Reef-Pi Controllers: New all-in-one with power bar, and the original Pico

OP
OP
theatrus

theatrus

Valuable Member
View Badges
Joined
Mar 26, 2016
Messages
1,941
Reaction score
3,329
Location
Sacramento, CA area
Rating - 0%
0   0   0
Just a small vignette on potted sensor development. The little sensor boards from earlier are soldered down, attached to a polyurethane jacketed cable (short for testing), and a quick little body for the sensor is made from acetal tube with a final OD of 0.5". The bottom is squared mainly out of prototypiness and used a pair of scissors on the acetal sheet - will get something proper cut later.

I made one earlier which used an old batch of MG Chemicals Epoxy I had on the Shelf of Dubious Chemicals and Glue, and is now in the hazardous waste disposal pile because it was total garbage, and even a bad thermal conductor. The final epoxy I'd like to use is a canister dispense unit but wanted to knock something up with a pot since I can use far less and not waste it in the mixing tubes for the cartridges.



IMG_7342.jpg


Trying to speed up the cure of the epoxy in my lab oven at 65C.

No I don't eat from those springform pans, they were donated to the cause ;)
 
Last edited:
OP
OP
theatrus

theatrus

Valuable Member
View Badges
Joined
Mar 26, 2016
Messages
1,941
Reaction score
3,329
Location
Sacramento, CA area
Rating - 0%
0   0   0
I've been doing a bunch of prototyping and sorting out my wire inlet problems. In short, they were a mess since there was nearly not enough space on the first shot boards to route cables. Cables and wires in any situation occupy some sort of non-euclidean space and defy logic, so its very hard to simulate these things without a ton of other work to figure out bend radius, possibility for a human to install, etc etc.

It turns out there isn't enough clearance through to the top of the case, and the quick connect tabs are simply in an un-routable position, AND using the PCB as the ground bonding point is contraindicated.

Example of a partially built board doing fit checking:

IMG_7354.jpg


Simply no good.

But, we have plenty of options, including this fake PCB I 3d-printed to check component positioning, and this new routing scheme:


IMG_7378.jpg


This uses 90 degree flag quick connects, and pushes the whole power supply left a good distance. The ground wire is now running to a ground ring, which will act as the ground bond point AND connect to the PCB, giving a lot of rigidity. The extra space also makes it easy to mount the CT for current measurements.

Untitled-1.jpg


The standoffs act to tie all the case work and PCB together here, and reduce board flex when inserting and removing connectors. I'll probably order a replacement set of boards this week while I finish designing the main logic board.

Some other updates:

- The initial design passes a variety of HiPot tests. The target voltage for this type of design is around 1.2kV of insulation to ground, which passes from the PCB and PCB in case perspective so there were no major goofs there:


IMG_7379.jpg


(Test was run over multiple minutes and different parameters, so don't fret for the 1s test time for this picture).

- I'm getting quotes to streamline machining the top of the enclosure - the first few will be done in house. But thankfully everything fits, thanks to this 3d printed mockup of a machined top:


IMG_7426.jpg


I also did some more testing for connectors, and still fell back on the JST XA series. The closest contender was the Molex Nano-Fit (actually used on the APEX for the power connectors). Very good connector, 4.5x the cost, and it presents a unique problem of giving a very wide "lever arm" on the PCB when on the strip which will inevitably lead to lots of broken board connectors. I'm going to keep it in the back pocket, but sticking with the slightly uglier but very functional XA series (top connector).

IMG_7371.jpg


Thanks for tagging along with this product design journal. I'm going to do a good write-up summary at some point, including development costs (not cheap). Maybe I'll need to do a Kickstarter to get good volume pricing for folks ;)

Also, I did a teardown of "that other brand" for interest value since the last "other brand" commercial controller I owned was last generation. You can find it on Twitter (same username as here).

Next update will be on sensor assemblies, and on the firmware architecture - which means more diagrams and less pictures. I'm still planning on keeping that open source, since that is the power of the community to built what _they_ want, not what commercial partnerships you want.

Also, _IMPORTANT QUESTION_ if you made it this far: What should I even name this thing? :-D
 
Last edited:

robsworld78

Well-Known Member
View Badges
Joined
Feb 14, 2020
Messages
952
Reaction score
1,280
Location
Edmonton, Canada
Rating - 0%
0   0   0
I've been doing a bunch of prototyping and sorting out my wire inlet problems. In short, they were a mess since there was nearly not enough space on the first shot boards to route cables. Cables and wires in any situation occupy some sort of non-euclidean space and defy logic, so its very hard to simulate these things without a ton of other work to figure out bend radius, possibility for a human to install, etc etc.

It turns out there isn't enough clearance through to the top of the case, and the quick connect tabs are simply in an un-routable position, AND using the PCB as the ground bonding point is contraindicated.

Example of a partially built board doing fit checking:

IMG_7354.jpg


Simply no good.

But, we have plenty of options, including this fake PCB I 3d-printed to check component positioning, and this new routing scheme:


IMG_7378.jpg


This uses 90 degree flag quick connects, and pushes the whole power supply left a good distance. The ground wire is now running to a ground ring, which will act as the ground bond point AND connect to the PCB, giving a lot of rigidity. The extra space also makes it easy to mount the CT for current measurements.

Untitled-1.jpg


The standoffs act to tie all the case work and PCB together here, and reduce board flex when inserting and removing connectors. I'll probably order a replacement set of boards this week while I finish designing the main logic board.

Some other updates:

- The initial design passes a variety of HiPot tests. The target voltage for this type of design is around 1.2kV of insulation to ground, which passes from the PCB and PCB in case perspective so there were no major goofs there:


IMG_7379.jpg


(Test was run over multiple minutes and different parameters, so don't fret for the 1s test time for this picture).

- I'm getting quotes to streamline machining the top of the enclosure - the first few will be done in house. But thankfully everything fits, thanks to this 3d printed mockup of a machined top:


IMG_7426.jpg


I also did some more testing for connectors, and still fell back on the JST XA series. The closest contender was the Molex Nano-Fit (actually used on the APEX for the power connectors). Very good connector, 4.5x the cost, and it presents a unique problem of giving a very wide "lever arm" on the PCB when on the strip which will inevitably lead to lots of broken board connectors. I'm going to keep it in the back pocket, but sticking with the slightly uglier but very functional XA series (top connector).

IMG_7371.jpg


Thanks for tagging along with this product design journal. I'm going to do a good write-up summary at some point, including development costs (not cheap). Maybe I'll need to do a Kickstarter to get good volume pricing for folks ;)

Also, I did a teardown of "that other brand" for interest value since the last "other brand" commercial controller I owned was last generation. You can find it on Twitter (same username as here).

Next update will be on sensor assemblies, and on the firmware architecture - which means more diagrams and less pictures. I'm still planning on keeping that open source, since that is the power of the community to built what _they_ want, not what commercial partnerships you want.

Also, _IMPORTANT QUESTION_ if you made it this far: What should I even name this thing? :-D

I'm truly in awe, if you do this for ***** and giggles I can only imagine what you can pull off. :)

If the hardware is a ready to go system like Apex you'll clean up, definitely make a nice presentation and run a kickstarter, it'll flourish. Thanks for sharing so much details, it's great to see how a pro handles things.
 
OP
OP
theatrus

theatrus

Valuable Member
View Badges
Joined
Mar 26, 2016
Messages
1,941
Reaction score
3,329
Location
Sacramento, CA area
Rating - 0%
0   0   0
Name it after the clam

What if it had no name? I'm bad a naming things apparently.

(For the record, earlier this year my derasa clam from 2006 succumbed to something or another. I just cleaned off the shell after some outside sun baking this week or last.)
 

Bigtrout

Valuable Member
View Badges
Joined
Dec 16, 2018
Messages
1,189
Reaction score
2,826
Rating - 0%
0   0   0
Looking good. Nice work so far!
I can tell alot of thought is going into the design.

As someome who gets into control panels, control boxes and various electrical connection compartments on a daily basis, the fact that you thought about cables beforehand gets you a step ahead. Seems thats the last thing most designers/engineers think about, cables and connectors. Usually it ends up being something cobbled together to just make it work, and this is on commercially available equipment.

When this part of design is overlooked, cable and connector failure becomes one of the major failure points on said equipment.
 

rushbattle

Valuable Member
View Badges
Joined
Oct 19, 2016
Messages
1,347
Reaction score
1,644
Location
Equality
Rating - 0%
0   0   0
Looking good. Nice work so far!
I can tell alot of thought is going into the design.

As someome who gets into control panels, control boxes and various electrical connection compartments on a daily basis, the fact that you thought about cables beforehand gets you a step ahead. Seems thats the last thing most designers/engineers think about, cables and connectors. Usually it ends up being something cobbled together to just make it work, and this is on commercially available equipment.

When this part of design is overlooked, cable and connector failure becomes one of the major failure points on said equipment.
I don’t do it every day anymore, but I made a good living diagnosing failures in electronics subject to high shock and vibration. It’s almost always a failure in the connectors/wiring harness. It’s always shocking to me how some engineers are comfortable wadding up slack cables and jamming the wad into a void somewhere.
 
OP
OP
theatrus

theatrus

Valuable Member
View Badges
Joined
Mar 26, 2016
Messages
1,941
Reaction score
3,329
Location
Sacramento, CA area
Rating - 0%
0   0   0
New sensor revision. The original potted tube has now been covered with 3:1 shrink heatshrink to clean up the edgesand give it some smooth looks. Its not intended as a water seal in any way (the epoxy does all of that).

I opted for white because.. why not. I could also use a variety of colors including purple, maybe as designer sensors ;)

I'll pop a very small lot on the website for existing Reef-Pi users with TRS or bare wire options as a trial run, at a discount for helping test.

IMG_7448.jpg
 
Last edited:
OP
OP
theatrus

theatrus

Valuable Member
View Badges
Joined
Mar 26, 2016
Messages
1,941
Reaction score
3,329
Location
Sacramento, CA area
Rating - 0%
0   0   0
1602140955546.png


I've been doing planning and buildout of the logic control board and decided to post an annotated diagram of whats on the board. There are a few key things missing (such as an analog stage for the current sensor), but the above shows the majority of the hardware on a (very very) empty board.

- 4 sensor ports, can be either temp, float, or an external pH module running on 1-wire (and more later). Each sensor port is power controlled (resets, etc), and the whole sensor bank has an active current limit switch.
- Two power stages, the second one for the OLED is optional as it may not needed (it wants 12.5V, I have 12.0V, does it really care about 500mV?) - more testing required. The OLED is a single color, 1.5" diag afair for displaying status, and will be the last item lit up by firmware when I get to it. Including it adds to the price by $15 or so, so is the first feature to be jettisoned when the time comes :)
- CAN and USB connectors, using RJ10 (4p4c) for CAN and large USB-B for USB as its the most robust variant out there. USB has a current limit switch as well, and can back-feed the Raspberry Pi for power.
- Two buttons for buttony things (bootloader, force outlets on/off, etc)
- The main MCU is a SAME51 series, chosen for software familiarity, CAN and USB, cost, and its dual bank flash which means firmware upgrades are a lot less risky and can be done live, since a new image is written to the B bank, and then rebooted into with minimal down time. Lots of RAM, lots of flash, and even 120MHz, which is totally overkill.

The board to board cable is a 1mm flat flex cable, which is plenty for the current needs of everything plus the relays.
 
OP
OP
theatrus

theatrus

Valuable Member
View Badges
Joined
Mar 26, 2016
Messages
1,941
Reaction score
3,329
Location
Sacramento, CA area
Rating - 0%
0   0   0
Some more design updates:

Doing some display testing to check for routing of the cables. I made a mock-print of the control board, as well as one half of the side panel to check for flat flex routing. Looks like I'm spot on now:

IMG_7465.jpg
IMG_7481.jpg


3D printers make for easy mechanical mockups :)

New relay boards should show up soon, but in the interim I spent cycles kitting out the "ridiculous load tester". I ordered a custom case from Protocase with the cutouts, which saved a ton of time and actually looks really nice (though very expensive!)
IMG_7484.jpg


I also built most of the 4 relay resistor module packs. My through hole construction technique needs work, and soldering to 3oz copper planes with no thermal relief is more difficult than you think. I had to get out the "big guns" (aka the giant JBC chisel tip, on a 245 series iron, 135W peak I think)

IMG_7493.jpg


IMG_7497.jpg


I also spent some time also syncing my PerfectCarbon design https://www.reef2reef.com/threads/p...-calcium-reactor-control.752448/#post-7887880 to work on the same processor family (SAM E51/54) and connectivity (CAN + USB), which lets it all sync up and work on the same code base (one person development means everything needs to be as similar as possible or I'll go crazy). I haven't touched the PerTempCo design yet.
 
OP
OP
theatrus

theatrus

Valuable Member
View Badges
Joined
Mar 26, 2016
Messages
1,941
Reaction score
3,329
Location
Sacramento, CA area
Rating - 0%
0   0   0
Project updates!

Good news! Control board prototypes ordered. Firmware development started. Not much to report yet except a gratuitous picture:

1602827750861.png


As you can see, there are literal acres of space for ... things. Nothing really since that would just increase the cost.

Next up, the relay board and guard board (secondary insulator to prevent any chance of the relay board making contact with the control board - just a chunk of fiberglass) revision two boards showed up (the relay board is a 2oz copper board so a bit specialized). No parts mounted yet (oh no, busy week), but I can validate the alignment and check that the new grounding screws and posts work. The black bottom board is the 3d-printed control board mock up - its super rough since my PETG settings need a lot of work and I ran out of ABS.

IMG_7505.jpg


Needed some good cutouts to make the earth lock nuts sit in the right spot:

IMG_7506.jpg


Next update: The George Foreman Mk2 tester made significant built progress. I'm still missing the control board for it, but the rest of the build is nearly done:

IMG_7520.jpg
IMG_7521.jpg


This was a pretty fun project, but I'm also glad I'm only building one :-D. Need to do some wire cleanup still, and I'm powering relays up externally owing to the lack of a control board.

Not everything is fantastic news, and that bummer is overall cost. Hitting the magic <$200 is not possible with any margin that makes it worth doing. We're still well under APEX pricing here if we bundle sensors and probes and a Raspberry Pi, so its not totally out of whack (I'm using a no-reseller APEX-EL MSRP as my reference point). Pricing factors include higher cost of end caps, and higher cost of the control board (some cost reduction options there, like moving to a 2 layer board form a 4 layer, as well as a few subcircuits), higher cost of the faceplate (literally only quoted one place to laser cut aluminum) and display addons.

1602827483474.png
 

geekengineer

Active Member
View Badges
Joined
Feb 10, 2019
Messages
427
Reaction score
538
Location
South Aurora
Rating - 0%
0   0   0
Well what did I miss? LOL.. I got distracted for a while after shifting jobs. So I was looking an at one point you were talking about a water level sensor using a pressure sensor. I was just curious if you got that working. It looks like I got ton to catch up on here. Thanks.
 
OP
OP
theatrus

theatrus

Valuable Member
View Badges
Joined
Mar 26, 2016
Messages
1,941
Reaction score
3,329
Location
Sacramento, CA area
Rating - 0%
0   0   0
Well what did I miss? LOL.. I got distracted for a while after shifting jobs. So I was looking an at one point you were talking about a water level sensor using a pressure sensor. I was just curious if you got that working. It looks like I got ton to catch up on here. Thanks.

:)

The pressure sensors do work! Sadly, the ones I was testing with (NXP MPXV7002 series) are totally unobtanium since about February. I suspect they were consumed by medical device manufacturers and nothing was sent to distribution - current estimates are Feb 2021. I use it as my pandemic gauge - everything will start returning back to normal when I can get some pressure sensors.
 
OP
OP
theatrus

theatrus

Valuable Member
View Badges
Joined
Mar 26, 2016
Messages
1,941
Reaction score
3,329
Location
Sacramento, CA area
Rating - 0%
0   0   0
Still waiting on first articles for the controller PCB. As part of the process I'm iterating on the end caps, which right now are planned to be a combo of laser cut metal and an overlay of PCB on top. First shot... had my cutouts way too wide (0.020") - the snap-in power socket very much doesn't snap in for this configuration. The display side one is also much "larger", as it acts as a holder for the display, and the final panel will cover larger portions of the non-active display cable, the FPC cable, and the works:

image0.jpeg


I've also been noodling on names, and so far have ended up with ReefVolt. I had a designer do some mockups:

1603582377268.png
 
OP
OP
theatrus

theatrus

Valuable Member
View Badges
Joined
Mar 26, 2016
Messages
1,941
Reaction score
3,329
Location
Sacramento, CA area
Rating - 0%
0   0   0
Small delays due to.. you know, everything around. But, some good news:

- Maybe we won't be heading down the road to a coup.
- Initial PCBs arrived and _work_. I'm busy in a hole with firmware, but will be able to have something to show off soon.
- I still need to setup and fixture and toolpath the milling of the top case. Stand by for final... metal looks.

More updates soon!
 
Last edited:

_AV

Well-Known Member
View Badges
Joined
Oct 3, 2019
Messages
677
Reaction score
694
Rating - 0%
0   0   0
The RJ10 port is the CAN bus link which can link sensors to other control points in an autonomous way (just broadcasting messages for sensor values), in lieu of using the USB port for active control.

You so should add Apex support to this ;)
 
Back
Top