Coding

DIY Standing Desk

If you work at your desk all day without getting up and moving around regularly, you’re not doing your body any favors! In the last couple years, I’ve tried moving around more while on conference calls and sitting in different locations (which is really easy at home, but in our new collaboration spaces at work it’s super easy too). Last year, while in the office, I used one of the sit/stand desks. I was surprised that after using it for an hour in stand position, I was still standing without a desire to sit (from my feet or my brain). I felt more focused standing, and when my feet and back told me it was time for the chair, I did it long enough to feel rested and did more standing. I didn’t like that the sit/stand desk had no keyboard tray, but it did go pretty low, so it wasn’t an ergonomic disaster-just not quite optimal.

The next day, I started researching buying a desk for home and found it’s pretty easy to spend between $600-900. I’m a person that always likes to make things my own, whether that involves customizing them or building them. I started looking for options to build a desk from a frame, and it turns out a few companies offer their desks without the top for people like me. I searched quite a bit to find a desk that could handle a decent weight, had good reviews online, and more than a 90-day warranty. After a lot of searching, I found a deal on Amazon. It was the… It was on sale for $279 and had an extra 20% cash back. I ordered it Friday and it appeared Sunday morning! I wasn’t ready with a top, so I cut a bit off a piece of plywood in my leftover stack in the garage to use in the meantime.

IMG_1508

IKEA Top with four coats of wipe-on poly

For the top, I wanted to go big. My current desk in my home office was a standard computer desk, about 4 feet wide and 20 or so inches deep. Most the ones online offered a 30×60 size which sounded amazing. I wanted a nice clean look – something with some nice grain patterns, not dark. I found my local IKEA store sold an unfinished solid wood butcher block top. I put three coats of poly on it to make the wood colors pop without a glassy finish. It was simple to mount (like < 30 minutes) and then I just had to attach the controls to the front of the desk. I found a great standing pad that I stick under the desk when standing and it’s worked out great. I found generally I like to stand 2-4 times a day for 45-60 minutes. I think that’s a great compromise for an information worker and I think the look turned out great.

IMG_1509

Since the picture above, I 3D printed some cable management guides, mounted the power supply, and printed an under mount hook for my over the ear headphones (I call them the focus-phones)

Here are the parts I went with if you want to follow my path. The total for everything was under $400.

Raspberry Pi – Under Voltage or Temperature Issues

IMG_0423Step number one for ANY Raspberry Pi issue is “check the power supply”! Many issues on the Pi come from improper power; SD card corruption is a popular one. Often, people, including me, try to use an adapter with a Micro USB cable. Even if the adapter is rated for 5+ volts and 2.5A (for Pi 3), the quality or length of the cable might limit the power it can draw. For this reason, some people recommend a corded power supply, typically one from a “trusted source” like Adafruit. Despite using the “right” power supply on my two Model 3B devices, I was seeing under-voltage warnings (rainbow block) and at one point, a temperature warning. One was using an Adafruit supply and the other an iPad supply (both higher rated and with a nice thick cord). Both are in a cheap case I got from MicroCenter.

The software I was using had a menu system with some video on the screen, and it was consuming about 50–60% of a CPU core. I was seeing temps in the 77-79 range with some CPU throttling. It was curious to me that this activity was drawing so much on the power supply that I’d get an under-voltage warning. I removed all USB devices to be sure nothing other than the Pi was using the power. Even when I swapped the SD card, Pi, and power supply, it was pretty predictable that after about 15 minutes of use, I’d see the rainbow block.

I decided to run sysbench to see if a much higher CPU utilization (maxing out all four cores) would generate much more heat. Interestingly, while it did heat up more quickly, the top temps were about 83-85 with some significant throttling to keep the temps at bay. Again, the under-voltage warnings appeared. To me, it seemed like the heat was triggering the under-voltage warnings.

Sysbench:

sysbench –num-threads=8—test=cpu—cpu-max-prime=10000000000 run

Temp/Throttle Monitoring:

watch ‘(vcgencmd measure_temp; vcgencmd measure_clock arm) ‘

Next, I ran a test at a much lower CPU utilization, but covered up the case to build up the heat, and after about 10 minutes, I got the rainbow block again. I wouldn’t think these minor changes in temperature would affect the voltage draw, especially given the throttling. I picked up a pair of cheap stick-on “Addacore” CPU stick-on heat sinks (yeah, I said stick-on). I repeated the same experiments and they clearly did their job. I saw a consistent drop in temps of about 4-5 degrees, which is more than enough to prevent the throttling and the under-voltage warning (which I’m not sure I believe). Even the system bench test had much lower temps and less throttle.

OK, it’s 3M tape. I would not mount this Pi vertically because I suppose if it got warm enough gravity might have its way with this adhesive and ruin the board, but it’s working great. No more high temperatures or under-voltage warnings. While I’m tempted to remove one of them and repeat the experiment and measure the voltage from the Pi with a meter, I’m going to leave well enough alone.

Anyway, if you are seeing rainbow blocks, check your power, but also check your temps and try a heat sink:)

Coding to Kaltura API using “testme” Console

I’m probably too easily impressed anytime a vendor shows me a web “playground” to experiment with their API, especially if it will generate code snippets for specific languages. IBM has a pretty neat tool for IBM Connections, but it requires too much setup/footprint in my opinion. Kaltura has something similar, but has the advantage of needing just a client library. I needed to get a list of owners and managers for channels in the KMS and it seemed like for a few calls, the test console could give me the basic code.

I started with JavaScript and ran into a lot of trouble, which I chalked up to my proxy for the request. I figured server-side Java would be a better choice since I wouldn’t have to deal with the cross-site scripting and it’s a language I know well. With the Kaltura API, you start by creating a “session” (also called a “ks” for Kaltura session). You can do this with your secret admin code, but it’s far better to do it with a user that has the privileges you need. I generated some of the basic code and pasted it into a runnable class.

Kaltura Generated Code

KalturaConfiguration config = new KalturaConfiguration();
config.setPartnerId(partnerId);
config.setEndpoint("https://www.kaltura.com/");
KalturaClient client = new KalturaClient(config);
String loginId = "userid";
String password = "password";
int partnerId = 999999;
int expiry = null;
String privileges = null;
Object result = client.getUserService().loginbyloginid(loginId, password, partnerId, expiry, privileges);

Immediately, I noticed some interesting things like assigning primitives to null and declaring variables after they are used, which are easily fixed, but the issues don’t end there. The return types are not usable (all objects) and the API calls don’t match the API you downloaded from the test application. Other code examples were worse-variables were declared when all that was needed was a simple assignment to something already created.

In hindsight, I expect it wasn’t my proxy or JavaScript skills, but similar code generation issues and API mismatches that were my downfall with JavaScript. My advice is to try the console to know what classes you’ll need and get a “hint” of what data is passed around-assume the code you get is just plain wrong. I posted my experience in the forums and the response was that Java isn’t as well supported as PHP and it’s just a guide to the type of code you might write. They have a developer site with code examples, but nothing in Java yet.

In case you were hoping to find some working sample code, I’ve attached code to create a session, get a list of categories, and print out all the users with manager access in case you’d find a complete code example useful.

Download ListChannelOwners Java Class

Diagnosing problems with First Robotics Simulator (FRCSim)

WPI in conjunction with the Open Source Robotics Foundation has developed a 3-D simulation tool for FRC teams coding their robots in C++ or Java with WPILib. It’s pretty much limited to the sample robots you get with the development package right now but it’s great for students to try out the code or use the command framework to build some new commands. This exposure to the API doesn’t require a robot – which in the first few weeks of the project is a great thing. [more details]

The underlying tool for the FRCSim, Gazebo, isn’t very friendly about telling you what the problem might be before it dies an ugly death. Most likely the problem is you’re seeing errors about the “Gazebo Transport” in Eclipse since it’s unable to connect to the dead Gazebo server. I’ve documented some steps to try and determine the cause of your problem.

1) Make sure you’re following the latest steps for running the tool and have a proper updated Ubuntu installation. You need to be on the 64-bit version of the OS. To verify this try typing the following on the command:

> uname output

You should see X86_64 in the output. If you don’t your machine isn’t compatible or you installed the 32bit OS version.

2) Try running the following command to launch the Gazebo server directly and look for any human readable errors in the log:
> frcsim -u –verbose /usr/share/frcsim/worlds/GearsBotDemo.world

You can look at the log files in ~/.gazebo for any useful error information. It will be cryptic but maybe you’ll find something useful. I wouldn’t spend a lot of time here but maybe you’ll see some obvious failure.

3) Determine if your card supports direct rendering. Try the following command. If you see output of “direct rendering: Yes” that’s good.
> glxinfo | grep “direct rendering”

If you don’t see the right output see step below.

4) If you’re still stuck your video card may not be supported. If you’re using an integrated controller you’ll likely need to try an ATI or nVidia dedicated card. I tried 2 linux laptops and a desktop all with integrated controllers from a variety of vendors (with correct support for direct rendering). Only after updating the desktop with a super cheap $30 nVideo card was I able to get the simulator to run – and runs pretty well from limited testing.

If you’re stuck hopefully some of the stuff above helps out. Thanks to Alex Henning for the tips!

leJOS with Eclipse on Mac with 32 bit

While trying to compile applications for the Mindstorm NXT with a modern version of Eclipse on a Mac you’ll likely run into the deadly error, “Error: This Java instance does not support a 32-bit JVM.”. The need is clear – you need 32 bit Java support – but how? I tried various JVM flags (-d32) and trying to change where Eclipse found Java but no luck! Below I’ve detailed what worked for me.

The first step is to figure out if you have the older 1.6 version of Java on your Mac. For this just look in the directory /System/Library/Java/JavaVirtualMachines/. If you’ve had your Mac for a while you likely have the directory “1.6.0.jdk” and you’re ready to move on to the next step. If you don’t you can find it here.

Changing Eclipse to use the JDK was tricky until I found the RIGHT way, then it was easy of course. Go to /Applications/Eclipse/Eclipse.app in Finder. Right click and choose “Show Package Contents”. Next open Contents/Info.plist with a text editor (not plist editor). You’ll notice at the bottom there are actually instructions on how to launch Eclipse with a different VM. I simply stole the commented out line below and pasted it under Eclipse/array.

<string>-vm</string><string>/System/Library/Java/JavaVirtualMachines/1.6.0.jdk/Contents/Commands/java</string>

All is working now! If you’re trying to get Eclipse on your Mac working with leJOS hope that helps!