RAMPS 1.4.2 fork: RD3D v1.0 (6 steppers, 24v, Due)

Welcome to the real world of 3D printer controller design!

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Quote
dc42
Welcome to the real world of 3D printer controller design!

haha... mwahahahah. very funny, dave

Now you spent a few nights designing a 2 layer board, just to find out soon a 4 layer board would've only cost a few bucks more
I hope NOT, please tell me you asked about the prices before?

Quote
o_lampe
Now you spent a few nights designing a 2 layer board, just to find out soon a 4 layer board would've only cost a few bucks more
I hope NOT, please tell me you asked about the prices before?

ha no, surprisingly, 4-layer boards are quite a jump, but also they take longer to get prototypes done. a 2-layer board (if you're prepared tto pay a bit more) can actually be done in 24-48 hours at reasonable cost. even eurocircuits.com have really good pricing for 2-layer PCBs (with certain limits like minimum 12mil copper and clearance, and only 24mil drill holes) which is very easy to get hold of low-cost manufacturing equipment to do that.

no the other main reason for doing a 2-layer board is because then people can etch them out themselves. making your own 4-layer sandwich can be done but i wouldn't want to try it.

dang i just can't seem to leave this alone. dc42 i spotted you have that mini z-probe thing on your .sig, so i went, "hmmm..." and of course ended up adding a fourth ADC this time with a VCC pin, so people could create a header which supplies power to the IR z-probe. i put UART back because there's room now. added a 2x6 header because... there's room for that. put in a power-selector header so it's possible to swap between 5V and 3.3v...

when does this end!!

"when does this end!! " About 60 days after you have been declared legally dead. Then someone else will take over


If your after ideas... (Evil cackle)

You should convert the aux4 and SPI to the standard 2 * 5x2 pin headers for lcd, eliminating the lcd daughter board. Just use pins no shell.. as some lcds need cables flipped 180.

Edited 1 time(s). Last edit at 09/21/2017 07:57AM by Dust.

Quote
Dust
"when does this end!! " About 60 days after you have been declared legally dead. Then someone else will take over

welcome to the reprap community. we'll be burying you tomorrow

Quote


If your after ideas... (Evil cackle)

You should convert the aux4 and SPI to the standard 2 * 5x2 pin headers for lcd, eliminating the lcd daughter board. Just use pins no shell.. as some lcds need cables flipped 180.

i was after _not_ doing any more! then i looked at the duet 0.8.5 and noticed some 10k resistors in series with the ADCs...

i looked at the RAMPS_FD, he copies the exact same pins into 2 places for SPI signals? worra? nehhh i think i'll leave it as what's on RAMPS. apart from anything i'd have to increase the PCB size to fit the larger connector. and the 1x18 fits along that edge: eagle can't handle more than 4in PCBs, and it's already... 4.01in....

l.

You're an animal... once you've stabilized the design some, and you meet your cost goals, let me know how to buy one I like low cost experimental ideas.

Quote
obelisk79
You're an animal... once you've stabilized the design some, and you meet your cost goals, let me know how to buy one I like low cost experimental ideas.

haha sure. you're welcome to one of the first batch of 10 if they work. the speed is because i'm a bit under pressure as i'm supposed to be doing other things.. sort-of, but also there's Shenzhen Maker Faire coming up on the 10th-12th November, i need a working board for then. PCBs arrived at my friend's factory, he's just getting the components, he asked "what colour LED do you need" it's like, "aiyaaa i don't care!" but i just have to calmly reply instead, "green please, mike"...

Quote

when does this end!!

I could engrave your tombstone with that question ,if you like
But seriously, professional project managers start with a tasklist and then stick to it. That's why describing the tasklist takes almost as long as the project itself.
Some people try to get along with the make or brake method, but that's not very efficient oftenly...

There's a saying: Why don't we have time to do it properly, but always have time to do it again?

Quote
o_lampe

Quote

when does this end!!

I could engrave your tombstone with that question ,if you like

only if you can also get metallica to turn it into a heavy-metal rock ballad, "MY Coffin Liiiiid *thrashing guitar riff* is a PEE CEE BEEEE *thrashing guitar riff*", which they have to play every year in the graveyard

Quote

But seriously, professional project managers start with a tasklist and then stick to it. That's why describing the tasklist takes almost as long as the project itself.
Some people try to get along with the make or brake method, but that's not very efficient oftenly...

There's a saying: Why don't we have time to do it properly, but always have time to do it again?

i appear to be taking a different approach: minimum amount of changes to RAMPS 1.4.2 in a bit of a rush, then pretend i have time until that revision arrives to make some mods. the cut-off point will be when those bits arrive and i have to start assembling the prototypes (middle of next week most like). that's the theory anyway. the actual deadline's SZMF.

ok by removing the PCB holes (they were really getting in the way) i managed to jam in another SN74HC125 for level-shifting the SPI signals to the MicroSD card. this will be faster than resistors (and more reliable) but it's still SPI... and that's okay.

also in the left part of the picture, an ADC z-probe thing. no 4.7k resistor on that. dc42 i really like the mini IR z-probe thing, might want to talk to you about that, getting some made in shenzhen and sending you some $. really don't want to duplicate the damn circuits or firmware: if i did have to i'd do them as GPLv3+... and i'd *still* send you some $.

oooo this is so frustrating! i did what i considered to be a perfect 2nd layout, ready to go for some factory quotes... and started doing a RepRapFirmware port..... only to realise that the SPI port on RADDS is an extra connector that's NOT pulled up from the Arduino 2560 PCB / RAMPS combination (because they use the *other* Alt-pins SPI interface on RAMPS...). and that would mean doing some modifications to the support libraries that i wasn't comfortable with (as well as putting load on the Due due to bit-banging SPI. yuk).

so i added in the 6-pin SPI connect to see what would be needed... it came up right in the middle, overlapping where i had neatly laid out the stepper modules, and of course there's no room to move things around: the steppers were incredibly tight as it was...

... so it's a near-total redesign job. FRICK.

so, the three lower steppers need to go *below* the line of arduino pins, the 8V regulator goes above (in the middle of the PCB ), thermistors likewise swap positions to go in the middle. this all requires an expansion of the board depth to 3.5in but is doable and actually allows much clearer space in he centre of the board.

soooo that meeeeeans... i might be able to get SPI pins to all the steppers so i can support SPI mode on the TMC2130 silent steppers, yay! so that's definitely worth trying.



Edited 1 time(s). Last edit at 09/25/2017 03:10AM by lkcl.

Enjoying watching your progress! Unfortunately as a technician for military electronic systems, PCB design is well outside of my comfort zone but clean and effective design is something I can certainly appreciate.

Quote
obelisk79
Enjoying watching your progress! Unfortunately as a technician for military electronic systems, PCB design is well outside of my comfort zone but clean and effective design is something I can certainly appreciate.

thanks obelisk79. making a PCB design "look beautiful" is apparently as good a way as any to get reduced EMI eagle however is so... lacking in professional CAD features, shall we say, that it is near-impossible to use effectively. i'm doing my best....

ok i need a bit of help reviewing this, the idea is:

* support standard driver stepper modules (3 jumpers)
* also support connecting SPI to the correct pins (including a dedicated CS - chip select - line) so that TMC2130 and other SPI-based modules can be used.

it's slightly complicated by the fact that "normal" stepper operation (A4988, TMC2100) have a pull-down resistor on MS1.

so.

to support that, i've added a *five* pin header, which despite how it looks in the schematic, goes in the MIDDLE between the two 4-pin ones.

for standard steppers:

* pin 1 of the 5-pin MUST be connected to the same header that has all the VCC-3v3 on them.
* if you want MS1, MS2, or MS3 HIGH you connect them in the standard way.
* if you want MS2 or MS3 LOW you leave them UNCONNECTED in the standard way
* if you want MS1 low you MUST connect pins 4 and 5 of the 5-pin header

for the TMC2130:

* bridge all 4 pins 1 through 4 of the 5-pin header to the SPI header.

does this look reasonable, according to the screenshots below? also to confirm: SDI is MOSI - master out *S*lave *I*N (MOSI) and SDO is MISO - master in *S*lave *O*ut ?

thx people

All that SPI-hassle for an underrated TMC2130? We have tested the TMC2100 and it seemed most people weren't satified with it's performance.
The 2130 is the same core, but with SPI?
I'd keep the design simple and compatible for Pololu drivers and maybe later add another sub-PCB that fits bigger SPI drivers like the ones on the newer Duet boards.

I have TMC2208's I think SPI would be a nice feature. But standard jumpers would still be functional. Performance information regarding the 2208 is a bit sparse but what I've seen so far looks rather positive.

Ikcl, your jumper configuration sounds slightly confusing.

Quote
obelisk79
I have TMC2208's I think SPI would be a nice feature. But standard jumpers would still be functional. Performance information regarding the 2208 is a bit sparse but what I've seen so far looks rather positive.

Ikcl, your jumper configuration sounds slightly confusing.

yehyeh, it's down to supporting two possible configurations. take a look at this: on the left is standard polulu, A4988, etc. etc. on the right is support for *both* A4988 *and* TMC2300. you have to connect *four* wires... whilst also-supporting-the-connect-the-pulldown-resistor-and-the-connection-of-sleep-to-reset-which-is-required-for-A4988 configuration:



so you have to BREAK ms1's pull-down resistor so that the MS1 pin can be used for SPI - but allow it to be re-connected to the pull-down resistor AND allow it to be set "HIGH" (marked in red)

AND you have to BREAK the connecttion between reset (CFG0 on the TMC2300) and sleep, so that again reset (CFG0) can be used for SPI on TMC2300.... but allow it to be re-connected to the reset pin when using A4988s etc.

all good fun

Quote
o_lampe
All that SPI-hassle for an underrated TMC2130? We have tested the TMC2100 and it seemed most people weren't satified with it's performance.
The 2130 is the same core, but with SPI?
I'd keep the design simple and compatible for Pololu drivers and maybe later add another sub-PCB that fits bigger SPI drivers like the ones on the newer Duet boards.

ok let's work through it, because what you say makes a lot of sense.. *and* there are definitely things i want to experiment with / support.

it's not just for the TMC2130, but for other drivers that do SPI (i'm thinking of making one with the TMC2660 and SEVERELY curtailing its maximum current).

more than that, with the SPI interface (i investigated TRAMS) you can set stallguard, change the pattern of the output to the steppers (yes, really!), change the micro-stepping and a hell of a lot more. they're really really sophisticated steppers, i don't think people quite realise how much they can do. [oh - they can also handle X-end and Y-end stops... *on the stepper*. so they will stop immediately rather than under GPIO control... and you can then get the status.... but the polulu design doesn't cater for that... oh well...)

anyway, a couple of things: Trinamic's offerings seem to be less power-efficient (you definitely need to use forced-air cooling). but, also: the peak current (2.5A) is almost certainly way beyond what the tracks on a standard RAMPS or other "modular" board's power tracks are designed to cope with. you *might* get better results with the RAMPS 1.4.2 because they used 2oz copper.

getting SPI to all six modules is a bitch. been trying for about... 12 hours now. what you say about doing an expansion board makes sense... i still want to at least try TMC2130s, see for myself what they're like. you know how it is

so. what do you *actually* need TMC2100s / TMC2130s for? X, Y and maybe extruder. that's where you want fast changes of direction (peak current usage) and you want a lot of oomph on the extruder. you certainly don't need them for Z.

so if i skip the 3 Zs entirely, that makes life a lot easier. you just don't need huge power or speed to drive Z axes.

so... 3 and 3... that would make sense. then if people want to use 3 SPI-based steppers for a Delta, that would work really well. i'd like to make it 4 SPI-based steppers... [day later, forgot to hit send on this: yes i managed 4 SPI-capable steppers]

okaaay so i did it. whew. managed to complete the routing, without it being too much of a dog's dinner. power is very clean... including to most of the steppers. GND to E0 (the lower middle) stepper, is... well, it's connected to the top left pin (near VCC) but the way that GND *gets* to that pin is very fragmented, and there's not a lot i can do about it. GND comes in from all directions, probably totals about... 1 amp. the other GND pin is fine: that has about... 3 maybe 4A worth of copper surrounding and leading to it (the pin is rated for 3A).

i'm being quite paranoid about power, here. i really don't like 2-layer PCBs that are only 4in x 3.25in and could be used to deliver 500W of power @ 24v. makes me nervous.

anyway on the main power +ve track (12/24v) which is highlighted in the picture below i've marked out approximate current carrying capacity if 1oz copper is used. it's mostly a 12A capacity, branching down to 10A for the first stepper, 8A by the time it reaches the 2nd, and down to 5A for the third. there are two such branches, covering each of the rows of steppers. for the MOSFETs that's approximately a 5A capacity to each, but for the top 2 (and the low-power MOSFET) it has to go through an appx 7A bottleneck. i MIGHT be able to mitigate that by putting a 3A track on BOTTOM layer, connected with some VIAs, stretching from the big area on TOP with plenty of capacity, over to where the fork occurs (to D10 and D11's MOSFETs)... have to think about it though as it would be at the expense of restricting GND capacity.



this one just shows how much of a dog's dinner mess the SPI routing is. the SPI connector is in between those two stepper boards, and had to go up and over to get to the right-hand one, then down the side to get to AUX2 (this is 3 tracks...), then carry on down to get to MicroSD and AUX3 in the corner, then go LEFT from there to get to the other 2 stepper boards.

What can I say other than "good booooy"

Quote
o_lampe
What can I say other than "good booooy"

i feel like i'm some sort of horse that is galloping madly and wants to throw its rider....

Quote
lkcl

Quote
o_lampe
What can I say other than "good booooy"

i feel like i'm some sort of horse that is galloping madly and wants to throw its rider....

You may have mentioned the "dogs dinner mess" too often, so I felt the urge to tap your head a bit

bwahaha!

Quote
obelisk79
bwahaha!

somehow i get the feeling we're not taking this whole thing seriously. i might be mistaken... but isn't this supposed to be... y'know... grumpy people online, shouting at each other not getting anything done, or... something!

anyway. components are on their way for the first revision... hopefully they will have gotten out of China right before the national holiday which starts on oct 1st. entire country shuts down...

Any updates on your initial prototypes?

Quote
obelisk79
Any updates on your initial prototypes?

bits made it into an SF-Express Depot... on saturday. 31st september. day before the holiday. the sf-express.com courier tracking service saaaays... it's been transferred to the shenzhen distribution centre. where there will be presumably people still working, during this big chinese holiday, just veery slowly. once it makes it out of china either to HK or direct to TW things should go a bit quicker.

2017-10-02 Monday
11:22:23	Shipment loaded at 【Shenzhen Distribution Center】, send to 【Hong Kong Tsing Yi Distribution Center】
10:17:37	Shipment arrive at 【Shenzhen Distribution Center】
00:35:35	Shipment loaded at 【Shenzhen Wuhe Distribution Center】, send to 【Shenzhen Distribution Center】

yay, out of SZ. whewwww

edit....

2017-10-03 Tuesday
01:24:33	Shipment loaded at 【Hong Kong Tsing Yi Distribution Center】, send to 【Taiwan Taoyuan Distribution Centre】
01:20:13	Shipment arrive at 【Hong Kong Tsing Yi Distribution Center】

on the way to taiwan...

Edited 2 time(s). Last edit at 10/02/2017 06:49PM by lkcl.

btw in case anyone's interested in what i've been doing for the past.. 5? 4? days, [www.stm32duino.com]

i kind take objection to chinese clones of the ATMega being only $5 but the Due is 3 times that @ $15 to $16. i mean.. what the hell?? what is wrong with Atmel - and the component selection on the Due - that it can't be had for $7? so with a bit of help from the stm32duino community i create the stl47o - pronounced S Tee El Four-Seven-Oh... don't all laugh at once...

Ok, I am hesitant to post this but WTH...

I have a thought here and its been in my head for a while... seems its a bit along the lines of the design aspects on this thread but as you guys here are WAY WAY past me on design I decided to throw this out there and see it get shot down in flames....
at least if it gets shot down I can get this thought out of my head

The six drivers with three stepsticks for the three extruders..... I cannot think of any time when all three extruder motors are being driven at the same time. Unless you use three separate hotends and I think its still unlikely to drive all three at once.
So would it be possible to use one driver and in some way switch it from motor to motor?
So E1 retracts and parks its filament.. the motor state is stored in variables and E1 is effectively disconnected, E2 is activated and its stored state is pulled and E2 is now actively being driven, while E1 and E3 are just off and waiting their turn...

If this is not just my naive ramblings, then it would save a lot of real estate on the pcb right? Also lower the cost of six stepsticks to the cost of four and still do the same job?

ok, my apologies for being a noob

Edited 1 time(s). Last edit at 10/02/2017 05:23PM by JustSumGuy.