layers.

Once I switched to Circuitmaker, I haven't looked back. Helps that my brother uses Altium professional daily and can ask him questions.

[jfong]

Eagle is now owned by Autodesk and not cadsoft. I'm sure if you asked Autodesk nicely, they may give you a premium license. That still doesn't change the fact that it is a yearly fee for most people requiring larger complicated boards.

I don't fall under personal/educational.

[RCaffin]

The purpose of open source is to aid rapid development by allowing a collective of developers to work on a project and to encourage development
Yes, but that is just your view of open source. It may not be everyone elses'.

At the end of the day it is HIS project to do with as he pleases, but as the files are all available
Precisely.

If macdowswr wishes to run HIS project under slightly different 'rules' from what you want, that is his perogative. He has generously made the sources available in the format he uses. That is 'open source'.

What I want to see is a report on how his completed PCBs work with a real load. If they pass that test - and the SW successfully handles various fault conditions, then it's all pretty amazing stuff.

As an aside: these modern digital MOSFET drives with uP cores give a performance which older CNC machines could only dream of achieving. These drives (by M) are not single chip drives for tin-can steppers: they are way, way beyond that. The idea that they have to be made available at $50 a pop is a bit humerous: they can replace older drive systems costing $10k or more.

My 2c.
Cheers
Roger

[websrvr]

The purpose of open source is to aid rapid development by allowing a collective of developers to work on a project and to encourage development
Yes, but that is just your view of open source. It may not be everyone elses'.

At the end of the day it is HIS project to do with as he pleases, but as the files are all available
Precisely.

If macdowswr wishes to run HIS project under slightly different 'rules' from what you want, that is his perogative. He has generously made the sources available in the format he uses. That is 'open source'.

What I want to see is a report on how his completed PCBs work with a real load. If they pass that test - and the SW successfully handles various fault conditions, then it's all pretty amazing stuff.

As an aside: these modern digital MOSFET drives with uP cores give a performance which older CNC machines could only dream of achieving. These drives (by M) are not single chip drives for tin-can steppers: they are way, way beyond that. The idea that they have to be made available at $50 a pop is a bit humerous: they can replace older drive systems costing $10k or more.

My 2c.
Cheers
Roger

At least a reply worth responding to.

Yes it's his project and he's the only real developer involved in it and mostly for the reasons I stated.

I'm in no rush, I have all the time in the world, if he abandons it I'm sure I can commission someone to finish it for me.

This driver will never replace an expensive servo system without increasing the voltage and current capabilities to handle significantly higher loads and larger motors and to think otherwise is foolishness.

This is a hobby product with some light commercial/industrial small machine applications but I wouldn't expect it to drive a CAT30 size machine when the power requirement is 2KW+ per axis and just slinging the 400lbs dry table around requires more power than an outrunner can provide.

I look at some of these 500W and 1KW outrunner BLDC motors and the peak current exceeds 50A that it's just not possible to see the rated power from the motor when the current is just not available from the drive and the efficiency of an outrunner isn't very high so don't htink input power equals output power cause it's not even close.

The way that some of these outrunner motors are rated is so far from reality that it's hard not to laugh, not one actually came close to the rated HP on a dyno and from what I can only imagine they did to get the power was, they have calculated the HP based on the voltage and current consumption which is no where near reality.

I've never seen a modern servo system costing $10K, our Okuma 8050 drive system costs $2700.00 and $1600.00 of this cost is the motor cost, we bought an extra to have in stock in case of a failure to reduce down time as advised by Okuma but our older 7040 machine the motors alone are $3500.00 and physically twice the size.


During development, cost is important, as a hobbyist you can't afford to go through $100.00 in fets on a weekly basis to work out a current saturation problem or burn up a $100.00 board trying to get the code right and trying to do it with one board is nothing shy of a waste of time cause your gonna blow one at some time during development.

Access to good quality BLDC motors designed for servo applications and not converting an outrunner reduces the risk of thermal overloading and I'm lucky to have a source of servo drives that meet this specific application for basically the same cost so I'm one step ahead of the game when the product is finally completed.


The OP picked his software and manufacturing sources and these alone are the largest contributors to the lack of willing developers to join in, unless he's offering them free boards as an incentive most look at the cost to be involved and find it unreasonable and move on.

The OP seems knowledgeable, I've tried to engage him in a conversation to work on a side project of a single axis SMT32F1 driver by providing the hardware for free and I didn't even get a courtesy reply so I'm not expecting anything miraculous from him.

I'm looking at his code base and reducing and adjusting it as required for a single axis application (STMCube software is useful to a point), PCBA production cost $31.00/pc, 25 board being produced as I write and I'll send a couple off to some interested developers so having a single axis solution in the interim will be available to me as I wait for the 2-axis, 3-axis and 4-axis drivers to be completed by the OP.

[macdowswe]

Hey guys, long time no post. Thought I'd show off the new demo video:

https://www.youtube.com/watch?v=FUh36RUHzdU

[aarggh]

Looks awesome!

Great stuff!

cheers, Ian

[NIC 77]

It's pretty cool what you're doing.

I quickly brought myself up to speed by skimming through the posts and trying to skim more quickly through the fight parts, although I admit I was rooting for one or two people (won't say who).

I was actually flying electric RC planes earlier today.

macdowswe,

First of all, very awesome, I wish I had your skill set sir.

A couple of questions for you (or anyone who knows)

The demo video (post #139), very nice, how long have you been able to continuously run that machine for? What acceleration values are you using? Looking for inches per second squared or meters per second squared or G's, something like that.

I had a look through this document you posted in a previous post:

https://docs.google.com/spreadsheets...H4Y/edit#gid=0

Is the torque in Nm (column H) valid across the entire speed range? If not, is there a torque vs speed graph? Also, do you know what the rotor inertia is for the different motors?

I ask, because I'm thinking of doing the math using some different gear reducers to predict the performance of larger, heavier machines using ball screws or rack and pinion including inertial effects. For example, calculating the maximum linear acceleration at different speeds (In mach 3 for example, you can only enter one acceleration value and you are stuck with that one value up to your top speed), the cutting force available at different speeds using a constant acceleration value, etc.

Also, not quite sure what you are doing for a power supply? Earlier there was some talk of using batteries to help out during peak demand? I can use up a 4 cell lipo in 7 minutes, but it takes 45 minutes to recharge, and that's just for one motor.

[macdowswe]

Hi,
Just wanted to let you guys know that there is an ODrive community getting started over at https://discourse.odriverobotics.com, with some discussion that you might find interesting.

Cheers!

[RCaffin]

Yeah, those tiny outrunner motors look real cute, and probably go great for 5 minutes on an RC plane with a LOT of air flow. Don't think I would try using one on a CNC though.

a CAT30 size machine when the power requirement is 2KW+ per axis
Dunno about the 2kW. My machine runs a BT30 spindle, and the motors are rated at only 300 W (XYZ) and 500 W (spindle). Ah, but the motors are Baldor industrial DC motors or equivalent, and they are rated for 24/7 at full power. And yeah, they are a lot heavier than a tiny outrunner. BIG lumps of machined cast iron I think.
The power rating is a bit deceptive too: I miscalculated when peck drilling some titanium and suddenly noticed I was putting about 2 kW into the 500 W motor for short periods (a few seconds each). Oh well - oops. No damage. Bottom line for me: show me the 6 hour rating. That is meaningful on a CNC.

as a hobbyist you can't afford to go through $100.00 in fets on a weekly basis
Been there, done that, but not as a hobbyist. Ended up with a large jam jar of blown TO220 FETS. That was before they put the intrinsic reverse diodes into the packages. The switching transients were murder.

Let us wait and see. Nothing like the enthusiasm of a start-up.

Cheers
Roger

[websrvr]

Nice of you to whinge about others responses but man you can't see just how rude in someone else's thread you are?

If all the aspects are so bad, and the design is so poor, and the hardware is laughable, and you expect it will fail dismally anyway, why on earth aren't you taking the source and forking it yourself?

And if/when you get the boards done, and they're not so outrageously high in cost, I might consider buying some!

cheers, Ian

I offered the OP an opportunity to produce boards at 1/2 the current cost, he rejected the offer so if you want to complain about the price complain to him, it was his decision to use an expensive solution and beyond my control.

It would help if you could read english, then you would see I praise the concept, the design and the efforts of the OP but unfortunately that where it ends and you can't seem to grasp what I'm talking about and clearly an example of someone looking for their five minutes in the lime-light as you try to make something out of nothing.

Why you insist on fabricating and twisting what I say to mean anything more than what I have stated is a common problem you seem to have repeatedly made and it's a good reason that you should refrain from involving yourself in matters way above your comprehension level, now I'm going back to just ignoring you.

Yeah, those tiny outrunner motors look real cute, and probably go great for 5 minutes on an RC plane with a LOT of air flow. Don't think I would try using one on a CNC though.

a CAT30 size machine when the power requirement is 2KW+ per axis
Dunno about the 2kW. My machine runs a BT30 spindle, and the motors are rated at only 300 W (XYZ) and 500 W (spindle). Ah, but the motors are Baldor industrial DC motors or equivalent, and they are rated for 24/7 at full power. And yeah, they are a lot heavier than a tiny outrunner. BIG lumps of machined cast iron I think.
The power rating is a bit deceptive too: I miscalculated when peck drilling some titanium and suddenly noticed I was putting about 2 kW into the 500 W motor for short periods (a few seconds each). Oh well - oops. No damage. Bottom line for me: show me the 6 hour rating. That is meaningful on a CNC.

as a hobbyist you can't afford to go through $100.00 in fets on a weekly basis
Been there, done that, but not as a hobbyist. Ended up with a large jam jar of blown TO220 FETS. That was before they put the intrinsic reverse diodes into the packages. The switching transients were murder.

Let us wait and see. Nothing like the enthusiasm of a start-up.

Cheers
Roger

Hard-rail machines require massive inertia motors and a CAT30 Okuma 6030 with a 400lbs table or our CAT40 Okuma 7040 with a table weighing in at over 600lbs, 500W wouldn't move it but our new Okuma 8050 with linear rails and medium inertia motprs is over-powered and 500W would probably move it but most likely not at reasonable speeds.

I've set aside a small budget for 75 motors and have 25 coming now as I start to build some frame around them, I'm thinking travels such as 16 x 12 x 16 in R8 / BT20 size tooling would make for a decent size hobby mill, I've got three BT20 ATC spindles already assembled and will be starting to assemble the 2HP output power 12,000RPM asynchronous servo spindle motors to drive them in the next month or two based on an existing design that has been in use for a couple of years now so I'm confident it's more than suitable for this application.

While I've mostly used steel and cast iron in frame construction, I might try my hand at a cast-able material platform in either a true polymer or real epoxy based with multi-sized aggregate and not that crappy (non real) epoxy which is polyester resin based or perhaps something along the lines of epucrete which is a polymer based concrete type material currently used in machine.beds by some machine manufacturers but may be not cost prohibitive as I'm still looking into it.

There's also some other concrete type material being used to cast countertops with which look promising and I'm talking with a local factory that produces some impressive looking countertops who is willing to consider a side line of casting small and medium sized machine frames if the materials holds up to some abusive vibration testing and some environment testing for durability and longevity.

One reason for considering a cast-able material is low production cost, the only real cost is in the mold itself and precision is only in certain surfaces like rail seats and column perches, everything non relevant is purely comsmetic, 15 or 20 gallons of casting material would be less than $250.00 including the cost of the inserts and once set, no post machining is required, all precision is done in key surfaces in the mold and the molds are designed so that the key surfaces can be calibrated so it does have some excellent inexpensive frame potential in quantity productions to swallow the cost of the mold but I am willing to produce the first mold at my own expense in a machine size suitable for most home/hobby applications with a proposed travel of 16 x 12 x 16.

We have access to some equipment at a nearby shop that has a large Nacamura-Tome 3080L composite lathe, I don't care what anyone says, a 24in 6-jaw power chuck is massive by any standard and with a 6.5in thru-hole makes larger material easier to manage, it looks like it's made of black granite and installation was done in a morning by simply pouring 4 puddles of concrete, setting the lathe on the puddles and leveling it before the concrete set and then powered up and ready for use the next morning, of course this particular shop has some impressively large toys and it's always a treat to go over and ogle some of the modern technology used in mold, stamp and die making equipment such as frame-less articulating arm milling machines that can machine in any 3D spatial position of the work area with extreme precision and Ezzel, a local grinding shop that does some insane precision grinding makes the hardest job look easy so I have some perks being in the industry.

[RCaffin]

our new Okuma 8050 with linear rails and medium inertia motors is over-powered and 500W would probably move it but most likely not at reasonable speeds.
I should have mentioned that my machine uses linear bearings everywhere. That does make it easy to move.
In fact, what was almost rather frightening was the time I stripped off the X axis ball nut and tilted the whole machine slightly. The carriage went flying at just a few degrees off horizontal. Serious oops moment there!

Speed - well, more power => more acceleration. Very true. But my machine is a 'hobby' machine: even though I do some production work (for sale), I am not dependant on it for a living. It is fast enough. A nice BT30 spindle, up to 3,500 RPM. It does a nice finish on Al and steel.

cheers
Roger

[websrvr]

our new Okuma 8050 with linear rails and medium inertia motors is over-powered and 500W would probably move it but most likely not at reasonable speeds.
I should have mentioned that my machine uses linear bearings everywhere. That does make it easy to move.
In fact, what was almost rather frightening was the time I stripped off the X axis ball nut and tilted the whole machine slightly. The carriage went flying at just a few degrees off horizontal. Serious oops moment there!

Speed - well, more power => more acceleration. Very true. But my machine is a 'hobby' machine: even though I do some production work (for sale), I am not dependant on it for a living. It is fast enough. A nice BT30 spindle, up to 3,500 RPM. It does a nice finish on Al and steel.

cheers
Roger

I've been manipulated into accepting the following concept, there is little difference from a machine for hobby use and a machine for commercial/production use, I should expect the same performance, repeatability, precision and power based on the platform size.

I would never build a wooden router or power one with a dremel or porter-cable wood router or a cheap water cooled china high speed ER20 spindle, if the spindle is not designed to be air cooled and used continuously and properly power rated such as actual output of 24,000RPM 3.0KW then the spindle is considered garbage just like those cheap china 24,000RPM water cooled spindle rated at 3.0KW, but are 220V/10A and really only outputs 1.43KW and for mills it better be an S1 rated 2.2KW (3HP) output (220V/12.75A) 100-10,000RPM asynchronous servo spindle motor driving a BT20 or R8 spindle or an S1 rated 3.0KW (4HP) output (220V/17.25A) 100-10,000RPM asynchronous servo spindle motor driving a BT30 spindle and these are considered my bare minimum power requirement if I expect to do some real work.

Yes I can make a machine that moves along with 100W DC servo motors but if I can't cut stainless at reasonable feeds because it stalls an axis then I've made a wrong choice in axis motion power and need to make a new choice immediately.


The cost of 220V AC servos with drivers is pretty cheap now, 750W motor and drivers can be had for about $300.00 new and 16in travel, 1/2in DIA 0.100in pitch ballscrews are about $150.00 and provide 300IPM rapids and 100IPM cut speeds so a small mill with sufficient power to cost steel and aluminum shouldn't be expensive but a 500W BLDC servo and driver should be doable for half the cost and still be within the axis motion power requirement.

If a 1in DIA 0.200in pitch ballscrew is employed then I would expect nothing less than 750W using a 2:1 reducer and still see 300IPM rapids and 150IPM cutting speeds without stalling an axis and yes, 300IPM rapids on 16in of axis travel looks fast, using 1.5KW motors and no gear reducer gives you 600IPM of rapids and 300IPM of cut speed without stalling and becomes mandatory when large mass comes into play like moving and stopping 400lbs tables.

[joeybagadonuts]

Macdowswe,

You are doing a most awesome job regardless of the rumors the guy from the sunshine state is spreading.
Got to get back to the bakery now.
JoeyB