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

Quote
Dust
I also wonder if you can get SMD .1" connector strips... that would remove a bunch of holes

Oh you can (not really surprising)

only thing is, you gotta watch it on the ampage. THREE amps (!) and one of the things i'm doing on RD3D - bear in mind it's a 2-layer board - is connecting both sides and making sure there's plenty of copper leading to them.

for the arduino pins.... yyeah they'd be on the underside, which makes it a double soldering jobbie, that means some different (slightly more costly) reflow technique or something. in some ways it's better to ship people a kit and say "here! you do it! it'll be fun, too!"

Quote
JustSumGuy
I am going to lurk out here for a while... learning lots and have little to input

thanx for letting me shoot some input tho

I think I want one of these when it gets a little more traction....

you're more than welcome. (oh.... great book "good omens" by neil gaiman and terry pratchett... reminds me of the quote about bruce springsteen's "born to lurk"....)

Edited 1 time(s). Last edit at 10/06/2017 01:11AM by lkcl.

3 MOSFETs are working, yet to put thermistors on and test those. Marlin-rc firmware appears to be buggy: it won't enable the fan but by swapping Fan and Extruder (D8 / D9 to D9 / D8) i was at least able to check the hardware. four out of six steppers tested, all good. one i had to re-seat the resistors. i ended up putting on the TSSOP-14 buffer IC, it seems to be working well, it's a bit small though for hand-soldering. i've replaced it with SO-14 on the newer revision board.

the power connector, dang it's terrible. how do people put up with it?? the wires don't stay in properly!

anyway, with the buffer IC on, i could actually try a due with a RepRapFirmware port. it'll be worthwhile putting the MicroSD card slot on, although it's not got a buffer IC on it on this revision of the board.

working assembled prototype! so that's 3 MOSFETs confirmed (fan, bed, extruder heater 1), thermistor 2, endstops, and 4 out of 6 steppers tested, two of them with TMC2100s and an A4988, the others with just an A4988. this is the version that can't do SPI, and only has 3 endstops. it's tested with Marlin-rc7 commit d97dc102, on both an italian and chinese ATMega 2560 r3. i'll set up the git repo soon, with the new header file for the pin-definitions.

next on the list is to do the pinouts for RepRapFirmware, put on the micro-sd card, and test with a Due. very carefully.

Those fuses looks suspiciously like wires

But well done

Quote
Dust
Those fuses looks suspiciously like wires

But well done

biig wires. yeh mike was in a rush (just before holidays) and couldn't find the fuses or fuse-holders. bit of a risk stuffing bits of wire in the holes instead, but hey...

Quote

the power connector, dang it's terrible. how do people put up with it?? the wires don't stay in properly!

The genuine Phoenix connectors are meant for these crimp on cable protector thingies... I don't know what you got from china?

arse! fek! that is an ecumenical matter! [www.youtube.com]

the SN74HC125 only switches on @ 3.15 volts for a 4.5v input, and the Due (ATSAM3X) supplies 2.95. frick! rrrr..... ok have to use a proper level shifter IC. frick.

edit: arse, arse - it has to be a SN74AHCT125D or ACT type: those respond @ 2v (but have to be supplied with 5V, which is ok, that's what they're getting). what a pain in the neck.

Edited 1 time(s). Last edit at 10/08/2017 02:05AM by lkcl.

If the Due is only supplying 2.95v output, then I suspect that either it is getting less than 3.3V supply, or you have a pulldown resistor with a rather low value connected to the output.

A 74HCT chip is good to use as a level shifter to 5V.

---

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
If the Due is only supplying 2.95v output, then I suspect that either it is getting less than 3.3V supply, or you have a pulldown resistor with a rather low value connected to the output.

A 74HCT chip is good to use as a level shifter to 5V.

yehhh i didn't use a 74HCT, i used a 74HC by mistake. the datasheet says that they accept variable input from 2v to 6v, and their input response characteristics (high and low threshold levels) change according to the power rail voltage. whoops....

3V3 is registering 3.28v so that's good...

board bring-up's on RRF is going well btw... apart from the wrong 125 buffer IC. steppers 1 through 3 work fine, just checking extruder. mailed you a bugreport (ternary operator, lack of brackets, bug is not noticeable of HIGH_ON is false)

ok so with RepRapFirmware, i have:

* Y and Z endstops working, no explanation as to why X isn't (despite it working on Marlin - i'll have to double-check that)
* A4988 steppers working in positions X, Y, Z and E1.
* A4988 stepper not working in position E0, worked in Marlin, will have to double-check (in case of a hardware fault or maybe the resistors need re-seating. the steppers are being driven from 3v3 not 5v0)
* position E2 (Z2) not yet tested.
* Thermistors are good.
* Micro-SD not testable without doing a software (bit-banging) version of SPI. oops. speed's not a priority.
* Heater PWM works but the voltage is too low for the (wrong) buffer IC: have to switch that out.

i think... these things i can put up with as far as making a decision to go with the 2nd revision right now is concerned. the steppers are all reworked anyway, the micro-sd card is on a hardware-driven SPI, i'm confident (thanks david for prompting me to investigate properly) that selecting the correct 125 IC will get the MOSFETs working.... yeah.

yyeah the resistors in positon E0 are measuring silly values. the 100k is registering 33k and the 10k about 4.8k. that's not a surprise: they were the first resistors i started putting on, used far too much solder paste and used the heat-gun as well (which i stopped doing very quickly) - it looks like not only the solder paste has created resistance but also the resistors themselves are damaged. i'll pull them out and replace them.

oof! i just realised, the micro-sd card's buffer IC (using a SN74HCT125) isn't going to work.

* power from the arduino/due can change between 3.3v and 5v
* the microsd card can only take 3.3v
* SN74HCT125 can only be supplied by 5V but its inputs (HIGH) can be signalled at 2V or above
* SN74HC125 can be supplied by 2-6V but its inputs (HIGH) *MUST* be over 4.5V when it's supplied by a 5V rail

so neither of those are going to work. does anyone have any suggestions? i've used a TXS0104 in the past (and a TXS0106 or close to it for MicroSD), it's quite smart, can detect push-pull and OD, is supplied by two different voltages, A and B. however it's bi-directional so is $USD 0.45 instead of... $0.15 or whatever. any other ideas that would be reasonable cost? uni-directional is all that's needed because level-shifting's not needed on the input from the SD card - not for slow SPI operation, at least, even though it's supposed to be push-pull not Open Drain, the Arduino 2560 seems to be ok.

[edit...]

ok so the SN74LV4T125 looks like it might do the trick. it can be supplied by 3.3v and has 5V tolerant inputs. soooOooo....

Atmega 2560
-------

* datasheet atmel-2549 section 31.1 says that input high voltage level is 0.6 times VCC for 2.4v-5.5v supply.
* output high voltage is 4.2v @ VCC=5v
* arduino mega 2560 datasheet says VCC = 5v.
* therefore input "high" voltage threshold is 3.0 volts.
* but... we don't care about *input* level-shifting.. only *outputs*. which will be 4.2v @ VCC=5v.
* input "low" voltage threshold says it's a maximum of 0.9v @ VCC=5v

* SN74LV4T125 datasheet says, according to section 7.5, that High-level input voltage @ 5V is 2V
* and that Low-level input voltage @ 5V is 0.9v at Ta=25C but can vary (0.85v) at other temperatures. it's just about okay.

therefore INPUT is okay (within tolerance... even if the temperature is off...) as even if the 2560 outputs 5V when the LV125 is supplied by a 3.3v rail that's perfectly okay.

the other way round (MISO, CD) we know from various break-out boards that you can simply wire those up direct to the arduino 2560, no problem, no level-shifting needed.

Due
-----

* Supplies are all 3.3v
* output high voltage according to section 45.2 of atmel-11057 SAM3X datasheet is: supply voltage - 0.4V (or VDDIO-0.2v for some ports)
* output low voltage is 0.4v (or 0.2v for some ports)

(hmm dc42, this explains why i was getting 2.95v on D8-D11: it's a different port.)

so, output low voltage is waaay below what's needed, all good there. output high is 3.3-0.4 = 2.9v

2.9 volts is greater than the 2.0v threshold.. WAIT... waitwaitwait.... i'm not going to be supplying the LV4T125 @ 5v, it's going to be *3.3v*.


argh.


ok. redo.

Mega
------

* LV4T125 high input threshold @ 3.3v is 1.3v
* LV4T125 low input threshold @ 3.3v is 0.85v (dropping to 0.75v @ extremes of temperature range, -40 to 125C)

* Mega's "low" voltage output level is... 0.9v max.

i'm not comfortable with that. it's too close. however..... people do seem to be using the HCT125 successfully, even though the datasheet says "0.8v" for the LOW input threshold.

also, taobao pricing is at piss-take levels. USD 1 in some cases.

anyone any other suggestions? the TXS0104E at least has sellers around $0.25...

Edited 1 time(s). Last edit at 10/09/2017 11:42PM by lkcl.

The 0.9V output low voltage of the Mega is probably quoted when the output is sinking a lot of current. It will be much closer to 0V when it is lightly loaded. AFAIR there are output voltage vs. sink current charts in the datasheet, and although they show typical rather than worst-case values, they should confirm this. However, ground noise generated by the stepper motors may be a bigger problem if you are not careful with PCB layout and routing.

Edited 2 time(s). Last edit at 10/10/2017 12:16AM by dc42.

---

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].

SN74AVCT774 looks like a good candidate: uni-directional but *selectable* uni-directional... [www.ti.com]

has 2 diferent supply voltages (like the TXS0104 range), so it would be ok to put 3.3v on the B side and either 3.3 or 5 on the A side... it would definitely do the job...

Quote
dc42
The 0.9V output low voltage of the Mega is probably quoted when the output is sinking a lot of current. It will be much closer to 0V when it is lightly loaded. AFAIR there are output voltage vs. sink current charts in the datasheet, and although they show typical rather than worst-case values, they should confirm this.

ah ha! excellent, that was the clue i was looking for.... section 32.7 shows a graph: the 0.9v is when VCC = 3.0... bizarrely when VCC=5.0 the output is always below 0.6v! doh...

ok so that LVT125 would actually do the job. whew. no change to the board layout needed.

Quote

However, ground noise generated by the stepper motors may be a bigger problem if you are not careful with PCB layout and routing.

yyeah... i've stuffed this 2-layer board with so many VIAs it's... mad. i've done larger (and SMT) 2-layer boards before, this is very much an experiment.

i'd really like to get a suitable inductor onto the VMOT supply, but it has to be through-hole: there's no way i'll get a large SMD inductor on the board... hmmm if i could use several of these ( BL03RN2R1M1B - 6A ) that might do the trick... [edit: or maybe the BL02RN2. 4 of them, to help bridging from the main 12/24v input. they also have the advantage of being quite cheap: $0.06 in 1k quantities even in the US off of digikey) ]

Edited 1 time(s). Last edit at 10/10/2017 01:04AM by lkcl.

..neverending story. I feel for you
I hope the new chosen chip fits all the other needs too, like latency. But I'm sure it's quick enough.

Quote
o_lampe
..neverending story. I feel for you
I hope the new chosen chip fits all the other needs too, like latency. But I'm sure it's quick enough.

i hadn't checked... but now that you mention it i will... and ooo it's a good job you brought it up, as the datasheet says it only does up to 50mhz @ 3.3v and 5v. luckily there's no way the SPI interface is going to realistically get 50mhz, probably not even 25mhz, so it's fine.

just spent the afternoon putting in those noise-suppression filters. yet more through-hole components, that's another twelve holes in the board. dang. swiss cheese has nothing on this...

btw o_lampe, dc42, everyone else: i really appreciate the interaction, here - it might not necessarily be an exact question but something that spurs me to properly investigate, and do a decent job where otherwise, on my own, i just... wouldn't bother. so.. thank you.

I've found it interesting to follow the development. Like I said, once the design is relatively stable I'll be interesting in trying one out. My latest project has me swapping crystals on my MEGA2560 to get a little um oomph compared to the arduino reference design since the chips run 20% slower than what they are rated at.

Quote
obelisk79
My latest project has me swapping crystals on my MEGA2560 to get a little um oomph compared to the arduino reference design since the chips run 20% slower than what they are rated at.

The rated frequency of the 2560 is only 16MHz. It's the smaller members of the ATMEGA family (e.g. the 328) that are rated at 20MHz.

Edited 1 time(s). Last edit at 10/10/2017 05:00PM by dc42.

---

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].

Thanks! I stand corrected, I misread the ATMEL website. Based on reading other user experiences they do still have a lot of headroom to allow for overclocking up to 24Mhz, I don't mind risking damage to some cheap hardware. Just because I have a bad habit of 'what if?' and of course, it was cheap.

Quote
dc42

Quote
obelisk79
My latest project has me swapping crystals on my MEGA2560 to get a little um oomph compared to the arduino reference design since the chips run 20% slower than what they are rated at.

The rated frequency of the 2560 is only 16MHz. It's the smaller members of the ATMEGA family (e.g. the 328) that are rated at 20MHz.

Edited 2 time(s). Last edit at 10/10/2017 07:45PM by obelisk79.

Quote
obelisk79
I've found it interesting to follow the development. Like I said, once the design is relatively stable I'll be interesting in trying one out. My latest project has me swapping crystals on my MEGA2560 to get a little um oomph compared to the arduino reference design since the chips run 20% slower than what they are rated at.

yyeah the allwinner a20 does that trick. variation in manufacturing where the doping on the silicon varies due to the particular IC being at the edge of the wafer etc. etc. you can get some ICs that will do a faster speed than others, have less tolerance than others etc. etc.

what intel and AMD do is, they test absolutely every single one of their ICs, and the ones that have all the cores working and pass the tests @ NNN x 20% clock speed, they put a 300 to 900% markup on them and sell them at you as the super-duper "high end" processors.

the ones that *fail* on certain cores, or the L1 or L2 cache doesn't quite work, or they're a bit dubious on some of the tests: they shut those cores off, and/or ramp the speed down a bit, they slap a different label on them and sell them for less.

sometimes they don't quite fill all the market demand so they shut off PERFECTLY WORKING cores and sell you those just so they don't have to go through the rigmarole (and cost) of making another batch and testing them again, only to find that the yields *still* didn't quite give them enough to fuflil market demand.

allwinner, atmel, ST etc. etc. at this much lower pricing bracket, the profit margins are far too low to do that... so they don't bother. they do a basic test (turns out that treating the entire wafer as a diode and if it reads 0.7 volts it's 98% likely to be good), throw it in a package, *then* they will do a basic systems functional test, and if that passes it goes out the door at the LOWEST common denominator speed.

so

bottom line is: you *MIGHT* be able to run *SOME* 2560 processors at higher rated clock speeds, but you almost certainly will not be able to run all of them at the higher speeds. and that's ok: like you said, if one blows up, or overheats, or gives you incorrect calculations, just use another one. you have time on your hands to do those kinds of tests: Atmel, who want to make money, don't.

Edited 2 time(s). Last edit at 10/10/2017 11:15PM by lkcl.

ok so thank you to dc42 for mentioning about the EMI from the motors, i'd been looking at the duet wifi schematics and noticed that you have a 10A inductor on VMOT. i'd been wondering about that, whether i should put in some suppression filters or not, and the tangle of swiss-cheese spaghetti that this board has become just made me realise ferrite beads really are needed.

so! this is what the +24v plane now looks like:



previously it was going all the way across 70% of the board, right in the middle and also along the bottom. huge areas taking up the majority of free space for routing because also there has to be a correspondingly large(r) area for GND underneath it. this is the primary cause of the routing tangle: the 5V line is particularly hairy and goes right along the edge of *two* sides of the PCB in order to get to the top left corner (SN74LV125 buffer) hmm i might change that to go round the other two edges... hmmm...



so this shows where the power inductors have been placed. if i only had the one (or two) it would mean having a 6 to 9 AMP track for the top three motors and ANOTHER 6 to 9 amp track for the bottom three. that means a triple split (MOSFETs, 1st bank of VMOT, 2nd bank of VMOT) and just it's too much, particularly as the inductors take up space on *BOTH* sides of the PCB, with the holes themselves taking up about 3 amps worth of board space. remember there's not enough room to fit a 10A SMT inductor on the TOP layer, so through-hole it has to be, and that means taking up space on BOTTOM as well, right where the corresponding critical GND plane has to be.

so instead of doing that i decided to put in SIX inductors - one for each. the ones i've selected are complete overkill: 6 amp rating. but, they're very very common and so are very very cheap. i suspect that they're extremely commonly used as EMI suppression filters in mass-produced power supplies.

with one separate inductor per VMOT, as they're through-hole the GND plane can fit in between the two pins. one thing i am a little paranoid about is making sure that there's copper on *both* sides of these through-hole components (the MOSFETs as well - nearly all of them). so, i put two VIAs down (one either side), a small piece of copper on BOTTOM to join the two VIAs and the component's through-hole pin, so that current can go down through the two VIAs from TOP and into the pin from the *other* side. the VIAs are rated at maybe.... 1A to 1.5A per VIA, and the copper surrounding each pin at maybe... 3A. so you get 3A from TOP, then with 2-3 VIAs you get enough current going down the VIAs to feed the *other* side with 3A, up the pin: total rated current, 6A. if it was only one side it would only be 3A. it's overkill but it makes me feel a little bit happier.

these screeenshots highlight the GND plane. it's a spectacular... mess. totally fragmented and there's really nothing that can be done about that. in red (non-highlighted) you can see where the +24v plane takes up a huge area, and then on the corresponding blue (BOTTOM) you can see that i've made sure that GND is clear underneath it. in the bottom left corner and right up the left edge i've drilled a *massive* number of VIAs to join the two areas together, in order to try to use both sides of the power (and fuse) pins. i swapped the order of the power pins (- + - +) compared to RAMPS in order to have the top-most GND pin directly onto the big GND plane on both sides of the PCB. basically it's damn complicated!

i'm also spending a considerable amount of time making sure that both GND pins of the steppers have a significant "join" to GND, no matter how fragmented the two planes are. this has proven to be relatively straightforward for all but the bottom right and bottom middle two stepper boards, in part because of SPI routing but also because 5V, 3V3, EN/STEP/DIR pins *and* the actual arduino headers *and* the thermistors are all in the way! luckily i just worked out that i can take the 5V line round the other two edges so this will help enormously.




Edited 1 time(s). Last edit at 10/10/2017 10:31PM by lkcl.

did a walk-through video: [youtu.be] ran out of space on the sd card...

The Duet WiFi design includes provision for that 10A inductor, but on production PCBs it is bypassed and not fitted.

---

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
The Duet WiFi design includes provision for that 10A inductor, but on production PCBs it is bypassed and not fitted.

ahh interesting. turned out it wasn't needed? hmmm.. TMC2660s would not generate as much noise, would they... people with this design could be using DRV8825s and other low-cost stepper ICs...

ok so yep i replaced the 100k and 10k resistors that were covered in black bits of gunk (crispy deep-fried solder paste...) and the stepper in position E0 works fine, now. the 100k resistor actually measured 33k and the 10k around 4.5k which tells you how much crud i'd managed to put on the PCB... whoops...

so that's at 3.3v using the Due, which is a good sign.

ok so i decided to go with it - sent the BOM off to the factory and the gerbers - otherwise i will just keep on making tweaks. need to stop doing that and move on to other things...

i'm sort-of in... shock. this is a 4in x 3.25in board and it has 900 drill-holes in it! whaaat the ferrrr??

Too many through-hole parts and/or vias? Although the connectors for the Arduino and the stepsticks no doubt account for a lot of them.

---

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].