Skip to content

Archive for July, 2017


3D Printer Power Supplies

Having built a dozen 3D printers, you get used to looking at parts that are top notch, some marginal, and some that you wonder why they were ever produced.

But there’s also some parts that you just take for granted. Case in point, the typical 12V, 20 or 30A (200-360 watt) power supplies.

Normal cost is in the $25 – $40 range (USD or CDN). Paying a premium for one, double or triple the cost, does not promise any higher quality than the budget ones in my experience.

The Tin Can

The typical tin can power supply takes 110-220 and drops it down to 12V (or whatever working voltage you need) that has a trim pot for fine adjustment. The majority of them look much like the one in the photo.

12V 30A Supply

The variation of these are ones that have a mesh covering and have no fan. Relying on convection to keep the wee beastie cool. For most of these types, 20A is usually the limit of their current rating, but I’ve seen a few 30A ones.

12V 30A Mesh

There’s a soldered fuse inside because using proper fuse clips would add cost and make no mistake about it, these things are cheaply made.


The mesh covered ones are self cooling. Or as I found, when you fry one, they remain permanently cool. As you’ll eventually see this may not be a bad thing either.

The tin can versions all sport a fan. Considering the PCB’s in the ones I’ve looked at from various suppliers are all oddly similar, the fan voltages range from 12 to 16V. Or perhaps whatever is in the parts bins for that days assembly.

Paying a premium for one doesn’t ensure better quality but occasionally you’ll find a temperature controlled relay so the fan doesn’t run all the time. I blogged about this previously

For the MOSFET’s that do all the current handling, cooling is either them clamped to the case itself, or with an aluminum bar between the MOSFET’s and the case. Usually with some silicon material for insulation and goop for heat transfer.

IMG 1155

It’s not pretty, but it’s at least marginally functional.


Like most end users, I just wire them up and away I go. However, since I wrote the blog about adding a relay for fan control, I’ve modified half a dozen power supplies and that means I have to take them apart to do it.

What I found in TOO many is damned scary. Seriously. If you have one of these power supplies, my advice is disconnect it, take it apart and INSPECT THE THING! Yes, do it. If you get a new one before you put it into service, take it apart, INSPECT IT!

And don’t worry about the warranty sticker covering one of the screws, it’s a tail light warranty at best. I.e. when they can’t see your tail lights the warranty is up.


There will be a maximum of SIX screws on the top piece holding the fan. The “warranty” sticker might cover one. Don’t be surprised if some screws are missing, or stripped. I’ve found both. Standard M3 screws are what they use.

IMG 1151

Once you get the top screws off, you’ll see something like this, if not exactly the same if you’re disassembling a 12V 30A supply.

IMG 1153

Unplug the fan connector from the PCB and set it aside.

Remove the four or six screws holding the MOSFET’s to the case from the one side and end. Be careful not to rip the insulting gasket. If your supply has a metal bar between the MOSFETs and the case, wiggle it a little and slide it up out of the case. Set all those parts aside aside.

There are five holes in the PCB for holding it on to the case. I’ve yet to find a supply that uses five screws. Screws cost money, You’ll no doubt find four screws (one in each corner, one rarely in the centre) so remove them. Set them aside.

The whole PCB should now lift almost straight up and out of the case. And the fun begins.


It doesn’t matter if you know zero about electronics or not. You’re not troubleshooting or doing component level repair.

What you’re looking for is potential hazards that will be pretty obvious once you know what to look for…

First, check the solder side of the PCB. You don’t want any untrimmed leads hanging down more than 1/8″ under the circuit board. On some of the supplies there is a plastic shield under the PCB, but I’ve also found a few that have nothing. If you see a long lead, don’t bend it over, trim it off. Use nail clippers if you don’t have wire nippers.

A reasonably good power supply PCB will look like below. All shiny, no dull solder joints, no leads poking out.

Good Supply

On the other hand, you can find something that looks like this one. See those BIG wire jumpers? That tells you SOMETHING is amiss with the PCB foil traces. Take a photo or two, put it back together, tell the seller it’s junk and don’t use it.

Not So Good

Moving on, flip the circuit board over so you’re looking at the component side. What you’re looking for is two things. Components that are bent over, shorting against other components and solder splashes.

I’ve only found one capacitor that was laying on top of a nearby resistor, which would have caused the supply to leak smoke when I powered it up.

On the other hand, solder splashes? LOTS..and LOTS. The trouble with solder is that it can stay fixed to where it settles, but at some point, heat, maybe a jar to the supply when you’re moving or transporting it and now it’s a shorting strip looking for a place to cause havoc.

Take a look at this photo…big thingie with coloured bands is a power resistor. But! What’s that “glint” of silver on the top of it? Solder splash. Get some tweezers and take it off. In some cases, you can flick it off with a finger nail. Just don’t flick it back on the circuit board.

Power Supply Solder Glob

Or look at this example on the last supply I checked…under the heavy jumper wires there is a piece of solder splash. Just. Sitting. There. I removed it with a pair of tweezers and it wasn’t attached to anything.

After which I thought I was done. Right….not! Take a look at the “400.”… that dot after the 400 silkscreen isn’t a dot, it’s a ball of solder. One touch and it moved.

Power Supply Glob 2

Tilt the board and look under any components you can. Loose solder loves to pin itself up against other component leads.


If you currently have a power supply in use and are under the impression that since you’ve been using it for “x” hours, it must be fine. Don’t assume, stop, check it.

The one I showed with the big jumpers on the PCB was one I’d used for over 400 hours. It worked fine. But every now and then the printer would do something odd and I blamed the firmware. So I took it apart. One of those jumpers wasn’t properly soldered. No, I didn’t fix it. I tossed it out.

I doubt these power supplies go through anything more than the briefest cursory glance for quality control. No doubt more about how fast they can be assembled, than can be they be assembled correctly. The old, time is money, problem.

Sadly, the end users, you and I, will end up the loser. Thus if you have one of these supplies, do yourself a favour and have a real good look at it. At least you’ll get a little piece of mind when you start one of those 33 hour print jobs.


3d Hot End Temperature Problem

Ever have one of those “problems” where you spend endless amounts of hours trying to solve it, doing everything thing you can, “By The Book”, only to be stumped at every turn?

Yeah, me neither.

Until 3 days ago.

In the Beginning

Last year we had a new fence put in, and over the winter and the heat this summer, the not so stellar post caps decided to split, curl and look more like pagoda roof lines than post caps.

IMG 1201

Which is not a big deal until you consider than the rain will end up on the post instead of being diverted by the cap.

I spent some time with Tinkercad designing a new post cap, hollow up inside, over hangs the edges of the post so there is no way the rain can land on the post.

IMG 1202

The Design

The design is pretty simple. I mean, it’s a post cap. How complex can it be…

IMG 1228

It prints upside down, top is about 86mm square, base is about 146mm square (when mounted on the post). The sides are 45 degrees so no support. But it is fairly big so it does take some time to print out.

The Ugly

When I started the first print, I was running the perimeters at 30mm/s, the infill (what little I used) was 50mm/s. After all nothing but a bunch of long straight runs.

And then I looked at the temp stability. Or more specifically the lack thereof. Crazy, 10C higher than set point, 12C lower, and oscillate like that for the entire print job. if it settled down it was only for two or three minutes. Then back to high and low sweeps.

I managed to get the print done, decided I’d use the other printer I have. Exact same symptoms. Large temp swings during the print.

Next I loaded up a different print as a test and printed it out. Temp was rock solid. Back to the fence cap. Wicked temp swings again. Do this enough and the result is male baldness.

The Solution Search

Off to the internet I go to see if anyone has the same problems. Only about 500,000 people do. Safety in numbers you know.

So I started reading about possible “solutions”…keyword, “Possible”…

Arduino AREF isn’t stable. Turns out if you have flaky 5V supply line this can happen. Mine is rock steady.

Shield thermistor leads. According to the theory, the PWM signal from the heater element can create “noise” on the thermistor leads. So I stuck my scope on it. My cables are all wrapped in a bundle so you’d think there’d be lots of “noise”. My conclusion after testing was the “noise” was caused by someone not knowing what the heck they’re talking about. In the one case it did fix, turned out it was a bad connection not noise. Moving on.

Heater damaged. As I happen to have spare parts for my printers, I replaced the heating element. No change.

Thermistor. Now we’re getting into the grey area. Damaged thermistor, loose wire, bad Dupont connector, RAMPS problem (should be using RAMBO or some other 32 bit control board), fan blowing on it, wrong setting(s) in firmware, not PID tuned, and broken leads. Except this does not explain why a different design prints with perfect temp and this stupid fence cap doesn’t. None the less, I tested three different thermistors. And got the same huge temp swings.

None the less I made a lot of adjustments to the firmware in the section that controls the PID, temp curve and heating variables. All to no better outcome.

Small design printed better. Fence cap, huge temps swings. No cure.

Then I thought, ah ha! Slic3r. Yeah, no. Made no difference. Sigh…

PLA! Yeah, something wrong with it. Change it! Yeah, did that, no change…ugh.

Nozzle! Yeah, no. Not that either.

Arduino! Swapped out an Arduino MEGA with a 16U2 for the el sleaze bucket ones with a CH340G. Nah. No change.

MUST BE THE RAMPS!!! Yeah, tried two different RAMPS boards. Same temp swings.

I pretty much gave up.

Two Heads, Better than None

I was ready to throw in the towel (along with the hot end and the rest of the printer) when the wife came in to see what I’d been mumbling to myself about.

I showed her the problem with the temp swings, told her about all the changes I’d made. She looked at the test prints, perfect, looked at the fence cap print, not so good.

Told me to start the fence cap print again and she watched the temps and called them out to me…

We let it run for about 3 layers and she said, “Cancel the print job”. I cancelled it, she said, “Okay, re-slice it and slow it down.”. I quickly said, “That won’t do anything, I can print far faster than I have it going anyway and the test prints were printing out fine at 50-60mm/s.

At this point I’d have peed on an electrical outlet if I’d thought it would help. So I set the speeds, infill, perimeter to 20mm/s. Or roughly half the speed of the last ice age.

Started the print, swing started the same and then…yeah. Stable. I was amazed.

So we did three quick tests. Speeds at 20, 25 and 30mm/s. You can see them in the graph below as test 1, 2, 3.

Fence Cap Temp

Time for a Guiness…

The Problem

When you look at a typical hot end, there’s a big chunk of aluminum with a heating element and thermistor in it. If and when you do some PID autotune, the head is stationary. All is right with the world.

When you print, the head moves, that big chunk of metal becomes a mobile heating element. Swing your arm, feel the wind rushing around your hand. Now pretend you’re the FLASH (oh go ahead you know you want to) and make your hand only vibrate. Very little air movement (if time starts to go backwards I suggest you stop…:-)

So the problem is simple. When the head is doing those LONG FAST lines, that heater block is dissipating heat like crazy, so much so the firmware can not maintain a control on it. Like tossing a stove top element around and trying to maintain a stable temp. Not. Going. To. Happen.

The Solution

The obvious solution is that if you’re going to print big stuff where the head is moving fast and long, plan on slowing it down. A lot. If you find big swings and you have a print job running, slow it down.

There are folks who say that a silicon boot on the heater block or wrapping the block in cotton/kapton will help, but I haven’t tried either of those to see if there is any merit there.

I suspect this is why a number of printers are enclosed in boxes so air in the box is warmer, thus easier heat control.

Insulating the head would be easier on a delta. Hence my thinking is it should be able to maintain a stable temp when the head is moving long distances fairly quickly.

Then again, if you read back through the initial problem “solving” list I had in the first place, you might just want to smile, and knowingly nod your head a little bit…

Later that same day…

Good thing I’m not a cat. My nine lives would have been used up years ago with my curiosity…

I decided to do a test by making a sock for the hot end. I used the cotton strips and some Kapton tape to make a boot..The print cooling fans were not used, the only fan on was for the heatsink tower.

IMG 1230

The hot end is wrapped all around except for the back where the wires for the heating element and thermistor are. Kapton tape is holding it all in place.

I loaded up the same fence post cap, set the speed for the perimeter to 30mm/s, infill to 50mm/s. And did three layers.

Insulated Hot End

if you look directly above the 7 min mark, you’ll see the temp drop and thats where the second layer starts. At the very right edge it looks as if the temp is coming under control. Except at that point I’d set the printer to 50% speed.

The cotton insulation does help, a bit. But certainly no where near what I imagined it might. The back of the heating element where the thermistor is no doubt very sensitive to air flowing past it. Thus if the whole end was wrapped up like a mummy, maybe it would be better. Or not.

For now, print at a slower speed, it works.


Gibson Robot Guitar – Revisited

I was in the studio the other day, went to tune up the Gibson Robot and…meh. Nothing. Plugged in the charger, and couple hours later. Nothing.

Swell. Just swell.

Back in 2014 I’d fixed it with a couple of El Cheapo rechargeable batteries I sourced off eBay. Fast forward three years and those batteries were toasted.

Of course this is pretty much standard for NiMH so…no big surprise. And those 2000 to 3000ma rated batteries off eBay, keep in mind those are just NUMBERS. For example I charged up some Eneloop’s, got 1.48V full charged. Did the same with some A3300 batteries (and some unmarked ones) off eBay and the best I got as 1.37V. Maybe you’re thinking 1/10th of a volt isn’t much, but trust me, in this case with a robot, it’s huge (for other reasons).

As it happened someone emailed me within the last week about that original blog post and since I had some time I decided to dive back into the Tronical murky waters.

The Examination

Instead of grabbing another set of junk batteries I decided that I needed to adjust my thinking for “long term”. If the batteries were going to be replaced every three years I might as well make it easier to do.

Next I didn’t want to use cheap batteries. My choice was to go with the Sanyo or Panasonic Eneloop batteries. The problem here is that the batteries don’t have solder tabs on them so one has to solder right on the battery terminal.

I started by looking for a dual AA battery holder than would fit in the confined area of the Gibson. I found a couple of these in my parts bin.


The one I had fit, the second one was wider than the Gibson cut out so no go with it. I soldered up the wires, plugged it in and was rewarded with a flashing sequence of blue LED’s around the MCK ring. I have NO idea what that indicated. But there’s no way the thing would respond or tune. Cool flashy LED’s though….

In trying to figure out why, I originally thought the electronics might have fried itself, I came across the specs for these holders. Specifically 1 AMP output MAX! Any more than that the contacts/wire can heat up. Even with the connection I did have there was a large voltage drop. Mostly they want FAR less than 1 amp to be happy.

Last time I tried to power the Gibson with a bench power supply, it failed miserably. Apparently my new bench supply is of higher quality and it powered up the robot just fine. My goal was to measure the current used at idle and when tuning.

At idle the robot used around 340ma (one third of an amp). That surprised me, I didn’t think it would be that demanding.

Kicking the robot into tune mode, wow. Anywhere from 800ma to 1.6 Amps. Depending on how many tuners were running.

No wonder the batteries were soldered in.

The Options

I could either replace the batteries with some new soldered ones or find a battery holder that would fit.

According to what I could find, Eneloop does make tabbed batteries but I could find no source in North America for them. And really, I wanted to avoid soldering in new batteries if I could.

Which lead me to decide on a battery holder.

Armed with some electrical data, I started looking for “spring less” battery holders. Springless holders typically have large metal tabs that rub against the ends of the battery. The current through these connections can be far higher than the typical brass button and spring units in the previous photo.

Try as I might, all the holders I found weren’t going to fit into the Gibson’s battery area.

So you think that would be the end of it right? Nah.

Unleash the Maker

Back in 2014 I didn’t have a 3D printer. I do now. Couple of them as a matter of fact. I used Tinkercad and started designing a box that would fit in the area. A tray is what I finally found worked best.

But first, I needed some UBER springy metal to make battery tabs with. Off I went to eBay, Banggood, etc and found a few things, but not what I thought I could use. So I started to look around the shop for springy tin and presto. It darned near jumped out at me…


When I was building guitar stomp box pedals, I was using a lot of these jacks on the PCB for connections. Nice springy tin connections. Yep. I cut the plastic off two of them to get four terminals. Bent them so they’d hold a battery.

Then it was off to add in the design to hold these. So the first couple of designs didn’t work but eventually I got a design that was going to work, at least good enough to test with.

IMG 1195

I couldn’t solder the wires on the tabs when they were in the battery housing I’d made or the PLA would melt. I decided to put the wiring on the bottom of the holder. Didn’t really matter any way.

IMG 1196

I used two terminals with a 16 gauge jumper wire. Took some wrangling to get it in but I got it.

IMG 1197

From the battery side it looks like this:

IMG 1198

Oddly enough, there is a LOT of tension in those little spring clips. So once the batteries are inserted you can see the force of them pressing against the ends.

IMG 1199

I finally put the whole thing back in the Gibson and it works absolutely perfect. As good as or better than the original Tronical design. When I plugged in the charger, it sync’d right up with the guitar and quickly charged the batteries to full.

IMG 1200


Is this perfect? Will it stand up? I don’t know.

I do know the batteries are wedged in the holder pretty darn good BUT, if they do come loose, I’ll put a zip tie around them and the holder so there’s no way they can budge.

Looking at the clips, I can also see another way to do this. A couple of M3 screws, some terminals, no spring clips needed. But then again, this might not work as good as the spring terminals…

All I know is that when the batteries go again in three years, I’ll be putting in another set of Panasonic Eneloop’s and it’s going to take me about two minutes to do it.