Collin Probst

Social Links Navigation

The Bluetti Elite 200 V2 is a fantastic power station for most people who want battery backup for long drives, camping, power outages, etc. It's also great for commercial use, reliable power on the job site, and as an actual UPS.

Why you can trust TechRadar We spend hours testing every product or service we review, so you can be sure you're buying the best. Find out more about how we test.

Bluetti is one of the more well-known power station companies. They make phenomenal offerings like the AC500, AC200L, and the Handsfree 2 backpack. The Elite 200 V2 is another offering in this quality lineup, expanding the form factor of the AC200 and AC200L with better output, smarter technology inside, and a slightly more compact design.

I've reviewed plenty of the best portable power stations, and this one has explicitly found great use in my workshop area, ensuring that all of my gear is readily charged and good to go. This charges drill batteries, vacuums, toys for the kids, etc. It got me thinking about how great of a solution this power station would be for a job site or in a utility capacity for charging tools out of your truck or van or even making sure you can top off power tools at a construction site.

Alternatively, perhaps you're not in that line of work; this would also be a good pick for a cabin, a shed, a camper van, or something similar, especially if it's paired with an all-season solar panel to keep you off the grid entirely.

(Image credit: Collin Probst // Future)

Bluetti Elite 200 V2: Price and Availability

The Elite 200 V2 is frequently discounted to under $1,000 from its regular MSRP of $1,999. That's a pretty good deal, and if you're even considering being in the market for a portable power station. You can buy these from Bluetti's website, Amazon, and a handful of other retailers.

(Image credit: Collin Probst // Future)

Bluetti Elite 200 V2: Unboxing & first impressions

The Elite 200 V2 comes in a simple box with only bare essentials. It has the power station itself, an AC Charging cable, solar cable attachments, a grounding screw, and a manual—nothing out of the ordinary there. The casing and handles help build into this power station's rigidity factor, which, paired with the overall heft, feels quite premium.

It doesn't have as much casing as some of the more "all-season" units that Bluetti has (review coming soon on a more weather-resistant model), but it still feels like you can knock it around, and it will be fine. The two integrated handles on either side of the top of this unit make moving the power station something I can do on my own, though it's not something that most would enjoy doing.

If there were one handle, this unit would be flat-out inaccessible to many, if not most, and that's not to say I'm even some sort of professional lifter of heavy things. My wife and a few co-workers struggled immensely trying to lift this unit into position, even with the two handles. If this is you, the two handles are positioned so you can easily buddy-carry and share the load with a friend (or stranger).

(Image credit: Collin Probst // Future)

Bluetti Elite 200 V2: Design & build quality

Specs Capacity: 2,073.6Wh

Output:

4x 120V AC (20A)

2x USB-A (15W)

2x USB-C (100W)

1x 12V/10A car outlet

Input:

AC (up to 1800W)

Solar (up to 1000W)

AC + Solar = 2400W max

DC car (12V or with Charger 1 accessory)

Dimensions: 13.8 × 9.8 × 12.7 in

Weight: 53.4 lbs

As I alluded to in an earlier statement, I am also testing the IP-rated variant of this power station. So, in comparison, this unit is much less rugged, but not in a bad way. I'd consider this a professional-looking power station, ready for the board room just as much as it's prepared to live in the back of a work truck.

All the ports are on the front of the unit, other than the power inputs (AC and Solar) and the grounding wire, which are all on the right side. The output ports are all on the front, beautifully surrounding the display without encouraging covering the screen.

(Image credit: Collin Probst // Future)

Bluetti Elite 200 V2: In use

I briefly mentioned above that I've found this machine useful in a workshop area, simultaneously charging up some of my gear. I have also taken this outside (when it wasn't raining) to help keep drills recharged while working on building a giant playset in my backyard. I brought this to a job site where I was helping run conduit, power, and CAT-6 Ethernet in a workspace at a company I work with. When the power was cut, I used this power station to shop lights and keep drills recharged for the team, throwing up a half-wall and drywall.

During this testing, I was impressed with the output capacity of this hefty yet mighty power station. Sometimes, I wish it had more than four ports, but in all of my testing, ID 3. So, while more feels better, I don't usually find myself needing more on the go. And, if I did, I could throw a power strip on there pretty straightforwardly, like what I am doing with a work cart powered by my EcoFlow River 3. This power station only has one AC port with a ground and just a two-prong AC port. Still, I run a robust cart with a dozen outlets mounted all around for ease of use, a Thunderbolt Dock, a monitor, a charging station, and a charger for a few radios. So, if I can run all of that off of a smaller unit, with only one and a half AC ports, you can run plenty with four full three-prong AC ports.

(Image credit: Collin Probst // Future)

Another excellent use case for this power station would be for emergency backup. It could power a fridge (the time it could run would depend a lot on the type of fridge) for hours, a router, a server, phones, computers, medical equipment, security equipment, and so on.

Swipe to scroll horizontally Attributes Notes Rating Design Compact and powerful ⭐⭐⭐⭐⭐ Ease of use Easy to use ⭐⭐⭐⭐⭐ Practicality Practical for most ⭐⭐⭐⭐ Price Decent price ⭐⭐⭐⭐

Bluetti Elite 200 V2: Final verdict

The Bluetti Elite 200 V2 is a fantastic portable power station for most individuals. It is a tremendous all-around Swiss army-knife power station that is great at many things. It's hefty, doesn't have waterproofing, does not come with solar panels, doesn't have wheels, and some features are only accessible with the app.

Still, outside of those few minor inconveniences, not even problems, this power station is excellent. It's dependable, durable, robust, purposefully designed, and all less than $1000. If you're looking to get into portable power stations, need a solid one to add to your fleet, or anywhere in between, this power station is worth considering, especially if it's on one of its sales.

The Bluetti Elite 200 V2 is a great power station with an impressive battery, fantastic output, and decent ports. Just be warned—it's more than a little cumbersome

The Bluetti Elite 200 V2 is a portable power station with an impressive 2,073Wh of capacity, 2,600W output, four AC outputs, two USB-C, two USB-A, A DC out, DC in, a grounding port, and an AC Input. All these ports in a small and powerful machine are excellent; just be warned that it is also 53.4 lbs.

I tested the Bluetti Elite 200 V2 for the last 160 days - an excellent portable power station that's absurdly heavy for its size

082e87735c3b9b6bb5b447103d6c9513820dd471

BLUETTI – a leader in portable off-grid power station solutions – has a launched a new loyalty rewards program delivering a way for customers to score deep deals and free credits towards future purchases. The brand has become well known for its portable battery systems to light up your camp site, provide much needed outlets […]

Save on your dream portable power station setup with new FREE BLUETTI rewards program

Nibbling Around the Sound Blaster's Lunch - Digital Sound in DOS Games without a Sound Blaster

As I watch Law and Order reruns on my business trip, first, a couple followups. The big note is that it looks like Intel and some ARM cores aren't the only ones vulnerable to Meltdown; Raptor Computer Systems confirms that Meltdown affects at least POWER7 through POWER9 as well, and the Talos II has already been patched. It's not clear if this is true for POWER4 (which would include the G5) through POWER6 as these processor generations have substantial microarchitectural differences. However, it doesn't change anything for the G3 and 7400, since because they appear to be immune to Spectre-type attacks means they must also be immune to Meltdown. As a practical matter, though, unless you're running an iffy program locally there is no known JavaScript vector that successfully works to exploit Spectre (let alone Meltdown) on Power Macs, even on the 7450 and G5 which are known to be vulnerable to Spectre.

Also, the TenFourFox Downloader is now live. After only a few days up with no other promotion, it's pulling down about 200 downloads a day. I note that some small number are current TenFourFox users, which isn't really what this is intended for: the Downloader is unavoidably -- and in this case, also unnecessarily -- less secure, and just consumes bandwidth on Floodgap downloading a tool to download something the browser can just download directly. If you're using TenFourFox already (at least 38 or later), please just download upgrades with the browser itself. In addition, some are Intel Mac users on 10.6 and earlier, which the Downloader intentionally won't grab for because we don't support them. Nevertheless, the Downloader is clearly accomplishing its goal, which is important given that many websites won't be accessible to Power Mac users anymore without it, so it will be a permanent addition to the site.

Anyway, let's talk about Power Macs and radios. I'm always fond of giving my beloved old Macs new things to do, so here's something you can think about for that little G4 Mac mini you tossed in the closet. Our 2,400 square foot house has a rather curious floor plan: it's a typical California single-floor ranch but configured as a highly elongated L-shape along the bottom and right legs of the property's quadrilateral. If I set something playing somewhere in the back of the house you probably won't hear it very well even just a couple rooms away. The usual solution is to buy something like a Sonos, which are convenient and easy to operate, but streaming devices like that can have synchronization issues and they are definitely not cheap.

But there's another solution: set up a house FM transmitter. With a little spare time and the cost of the transmitter (mine cost $125), you can devise a scheme that turns any FM radio inside your house into a remote speaker with decent audio quality. Larger and better engineered than those cheapo little FM transmitters you might use in a car, the additional power allows the signal to travel through walls and with careful calibration can cover even a relatively large property. Best of all, adding additional drops is just the cost of another radio (instead of an expensive dedicated receiver), and because it's broadcast everything is in perfect sync. If your phone has an FM radio you can even listen to your home transmitter on that!

There are some downsides to this approach, of course. One minor downside is because it's broadcast, your neighbours could tune in (don't play your potentially embarrassing, uh, "home movie" audio soundtracks this way). Another minor downside is that the audio quality is decent but not perfect. The transmitter is in your house, so interference is likely to be less, but things as simple as intermittently energized electrical circuits, bad antenna positioning, etc., can all make reception sometimes maddeningly unpredictable. If you're an uncompromising audiophile, or you need more than two-channel audio, you're just going to have to get a dedicated streaming system.

The big one, though, is that you are now transmitting on a legally regulated audio band without a license. The US Federal Communications Commission has provisions under Part 15 for unlicensed AM/FM transmission which limit your signal to an effective distance of just 200 feet. There are more specific regulations about radiated signal strength, but the rule of thumb I use is that if you can detect a usable signal at your property line you are probably already in violation (and you can bet I took a lot of samples when I was setting this up). The FCC doesn't generally drive around residential neighbourhoods with a radio detector van and no one's going to track down a signal no one but you can hear, but if your signal leaks off your property it only takes one neighbourhood busybody with a scanner and nothing better to do to complain and initiate an investigation. Worse, if you transmit on the same frequency as an actually licensed local station and meaningfully interfere with their signal, and they detect it (and if it's meaningful interference, I guarantee you they will sooner or later), you're in serious trouble. The higher the rated wattage for your transmitter, the greater the risk you run of getting busted, especially if you are in a densely populated area. If you ever get a notice of violation, take it seriously, take your transmitter completely offline immediately, and make sure you tell the FCC in writing you turned it off. Don't turn it back on again until you're sure you're in compliance or you may be looking at a fine of up to $75,000. If you're not in the United States, you'd better know what the law is there too.

So let's assume you're confident you're in (or can be in) compliance with your new transmitter, which you can be easily with some reasonable precautions I'll discuss in a moment. You could just plug the transmitter into a dedicated playback device, and some people do just that, but by connecting the transmitter to a handy computer you can do so many other useful things. So I plugged it into my Sawtooth G4 file server, which lives approximately in the middle of the house in the dedicated home server room:

There it is, the slim black box with the whip antenna coming off the top sandwiched between the FireWire hub (a very, very useful device and much more reliable than multiple FireWire controllers) and the plastic strut the power strip is mounted on. This is the Whole House FM Transmitter 3.0"WHFT3" which can be powered off USB or batteries (portable!), has mic and line-level inputs (though in this application only line input is connected), includes both rubber duck and whip antennas (a note about this presently) and retails for about $125. Amazon carries it too (I don't get a piece of any sales, I'm just a satisfied customer). It can crank up to around 300 milliwatts, which may not seem like much to the uninitiated, but easily covers the 100 foot range of my house and is less likely to be picked up by nosy listeners than some of the multi-watt Chinese import RF blowtorches they sell on eBay (for a point of comparison, a typical ham mobile radio emits around 5 watts). It also has relatively little leakage, meaning it is unlikely to be a source of detectable RF interference when properly tuned.

By doing it this way, the G4, which is ordinarily just acting as an FTP and AFP server, now plays music from playlists and the audio is broadcast over the FM transmitter. How you decide to do this is where the little bit of work comes in, but I can well imagine just having MacAmp Lite X or Audion running on it and you can change what's playing over Screen Sharing or VNC. In my case, I wrote up a daemon to manage playlists and a command-line client to manipulate it. 10.5+ offers a built-in tool called afplay to play audio files from the command line, or you can use this command line playback tool for 10.2 through 10.4. The radio daemon uses this tool (the G4 server runs Tiger) to play each file in the selected folder in order. I'll leave writing such a thing to the reader since my radio daemon has some dependencies on the way my network is configured, but it's not very complex to devise in general.

Either way works fine, but you also need to make sure that the device has appropriate signal strength and input levels. The WHFT3 allows you to independently adjust how much strength it transmits with a simple control on the side; you can also adjust the relative levels for the mic and line input if you are using both. (There is a sorta secret high-level transmission mode you can enable which I strongly recommend you do not: you will almost certainly be out of FCC compliance if you do. Mine didn't need this.) You should set this only as high as necessary to get good reception where you need it, which brings us to making sure the input level is also correct, as the WHFT3 is somewhat more prone to a phenomenon called over-modulation than some other devices. This occurs when the input level is too high and manifests as distortion or clipping but only when audio is actually playing.

To calibrate my system, I first started with a silent signal. Since the frequency I chose had no receivable FM station in my region of greater Los Angeles (and believe me, finding a clear spot on the FM dial is tough in the Los Angeles area), I knew that I would only hear static on that frequency. I turned on the transmitter with no input using the "default" rubber duck antenna and went around the house with an FM radio with its antenna fully retracted. When I heard static instead of nothing, I knew I was exceeding the transmission range, which gave me an approximate "worst case" distance for inside the house. I then walked around the property line with the FM radio and its antenna fully extended this time for a "within compliance" test. I only picked up static outside the house, but inside I couldn't get enough range in the kitchen even with the transmitter cranked up all the way, so I ended up switching the rubber duck antenna for the included whip antenna. The whip is not the FCC-approved configuration (you are warned), but got me the additional extra range, and I was able to back down the transmitter strength and still be "neighbour proof" at the property line. This is also important for audio quality since if you have the transmitter power all the way up the WHFT3 tends to introduce additional distortion no matter what your input level is.

Next was to figure out the appropriate input level. I blasted Bucko and Champs Australian Christmas music and backed down the system volume on the G4 until there was no distortion for the entire album (insert your own choice of high volume audio here such as Spice Girls or Anthrax), and checked the new level a few times with a couple other albums until I was satisfied that distortion and overmodulation was at a minimum. Interestingly, while you can AppleScript setting the volume in future, what you get from osascript -e 'set ovol to output volume of (get volume settings)' is in different units than what you feed to osascript -e 'set volume X ': the first returns a number from 0-100 with 14 unit steps, but the second expects a number from 1-10 in 0.1 unit steps. The volume on my G4 is reported by AppleScript as "56" but I set that on startup in a launchd startup item with a volume value of 4.0 (i.e., 4 times 14 equals 56). Don't ask me why Apple did it this way.

There were two things left to do. First was to build up a sufficient library of music to play from the file server, which (you may find this hard to believe) really is just a file server and handles things like backups and staging folders, not a media server. There are many tools like the most excellent X Lossless Decoder utility -- still Tiger and PowerPC compatible! -- which will rip your CDs into any format you like. I decided on MP3 since the audio didn't need to be lossless and they were smaller, but most of the discs I cared about were already ripped in lossless format on the G5, so it was more a matter of transcoding them quickly. The author of XLD makes the AltiVec-accelerated LAME encoder he uses available separately, but this didn't work right on 10.4, so I took his patches against LAME 3.100, tweaked them further, restored G3 and 10.4 compatibility, and generated a three-headed binary that selects for G3, G4 and a special optimized version for G5. You can download LAMEVMX here, or get the source code from Github.

On the G5 LAMEVMX just tears through music at around 25x to as much as 30x playback speed, over three times as fast as the non-SIMD version. I stuck the MP3 files on a USB drive and plugged that in the Sawtooth so I didn't have to take up space on its main RAID, and the radio daemon iterates off that.

The second was figuring out some way to use my radios as, well, radios. Yes, you could just tune them to another station and then tune them back, but I was lazy, and when you get an analogue tuner set at that perfect point you really don't want to have to do it again over and over. Moreover, I usually listen to AM radio, not FM. One option is to see if they stream over the Internet, which may even be better quality, though receiving them over the radio eliminates having to have a compatible client and any irregularities with your network. With a little help from an unusual USB device, you can do that too:

This is the Griffin radioSHARK, which is nothing less than a terrestrial radio receiver bolted onto a USB HID. It receives AM and FM and transmits back to the Mac over USB audio or analogue line-level out. How do we hook this up to our Mac radio station? One option is to just connect its audio output directly, but you should have already guessed I'd rather use the digital output over USB. While you can use Griffin's software to tune the radio and play it through (which is even AppleScript-able, at least version 2), it's PowerPC-only and won't run on 10.7+ if you're using an old Intel Mac for this purpose, and I always prefer to do this kind of thing programmatically anyhow.

For the tuner side, enterprising people on the Linux side eventually figured out how to talk to the HID directly and thus tune the radio manually (there are two different protocols for the two versions of the radioSHARK; more on this in a moment). I combined both protocols together and merged it with an earlier but more limited OS X utility, and the result is radioSH, a commandline radio tuner. (You can also set the radioSHARK's fun blue and red LEDs with this tool and use it as a cheapo annunciator device. Read the radioSH page for more on that.) I compiled it for PowerPC and 32-bit Intel, and the binary runs on anything from 10.4 to 10.13 until Apple cuts off 32-bit binary compatibility. The source code is available too.

For USB audio playthru, any USB audio utility will suffice, such as LineIn (free, PowerPC compatible) or SoundSource (not free, not PowerPC compatible), or even QuickTime Player with a New Audio Recording and the radioSHARK's USB audio output as source. Again, I prefer to do this under automatic control, so I wrote a utility using the MTCoreAudio framework to do the playback in the background. (Use this source file and tweak appropriately for your radioSHARK's USB audio endpoint UID.) At this point, getting the G4 radio station to play the radio was as simple as adding code to the radio daemon to tune the radio with radioSH and play the USB audio stream through the main audio output using that background tool when a playlist wasn't active (and to turn off the background streamer when a playlist was running). Fortunately, USB playthru uses very little CPU even on this 450MHz machine.

I mentioned there are two versions of the radioSHARK, white (v1) and black (v2), which have nearly completely different hardware (belied by their completely different HID protocols). The black radioSHARK is very uncommon. I've seen some reports that there are v1 white units with v2 black internals, but of the three white radioSHARKs I own, all of them are detected as v1 devices. This makes a difference because while neither unit tunes AM stations particularly well, the v1 seems to have poorer AM reception and more distortion, and the v2 is less prone to carrier hum. To get the AM stations I listen to more reliably with better quality, I managed to track down a black radioSHARK and stuck it in the attic:

To improve AM reception really all you can do is rotate or reposition the receiver and the attic seemed to get these stations best. A 12-foot USB extension cable routes back to the G4 radio station. The radioSHARK is USB-powered, so that's the only connection I had to run.

To receive the radio on the Quad G5 while I'm working, I connected one of the white radioSHARKs (since it's receiving FM, there wasn't much advantage to trying to find another black unit). I tune it on startup with radioSH to the G4 and listen with LineIn. Note that because it's receiving the radio signal over USB there is a tiny delay and the audio is just a hair out of sync with the "live" analogue radios in the house. If you're mostly an Intel Mac house, you can of course do the same thing with the same device in the same way (on my MacBook Air, I use radioSH to tune and play the audio in QuickTime Player).

For a little silliness I added a "call sign" cron job that uses /usr/bin/say to speak a "station ID" every hour on the hour. The system just mixes it over the radio daemon's audio output, so no other code changes were necessary. There you go, your very own automatic G4 radio station in your very own house. Another great use for your trusty old Power Mac!

Oh, one more followup, this time on Because I Got High Sierra. My mother's Mac mini, originally running Mavericks, somehow got upgraded to High Sierra without her realizing it. The immediate effect was to make Microsoft Word 2011 crash on startup (I migrated her to LibreOffice), but the delayed effect was, on the next reboot (for the point update to 10.13.2), this alarming screen:

The system wouldn't boot! On every startup it would complain that "macOS could not be installed on your computer" and "The path /System/Installation/Packages/OSInstall.mpkg appears to be missing or damaged." Clicking Restart just caused the same message to appear.

After some cussing and checking that the drive was okay in the Recovery partition, the solution was to start in Safe Mode, go to the App Store and force another system update. After about 40 minutes of chugging away, the system grudgingly came up after everything was (apparently) refreshed. Although some people with this error message reported that they could copy the OSInstall.mpkg file from some other partition on their drive, I couldn't find such a file even in the Recovery partition or anywhere else. I suspect the difference is that these people encountered this error immediately after "upgrading" to Because I Got High Sierra, while my mother's computer encountered this after a subsequent update. This problem does not appear to be rare. It doesn't seem to have been due to insufficient disk space or a hardware failure and I can't find anything that she did wrong (other than allowing High Sierra to install in the first place). What would she have done if I hadn't been visiting that weekend, I wonder? On top of all the other stupid stuff in High Sierra, why do I continue to waste my time with this idiocy?

Does Apple even give a damn anymore?

And now for something completely different: Make that Power Mac into a radio station (plus: the radioSHARK tank and AltiVec + LAME = awesome)

Long time, no posts!

Finally wrote up this little project from last year: Repurposing my SHA2017 hacker camp badge into a solar energy and power consumption monitor.

Background

Lately I've really appreciated systems that show passive information - learning something useful without needing to parse a flashy Dashboard or interact with a touch screen.

For example, I like our weather station. It might be ugly but it's always there in the kitchen, quietly showing you the temperature:

To experiment more with this kind of interface, I backed the Inkplate 10 programmable e-ink display on CrowdSupply last year.

In the meantime, I noticed this neat project from the Netherlands that repurposed the SHA2017 Hacker Camp badge as a CO2 sensor:

https://polyfloyd.net/post/co2-sensors-against-covid /

"Oh, I have one of those badges too!"

What was SHA2017?

Almost five years ago a group of hackers and other technically minded oddballs camped in a field in the Netherlands to have fun and play with LEDs:

I was lucky enough to be in Europe at the time, so I came along.

Every attendee received an open source hardware badge. It has an e-ink display, RGB LEDs, touch pads in a controller layout, an ESP32 microcontroller for Wi-Fi/Bluetooth, and capacity for expansion boards called scarves. There was a badge hacking tent where attendees could solder LEDs onto their badges or do other modifications.

A "Hatchery" app store let everyone download and share new apps. The badge was so successful that it even suffered a ransomware outbreak! This Hackaday article has more.

At the camp, my own contribution was a Wi-Fi signal strength monitoring app that graphed the RSSI value. Four years later it's still available to install on your badge.

badge.team

The original SHA2017 badge infrastructure was still online in 2021, but essentially abandoned. Instead, there is something better: badge.team. The badge.team has taken up the task of supporting over a dozen different con/camp badges, providing a common firmware base and a hatchery to share apps for any badge.

This is such an awesome example of open source allowing a project to stay useful and practical over time, and a community working together to keep it like that.

Solar Monitor

Our rooftop solar system has 6kW of panels with a 5kW grid-connected inverter. Because the feed-in tariff for selling electricity to the grid is so low, it's better for us to consume our own solar power rather than sell it back to the electricity grid and then consume power from the grid later.

Therefore, it's great to know how much electricity we are generating and how much we are using at any given time. This can be hard to guess, especially on a cloudy day when your eye is not a good judge of exactly how much sunlight there is.

The goals of the Solar Usage Monitor project:

Be a passive interface that doesn't need any interaction.
Show how much electricity we are using in the house and how much solar electricity we are generating.
Provide enough information to answer questions like "should I turn the heater up right now?" or "is it a good time to run the dishwasher?".
Not buy anything new. (After all, I already had an InkPlate 10 on the way...)

Solar Infrastructure

Our solar inverter (Fronius Primo) has a Wi-Fi interface and a JSON web API. A Python module exists, so with a few lines of Python it's possible to shovel the current state of power flow into an InfluxDB time series database and view it in Grafana:

However, a web page dashboard is exactly the kind of interface I want to avoid. Time to show this information on the badge!

MicroPython firmware

The language of choice for badge programmers is MicroPython. I have a lot of respect for MicroPython: out of the many "run this dynamic garbage collected language on a tiny microcontroller!" environments, it delivers the most in my experience.

The solar monitor firmware has a pretty simple use case:

Connect to Wi-Fi
Query InfluxDB for a JSON payload with the solar usage data
Draw a graph and a numeric summary of the data on the e-ink display
Sleep for a bit
Go back to step 2

MicroPython Performance

Any time I've written MicroPython, I've followed these basic steps:

Write code as if it was for CPython on a "big" computer.
Run that in MicroPython
Rewrite the part which was too slow or ran out of memory
Repeat until all is well

This doesn't necessarily scale to big projects, projects that need to be very reliable, or library code where you can't anticipate the user's constraints. However it's worked well for my little projects, I'm always pleasantly surprised by how little code needs rewriting in step 3.

For the solar monitor, it's very unlikely the firmware will be "too slow":

e-ink displays take a second or two to physically update their pixels.
Even updating every five or ten seconds would be plenty.
The badge is going to be permanently mains powered, and it only uses a few Watts maximum, so I won't notice any differences in power consumption.

However, there is definitely a chance for "ran out of memory". Dynamic garbage collected languages are prone to "allocation bubbles": some nice neat readable code that looks innocent but allocates multiple short-lived temporary copies of a memory buffer.

These can be very hard to spot - especially if you're not familiar with the details of the language, it can be hard to tell exactly what's going on. When I think about avoiding "allocation bubbles" in Python I often think of Alex Gaynor's 2014 talk Fast Python, Slow Python where he talks about how idiomatic Python code will often hide memory allocations. (His focus is speed, but it turns out that memory allocations are also slow!)

The challenge here is to keep the benefits of Python while avoiding the worst performance or memory issues. It's alternatively possible to write MicroPython as "C without curly braces", but this seems like a waste to me.

String bubble!

After some debugging, my naive Python code worked fine apart from this sequence:

resp = urequests.post('http://192.168.66.1:8086/query?db=sensors&epoch=s', data=b'q='+query, headers={ 'Content-Type': 'application/x-www-form-urlencoded' }) text = resp.text resp.close() if not len(text): return [] data = json.loads(text) result = [Sample(*x) for x in data['results'][0]['series'][0]['values']] result = [s for s in result if not s.is_empty] # remove all the empty samples

(Please excuse the hard-coded IP address in this early experiment, and the lack of API authentication...)

First we read the HTTP response as a string (resp.text), which allocates an entire second copy of the response data!

Then we allocate an entire JSON object tree from the string before making two separate lists of sample data via the list comprehensions. This is a lot of memory allocations, and unsurprisingly the program would often crash here depending on how large the JSON response was.

The final code ended up being very close to the original:

try: resp = urequests.post('{}/query?db=sensors&epoch=s'.format(influxdb_url), data=b'q='+query, headers={ 'Content-Type': 'application/x-www-form-urlencoded' }) except OSError: print(Failed to connect to InfluxDB server) return [] if resp.status_code != 200: print(InfluxDB returned error code {}.format(resp.status_code)) resp.close() return [] data = json.load(resp.raw) resp.close() gc.collect() result = [Sample(*x) for x in data['results'][0]['series'][0]['values']] result = [s for s in result if not s.is_empty()] # remove all the empty samples return result

The only strictly necessary difference is using the file-like interface (json.load(resp.raw)) to read the response into objects, and then forcing a garbage collection immediately after we close the connection and free up the connection buffers. This removes the need to keep two large string buffers in memory simultaneously, and makes sure that the connection memory is freed before we start creating Sample objects.

No doubt there are other improvements that could be made to reduce the memory overhead, but this turned out to be enough.

Powering

The badge has a USB port, but it's in an inconvenient place for hanging on the wall:

How dare the badge designers not think of my very obvious use case!

I decided to remove the battery and provide power directly to the battery connector. Sadly, despite having hoarded an enormous box of random plug adapters I didn't have anything in the 3.5V to 4.1V voltage range. Let this be a lesson to hoard more comprehensively!

Bending the "not buy anything new" rule a little, a local junk shop exchanged $5 for this 100% Not New travel charger for an LG Flip Phone, circa 2004:

A quick session with the Reload:Pro confirmed that after almost two decades it could still supply its rated 800mA current with only a small drop in voltage:

4.6V is still higher than the maximum Lithium Ion battery voltage of 4.2V... I wasn't certain if the charger chip on the badge might be damaged or confused by this.

To be sure, I soldered a single diode in series with the adapter output for a 0.7V drop:

Unfortunately, things didn't go well when I plugged the travel adapter into the badge. About 50% of the time the ESP32 would suddenly reset when connecting to Wi-Fi! This was the very familiar problem of not having enough stable power on the board.

The travel adapter may be able to deliver 800mA but the lead is pretty long (resistive drop), and it was probably never designed to stay stable over large current transients...

Continuing with the theme of quick hacks, I soldered some extra surface mount ceramic capacitors and an electrolytic capacitor to the board to help smooth over any power spikes.

Just like that, a rock solid (dodgy) power supply!

Mounting

The designers of the SHA2017 badge thoughtfully provided some M3 mounting holes on the PCB.

A bit of scrap wire and some standoffs let me create the world's sketchiest picture frame:

Results

The badge has sat on our kitchen wall next to the weather station for almost six months now.

It's handy to be able to see at a glance what the solar system is doing, and how much power we are using:

The numbers at the top show the instantaneous solar output and the total power consumption of the house.
The line graph shows the solar usage (dotted line) and the power usage (solid line) over the past hour. Every 15 minutes the graph scrolls one segment to the left.
Only minutes are shown on the X axis legend because otherwise the firmware needs to know about timezones and daylight saving time...

It's also a lot of fun to have friends visit and say "do you have... a video game controller on your wall?"

Being able to put this together so easily is really due to the fantastic tools from MicroPython and badge.team. It's so much faster to start from a working execution environment and a working graphics display, rather than building everything up from first principles.

Because the solar usage monitor is so specific to our home's setup (at minimum, the InfluxDB schema has to match), I haven't shared it on the badge.team hatchery. I have put the code up on GitHub, in case anyone wants to crib from it - please excuse the naive coding style and the (still) lack of any InfluxDB credential support (eek!)

Future Plans

The next iteration of this project is to use the InkPlate 10 to display this information along with the weather station data and some other information, in a single attractive display. For now, however, the SHA2017 badge is doing a fantastic job!

(Interest Disclosure: For five years I worked for Espressif who make the ESP32 microcontroller used in both the SHA2017 badge and the InkPlate 10. However, this is entirely a personal project.)