85 replies to this topic

[Steve Scheifler]

It's a good idea. I've been involved in the open hardware community for about 7 years now, the last 5 owning an open hardware based fabrication company. For you guys looking just to get something on the cheap this isn't the project for you. Designing and building open hardware is more than just looking to get something on the cheap. It's a process and until it gets past the first several iterations and becomes more mature it's not for end users but early adopters. For guys like Benchy and Steve, this isn't for them. These sorts of projects are for people that want to tinker and experiment and do their own thing.
...

Actually, exactly my kind of thing normally, but as I said more as an interesting project than a way to flll a real need on the cheap. I have always been the tinkerer. In the 70s I was building tuned-port speaker enclosures and passive crossovers making circuit boards with essentially a crayon and etching solution, then graduating to active multi-amp crossovers and double-sided boards using transfer sheets instead of wax pencils.

Anyway, other than that I agree with your post and the ones that followed.

And Shapiros is racer-friendly. Not the absolute cheapest but a great place to walk around picking through cut offs, scaps, etc and pay cash by the pound. Been doing that since I made my own sway bars for a 1969 Datsun 510, in 1975.

[Bench Racer]

Speaking of the Scheifler's, did you guys know there's a race 6 hours northeast of you this weekend?

[Andy Mitchell]

Been watching this thread with interest, mainly because I built something similar last winter (I was trying to wean myself off of borrowing from a friend all the time).

 

I used Chinese 500 kg load cells with individual off-the-shelf Sparkfun amplifiers. Each pad connected to an Arduino via a 15 meter plug-in network cable. The electronics worked great, no problem, but I my welded-up 15"x15" scale pads (I admit, made out of scrap angle iron laying around the shop) turned out to be a lot more flexy that I had imagined they would be. I have access to an Instron tester for calibration/testing, and got good accuracy from the electronics, but had a 'spring constant' for my pads of about 4600 pounds/inch. That meant that 560 lbs (my typical front wheel load) would deflect my home-built scales about 0.12 inches... which I'm thinking is not ideal. Thus my earlier comments about the design of the pads themselves being something important to think about.    

[Andy Mitchell]

I also, I had to mill the load cell mounting pads to get the pads to sit acceptably parallel.. just something else to keep in mind.

 

PS - I would be happy to post pics, graphs etc. of how not to design scale pads, if anyone is interested. Probably be the first of the week before I'd be back in the shop to do it, though.

[Andy Mitchell]

Don't know if this works, but here's a link to some pics of my home-built scales that I put up on facebook:

 

https://www.facebook...=1&l=b3d45c53ad

 

Please have mercy on me, I know my fab skills suck. And as I said, they're pretty flexy. I'm mainly showing you this so that everyone will know what not to do, lol. 

[Steve Scheifler]

Andy, you are too hard on yourself!

But correct me if I'm mistaken, it appears that based on the Instron the scale weight is off by 8-10 lbs in the range we are using. If so, that certainly won't fly.

[OrangeCrush86]

Don't know if this works, but here's a link to some pics of my home-built scales that I put up on facebook:

 

https://www.facebook...=1&l=b3d45c53ad

 

Please have mercy on me, I know my fab skills suck. And as I said, they're pretty flexy. I'm mainly showing you this so that everyone will know what not to do, lol. 

 

Interesting. Did you happen to test your pad with a loadcell where you didn't cut the wire? In general you should not cut loadcell leads and avoid splicing them. I'm curious if your accuracy goes up if you don't cut and direct wire.

[Andy Mitchell]

Ah, I'm using a chunk of open source code to process the data, and it (in turn) wants you to supply a software 'calibration constant'. It appears I'm a little off on the number I supplied. A tweak or two and I think they'd be pretty much bang-on, as the % error between the two outputs is pretty much constant. The electrical stuff for a project like this is actually easy, it's the pads that require some thought and engineering.

 

In my own defense, I built these things out of scraps back in the break between Christmas Day and New Year's, and just sort of lost interest when they didn't to do what I wanted them to first time out. I think I could build a Mark II version now that would be better. 

[Andy Mitchell]

Interesting. Did you happen to test your pad with a loadcell where you didn't cut the wire? In general you should not cut loadcell leads and avoid splicing them. I'm curious if your accuracy goes up if you don't cut and direct wire.

I'm unaware of any reason to preserve load cell leads as supplied. Maybe there is one and I just don't know it though. There's only a bunch of strain gauges in there, and I wire those up all the time with random lengths of leads without any ill effects. 

 

Anyway, I didn't want to talk to the load cells over long, noisy analog wires so I shortened them. Those little amps talk to the arduino via a digital connection, and they work fine (noise free) over the 15 meter Ethernet cables I picked up at the dollar store. Seemed like a practical way to do things. 

[Bench Racer]

Not slamming Andy or anyone else design/fab, but just maybe there's very good reason to look again at what scale manufacture's do with their casting design. Shown is the bottom side for a dual cell scale, single cell scales are very similar.

 

http://www.usracegea...oaded/72725.jpg

[BNaumann]

I initiated a pull request that has rs format gerbers generated from Fritzing for your repo. I ran them through the gerber viewer at Seeed and they passed. I almost bought a batch last night but decided to wait to see if that rev was rtm. If you put it in another branch when you finish I'll merge it to my repo and generate some gerbers and place an order. Cost (which I'll gladly pick up) is about $5 for the parts (we get a nominal 10 parts per order) and $20 for DHL from Seeed. Dangerous Prototypes is about $17 including shipping but they'll take a week or two longer via China post.

I was going to convert to Eagle last night but I don't have a footprint for the NX711 carriers. I'll have to make one. It was my first time using Fritzing. Pretty cool, much more maker friendly than Eagle. Sparkfun has a Fritzing library.

I haven't put those python scripts on a Pi yet. I've only got 1 Pi3, a few Pi2 and those are being used in production but I've got some B+ laying around (and a Pi TFT). That should be enough to demonstrate the concept. Once I get that up and working I'd like to try it with a Pi ZeroW.

I didn't understand one word of that. And that's why I paid Longacre 800 bucks.

[Bench Racer]

^ I really hate to pick on Mr. Stevens, BUT, I whole hartedley agree with you. No one wants to talk about the platform casting design and why all scale manufactures today have same platform designs.

 

I get what OrangeCrush is wanting to do, he's into an ingenious design per say come hell or high water and I wish him well.

[BNaumann]

That was more a compliment that those guys are way above my level. As far as the mechanical side goes, I'm no engineer, but I'm sure the pads will be fine as long as they use a baseplate of prefamulated Amulite, surmounted by a malleable logarithmic casing.

[dstevens]

^ I really hate to pick on Mr. Stevens, BUT, I whole hartedley agree with you. No one wants to talk about the platform casting design and why all scale manufactures today have same platform designs.

 

I get what OrangeCrush is wanting to do, he's into an ingenious design per say come hell or high water and I wish him well.

Dewey I was talking about the pads being the most challenging part of it in a low cost build.   You can build them but they're going to be pretty heavy from steel and if you're building one offs AL will be a bit spendy.  In the end it's a scale, not rocket science.  The reason those pads are cast that way is because at production scale it's the most cost effective way to build them not because they are necessarily better.  

 

I'm waiting on the load cells and amps, Andrew is checking the amp footprint and when he releases that I'll have some boards made.  With this electronics package and the app you should be able to take your entry level limited feature system and upgrade it with more features.  Speedway is already selling import off brand scales closer to $500.

[Steve Scheifler]

Perhaps the competition will bring prices down. What has always bothered me the most has been the upcharge for dual-cell pads. I would glady pay an extra few hundred, but they charge closer to $1000. Twice now they've missed an easy up-sell.

[Andy Mitchell]

I gotta say thanks to Mr. Orange (and dstevens) for reviving my interest in this. I put the project on the shelf a couple of months ago, and only tacked on that square stock brace yesterday afternoon as an afterthought after I promised to show some pics and data. It made quite a big difference to the stiffness of the bottom pad though, so now I'm thinking maybe I'll actually try to use them some time.

 

The bottom pad stiffness with the brace went up to about 6950 pounds//inch, meaning that you'd see less than 0.1" deflection at any corner when weighing a Spec Miata or similar. I have no idea of how that would compare to a commercial scale (but I think I'll try to measure next time I borrow the set I've been using all  along).

 

With regard to Steve S.'s comment on accuracy, I think you would have to individually calibrate individual pads to get accuracy within the pound or two he was seeking. The load cells I bought were mass-produced things, they weren't individually calibrated, and I'm guessing you would see some scatter from example to example. The numbers I put up were just using the nominal sensitivity printed on the cells from the OEM, and appear to be off about 1.4% or so.  Here's a better look at my chart and numbers:

 

https://www.facebook...60&l=c549e095a9

 

The calibration issue could be fixed easily, but would mean labeling pads for specific locations and using them accordingly.

 

Now somebody else post something and show me how to do it better! I get so many great tips from this site, it's enjoyable to share something back for a change.

[Steve Scheifler]

I assume that the proposed software will provide calibration per pad, something I wish we had access to with the commercial products. Having them recalibrated is a hassle and costly. After that it comes down to repeatability, which is where pad construction becomes important. Even the "better" of the available single-cell sets require some care to get consistent results.

[OrangeCrush86]

I assume that the proposed software will provide calibration per pad, something I wish we had access to with the commercial products. Having them recalibrated is a hassle and costly. After that it comes down to repeatability, which is where pad construction becomes important. Even the "better" of the available single-cell sets require some care to get consistent results.

 

The option to calibrate 0 and span per pad is in my backlog. In my opinion it's a requirement and I'm surprised that commercial scales don't make it easy to calibrate.

[OrangeCrush86]

I gotta say thanks to Mr. Orange (and dstevens) for reviving my interest in this. I put the project on the shelf a couple of months ago, and only tacked on that square stock brace yesterday afternoon as an afterthought after I promised to show some pics and data. It made quite a big difference to the stiffness of the bottom pad though, so now I'm thinking maybe I'll actually try to use them some time.

 

The bottom pad stiffness with the brace went up to about 6950 pounds//inch, meaning that you'd see less than 0.1" deflection at any corner when weighing a Spec Miata or similar. I have no idea of how that would compare to a commercial scale (but I think I'll try to measure next time I borrow the set I've been using all  along).

 

With regard to Steve S.'s comment on accuracy, I think you would have to individually calibrate individual pads to get accuracy within the pound or two he was seeking. The load cells I bought were mass-produced things, they weren't individually calibrated, and I'm guessing you would see some scatter from example to example. The numbers I put up were just using the nominal sensitivity printed on the cells from the OEM, and appear to be off about 1.4% or so.  Here's a better look at my chart and numbers:

 

https://www.facebook...60&l=c549e095a9

 

The calibration issue could be fixed easily, but would mean labeling pads for specific locations and using them accordingly.

 

Now somebody else post something and show me how to do it better! I get so many great tips from this site, it's enjoyable to share something back for a change.

 

Nice work. By the way, if I can get a little more time to work on the software, you should be able to use the code with your configuration. The sparkfun loadcell amps use the HX711 so I think it will be compatible. If you know Python (or want to learn) I invite you to contribute no Github.

 

I think you said you are currently using an Arduino? You would have to swap this for a Raspberry Pi 3.

[Steve Scheifler]

The option to calibrate 0 and span per pad is in my backlog. In my opinion it's a requirement and I'm surprised that commercial scales don't make it easy to calibrate.

They have a zero function of course, but not the ability to calibrate each to a reference weight. I could of course accomplish the same with small weights and the zero function but it should be simple.