Old Project Progress Pictures: The Laser Fist of Heaven

I've built up quite a pile of project pics I took meaning to post, then never got around to it.

OK, so, laser cutter.

China sells a lot of cheap cheap cheap "It's so cheap I can't believe that price covers shipping let alone materials" maker tools. Some are better than others, and many barely work for their intended purpose out of the box. Overall, plan to get what you pay for, and I mean that in both the good and bad senses.

Most of these tools make decent starting kits though! Research a bit, then choose the kit with the most parts you want to keep (or the fewest you plan to toss).

I bought a K40 variant (SL320) on AliExpress, and I'm building it into a complete cutter now. It has a decent X/Y stage, came with a good tube (which I promptly broke while being an idiot), a larger than normal case, tubing, flow sensor, air assist head, and a better than average power supply. It also came with an exhaust fan and aquarium pump I don't plan to use, a useless Z-stage, and an air pump that makes half of the ideal.

Others are already doing a good, detailed job of documenting their K40 builds. I've been referring to Dons Things and Tims Machines extensively for tips in my own build. So, I'm going to stick to pictures, terse captions, and the few things I've uncovered that I haven't seen solved elsewhere.

I don't want my water cooling tubing dangling out back to an open bucket and a failure-prone aquarium pump, and I'm pretty sure an external chiller is overkill. I want a fully internal, closed water cooling loop. Water cooling may be useless (if still cool) for PCs these days (ha ha, see what I did there), but it does mean nice cooling parts are cheap and plentiful for other uses.

So, step 1: shoehorn in the biggest possible radiator core. And by biggest possible, I mean, "move everything else as far out of the way as I can to make more room". That starts with lifting the laser tube tray up by about 2cm. The original mounting flange is just tack-welded on. It's easy to pop it right off.

The limitation to how much higher I can go is actually the clearance of the mirrors along the left side, and the XY stage fitting under the door opening lip in the front. You can see I also nibbled off about half the depth of the left side of the door opening lip.

Aside: freehand nibbling is really hard to get straight.


George makes it his own

Last summer I bought a house in New Hampshire for living as well as greatly expanded workshop space. The kids are up here about half the time, and it's a holiday getaway for Camilla as well. I've been moving in slowly since July (Priusful by Priusful, with the occasional Minivan Marathon) and finishing the unfinished space inside one room at a time.

A few weeks ago, George apparently claimed one of the as-yet-unused rooms for his own workshop. (The other side says 'Do Not Disturb').

Tags: ,


Shhh sh sh shhhh..... I'm making..... TOAST!

It's an old pic at this point (taken while TD, aka, @enginetankard was still here in January) but I can't mention 'laser cutter progress' without following up at least a little.

This was one of the first test runs, and it revealed a problem with laser arc stability. More about that soon, because I think quite a number of K40/D40/SL3020 users are putting up with it, unaware of the cause or the fix.



Oh right, about those LEDs

spoiler: Nichia Rocks. But first a flashback.

Remember those crazy Chinese LEDs I was looking for? They had kind of iffy construction, but up to that point they were the closest color match I could find to what I needed for my Thinkpad backlight project. I found more of the same type from random Chinese resellers, but none with exacly the same whitepoint as the original. I never determined the manufacturer.

I was looking for these dodgy Chinese LEDs because I couldn't get any major manufacturers or resellers to sell me the LEDs I needed in any kind of reasonable quantity. The smallest amount any factory quoted was a MOQ of 100,000 and that was only because a friend of a friend was willing to call in a favor to an unnamed factory in Hebei. The name brands you'd recognize weren't willing to discuss anything less than 1,000,000 units.

Or so I thought. I'd stopped looking slightly too early.

The AFFS LCDs I'm retrofitting have red and green filters with a lot of overlap in the yellow region because the original CCFL backlight doesn't put out much yellow. The broad yellow peak from typical white LEDs pollutes the red and green primaries. That messes up the saturation and color reproduction, even if the white point is correct.

But last year Nichia released a new line of component phosphor LEDs specifically designed for backlighting that don't use a broad yellow phosphor; they use separate red and green phosphors.

The specs also claimed tight binning, surprisingly high lumens/watt, and a weirdly low forward voltage. Oh, and they came in a low-profile 3014 package rated for mid-power output.

With no great optimism, I called up Nichia USA and asked for some samples. After being redirected a few times, I talked to a nice lady in Detroit who was happy to send me some LEDs for testing. That was farther than I expected to get.

The LED samples were everything the specs claimed them to be. I had never tested an LED as efficient, or with color binning so tight. They blew everything else out of the water.

So I called back. "What was the minimum order?" The answer: 1 reel, only 5,000 LEDs. Too good to be true.

The next question I was sure would undo me: "Can you tell me if I order now, what kind of bin I'll get?" I was still expecting to get something in the rough ballpark of what I really wanted that I could then tweak a bit with a filter. The reply: "What do you mean? You can have any bin mentioned on the spec sheet."

I could select specific whitepoint, forward voltage, and luminosity bin. At only 5,000 quantity. Jaw on floor.

It still felt too good to be true, but Nichia made the LEDs on-demand in Japan when I placed the order, and they arrived in the mail 15 calendar days later.

So as it turns out, I was able to get my dream LED for this silly little custom project. I'm reminded to post this now as there's been steady demand for the resulting Thinkpad backlight retrofit kits, and it's time to order another reel.

Tags: ,


Notes to myself: [Re]discovering/correcting a ThinkPad supervisor password crack

You can bypass and remove the supervisor password from ThinkPads by shorting two pins on an EEPROM at the right time during boot.

Don't believe it? I didn't either; it never worked for me. It turns out that's only because the contemporary instructions for how to do it are wrong, or rather, they've mutated into a form that only works on some machines. As originally discovered, the hack reliably unlocks any* ThinkPad up to and including the Ivy Bridge models.

*The SL300, SL400, SL500, and G550 do not store the supervisor password in EEPROM, so they're the only exceptions of which I'm aware.

DISCLAIMER: Any discussion of how to crack security on even vintage machines is banned on most ThinkPad forums. So much as mentioning this page can get you banned in some places.

ThinkPads have three types of security passwords.

A 'Supervisor' password (SVP) locks access to portions of the BIOS setup. The machine will still boot to OS., but if the CMOS battery is temporarily removed or dies, the SVP will also lock all power-on access to the machine. It's stored encrypted in a non-volatile EEPROM along with other asset information.

The 'Hard Drive' password protects access to the hard drive. The drive will be physically inaccessible to any machine without the password. [I don't address the Hard Drive password at all here.]

Lastly, a 'Power On' aka 'User' password (UP) locks boot access the machine. It's stored in volatile CMOS settings. Disconnecting the round yellow CMOS battery clears it.

A machine with a User password will almost certainly also have a Supervisor password. Clearing the UP by disconnecting the CMOS battery will cause the machine to demand the SVP at next boot. The password prompt icons for the UP and SVP are the same on older machines, so if it seems like clearing the User password didn't work, it did-- but now the machine is asking for the Supervisor password instead.


Step 1: You might as well make life easier by disconnecting the CMOS battery for a few seconds to clear any user password that may be present. Count to 15, reconnect the battery and proceed...

Step 2: Find the physical EEPROM that holds the supervisor password. These EEPROMs have varying numbers of pins. A few examples: On the T23, the chip is a 14-pin Atmel 28RF08. On the X31, the EEPROM is an 8-pin Atmel AT28RF08. On an X61, it's an 8-pin Philips 24S08. On a T60 it's a 40-pin Atmel 8356908.

On some machines you can access the EEPROM just by lifting the keyboard. On other machines, you'll have to disassemble the entire laptop. Locations of many ThinkPad EEPROM chips can be found at http://www.ja.axxs.net/eeprom_location.htm if you don't want to hunt for it.

Step 3: Locate the PROT and GND pins. These are the pins we'll short to subvert the ThinkPad's access to the supervisor password. On a 14-pin Atmel 28RF08, these are pins 5 and 6. On (I think all) 8-pin EEPROMs, these are pins 3 and 4. On the 8356908, it's pins 31 and 32 (pin 32 is actually the write-protect pin, but it will be grounded).

Personally I solder small-gauge wires to the pins to avoid any fumbling around while bypassing the password. If the chip is easily accessible, you can just short the pins with tweezers or a precision screwdriver. Either way, be careful not to short anything you don't mean to short.

[Note: I know I'm telling you to use different pins than virtually all the other instructions on the web. When this hack was first discovered, descriptions simply listed pin numbers for a few models, but it's clear from context that the intended pins were PROT and GND.

At some point, for some reason, reposts started saying to short SCL and SDA instead, possibly due to mixing up pins between 8 and 14 pin EEPROM variants. The mistake was probably cemented by the fact that shorting SCL and SDA does work on some models.]

Step 4: Boot the machine. Upon the 'ThinkPad' logo boot screen appearing, immediately short the pins and press the appropriate key to enter the BIOS (F1 on most models). The BIOS may take longer to appear than normal, and the machine will throw one or more errors; that's expected. It may also ask you to press F1 again (do so). Keep the pins shorted until you're in the BIOS screen, then release the short.

Step 5: Navigate to the BIOS 'Security' menu, then arrow down to the 'Supervisor password' selection, which should currently read 'Enabled'.

Step 6: Short the pins again. While holding the short, press Enter to select the Supervisor password entry, then release the short. It should not ask for the old password, only to enter a new password twice. After releasing the short, press Enter twice to disable the password.

Step 7: Press F10 to save and exit.

Step 8: Before reassembling or desoldering any temporary wires, reboot and test that you can enter the BIOS without any password.




An image forwarding test. With a duck.

Too slow! The fish already ate the duck.


King Creeper and his minion, George

For Halloween this year, both kids wanted Minecraft costumes. Denton went as King Creeper (his usual Minecraft skin), and George wanted to be a regular creeper as his minion.

The costumes got mostly finished this year; I ran out of time while painting the pixels on the shells. But a good time was had by all, and the kids even got to attack a Steve while trick-or-treating. Which is to say, they stood there and hissed at him while 'Steve's parents admonished him not to whack my kids with his sword (much appreciated).


Halloween 2: The Cleanuppening

That's much better.


Happy Belated Halloween!

Things went off better than the mess would indicate. Stay tuned for _Halloween 2: The Clean-uppening_, coming as soon as my left index finger heals well enough that I can type normally again.


Think DSP

A few months ago, Professor Allen Downey of Olin College wrote to me about a DSP textbook he's working on at think-dsp.com. He sent me links to his chapter about the sampling theorem, plus a blog post with an iPython notebook for playing with sampling a bit.

[I've been meaning to write back ever since, but I royally suck at email. So let me blog about it first, then perhaps extend the minimal courtesy of actually writing back.]

As he puts it, his approach is computational. Where many books on sampling and DSP in general throw a lot of dry equations at you without explaining much practical context (and I take the opposite approach, providing intuition and context, leaving the math out almost completely), he explores the practical arithmetic that gets us from point A to point B. "What happens if I do this..."

It's a slightly sneaky approach that builds up toward sampling through AM modulation (yes, as in AM radio), but this works toward a great illustration of how the Nyquist rate, the sampling theorem, and time-frequency duality aren't just sideshow theory, or a bag of cute tricks. They're practical truth that underlies nearly everything we do with signals and audio, even when we're neither sampling, nor working digitally.

Tags: , ,


Log in