OctoPrint On Air #21


#1

Topics in this episode which was recorded on November 16th:

  • 1.3.10rc3 and 1.3.10rc4
  • 1.3.10 stable release
  • A longer look at the current output of the Anonymous Usage Tracking
  • Q&A session touching on
    • the importance of full bug reports

Full contents with jump marks available in the YouTube description.

You can find all public past episodes in the OctoPrint On Air Playlist.

OctoPrint On Air is a roughly monthly live broadcast done by me (Gina Häußge) for all Patrons on the Patreon campaign pledging at the $5 perk or above. Attendees can ask questions which will be answered on air and I also report what’s been going on in the background, what happened, what the current challenges are and what the near future holds, plus talk about whatever else might be interesting in the world of OctoPrint.


This is a companion discussion topic for the original entry at https://octoprint.org/blog/2018/12/18/octoprint-on-air-21/

#2

Regarding the 1K (maybe 10% of the users) running their printers with active undervoltage conditions, a number of things could be contributing to this.

  1. of course, it could be the power adapter that powers the Raspberry
  2. it could be the line voltage at the wall which is under-powered (this routinely happens in office buildings); a UPS would help in cases like this
  3. it could be that too many things have been plugged into the Raspberry (webcam, TFT or possibly an HDMI monitor, keyboard, mouse)
  4. but perhaps we might have overlooked that 5V could be bleeding out over the serial connection to some of these printer boards

The 4th one isn't as crazy as it seems. My recent autonomous tank project taught me a lot about this.

Raspberry Pi Zero W -> non-USB serial -> Arduino Mega 2560 board

The power LED on the Arduino in this case would try to light up when the Raspberry was powered. The voltage on the serial line was being felt over on the other board, acting as a power sink in this case.

I know that my Robo controller board has a jumper which would prevent the board from sinking 5V from the serial connection but I couldn't tell you about other boards.

We have a few times when people have added tape to the 5V pin of their USB, so I think this is a possibility:
weird C270 webcam issue
printing just stop in middle of part
Octoprint loses USB connection with printer repeatedly

Also, there's a difference between a DC charger and a DC adapter. The output voltage on an adapter looks like a horizontal line if you see it on an oscilloscope. The output from a charger isn't expected to be a clean horizontal line; it's expected to be something which wobbles in the approximate area of the output voltage.


#3

You're right, knowing the RC participant numbers is big. That's cool. Congrats Gina.

The modern Pis don't run on 2.4A and don't like long USB cords, I'm not surprised we're seeing so much undervoltage. Line voltage fail seems weird, 'guru, generally a wall wart will work down to 80v or so. More likely, voltage-wise, people should be using 5.2v adapters. Still, wonder if at some point we'll have to disable the 'print' button for undervoltage.

I see stats.octoprint.org is online. But it has a password :confused: oh well.

I'm really curious about the 24h stats, about turning off/on. And that would hurt with SD cards, I'd guess. I assume the heatmap below the geomap will make it more clear. Perhaps we should send uptime in pisupport, perhaps some disk info too.

Safety 'screaming'- if you decide to allow prints, don't do it with a control panel checkbox. Require approval for every print.


#4

Sorry :wink: As I said, I really need a secondary instance to push snapshots too, I don't want to give public access to the connected instance. I'll look into that when I'm back from my break.

I just added OctoPrint's own uptime to the ping payload for 1.3.11 the other day. Not much to see yet though :wink:

I'm also thinking about how to detected unclean shutdowns of the server and popping up a warning on next start about that. Something like setting a flag in the shutdown routine should do it. The only problem here are windows systems since there I have run into issues with the shutdown routine not being called reliably, but if push comes to shove I'll just ignore windows for that functionality. It's only in 1% of installs anyhow:

image


#5

Just use an active usb hub on all arm devices and all is fine.


#6

What about making a static site out of it- e.g., wget to s3.


#7

That would be the fallback solution. Getting another public Grafana instance up and running and pushing snapshots of the various dashboard there via cronjob shouldn't be that tricky though - I just have to find half a day or so for it :slight_smile:


#8

In businesses, I'm usually the guy who volunteers to review the building's power, remove wall/box plates and then make a sad face when I see what's left of the original wiring.

  • common mismatch between the metal used in the wiring and that of the connecting screws for the outlets themselves resulting in galvanic corrosion, added resistance and dropped voltage...
    • ...which then leads to a cycle of added heat, more resistance, more heat and eventually a catastrophic failure of the remaining wire, arcing and more heat
    • this would effect anything on a particular circuit but most significantly the outlets which are arcing
  • similar problem related to damp environments like garages and basements (typical homes for 3D printers)
  • over-populated circuit breaker boxes (building designs 20 years ago could never have imagined the needs of today) which then deliver less than the expected voltage to all the outlets downstream
  • co-existence of compressors (refrigerator, air conditioner) and halogen lighting (startup power needs) on the same system

For a 115VAC system, I routinely see anything in the 106V-109V range here in California businesses as delivered to the wall plate. To a 5V power adapter that means the low end sees 4V output instead, so it's significant.


#9

You're seeing a lot of retro stuff. I checked a couple of adapters nearby, which are rated at 100-240v. Older amp or less PSes are probably linear, but modern stuff would put out too much heat. Quick glance gives the same range for the CanaKit power adapter, as close to a "standard" Pi power supply as I can imagine.

I'm so used to everything being multi-voltage that I don't even check before plugging it in when I'm in Europe. Easy rule of thumb is anything multi-voltage is using a SMPS and doesn't care about wide variation.

Definitely, though, lots of AC voltage headaches.


#10

"...as measured at the wall outlet"

Basically, the internal wiring of the building is trashed. I'd go over to Home Depot, buy a dozen outlets and replace them (cutting off two inches of wiring at the ends of runs).


#11

:slight_smile: Just had a recurring undervoltage errors:

  1. Everything normal with a Raspi 3B with a 2.8" Adafruit capacitive TFT screen, 2.5A adapter on a UPS
  2. Backed up and restored the microSD from the 8GB microSD to a 16GB microSD to develop a different prototype
    • Immediately began seeing undervoltage indications on the screen (the only change was the microSD!) So I was thinking "you've got to be kidding..."
    • Put a meter on one of the jacks for the UPS: solid 120VAC
    • Put a meter on the 5V line of the Raspi: 5V but occasional dips down to 4.5V or similar
    • Removed the TFT and rebooted, no errors
    • Meter on the Raspi: solid 5V with little deviation
    • TFT back on again but this time it's barely connected in the 40-pin socket (not all the way pushed down)
    • No undervoltage seen from watch grep --text Under /var/log/kern.log or the screen itself

The theory then is that the bottom side of the TFT daughterboard was grounding against the RJ45 connector on the Raspi itself.


#12

i am just interested in this discourse integration with wordpress and want to see how it is configured :slight_smile:


#13

It isn't Wordpress :wink: It is a static site built with Jekyll and the comments are embedded via Discourse's own embed feature.