[joeavaerage]

Hi Ardenum,

I wanted to open source everything so that anyone can make them,

A very laudable goal, but lets say there was a thousand hours design work just to get a workable design.....and I suspect that would be realistic, not a hugely polished commercial design,
but a workable drive of good reliability, that's a big investment for an individual to make for an open source project. Were such a project distributed amongst ten committed hobbyists then
the project becomes much more attainable, but with one or two contributors that becomes quite a hurdle.

Another way might be, if we were able to access an existing open source design which could be adapted to a linear drive with something like 100 hours, that too would be achievable. Do you know
of any such open source designs that might be adapted?

Craig

[ardenum2]

Another way might be, if we were able to access an existing open source design which could be adapted to a linear drive with something like 100 hours, that too would be achievable. Do you know
of any such open source designs that might be adapted?

there was argon that used SimpleMotion which is supposedly open source but the board itself doesn't appear to be open source.

https://granitedevices.com/wiki/SimpleMotion_V2
https://granitedevices.com/digital-servo-drive-argon/
About Argon servo drive project - Servo drive developerServo drive developer

Argon was $400 and it's been discontinued. They have another board called ioni for $200 but that one is low voltage. Both argon and ioni support linear motors.

I'll have a look around for others.

[joeavaerage]

Hi Ardenum,

the attached pic is a MOSFET switching module of my own design. What may not be apparent is that the PCB is very special, in particular it has VERY thick copper layers top and bottom.

Most PCB is called '1oz'. That is to say the copper cladding is 1oz per square yard which works out to be 35um thick. The board I have used for this module is 12oz board, and the copper layers
are 420um or 0.42mm thick! This board is very rare, usually only made to special order. I was able to get two sheets from a specialty supplier in the US......but it seems it is now out of stock.

The reason I used it is because I expect this MOSFET switch to operate at 48A for periods of seconds at a time. Thin copper would just vaporise. Most commercially made PCBs are etched, maybe 1oz or 2 oz
copper, and then electroplated up to say 4oz thickness. That's fine if you use a regular production process like PCBWay, but I use my mill to 'isolation etch' the boards. So for my purpose I want thick copper layers
without electroplating....ergo the 12oz board I have pictured.

Just a few notes to clarify what is going on.....you see the top side of the board only, that is the signal layer. The high current traces are on the underside. The isolation IC allows the micro controller
which has 3.3V outputs to drive the MOSFETs, and the MOSFETS will be anything up to 600VDC above earth. As such I use six of these modules for the three phase inverter, two modules for the
braking resistor switches and another two for the PFC boost circuit. I should also point out the the snubbers for each switch are NOT on this board, but are intended to be on the 'main board' above which these
switch modules sit. The switches are all identical but the snubbers will vary depending on what part of the circuit it is.

There is no way you could use this as an open circuit design, because few if any one could get the circuit board material I have used. I still have about 1 1/2 sheets left which will do me for years to come,
but could not hope to make them for anyone else. That was not a shortcoming when I designed it, I was only ever going to make one or two and for myself, so there was no need to accommodate the
need for it to be manufactured by others.

Craig

[ardenum2]

Another way might be, if we were able to access an existing open source design which could be adapted to a linear drive with something like 100 hours, that too would be achievable. Do you know
of any such open source designs that might be adapted?

found two more:

https://github.com/rene-dev/stmbl
https://github.com/odriverobotics/ODrive

[ardenum2]

found two more:

https://github.com/rene-dev/stmbl
https://github.com/odriverobotics/ODrive

apparently texas instruments also has a drive reference design and its with gerber file too. I don't know how it stacks to the one you posted earlier but the price is 10x of that one.

https://www.ti.com/tool/TIDM-SERVODRIVE

[joeavaerage]

Hi Ardenum,
the first two links (GitHub) are just software, which is in itself quite handy, but is not the complete drive. That just allows you program the micro-controller, there are still all the power
supplies, rectifier/inverter, current sense etc to design and build.

The TI link is to a board that has the whole shooting match, the micro-controller and the hardware to drive a real motor. It is based on a C2000 TI micro-controller,
another in the same family as the one I use, but its on a board that accepts 110VAC-220VAC and can drive motor of up to 1 hp, say 750W. Very useful, although
still well short of the 600VDC you suggested was required.

Next issue is that its cost. This is the listing from DigiKey, and is in New Zealand dollars, at $2169NZD (plus GST) or $2494NZD including local tax!!!! That's not cheap!!!

Craig

PS: I investigated the two GitHub links more closely and they do have hardware files....so I take back the comment I made about being software only.....there is a design for
the hardware as well. Interesting.

[peteeng]

Hi Ard - around 3 years ago I tried to buy Durcrete from germany but even though it says sold around the world they said they would not supply outside of CE. But yesterday I sent a note to Bernhard and he responded today.

Hello Peter,
you can calculate linear elastic, isotropic so it is quite easy.
Attached you find the material data, I use in FEM.

Please look into webinar 3, if you want to learn more about FEM analyses. There are English subtitles.

Design and analyses of solid machine beds is easy! Learn the principles for robust, production-ready design in our practical 8-part webinar series on YouTube. The videos inlude factory insights, standard embedded parts and casting of the cement-bonded mineral casting. In about 90 minutes, you will learn the essential technical aspects about components with Nanodur concrete out of a mould and sandwich constructions with the casting compound durfill.

Part #1: Design of solid machine elements made of UHPC
German with English subtitles, 24 minutes
https://youtu.be/au3eD3ZmAvI

Part #2: Analyses of lifting anchors in massive machine beds made of UHPC
German with English subtitles, 4 minutes
https://youtu.be/zwqO-NKMyf4

Part #3: FE-Analyses of massive machine elements made of UHPC
German with English subtitles, 13 minutes
https://youtu.be/Py9HYSlPGd0

Part #4: Dynamic analyses and measurements of solid machine elements made of UHPC
German with English subtitles, 10 minutes
https://youtu.be/YSAo3OLCnYg

Part #5: Embbeded parts for and production of solid machine elements made of UHPC
German, including factory videos, 8 minutes
https://youtu.be/sTqxwTpBjLM

Part #6: Production of self compacting Nanodur Concrete
German, including mixing and testing videos, 6 minutes
https://youtu.be/gKjzq4QpeBk

Part #10: Temperature
German with english subtitles, 21 minutes
https://youtu.be/_XGF2FZUTms

[joeavaerage]

Hi,
I've just got confirmation that my purchase of a renewal for my Fusion 360 Machining Extensions has completed. I paid for the Fusion 360 (Basic) subscription about
six weeks ago and that cost $778NZD (including local tax), but this is for the Extensions that allow simultaneous four and five axis, collision avoidance, toolpath editing etc.
It cost me $2034NZD (including local tax).

I should point out, and to Autodesk's credit, they knocked 20% off the price of the Machining Extensions otherwise it would have been even more.

Between both bits of software I've spent $2813NZD in the last six weeks. I think that is enough on my hobby for the moment!!!

Craig

[peteeng]

Hi all you electronics people out there. We have been discussing elastic modulus of materials for years and its usually determined via mechanical test which is complicated. You can do it by acoustic methods. You probably have an oscilliscope, wave generator microphone out there so get to it and make a rig!! We want to know the answers. This then could give us data on made machines as its non destructive. Get to it guys and gals... Its not my area....Peter


Materials | Free Full-Text | Measurement of Dynamic Elastic Modulus and Poisson’s Ratio of Chemically Strengthened Glass (mdpi.com)

Speed of sound - Wikipedia

Speed of Sound in Solids Calculator - Sound Wave Speed in Different Materials (omnicalculator.com)

[peteeng]

Hi Craig - The cheapest 5 axis software I could find a year ago was Fusion. The next step was Mecsoft at $11000USD plus $900USD /year Maintenace. Others then were much higher. I think Fusion is really good value. Just over a year ago my main computer died and I had to reinstall everything. I had to get passwords and various keys sorted. Mecsoft said sorry you are two years out of Maintenace so we want $150USD for the keys (plus a letter from my computer tech describing that the computer died!) and half price to get my subscription up to date another $600USD. So I said thanks but I'll try Fusion for a year at $700AUD. Been great so far... Peter

[joeavaerage]

Hi Peter,

The cheapest 5 axis software I could find a year ago was Fusion. The next step was Mecsoft at $11000USD plus $900USD /year Maintenace. Others then were much higher. I think Fusion is really good value

I agree. I already had Fusion Basic because of the electronics module (based on EAGLE which I've been using for over a decade) for work. So I already have Fusion Basic and have a definite use case for it.
When I started looking around for five axis....and lets be clear here I'm talking simultaneous five axis, not 3+2 or indexing 5 axis, the genuine deal....and you are 100% correct all the commercial offerings
are in the region of $10,000USD and upwards.

I really don't much like subscription software, neither do I particularly like the cloud based model either, but I just cannot refute the value for money that Fusion represents. If I exclude the cost of Fusion Basic,
the cost of the Machining Extensions has cost me $2500USD for two years....and if Autodesk keep offering discounts like that that works out to $1250USD/year.......or eight years worth of subscriptions
just to match Mecsoft for instance, and that is WITHOUT the maintenance of $900USD/year.

What it really means is I have to 'pull finger' and get some use out of it, and generate a profitable use case for Machining Extensions, which in turn means I really need to finish my five axis trunnion!!!

I've just had a customer through the door whom requires an unusual drive key for a Dennis Fire Engine that will require a simultaneous four axis tool path. So I am generating income from my fourth axis,
not enough to be considered profitable, but highly useful capability that I am able to exploit.

Craig

[Mr-MaW]

[QUOTE=peteeng;2559484]Hi Ard - around 3 years ago I tried to buy Durcrete from germany but even though it says sold around the world they said they would not supply outside of CE. But yesterday I sent a note to Bernhard and he responded today.

Hello Peter,
you can calculate linear elastic, isotropic so it is quite easy.
Attached you find the material data, I use in FEM.

Please look into webinar 3, if you want to learn more about FEM analyses. There are English subtitles.

Design and analyses of solid machine beds is easy! Learn the principles for robust, production-ready design in our practical 8-part webinar series on YouTube. The videos inlude factory insights, standard embedded parts and casting of the cement-bonded mineral casting. In about 90 minutes, you will learn the essential technical aspects about components with Nanodur concrete out of a mould and sandwich constructions with the casting compound durfill.

Sorry, I can't use it in China (Youtube).But if you really want to make a product, I can help you analyze it.

[peteeng]

Hello Mr-Maw - Thank you for the offer. I have been doing FEA & designing machines for around 35 years so I'm quite comfortable doing this sort of thing. The basic question that the forum is interested in is; what is the modulus of mineral castings & UHPC as there is a very big range of materials and quoted values out there. Once you know the modulus the rest is very straightforward. If you have a recipe to achieve E=70-80GPa like the Durcrete people have that would be very useful to many of us. Peter

[ardenum2]

The TI link is to a board that has the whole shooting match, the micro-controller and the hardware to drive a real motor. It is based on a C2000 TI micro-controller, another in the same family as the one I use, but its on a board that accepts 110VAC-220VAC and can drive motor of up to 1 hp, say 750W. Very useful, although still well short of the 600VDC you suggested was required.

You're right, we'd need a ~5kW drive for a max. 2000N forcer. but isn't it just a matter of changing the components to higher rated ones and changing the thickness of the traces? when I was in high school I was tasked with modifying a board, adding new components and changing trace lines in kicad, I knew my way around photoshop but had zero idea of electronics, it was more like placing the parts and drawing traces than anything but I do remember my changes changed the power specs while leaving the logic mostly intact(there was some added function to it too) I bet I'm way oversimplifying it...

[joeavaerage]

Hi,
TI are weird....they seem to sell some of their 'development boards' really cheaply and lay on all the software for free. For instance the development board I use (LaunchPad XL),
costs about $60USD.....and that is quite a price increase from when I first started using them.......$47NZD or just over $30USD. I bought another one (I have two now) about six months
ago and it cost $52NZD (plus 15% local tax) or $40USD. Just the micro controller IC alone is worth over $20USD. So, this board they are almost giving away, and that certainly encourages people
like me to use them...and I do!!

But then they have this TIDM-SERVODRIVE board for $2400NZD (including local tax)!!! Its priced right out of my range....and I have to question what they think they are doing?
In the case of the LaunchPad XL they are pricing it to give away and encourage customer uptake, I understand and applaud it, and yet they price this board exorbitantly to the extent
that no one will use it??? Can you think of a reason they would do something so contrary?

Anyway, given the cost of the TI board there is no point in trying to modify it. If you wanted to fit a bigger rectifier for instance, well you just could not, there is not the real-estate on the board to
fit a bigger component.

The other two designs however may well be amenable to the upgrades you are talking about.

I suspect that your linear servo drive will require a much higher voltage and much greater peak currents to match the forcers you have in mind. Thus if you take the circuit diagram for one
or other of these designs and uprated all the power components....you'd be approaching a solution. It would require a larger PCB, you cant just put bigger and beefier components on a board and NOT expect it to get bigger.
Likewise if you have available only standard copper thickness boards, or can only reasonably manufacture standard thickness boards then the traces must be much wider which again increases the board size.
So if you want to re-design the board, I mean all the current board is a Gerber file right?, its not like you are throwing something away, that is plausible.

As I've posted before I'm building my own servo drive for a servo motor I acquired many years ago off Ebay. The servo drive needs to be 3kW continuous and 12kW peak. The DCLink voltage
is to be 400VDC using a PFCboost circuit from 240VAC. So even my design is probably not suitable for you if you indeed require 600VDC for the forcers. There are higher voltage MOSFETs and so
on but just substituting a higher voltage part does not mean that all the performance parameters of the device increase. Thus you would find that you can increase the voltage but only by reducing the current.
Your idea of just substituting higher rated parts is somewhat simplistic.......not unreasonable, but does require some careful study.

My design is big by comparison to this that you have linked to. It has a high current (thick copper) mother board on which are the rectifier and power supply components, the bulk of the three phase inverter
traces (including the snubbers and Hall flux-gates) and various glue logic , although this latter part is a very small proportion of the total. Each of the high voltage/high current MOSFET switches as I have pictured previously
are mounted above the main board as sub-modules. The LaunchPad XL just plugs into the mother board.

My reasoning for breaking the design down this way was that if any one part of the overall device failed I could replace just that part rather than throw the whole thing out and start again. If I want to fiddle with
the micro-controller I just unplug the LaunchPad XL and can take it away to plug into my PC to reprogram it say. Likewise if one of the switches blows up then I just plug in a new module. The downside is that the
mother board was 250mm x 250mm.

I started on this design nearly eight years ago, and I advanced it to a bread-boarded but working state. Then I sold my house...and then.....and a different job....and then... All in all this project has been left alone
for nearly six years. Every once in awhile I pull it out and do something with it, but it has not progressed much for ages. In fact one part that I did have working really well was the PFC boost circuit, and that was
on its own high current (thick copper) board....and I ended up selling it to a guy who needed a PFC circuit to run metal halide lamps....for what purpose you ask??? So you might say that the project has gone
backwards, the one part that was really polished, and I was quite proud of to be honest, I've sold.

I have a much bigger machine now than way back then and would really benefit from this 3kW servo driven spindle. As I have nearly completed my trunnion fifth axis and have largely designed
a cast iron headstock so the way is clear for this project to advance over the next 18 moths or so.

Just as a matter interest I followed up on the US supplier of PCB material:

https://www.pcbmaterialworld.com/inventory.php

So while the 12oz board is now out of stock there is 31 sheets of 11oz board. Note that this thick copper board is not really suitable for common phot-etching processes, the etching undercuts
with this thick copper, it really needs to be isolation routed as I do with my mill. It would not be impossible for others to follow and replicate my design even if I do use thick copper board.

Craig

[peteeng]

Hi Ard et al - I have been corresponding with Durcrete and the inventor of the E80 material over the last couple of days. The E80 is accurate for FE they have done extensive flexure, compressive and acoustic testing to determine these numbers. They use E80 for FE work although the acoustic testing shows its a bit higher than this. The Rampf material I'd say the 40GPa is accurate. The sika3350 is tested to the same EN spec as the E80 material so lets assume its good data. I have dug into the E80 recipe and its 50/50 (by volume) nanodur 5941 / alox so since I have the same 1-3mm alox here as durigid uses I'll make a 50/50 sika/alox mix for test. This will be my final attempt!! The supplier of E80 will ship container loads of it anywhere in the world but not small qtys. The small sales shop won't send it out of the CE so are you in the CE?

The sika3350 has similiar mechanical properties to nanodur so maybe the 50/50 mix will score well. Rule of mixtures, back calculating the required modulus puts the alox at E110Gpa. Since it is E300GPa seems the efficiency is OK... if it bonds well to the grout.... We shall see.... Peter

My concern is that the 65% by volume alox with epoxy only scored 17GPa I thought it would do much better....

[pippin88]

Alox cost a fair bit more than Sika3350?
In non-space-constrained parts, better to just go bigger with the Sika3350

Interested in how you go

[peteeng]

Hi Pippin - I'll have to look it up. I have 20kg of alox to use up! in something. Since the sika and E80 material use the same test and they are confident in the numbers I'll make a lankor coupon and a sika coupon to see how that goes. The lankor is half the price of sika and its stocked by Bunnings. The alox lifts the Nanodur from 50 to 80 so the alox should lift the lankor from 40 to 70. E70 I'll be very happy with...

I'm always geometry limited doing small machines! I feel the laminated aluminium is my best bet at the moment. But these tests will tell the tale.... Peter

[ardenum2]

So if you started production, E80 is possible to get? I bet minimum order is 25-50 tons or similar but hopefully the price is also wholesale which would make it cheaper per machine than getting a bag or two here and there for each machine separately. Now what machine to built?

laminated al would also make sense, you have a router that can machine al and you are fluent in resins. I think you're down to these two in the end. laminated al does allow you more freedom as to what kind of a machine to build. When maximizing column travel, the juncture between the front rail surface and bottom carriage surface is impossible to make with eg/uhpc. To make it work you have to move the rail surface up by at least 120mm and that is a major problem.

[ardenum2]

Anyway, given the cost of the TI board there is no point in trying to modify it. If you wanted to fit a bigger rectifier for instance, well you just could not, there is not the real-estate on the board to
fit a bigger component.

the question would be if the sum of its components really cost that much? that board has $999 price list over here...something doesn't add up

extending the board is very simple, in kicad for example you press and hold one of the edges and move it as if you're cropping a regular image and I hardly care about the size of a drive. I'm thinking a 150x300mm or 200x400mm would be a good size, same depth of a regular servodrive but double the length(in Z). I also very like your approach with modularity in mind, keeping components in designated 'modular like' places. Keeping things organized like that in exchange for slightly longer and thicker traces sounds like a good trade off.

ironcores are usually 500-600vdc and between 4-8amps. max. 5kW, your drive 400vdc and 3kW? almost there...