ATX power supply and Sanguinololu controller

I'm the new owner to a partially complete kit and have some new build questions.

I believe that a 450W ATX power supply should be sufficient to power a mendel reprap, but I've heard that you should have at least 20A (with heated bed and such). The sticker on the power supply shows 18A for +12V and I assumed I could run the electronics on the 5V supply and save the 12 for the steppers and heaters. But looking at the wiring diagram for the Sanguinololu on [reprap.org] it looks like there is only one power connection for everything.

So is the 18A +12V supply from teh ATX suppyl sufficient or do I need to do something else? (I figured I'd run any fans off of the -12V supply which is 2A to help a little.)

Thanks,
Bret

you will want another supply. it is quite common for people to use 2 supplies for reprap. 1 for bed heater and one for the electronics. if you are mind set on using just one supply and using a heated bed, then you'll need another power supply with greater current on 12 rails. verify 12 rails ratings, and look for at least 30amps, 20amps are for the BED heater alone.

Edited 1 time(s). Last edit at 03/01/2013 11:37PM by jamesdanielv.

Use a smaller bed heater, say one rated at 10A, and you'll have plenty of overhead.

Ah, that makes some sense. Bummer. I really don't want to have to use 2 power supplies for this. Seems a little silly and I'd have to find a second ATX power supply. What power supply would you recommend?

Would something like this work: [www.ebay.com]

Thanks for helping me avoid a mistake.
-Bret

yes this supply should work.

Thanks. I am getting some conflicting information. If you look at the Heated bed PCB page it says the following ( [reprap.org] ):

Sanguinolou The MK2 PCB heatbed heats up to 110C when powered through the heated bed connection on Sanguinolou, your power supply should be 300W and ensure your wires from your power supply to the Sanguinololu should be capable of handling the total draw of 20A+. Using a 300W ATX power supply with the 4wire ATX dual 12V connector is working well for me.

So whoever posted this is getting by with the 300W power supply to source his whole Sanguinololu board and I've got the 450W version so it seems like it should be OK. I used a cheap ohmeter to measure the resistance and it came out as 1ohm which would put it at a 12A draw if it measured correctly (I'll check it again with a better meter later today).

The way the Sanguinololu is setup you supply it with 12V and then it drives everything including the heated bed. That little dinkly board with those small pins seems awefully small to be carrying 30A if it really requires that. Am I missing something here?

Thanks,
Bret

cobrageek Wrote:
-------------------------------------------------------
> The way the Sanguinololu is setup you supply it
> with 12V and then it drives everything including
> the heated bed. That little dinkly board with
> those small pins seems awefully small to be
> carrying 30A if it really requires that. Am I
> missing something here?

I used a 15A heat bed with Sanguinololu and blew up two bed control FETs and melted the bed power connector. So yeah, that dinky board can't handle currents like that. I had to get a better FET, replace the connector with something that doesn't melt that easily and improve the traces on the board.

Then again, the RAMPS board can't handle it either, the bed will trip the PTC fuse if I try to heat the bed all the way to 105 C in one go... have to take a break at 70 C.

This is why it's a good idea to power the bed from a separate 24V power supply if the bed current would be significantly over 10A at 12V.

It's interesting the conflicting information I'm getting on power supply. The PCB is 1 to 1.2 ohms so it should be around 10-12A when heating. On the Sanguinololu page it suggests wiring a bypass for the traces for the heated bed because the board can't handle it. They also suggest a heat sink for the FET (I'll scrounge one from the lab or make one our of some scrap aluminum I have). So I'm going to try these modifications and see if I can make it work.

If you assum 10A for the heated bed and ~5 ish for the extruder tip, that doesn't leave a lot for the rest of the system. But then from experience there are several people who run this setup and say they've never had problems. Maybe only heating to 60c (PLA) reduces the problems and the fact that only the 12V output from the supplies is used allows it to supply more. I also see a lot of other pages which say you need to have 20A for your full system with heated PCB (not 30A). I'll probably just try it and see how things go. I won't print unattended until I have confidence it isn't going to burn down the house :-).

1. An ATX power supply with 18A on the +12V leg would be sufficient to power your printer with the heated bed for printing ABS. The bed would use a around 11A, the hot end would use about 2 A (resistor around 6 ohm) which leaves 5 A for he motors and electronics. I've run a Mendel for a couple of years with such a power supply.

2. I think the Sanguinololu board traces cannot handle the current for the heated bed. Nophead has a solution that he claims to work well: connect the heated bed lead to the metal tab on the MOSFET and the other lead directly to the +12V teminal on the power supply.

3. It's not really that common for people to have separate PS for the heat bed. It's more common to have the same 12V supply for everything.

4. RAMPS 1.3 and 1.4 has separate traces for the heated bed power so it should be fine. But, depending on the MOSFET used, it might need a heatsink on the MOSFET. There are MOSFETs that have very low Rds so that it doesn't need a heatsink. I have been running my Mendel with RAMPS 1.2. I had to install a heatsink on the MOSFET and I ran the +12V directly to the heat bed and then added extra wires between the - heat bed terminal (D8) and the MOSFET.

Edited 1 time(s). Last edit at 03/06/2013 12:07AM by brnrd.

Thanks brnrd. That sounds like the direction I was going. It's good to have confirmation. Nice to be clear on the tip for bypassing the traces on the heated bed for the sanguinololu board. Sounds like I should be in good shape with the ATX supply then.
-Bret

18A for +12V is this the entire system or a single rail? supplies include 2 or 3 12 volt rails. typically 1 high current rail and 1 or two lower current.

you will want the high current 12v rail for heater bed, then the other one for stepper drivers and nozzle heater.

Edited 1 time(s). Last edit at 03/06/2013 07:16AM by jamesdanielv.

James, I'm not sure if 18A is for all the 12V's or just the high current one. It's just listed as 18A on the label for the 450W ATX supply.

On the Sanguinilolu how do I wire up the different supplies? If I do the trace bypassing idea, I guess I can take the high current supply and tie it into the bypass to the heater bed. Then use the other 12V supply(s) for the main supply connection. Is that what you had in mind?

I assume I can find specs for ATX supplies online and I was going to use that to figure out which wires where high/low voltage. Any good sources for ATX supply specs?

Thanks,
Bret

If you have a multimeter, then you can turn off your ATX power supply, unplug it and check for continuity between all the +12V wires (yellow is standard). I'm guessing that they will all be connected to each other since the sticker only lists one +12V rating. That would confirm that you only have one +12V rail.

Then you can run 2 of the yellow wires to one terminal of your Mk2. If you need to use additional wires, then use at least a 14 G to minimize voltage drop. If you use 16 gauge, keep it as short as possible or use two. I've used speaker wire for this purpose. Then connect the other Mk2 terminal to the heated bed MOSFET tab in the Sanguinololu.

You want to run 1 yellow wire to the Sanguinololu +12V input terminal and 3 ground wires (black) to the gnd terminal.

You' may also need to connect the +5V to a dummy load (10 ohm, 10 Watt wirewound resistor is fine) or a fan or light that draws about 0.5A. Most ATX supplies need a load on the +5V rail or the +12V will be too low or it won't stay on when you put the motors on idle and turn of all your heaters.

Thanks for the tips brnrd. Is there a way to power the Sanguinololu electronics with the +5V supply but feed the +12V for all the mosfets and controls? That would help slightly reduce the load on the 12V supply but also provide that minimal load on the +5V that ATX supplies require. I'm just not sure how or if it can be done.
Thanks,
Bret

I don't think it';s worth the trouble to go through all that. The only thing in the board that uses 5V is the microcontroller which doesn't use much power as far as i know. The stepper controllers use 12V. Better to use the +5V for cooling fans or LED so that you can see what you're printing.

We use the Sanguinololu 1.3b board with our kits, the Mk2 heat bed, and a single 360 watt power supply (non-ATX) to power everything. We power the board with 12 volts from the power supply.

I recommend that you use a relay to control the current being sent to the Mk2 heatbed. You are right, the Sanguinololu cannot handle that much current. We also recommend adding a fuse between the relay and the heatbed. We use a 12 volt, 20amp automotive fuse. This simplifies things because female spade automotive terminals can be used with these fuses (no need for a fancy fuse receptacle). Another recomendation is to add a 10A 12V automotive fuse between the power supply and the Sanguinololu.

Thanks 3dbotic. The suggestion on the main Sanguinololu page on the wiki is to run a bypass wire to the FET and use a heat sink. Do you not think that is sufficient? Have you seen issues from this kind of setup or just from running only through the PCB traces? Is the FET insufficient for the current? (anyone look at the specs for the FET?)

I assume your relay play is to use an automotive relay to switch the 12V and control it by the output of the FET? I think I may have a spare relay in the garage (just finished a car project and may have a spare that I didn't use).

Thanks for the input.
-Bret

(Here's the quote from the main wiki entry)

The pads and traces on the board are not robust enough to handle the high current involved when controlling the heated bed with the sanguinololu's dedicated mosfet. If left as is, the board will heat up in this area, could be damaged and the plastic connectors discolored. To fix, simply add a direct current path by way of (preferably insulated) wire from the 12v input to the 12v pins on the Heated Bed (H connector; from the center pin of the HB mosfet to the two ground pins on the HB connector; and from the ground pin on the mosfet to the ground input of the board (probably the ground side of the 12v connector). Find the appropriate traces using continuity check or resistance mode on your multimeter, and check after to ensure you didn't create a short between any of these components in the process. It would be prudent to check for shorts before starting so you don't waste effort trying to fix a short you think you just created but was already there before you applied this fix. To understand this fix, pretend that no traces exist between the high current pathways of the 12v input and 12v pins of the HB connector; the ground input and the ground pin on the mosfet, and the center pin of the mosfet and ground pins on the HB connector. Assume you must create these pathways, then do so!

Edited 1 time(s). Last edit at 03/17/2013 11:53PM by cobrageek.

From what we have seen, you can get 12-14 amps going to the heat bed (assuming 12v)

We don't recommend it but, you can run a bypass wire and heatsink the MOSFET, if you can find the room to get a sink on there. We use straight header style connectors. If your Sang board uses the 90 degree style, you might be able to put a sink on there. Also, you will need heavier gauge wires from the Molex connector to your heat bed, along with use 2 pins (in parallel) for + and 2 pins for - just to distribute the current load over the 4 pins. We just felt it was more elegant not to mention safer to use a Omron relaya, and have the heavy wires going to it from the power supply and heat bed, instead of taxing the Sang board.

An automotive relay would probably work just as well, but we don't use an automotive type relay, as the relays we use have clear cover, thinner spades and a LED that lights up when the relay is on. The automotive relay will have thicker spades so crimp on female spade terminals would work really well with it.

Yes, we are using the MOSFET to control the relay in our setups.

Here is the datasheet for the P55NF06L MOSFET we have on our sang boards (not sure what your PCB has on it). [www.datasheetarchive.com]#

As you can see, the MOSFET should be able to handle the current, even at the higher temps as the MOSFET heats up, its current rating goes down.

Nophead connects the other lead from the heated bed to the metal tab of the power MOSFET. I wouldn't use an electromechanical relay. It won't last long if the firmware uses PWM to control the heatbed. Relays are not necessary for 12V. There are MOSFETs with low Rds that are more commonly used now. If heat is an issue, you can change the MOSFET to one of those in addition to adding a heat sink. Also, about 10-11 amps is the most that the Prusa PCB heat bed uses since it's around 1.1-1.2V at room temperature and resistance goes up with temperature.

I have the same MOSFET (P55NF06L) so it should be able to handle the current. The suggestion is to use wires to replace (or replicate) the traces on the board, but I was actually thinking to run a separate wire directly from the 12V supply to the bed and ground to the MOSFET. And running double connectors on the 4 pin connector seems to be standard (at least shown in the wiring diagram).

Good point about the heat sink not fitting. There isn't much room is there. Maybe it will help if it's pinned onto the back? Maybe I'll fabricate something that is skinny at the bottom with fingers up on the top so it would fit.

brnrd, what do you mean by the "metal tab"? Are you saying that he connects to the metal back piece with the hole that comes out the top? Is this the same as the "ground" or return from the heat bed? (can you help spell out the whole wiring setup he uses? Does he have it on his blog somewhere or something where I can just go check it out?)

Thanks!

[forums.reprap.org]

[www.soliforum.com]
[1.bp.blogspot.com]

Thanks brnrd. I'll read through the material during lunch today. Looks like great stuff and I'm sure I'll get it setup safely after all that help.

For a heat sink, I was ripping apart an old PC power supply and found some perfect heat sink's that are just a flat piece of sheet aluminum at the bottom where they bolt to the regulators (which are shaped just like the MOSFETs) and then have fins above the devices. Effectively vertical heat sinks that would fit perfectly on all the MOSFETs on the Sanguinololu. I'm going to cut them to the right width and put one on each of the MOSFETs on the Sanguinololu. Can't be too cold :-).

Thanks for all the help,
Bret

brnrd, thanks for the excellent links. That post had a really good discussion with some very nice ideas and pictures.

It looks like nophead's original fix was mainly to avoid the smal connector problems. He has taken care of the connections to the heat bed (+12 directly from PS and common back to the mosfet drain/ring), but the source connection back to the ground on the PS doesn't appear to be enhanced (no discussion in his mendel 90 manual and no connection shown in the blogspot picture). It's also interesting that while he doesn't talk about heat sinks, the bolt he used in his is likely acting like a heat sink.

Here's my plan. See if you think it covers things:

1. I'm a bit nervous about the power going through the heat sink and the one I stole from an old PC PSU has insulator material so I think I'll use that cut down to the right width.
2. I'll pull the 4 pin connector off to get it out of the way.
3. Solder a thicker wire to the source and run directly to ground on the PS, but leave it soldered to the board as well. I can't see how that would be a problem as the bulk of the current will travel the easeist path (wide wire directly to the PSU).
4. Wire from the common on the heat bed to the mosfet tab/ring as the drain connection. Leaving the pin connected to the board shoudl have no effect because I've removed the 4 pin connector.
5. Wire from the +12V PSU directly to the heat bed.

It seems like this should remove the connector and the board traces from the path and provide heat dissapation for the MOSFET. Do you see any issues with this plan?

Thanks for all the help on this.

Yes seems like a sound plan.

The power connector on my Melzi is a screw terminal block and that seems to handle the ground current and so do the PCB traces it was just the 4 pin Molex connector that is nowhere near rated for 10A.

Yes I think the ring terminal and thick copper wire attached to the MOSFET tab do act a heatsink to some extent. It gets warm but not too hot.

---

[www.hydraraptor.blogspot.com]

Thanks nophead. Great Blog BTW. I may have to copy your fan hack too, but for now I'll just have it run all the time (or maybe wire it through the Z switch so it is on any time the head is off the first layer).

Can you help me answer an important question from my wife. How do you pronounce "nophead"? Is it 'nop-head' or is it 'nofed' or 'nofeed'? :-).

Thanks for all the help and sharing your work freely.
-Bret

It's "nop-head" as in NOP, an assembler instruction mnemonic all CPUs have that does nothing. At the beginning of my career I was in a small team of assembly language programmers and we used it as a term of abuse when one of us did something dim.

---

[www.hydraraptor.blogspot.com]

I thought as much. I have a background in CPU's and ISA's in particular as well so NOP is a part of my daily vocabulary :-). Great name. NOP-head. I like it and may have to use it on occasion. Thanks again for all the help.