Neal Wilson ✅ @neal@social.typica.us

Roasted a few batches of coffee with the new debug labels and that does seem to be opening up both a new avenue of investigation and gives me a little more confidence that I can use math to fix the problem I'm looking into. There's enough extra compute capability that I can run the original algorithm and what I think might work better at the same time and maybe I can just temporarily write a CSV with the output from both to verify that I'm getting the improvements I expect are possible.

Star Trek Voyager is a magical girl show. See the doctor's transformation when he becomes the emergency command hologram.

In last night's dream I got home to find a couple people working on setting up an eclectic lab of 80s/90s Macs from an alternate reality in which Apple stuck with and scaled up the boxy all in one design of the Classic. Several of the machines had been personalized with many stickers. One had a big mouse basket mounted to the side. There was also an all black one called the Mac VR with two pairs of glasses.

I think I've figured out how to express what should be a more robust approach as well, so maybe I'll let the device run and display the results from both algorithms and see if mine really is an improvement. If it is, I can upstream that. If it isn't, well, I'll have learned something.

Added a stack of debug labels off to the side of the screen so I can see the data that's getting used for current value calculations and while I still don't like the algorithm that's being used, I'm less confident that it's the source of the problem I've been noticing now that I understand the data flow a little more fully. Still, I'll take a look at that on my next roasting session and see if I notice anything.

Apparently the postal service has a loyalty program now?

Doing that cross checking, with the latest hardware and software updates most of the batches are checking out well, but the instrument seems to be biased toward producing a darker measurement. There's a suspicious operation in the code on the instrument that I think could be responsible for this, but I'd like to add some additional debug capabilities to look specifically at how the values in one array change during real batches and see if my intuition on that problem pans out.

This doesn't check that the readings are in any way accurate (that's what cross actual roasted coffees against something that's not a prototype is for), just that the sending and receiving systems agree on what the signal represents.

My software displays the measurements at a higher precision while the instrument generating the signal rounds its display to the nearest integer while still outputting the full precision measurement, but everything was within rounding distance so I'm going to just call that good.

Yesterday I was out buying more cleaning supplies so I also picked up a bunch of those cards that you can take to see how different paint colors might look under your lighting in assorted shades of brown. These were used to check that my software is producing the same readings as what the roaster cam is generating (hold color in front of camera, wait for reading to stabilize, compare the readings on the two displays).

Call the FCC on C-SPAN for the obscenity that is the U.S. Senate.

Sadly, Linux support for one of the data acquisition devices in use is only for one particularly ancient version of RedHat and another device has no official Linux support (but it does have technically incorrect protocol documentation) and I'd rather not rip out well tuned data acquisition hardware just to get rid of Windows.

Windows 10: Hey, I know you're trying to roast coffee right now, which I think makes this the perfect time to make you stop working so I can install my updates.
Me: Remind me tomorrow.
*20 minutes later*
Windows 10: So, about that update...
Me: Tomorrow was your idea. What part of that don't you understand?
*20 minutes later*
Windows 10: But, my update.
Me: My work is more important than your update.

Since the device generating those readings is already doing a conversion to L*a*b* and I have the ability to modify its source code, I'm tempted to add another label to the display to see if the a* reading reliably tracks the transition from green to yellow on coffees that haven't been decaffeinated (or if I can do something to make it track).

One of the limitations of the new device I've been using is that it constrains its output to a range ~25-95 Agtron gourmet scale, but one of the first significant events in coffee roasting involves that measurement going off the scale (my benchtop analyzer would put it around 130). Decaf coffees, however, often start out brown so they reach their brightest at a reading that's still in range. That inflection from coffee getting brighter to coffee getting darker is where tasteable chemistry starts

This game has the worst text based overworld map I've ever seen. #tootctl

Since @ryanprior and @yomimono were the only ones who answered, I've tried installing @elementary on Dad's old laptop. The installer left a system that showed logo, briefly flashed a login on tty1, and then black screened, but I could go back to tty1 and fix things so I can do a graphical login now. Not a great first impression, but I'll let it have its updates, install the stuff my sister wants to use, and see how things go.

Removing locally cached copies of media attachments from other servers has resulted in a terminal window completely filled with ............................................................... and it's taking a long time, but disk usage has dropped from 100% to 41% so far, so I should probably run that more often.