Fun with Pi

There have been a number of interesting things happening in the Raspberry Pi community lately.

One new hotness is the RPi 4 with 8Gig of RAM, which in addition to having more RAM than you will likely ever need on a machine like this, also has a slightly updated power module that improves the behavior with some USB-C power supplies. It does mean that you can now open too many Chromium tabs to keep track of on this platform too. 😛

Another interesting update is the support for booting off of USB drives, instead of the MicroSD card. This lets you use an SSD over USB3 as your main drive, which is both much faster than the built-in MicroSD, and will have a significantly longer lifespan.

Finally, there is the 64-bit version of Raspian (now renamed as Raspberry Pi OS), which actually matters given you can get a Pi with more than 32bits of address space, but it will likely also improve performance for many tasks once the beta kinks are ironed out.

Anyway, being the RPi geek that I am, here is my latest desktop…

An 8Gig RPi4 with a 500Gig Samsung SSD, in the awesome Argon ONE Pi 4 case, which has both a temperature controlled fan (that is almost always off under normal workloads) and, believe it or not, a working power button that does a safe shutdown! This is definitely the slickest Pi case I’ve used, even if moving all the ports to the back meant making it nearly 50% larger than the Flirc case, which was my previous favorite.

Anyway, even with an SSD that cost more than the Pi and the Argon ONE case together, and the Pi overclocked to 2GHz, it’s still too slow to be a great desktop experience, but it works well enough that I was able to write this post, while watching youtube videos, with several other apps open, including LibreOffice and Gimp, so it’s definitely real.

Google was wrong…

no one wants a pure web laptop.

As someone who has been using Chromebooks for a while…

… I have always tried to push the limits of what you can do with them. More often than not, the limits I hit were caused by one of the core tenets of the platform, that everything you need to do will be done with web apps. It turns out, this was never true, and it’s looking increasingly less likely that it is going to ever become true.

I read a web article a while back called “Is Chrome OS right for you? A 3-question quiz to find out“. The first two questions in the author’s quiz are:

  1. Do you spend most of your time using the web and web-centric services?
  2. Do you have specific local programs that you absolutely need, or could most of the things you do on a computer be accomplished with web-centric equivalents — along with Android and/or Linux apps to fill in any gaps?

At some level, these two questions capture both the history, and the problem with Chromebooks. When Chromebooks were first being developed, the notion that we might end up doing everything on the web was aspirational at best and, more cynically, was an attempt by Google to force people onto a platform they “controlled”. It rapidly became clear that the under-powered, first generation devices that were being built were too limited for day-to-day use, and as the complexity of applications that were being built on the web increased, they couldn’t even handle those web apps well.

Even more telling is that, even if we believed a web-based future would eventually come, what we have actually seen is the world going in a different direction: It’s not web apps that people run, but rather mobile apps. Question two above suggested we would use Android apps to “fill in any gaps”, but it’s Android (or iOS) apps that people typically use now, even if there are web app alternatives.

So what happened? Google added support for Android apps to Chrome, and Apple is in the process of doing the same on their platform. Want proof that it’s a mobile world? Just go take a look at the relative sizes of the Google Play Store and the Chrome Web Store. Go ahead, I’ll wait…

And the funny think about all this, is that yes, Chromebooks are useful now, because they can run the applications people are using, and have large screens and input hardware that makes using those applications easier than running them on phones. Oh well.

Sony WF-1000XM3

I have been a wireless headphone geek for a long time. To give you some idea how bad it is, I have all of:

  • Beats Solo3
  • Marley Smile Jamaica
  • Trekz Titanium (bone conduction)
  • Plantronix Backbeat Fit
  • Soundcore Liberty Air
  • Corsair Void Pro (times 2)
  • Steelseries Arctis 7

… and that’s only the ones that I still currently use.

When the Apple AirPods Pro and Sony WF-1000XM3 came out, I tried them both and decided to pick up the Sony’s based on a perceived slightly better sound quality. Since then, I’ve seen several reviews of them both, with some level of agreement on the sound, and indications that the noise cancelling is also better.

However, my overall experience with the Sony’s has been quite poor, to the point that in retrospect I’m confident the AirPods would have been the better answer. Honestly, the UX on the WF-1000XM3 is so bad that I can’t help but think Steve Jobs would have fired any team that came forward with something in the same state.

The first issue is the companion app which, despite several updates since I first ran it, is still frequently unable to connect to the headphones, or drops the connection at some point after it gets connected, or simply crashes completely. It also includes something called “Adaptive Sound Control”, which has the following Engrish description:

The app detects your actions and [Ambient Sound Control] is switched.

Whatever this is supposed to do, the behavior manages to be simultaneously be both intrusive and not useful. I turned it off almost immediately.

The headphones themselves are also unreliable. Sometimes when you take them out of the charge case, they don’t BT connect at all. Sometimes they connect (i.e. show up as connected in the BT device list) but don’t become available as headphones (doh!). If they do get connected, the connection will later sometimes drop for one or both of them.

And speaking of “one or both”, the strangest thing about them is the way they start up. They behave a bit like they are two different bluetooth devices that notice while they are connecting that they can work together. The start up sequence has three voice prompts:

  1. Indicate they are powered on (“Power on.”)
  2. Indicate the current battery level (“Battery fully charged.”)
  3. Indicate they are connected (“Bluetooth connected.”)

However, because they sync up part way through the start up, you will hear somewhere between zero and three of those messages in each ear, and by that I mean a *different* number of messages in each ear. It’s a terrible experience.

[Aside: Hey Sony, there is absolutely no reason to ever say “Power on.”. If you say any of the other prompts, we know the power is on. 😉 ]

One final comment: the WF-1000XM3 — btw, who thought that was a good name for *anything*? — are quite a bit heavier on the ear than, for example, the SoundCore Liberty Airs. I didn’t realize how much of a difference this was until I used the Airs again recently. I now find myself going back to the Airs often, even though the WF-1000XM3 have clearly better sound.

Rebuilding an old MBP

I had an old, broken 2009 MacBook Pro sitting on a shelf because I had never gotten around to trying to fix it. It had three significant issues:

  1. It randomly kernel panicked every once in a while
  2. It had an SSD drive in it that no longer worked
  3. Three of the keys on the keyboard didn’t work.

Last weekend, I finally decided to see if I could get it into working shape again. To start, I took the back off and removed the failed SSD drive. The MBP is too old to run a current version of MacOS, but it still has a working HD drive, so I decided to install Linux there.

To fix the keys, I tried removing the key caps and cleaning the switches, but after putting them back together they still failed. By chance however, I noticed that if I *smashed* the key beside one of the failed ones (in frustration? 🙂 ) then the original keys would start working for a bit. This led me to pulling that other key cap off and cleaning it, and after putting it back together all the keys worked.

This only left me with the random kernel panics, so I booted the Mac into hardware test mode and let it do a full test. Even after 10 minutes of hammering RAM and CPU (with the fans screaming), there were no failures, so I chalked the problem up to either a Mac driver issue, or something to do with the now removed SSD.

For linux, I just went with Ubuntu, mostly because there was a tutorial about installing on Macs that seemed pretty reasonable. The only wrinkle was the lack of a driver for the MBP’s discrete graphics card. I went for the easy option of doing nomodeset in grub, and ended up with…

I don’t think it’s worth putting a new battery in it, but if I needed a home computer for email and web browsing this would be perfectly useful. Nice!

I had a camera.

When the Micro Four Thirds mirrorless digital cameras first came out, I decided I was going to take the plunge and get myself a “good” camera. The one I ended up buying was an Olympus E-PL1. I guess camera geeks recognize that model as not being great, but at the time, it took what I thought were pretty amazing photos.

Over the years though, I didn’t really spend as much time as I thought I would taking pictures, and once cell phones started to have decent cameras, the E-PL1 got relegated to a shelf, except when Deb was using it to do product shots for Sheep’s Ahoy.

Today when I was cleaning up my stash of old USB cables — a bigger task than it might seem; I ended throwing out close to 50(!) — I actually found the camera’s custom USB cable, which had been missing for years. This sparked my interest in the old beast enough to recharge it’s battery and take some shots, to see how it compares to my iPhone’s camera.

I won’t keep you in suspense: The iPhone 8P’s photos were better. I thought that the larger sensor in the E-PL1 might let it take better low light photos, but the images were just as noisy as the iPhone’s, if not worse.

As a quick comparison, here’s one representative photo from each:

By just about any measure, the iPhone picture is better: warmer, less noise, sharper. Oh well. And Deb’s XR takes even better pictures than the 8P! I can’t even say that the E-PL1 had better manual shooting features, since at least for my amateur needs Halide does just as much, and I could even get additional lenses if I wanted to.

Don’t get me wrong. I realize there are still digital cameras out there that take much better photos than any cell phone. They’re just not priced to make sense for anyone who isn’t going to make a serious commitment to photography (like my friend John). I’m happy with the pictures I take on my phone, and I’m sure when I upgrade to the iPhone 12 Pro (or whatever it ends up being called) they’ll be even better. 😉

Who says the iPad isn’t a laptop replacement?

Imagine you’re sitting in a coffee shop, and you suddenly feel the need to do some coding. Your iPad has some decent code editors, but you’d like to execute what you wrote. How about connecting a Raspberry Pi and running it there?

The above is my new fave home-away setup:

  • 11″ iPad Pro
  • Brydge keyboard
  • Magic Mouse
  • Raspberry Pi 4 / 4Gig

What makes this work is that the Pi is both powered and gets a network connection over the USB-C port. To enable that, all you have to do on recent versions of Raspian is:

  • Add dtoverlay=dwc2 to /boot/config.txt
  • Add modules-load=dwc2,g_ether to /boot/cmdline.txt
  • Reboot

Assuming you’ve got SSH on your RPi, at this point you should be able to connect it directly to the iPad with a USB-C to USB-C cable, then use your favourite iOS SSH client to do the equivalent of ssh pi@hostname-of-pi.local, with “hostname-of-pi” being whatever you chose (“raspberrypi” by default). Usually, I use Remoter VNC for SSH but that’s mostly because I’ve owned the full version of it forever.

If you’re astute, you’ll have noticed that it’s not ssh running on the screen above. 🙂 There are many ways to get an actual GUI desktop attached to the RPi, but the one I’ve found works the best for me is (believe it or not) using Windows Remote Desktop. To enable this, on the RPi, you can sudo apt-get install xrdp, and on the iPad install Microsoft Remote Desktop or whatever RDP client you like best. When you create the connection, use hostname-of-pi.local just like you would have for SSH.

There are a couple of downsides to this setup.

  1. The microsoft remote desktop client, at least, is tuned for using a finger on the screen, rather than a mouse. In fact, support for mice on iPad OS is still in its infancy, though I expect it will improve over time.
  2. The RPi is being powered from the iPad, so you can expect the battery life to be reduced — it’s not impossibly bad, but it’s visible. (Btw, if anyone knows of a good USB-C hub that will power multiple devices with a data passthrough, let me know.)

Anyway, if you’re looking to try this, there are lots of good tutorials out there on YouTube and elsewhere, by people who have invested more effort than me. Regardless, I’m happy to try and help if you get stuck, so leave a comment.

Overclocking

In my post about the new home of GCW, I had a picture of the Raspberry Pi based version of the server. What probably stood out the most in that picture was the cool (sic!) ICE Tower active cooling system. That Pi has been rock solid since I turned it on, so given the insane cooler, I thought I’d try playing around with a bit o’ the old overclocking.

The stock CPU and graphics clock speeds on a RPi 4 are 1.5 GHz and 400 MHz respectively. In order to get the CPU clock past 1.75 GHz, you need to install firmware that is in advance of what comes with the standard Raspbian distro, so I figured I’d start with 1.75 GHz. For the graphics clock, I tried 600 MHz, which I’ve seen in a couple of how-to videos.

With those speeds, the Pi booted just fine, but after some heavy use (i.e. 3 simultaneous video playbacks) the graphics became unstable. I backed the graphics clock off to 550 MHz and the performance stablized.

With the ICE Tower running, the core temp never went above 40 degrees Celcius even under the heaviest load — compare that to the Pi thermal throttling at 80 degrees showing a single video with no cooling.

Overall, I’m quite pleased with the result. The combined boosts in clock speeds have made an immediately noticeable improvement in performance. It’s entirely possible that this has reduced the lifespan of the board, but honestly at the price I paid for it, I can live with that.