3D-Printed Automatic Weather Station | Hacker News Reader

3D-Printed Automatic Weather Station

3dpaws.comet.ucar.edu

109 pointshyperbovine24 comments

While everyone's right that it doesn't look cheap compared to the consumer AIO weather stations on the market, that's not a fair comparison. The reason is that this project aims for a meteorologic quality implementation, where the consumer ones, even the so called "Pro" are definitely not. The requirements of the mounting of the different sensors are quite different - off-the-shelf all-in-one are not meeting those. Wind vanes must be at a certain height, which is a different than the one for the temperature and for rainfall measurement. The quality of the sensors is very important too - for stable, trustworthy measurements. And looking at the cost from the perspective of a trusted source of meteo-data installation, one can see that the real professional offerings are in the range of thousands ($$$$). Having said that, I'm pretty sure there's a room for further cost optimization without sacrificing data quality.

The 3D-PAWS project is interesting, but I don’t see the cost advantage. Commercial stations like the Ambient Weather WS-2902 are cheaper, last just as long, and already integrate with Weather Underground. What makes this design worth the extra expense?

> What makes this design worth the extra expense?

3d printable means that parts are never out of stock and upgrades are relatively straight-forward to implement (assuming you have the cad/STEP files. modifying STL/Mesh files is a pain in the ass)

Agree, though; for not that much $, you can get a pretty nice kit from Ali Express that'll spit out measurements over a serial port for use with APRS / LoRA or even just an arduino / raspbery pi ... etc.

Right, my Ambient Weather Station lasted 7 years through severe weather. I understand the self-repair angle, but 7 years seems like solid longevity to me.

Some of my personal anecdata on this. I had an Ambient weather station that lasted only a couple years. First the wind speed sensor failed, and they replaced it under warranty. Then it failed again outside the warranty period. Next I believe the temperature sensor failed, and then finally the indoor console completely stopped working (the display no longer showed anything). Even when it worked, it had intermittent connectivity outages despite being <100 ft from the sensor array (and more-or-less line of sight through a window).

After that experience, I replaced it with a Davis Instruments Vantage Vue, which has been running for 10 years without a single issue. Just needs periodic cleaning and battery replacements. Someone also made a device that goes inside the console to provide local access to the weather station (for Home Assistant integration, for example). It's called WifiLogger2. Pricy, but recommended if you want to pull the data from the console easily. I've also set up two more of these weather stations for family, and they've been similarly stable. So my experience with Ambient Weather left a bad taste. I'd rather pay the higher cost for Davis Instruments hardware that I know will work for a long time, than roll the dice with Ambient Weather ever again.

It's also telling that a search for "davis instruments weather station" on Amazon suggests the WS-2902, followed by actual Davis Instruments products. I imagine Ambient Weather is spending a lot of money for that search placement.

This Davis unit?

https://www.davisinstruments.com/products/vantage-vue-wirele...

Looks really nice! Thank you for the suggestion

Yep. Although I only have experience with the classic console (the one with physical buttons), I haven't used the touch screen model that superseded it.

Mines going on 5 now. Haven't maintained it at all. What failed on yours?

It started with the temperature then rain gauge then wind. And went altogether

> modifying STL/Mesh files is a pain in the ass

During COVID lockdown I learnt 3D modelling via Blender and so STLs are my lifeblood for 3D printing!

It’s a bit like having a compiled executable. Yes, you can technically modify an STL, but you’d be much better off if you can edit the source that the STL was generated from.

> During COVID lockdown I learnt 3D modelling via Blender and so STLs are my lifeblood for 3D printing!

Blender is a good skill to have, no doubt. 99.9% of all the 3d modeling / printing that I do is "functional" and you need an entirely different set of skills/techniques/tools to build a functional part compared to the models that blender spits out. At the end of the day, all techniques do generate a list of triangles but "going backwards" from that list of triangles can be a massive pain in the ass especially if you have a precise modification that you need to make.

If you want to share your models _and_ make life easier for anybody that's going to tweak them, STEP files are the best bet.

I won't discourage anyone that wants to take things into their own hands, but this level of DIY seems too deep for me. I've gotten a great balance of reliability and open data/DIY spirit by using an off-the-shelf consumer weather station and just intercepting its signals with an RTL SDR running RTL_433 and piping the data into a prometheus instance.

Honestly, if you're in town, you're probably within radio range of someone else's station already for free.

Can anyone more knowledgeable about this topic explain why some personal weather stations like this or the Ambien Weather one mentioned below are those pretty big things that have to be installed on a pole, while some like the Netatmo weather station are just those little things you place outside? Is the difference just in being able to measure wind speed?

Weather stations do usually come with a calibration problem: If you are on the sunny side of your garden, temperatures will be higher. If you are above concrete, temperatures will be higher, above grass, lower. Height matters. Wind and shielding from wind matters. Same for most of the other sensors, data is easily skewed by location, color, size, height and ventilation of the housing.

Therefore for comparable, calibrated data for real meteorological use, there are standards (unfortunately national, so there might be differences in some places) for all the above. The sensor housing has to be white, provide shade and total rain protection, but well-ventilated by grills on all sides. It has to be mounted above a patch of lawn of at least 10x10m, no shade, no significant nearby wind barriers, at a height of 1.2m. All sensors have to be at least 10cm from the internal walls of the housing.

If you don't care about strictly calibrated scientific data to use in your weather model, of course a simpler device will do. Also, nowadays, it is possible to account for deviations of certain stations due to the aforementioned factors.

Another factor is also serviceability. A cramped tiny housing that is hard to open and inspect isn't ideal when you have people touring the country servicing weather stations all year.

I'd like to set-up/buy a weather station, but I want to avoid vendor lock-in. Preferably integrate with home assistant if possible. Any suggestions?

Almost all of the cheap ones broadcast on 433mhz and can be read with a SDR

Thanks!

This is a cool project but that total cost is prohibitively expensive. This feels more like an academic proof of concept than a viable project.

It would be interesting if they could get some input from some industry experts to cost reduce and simplify for a few iterations. From a quick look at the files this appears to be what I’d expect from the dev board proof of concept stage of a project, before it was iterated on to be simplified for wider use. Hopefully they can get some time to do a few more cycles

It's competitive with Davis' lower-end offerings with the benefit of being open source and modifiable. I'd feel better depending on data from this 3D-PAWS kit versus some of the cheaper proprietary options.

I've spent several years trying to get a drip water logging system working (for cave research). I have a long string of failures under my belt :))

Speaking from my experience.. the actual plastic body of the equipment is the easiest and most irrelevant part. You can 3D print it.. or even just get a bottle and cut it in half? The actual tipping bucket is also kind of ridiculous to make from scratch. There are plastic ones that are injection molded and made for pennies in China. (you'd probably need to contact the manufacturer and make an order of >1K of them). Much more interesting would be a system for calibrating them (hat's not as labor intensive as what they present

Real concerns that need solutions are going to sounds ridiculous but are actually really challenging:

- Moisture will penetrate any enclosure and fry your electronics relatively quickly if you're in the tropics. It's very very challenging to seal things completely. All sealing solutions degrade over time

- People will steal the solar panel within 48 hours. The reality is that nothing about this station needs more than a coin cell (depends a bit on how often you measure the anemometer I guess). Power solutions are not easy or cheap. Weird power issues and making your logging equipment resilient to that is hard

- Most plastics are not happy in direct sunlight. We tried some small 3D printed parts. They all "melted" and fell apart quite quickly (I'm sure some alternate 3D printing material would hold up better)

- A bird takes a massive shit in your rain gauge, what now?

- cockroach crawls into your tipping bucket mechanism and dies. what now?

- Electronics? You know loves electronics? Ants! You're gunna have a nest in all that void space

- free standing plastic enclosures. You will likely have a wasp nest within 6 months

- Exposed cables? Rats will chew through them. Plastic box? Rats will chew through them. Metal box? It'll rust through in 4 months. Where I've tested the largest animal we had to deal with were monitor lizards. I imagine in Africa you have larger animals...

Making an actual low power logger? Surprisingly challenging. The ones you can buy online are 100 USD minimum and use shitty proprietary cables/software. Have terrible power performance and aren't very waterproof

For a more serious look at this space, I'd look here:

https://thecavepearlproject.org/

Also, the most popular rain gauge manufacturer has shifted to tipping buckets to a tipping spoon design.

https://www.davisinstruments.com/products/rain-collector-tip...

I have no special insight in to why. I've look at the internals of both designs and the new mechanism is more complicated (has more component parts). But they no longer sell the tipping bucket design, so I'm guessing there is good reason they did that and there is some advantage (maybe in terms of calibration?)

> Also, the most popular rain gauge manufacturer has shifted to tipping buckets to a tipping spoon design. [...] they no longer sell the tipping bucket design, so I'm guessing there is good reason they did that and there is some advantage (maybe in terms of calibration?)

Yes, tipping buckets are less exact than tipping spoons, especially in low-rain season. The bucket is never perfectly symmetrical and the mounting is never perfectly level, so one side of the bucket will hold more water than the other before tipping, leading to asymmetric tips.

Their design doesn't look particularly amenable to tuning though. I'd imagine some design with the weight on a screw that can be screwed in and out to adjust the centerpoint.

The two bucket design doesn't seem overly problematic though. Often you can see this in the record as time-to-tick will follow a sawtooth pattern. I haven't tried it myself, but .. assuming you don't have any skips in your record.. I imagine you could periodically use a long syringe and measure the exact volume that triggers a tick on each side. And then correct the record in software.

A bit of a tangent.. but I've also not found any software or algorithm for correct interpolation of rain volume. Maybe you know of something? ChatGPT suggests a few things :)) but I wonder if there is some standard method people use

Yes, for larger volumes of rain, that sawtooth pattern is visible and can be corrected for, or even ignored and averaged out. But for low volumes of rain, there will only be one tick. Usually you don't know which side of the bucket that tick was on, because most sensors just count the tipping of the bucket, not the direction or resting position. So for the one-tick-per-day case, you will very often not know whether to count the "large" or "small" bucket amount.

Calibration will also be drifting off for dirt accumulation and evaporation. Different fill states of each bucket lead to different evaporation in each bucket (smaller will evaporate faster, deviating towards even smaller amounts), and different resting times lead to more dust and algae settling in the bigger bucket, making it even heavier.

Reply on news.ycombinator.com