reef-pi :: An opensource reef tank controller based on Raspberry Pi.

[PaulJ2303]

Thank you for reporting this. I think I understand the problem a bit better now. Can you confirm following:
- Does this only happen if you reboot? What if you just reload reef-pi ?

it still happens when i just reload

- When you next encounter this, instead of deleting /adding them, can you just wait few seconds and reload reef-pi, and see if that resolves this.

no it didn't resolve it.

mine is a pi 3

 

[PaulJ2303]

Just to try and add to the knowledge about this I did notice that when I reloaded reef-pi the first time it lost one sensor then after I waited some time and reloaded it again to see if it resolved the issue it lost the second sensor.

 

[Ranjib]

One thing that's consistent across all the temp sensor issues is they are beneath reef-pi. i.e. none of the issues indicate any bug on reef-pi code directly, yet. Almost all of them are related to sensor detection.

 

[hhaase]

I'm starting to build up my component libraries to work on schematics, and the adafruit guides are all pretty clear and easy to follow. So I doubt I'll have any questions on following the designs exactly, but I'll probably start harassing people when I come to design questions I have.

First one is if you guys are beholden to the LM2596 buck converter? There are a few small switching regulators that are more board friendly, such as the PSU5A from EzSBC. Some of the reefpi designs have such low component count and draw that I'd wonder if just an LM7805 might work. These other regulators would also eliminate the need to adjust the output. Has the actual amperage draw been measured?

For my own build I think I'll go down that route on the smaller boards to see how it plays out. Would be easy enough to add some extra holes so that either method could be used.

 

[MikeSpike]

Hi @Ranjib presume v3 is going to be released soon? If so (or at least in the 3 beta) I believe third party wireless sockets are supported. Is there a guide on how to do this?

 

[brandon clow]

Tried the new release today (reef-pi-3.0.0-rc0-pi3.deb), temp sensors seem to be working better this go. I did notice that one of my lights instantly pegged to 100% across all channels (using a PCA9685 with ldd-H meanwells). The config I had works correctly on 3.0 alpha and the 2.x builds but the latest two 3.0 builds have behaved this way. I did attempt to see if the active high/low was the issue but flipping it to active low goes to light out. The other two changed when I edited the profile but did not modify an actual value. I caught this when viewing the systemctl for reef-pi. The start times are correct but the "15:04:05" time is unknown to me as any gui entry. The profiles are using diurnal and should be outputting a value in the 40's roughly (blues at least). Has anyone seen this on the latest build?


Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 ERROR: lighting subsystem. Profile value computation error. Light: RightLEDLight channel: Violets Error: Failed to parse start time. Errorarsing time "09:15" as "15:04:05":
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 lighting subsystem: Setting Light: RightLEDLight Channel: Violets Value: 0
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 lighting-subsystem: Setting PWM value: 0 at channel: 8
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 onTime 0 offTime 4095
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 lighting-subsystem: Loading profile for channel Blues
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 ERROR: lighting subsystem. Profile value computation error. Light: RightLEDLight channel: Blues Error: Failed to parse start time. Errorarsing time "08:15" as "15:04:05": c
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 lighting subsystem: Setting Light: RightLEDLight Channel: Blues Value: 0
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 lighting-subsystem: Setting PWM value: 0 at channel: 6
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 onTime 0 offTime 4095

Rolled back to 3.0 alpha and lights restored to the correct values.

 

[Tplummer]

You t

Welcome to the development thread of reef-pi, an open-source, affordable, modular DIY reef-tank controller based on Raspberry Pi.

If you are trying to build a reef-pi controller, please check out the build guides linked below. You can use this thread to seek help and to stay updated with the latest development.

reef-pi provides following functionalities:

• Equipment control or automated power strip (including digital timers to switch on/off equipment at periodic intervals)
• LED light control (24 hour cycle), diurnal cycle, fixed dimmig
• Auto Top Off (based on photoelectric water level sensor or float switch)
• Temperature monitoring and control (switch on/off heater & chiller)
• pH monitoring
• Dosing automation
• Macro (feed mode etc)
• And several other features like camera controls (periodic image capture and upload on google drive), telemetry (dashboard and alerting based on sensor values using adafruit.io), mobile friendly UI etc.

A beginner friendly list of reef-pi build guides are available on adafruit.io, if you are planning to build a reef-pi controller start with them.

1. Setup and installation
2. Power controller
3. Temperature controller
4. Auto Top Off
5. Light controller
6. pH monitor

Since reef-pi is modular and customizable, the cost of individual builds varies. In our experience reef-pi controller are almost always cheaper than the commercial alternatives and costs anywhere between 100 to 550 USD. An example bill of materials can be found here as amazon list. If you are building a reef-pi controller, I highly encorage you start a build thread and tag it with #reefpi to seek help with your build


--------------------------- Original thread start ---------------------
Hello friends,
I am starting a master thread for a DIY reef tank controller that is based on all opensource software and hardware. I had started with this project with following goals:
- Build an opensource software and hardware based reeftank controller that anyone can use and hack
- Keep the controller well tested against a set of standard equipment I use, while make it possible for other equipment to be used
- Maintain my own tank build threads and their evolution, backed by this controller.
- Learn electronics and reef keeping along the way (I am biologist by education & software engineer by occupation)

Following are the details of what I have developed & tested till now:
Controller software:
- The main controller software is called reef-pi. It is written in go, its fast and performant. It used embd for hardware communication. It also runs a little web server which provides the UI for the controller. UI is written in React , and can be accessed from anything that has web browser. As of now, the reef-pi supports following things:
- AC 110/220 V socket on/off (uses optocoupled relays underneath). This is used to remotely switch on/off return pump, LED bulb, heater etc.
- 0-20 Volt DC PWM. This is used to control things like dimmable LEDs, DC pumps. I use it to control my Kessil A80, A360. As well as a 5V DC pump (i use it as wave maker in one of my pico with SPS). PWM is done using PCA9685 breakout board from adafruit. This board supports 16 channels. I am using only 5 at max, as of now.
- Analog sensor support using MCP3008 analog to digital converter. This supports 8 channels. i.e. I can use up to 8 sensors/probes. As of now, I have only tested temperature sensor with it.
Other than the following basic hardware/equipment support, the controller software provides following additional features:
- Setting up daily/weekly/monthly scheduled jobs. Like turning on/off certain equipment. I use this feature to run my AC20 HOB filter nightly.
- Setting up dusk-to-dawn like lighting using PWM. The UI gives 12 vertical slider based 0-24 hours (each slider representing 2 hours gaps). I combine two of these two control the kessil LEDs.
- Authentication using Google OAuth. Since the whole controller runs on raspberry pi and provides an web frontend, security of one of my concern. To mitigate this, I have added google authentication support, which when configured will allow only certain users to log in (email specified in the configuration value).

This is the summary of the core controller software and what features it has as of now. I am constantly writing/improving things around it. As of now, I have two reef tank powered by this. I 'll have seperate threads on the individual builds that will include the housing and ancillary hardware in nano-reef.com, as my tank build threads are there, but I'll share all the controller specific common bits here, including UI screen shots, and tank pics

you take pull request?

 

[Ranjib]

You t


you take pull request?

Yes, and give free hugs in return

 

[Ranjib]

Tried the new release today (reef-pi-3.0.0-rc0-pi3.deb), temp sensors seem to be working better this go. I did notice that one of my lights instantly pegged to 100% across all channels (using a PCA9685 with ldd-H meanwells). The config I had works correctly on 3.0 alpha and the 2.x builds but the latest two 3.0 builds have behaved this way. I did attempt to see if the active high/low was the issue but flipping it to active low goes to light out. The other two changed when I edited the profile but did not modify an actual value. I caught this when viewing the systemctl for reef-pi. The start times are correct but the "15:04:05" time is unknown to me as any gui entry. The profiles are using diurnal and should be outputting a value in the 40's roughly (blues at least). Has anyone seen this on the latest build?


Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 ERROR: lighting subsystem. Profile value computation error. Light: RightLEDLight channel: Violets Error: Failed to parse start time. Errorarsing time "09:15" as "15:04:05":
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 lighting subsystem: Setting Light: RightLEDLight Channel: Violets Value: 0
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 lighting-subsystem: Setting PWM value: 0 at channel: 8
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 onTime 0 offTime 4095
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 lighting-subsystem: Loading profile for channel Blues
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 ERROR: lighting subsystem. Profile value computation error. Light: RightLEDLight channel: Blues Error: Failed to parse start time. Errorarsing time "08:15" as "15:04:05": c
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 lighting subsystem: Setting Light: RightLEDLight Channel: Blues Value: 0
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 lighting-subsystem: Setting PWM value: 0 at channel: 6
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 onTime 0 offTime 4095

Rolled back to 3.0 alpha and lights restored to the correct values.

Thank you so much for reporting this. I think this is a driver bug :-( . I’ll start working on a fix

 

[hhaase]

Hmmmm, main board really isn't too bad at all. Even managed to get the level sensor circuits integrated into it, so won't need a breakout board.

If I can get a few things cleaned up I'll get some prototypes ordered tonight.

 

[Ranjib]

I'm starting to build up my component libraries to work on schematics, and the adafruit guides are all pretty clear and easy to follow. So I doubt I'll have any questions on following the designs exactly, but I'll probably start harassing people when I come to design questions I have.

First one is if you guys are beholden to the LM2596 buck converter? There are a few small switching regulators that are more board friendly, such as the PSU5A from EzSBC. Some of the reefpi designs have such low component count and draw that I'd wonder if just an LM7805 might work. These other regulators would also eliminate the need to adjust the output. Has the actual amperage draw been measured?

For my own build I think I'll go down that route on the smaller boards to see how it plays out. Would be easy enough to add some extra holes so that either method could be used.

I went with those because I didn’t know how the circuit will end up at the end. I can use them to feed 24/12/5v and get 10,12,5,3.3 out which I covers my entire range of use case. And their 3a rating helps as well. I definitely think you should tinker with better more appropriate converters. Unless we tinker we’ll never know. If you ask me , our current power circuitry is not good and I know for a fact things get dodgy (10v pwm) in all in one circuits.
My theory is that we need lots of passive (resistors and capacitors) to make things smooth , other than better converters . From power distribution perspective.

Any help in this regard will be tremendously useful

 

[Ranjib]

Hi @Ranjib presume v3 is going to be released soon? If so (or at least in the 3 beta) I believe third party wireless sockets are supported. Is there a guide on how to do this?

Yes , thanksgiving is the target date . I am in process of updating the adafruit guides for that reason. Expect those to go live around same time (may be couple of days before or after)

 

[Ranjib]

Yes , thanksgiving is the target date . I am in process of updating the adafruit guides for that reason. Expect those to go live around same time (may be couple of days before or after)

I think I have posted it here. To use the Wifi outlets, add a driver with the proper type (only hs103,hs110,hs300 supported) and add IP:9999 as address. Then proceed to create outlets with that driver

 

[hhaase]

I went with those because I didn’t know how the circuit will end up at the end. I can use them to feed 24/12/5v and get 10,12,5,3.3 out which I covers my entire range of use case. And their 3a rating helps as well. I definitely think you should tinker with better more appropriate converters. Unless we tinker we’ll never know. If you ask me , our current power circuitry is not good and I know for a fact things get dodgy (10v pwm) in all in one circuits.
My theory is that we need lots of passive (resistors and capacitors) to make things smooth , other than better converters . From power distribution perspective.

Any help in this regard will be tremendously useful

A lot of the issue may also come from using a switching regulator, as opposed to a linear. Switchers can be noisy, but I can start testing circuits with the scope to see what might be noisy. A weak grounding circuit can also play a lot into that. Tying everything into a single PCB will help with grounding quite a lot, since there is such a robust ground between the Pi itself and any hat boards.

I'll know better in about two weeks. Generating the files now, ordering shortly. I don't think I have enough posts yet to link images. Just shooting for a workable prototype at the moment anyway but I think it's looking clean for a quick night of work.

 

[Bigtrout]

Tried the new release today (reef-pi-3.0.0-rc0-pi3.deb), temp sensors seem to be working better this go. I did notice that one of my lights instantly pegged to 100% across all channels (using a PCA9685 with ldd-H meanwells). The config I had works correctly on 3.0 alpha and the 2.x builds but the latest two 3.0 builds have behaved this way. I did attempt to see if the active high/low was the issue but flipping it to active low goes to light out. The other two changed when I edited the profile but did not modify an actual value. I caught this when viewing the systemctl for reef-pi. The start times are correct but the "15:04:05" time is unknown to me as any gui entry. The profiles are using diurnal and should be outputting a value in the 40's roughly (blues at least). Has anyone seen this on the latest build?


Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 ERROR: lighting subsystem. Profile value computation error. Light: RightLEDLight channel: Violets Error: Failed to parse start time. Errorarsing time "09:15" as "15:04:05":
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 lighting subsystem: Setting Light: RightLEDLight Channel: Violets Value: 0
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 lighting-subsystem: Setting PWM value: 0 at channel: 8
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 onTime 0 offTime 4095
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 lighting-subsystem: Loading profile for channel Blues
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 ERROR: lighting subsystem. Profile value computation error. Light: RightLEDLight channel: Blues Error: Failed to parse start time. Errorarsing time "08:15" as "15:04:05": c
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 lighting subsystem: Setting Light: RightLEDLight Channel: Blues Value: 0
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 lighting-subsystem: Setting PWM value: 0 at channel: 6
Nov 13 18:39:44 raspberrypi reef-pi[14375]: 2019/11/13 18:39:44 onTime 0 offTime 4095

Rolled back to 3.0 alpha and lights restored to the correct values.

I have seen this behavior as well. I run 4 channels of lighting off the pca9685 and half my channels are unusable on 3.0 beta. 3.0 alpha works fine. Thanks for the heads up on the latest 3.0!

 

[Bigtrout]

A lot of the issue may also come from using a switching regulator, as opposed to a linear. Switchers can be noisy, but I can start testing circuits with the scope to see what might be noisy. A weak grounding circuit can also play a lot into that. Tying everything into a single PCB will help with grounding quite a lot, since there is such a robust ground between the Pi itself and any hat boards.

I'll know better in about two weeks. Generating the files now, ordering shortly. I don't think I have enough posts yet to link images. Just shooting for a workable prototype at the moment anyway but I think it's looking clean for a quick night of work.

I have been using meanwell RS series 5v power supplies exclusively one for the pi and one for my relay boards without a hiccup for 8 months now.

I did however have big noise issues with the cheap power bricks supplying my beamswork LED lights. They were very noisy and causing temp probe and ph probe errors. Switching them out with quality power bricks rated at double the current needed fixed the problem.

When choosing a power supply its never a wise idea to run at more than 80 percent of the rated load. 50 percent is even better. When switching power supplies approach 100 percent of rated load weird things happen with voltages, noise, waveforms and frequencies, etc.

The buck convertors used for reef pi, the lm2596 are a good circuit...with one caution!
The cheap boards from.china are very variable. Some come with properly sized and speced components and work fine. Others are copies and the passive components seem to be sized at whatever the factory had a bunch of and are very unreliable. I chose meanwell power supplies because they are at least UL listed and good quality.

As an example of the china electronic board variability, i used mosfet modules to interface my beamswork lights to the pca 9685 pwm. The board I used works great. But if you look on amazon and ebay and the chinese good sites, they all have similar modules but careful looking at the pics of them, half of them use passive components which would not allow you to use a pwm frequency much above 100hz. The ones I got allow me to use up to 1000hz. All this because they used 2
resistors that were not the proper resistance for the circuit design.

 

[hhaase]

That pretty much nails my concern with the LM2596 modules, cheap switching power supplies can have a huge negative effect on a processor system. Particularly ones that are made to such a vague standard with so little documentation. Given how prolific and varied the LM2596's are out there, I'd assume most of the IC's are counterfeit's on them, and god knows what the capacitor tolerances are. Switchers are just too easy to make poorly. I'm not great at designing power supplies, in fact I try to avoid it due to liability risks, but I can spot bad ones pretty easily.

That's really one of the hiccups in open-source systems in general. Once a design hits the Shenzhen / Amazon / Ebay pipeline, you have no idea what you're going to receive. Just look at all the knockoff arduino's, Pi's, and 3D printer controllers out there. Definitely something to consider in this project if it gets enough traction.

 

[Bigtrout]

That pretty much nails my concern with the LM2596 modules, cheap switching power supplies can have a huge negative effect on a processor system. Particularly ones that are made to such a vague standard with so little documentation. Given how prolific and varied the LM2596's are out there, I'd assume most of the IC's are counterfeit's on them, and god knows what the capacitor tolerances are. Switchers are just too easy to make poorly. I'm not great at designing power supplies, in fact I try to avoid it due to liability risks, but I can spot bad ones pretty easily.

That's really one of the hiccups in open-source systems in general. Once a design hits the Shenzhen / Amazon / Ebay pipeline, you have no idea what you're going to receive. Just look at all the knockoff arduino's, Pi's, and 3D printer controllers out there. Definitely something to consider in this project if it gets enough traction.

Exactly my thoughts...I know reef-pi is a open source DIY project but sometimes a few dollars cheaper is not worth it. Especially with power supplies. At least when you get a Meanwell you get a UL listed part and something with well documented performance specifications.

 

[Ranjib]

That pretty much nails my concern with the LM2596 modules, cheap switching power supplies can have a huge negative effect on a processor system. Particularly ones that are made to such a vague standard with so little documentation. Given how prolific and varied the LM2596's are out there, I'd assume most of the IC's are counterfeit's on them, and god knows what the capacitor tolerances are. Switchers are just too easy to make poorly. I'm not great at designing power supplies, in fact I try to avoid it due to liability risks, but I can spot bad ones pretty easily.

That's really one of the hiccups in open-source systems in general. Once a design hits the Shenzhen / Amazon / Ebay pipeline, you have no idea what you're going to receive. Just look at all the knockoff arduino's, Pi's, and 3D printer controllers out there. Definitely something to consider in this project if it gets enough traction.

For the right reason, we can definitely recommend quality parts. In fact raspberry pi and adafruit parts are kinda expensive compared to their alternatives. I agree amazon provides a lot more options, including the bad ones, but its up to us to reduce that set to few proven ones as and when we learn from using them. In general, I think Amazon is much easier and safer place to recommend as you have a single place to get everything, most things can be returned, its not as wild as ebay where you are totally on your own judgment. Like these power supplies, I don't think they are the worst. We are seeing issues with AIO circuits and as we know more about their working state, but for all other things they are working. As in I am using multiple lm2596 converters for over three years across temperature, ato, ph, power controller without any issue.. all of them run pi as well.
If you have any good recommendations, do let us know, happy to include them if they are better, even if its slightly more expensive. Price is not the main benefit for amazon, its easy and relatively better quality controlled and available to much wider audiance. And I am confident that if we come up with very good guides and recommended parts, we can contain the hardware quality issue to a large extent. Its an opensource, hacking friendly system, so folks will definitely hotrod it (and i love that too), in those cases they are on their own..

 

[Bigtrout]

And I am confident that if we come up with very good guides and recommended parts, we can contain the hardware quality issue to a large extent.

This would solve alot of issues that happen...if the hardware is good it makes troubleshooting much easier when someone new decides to build a reef-pi.