[SOLVED] Load cell issue may need to rethink

I recently ordered and received a 1kg load cell with the hx711 ADC Module (cheapy from eBay) to possibly use to weigh out small amounts of fish food. But this thing returns very erratic numbers.

I'm using the HX711_ADC library by Olav Kallhovd. I used the included sample sketch to "calibrate" this load cell. And then wrote a sketch that takes measurements over a 2 minute period and stores the highest reading and the lowest reading with no additional weight added.

It reports readings as low as -0.28 all the way up to 1.52. A variation of 1.8 grams. Watching the readings scroll by, I can see it jump from negative numbers to 0.50 (or more) in an instant.

After researching, I am convinced the issue is electrical noise. However, I don't have a way to prove that it is. I read that a 0.1uF capacitor could be placed between the data and clock lines that would help "smooth" the signal. Well, I tried that with a ceramic 0.1uf and get 0s regardless of weight added. Maybe I misunderstood where to place the cap?

I'm looking for a cost effective way to measure amounts of flake in the ballpark of 0.3 to 0.5 grams. But with a swing as wide as this thing reports, it's not possible as it is.

If anyone knows a way to gain much better accuracy/consistency out of one of these cheap load cells, I'm all ears (or eyes as is this case when reading).

Otherwise, does anyone have an opinion on a different approach to weighing small amounts that is in the typical hobbyist's price range?

And upon further reading, I think I bought a poorly made HX711 module.

If anyone knows a place to order a quality HX711 module, I'll try that route before I give up on using this load cell/module.

I bought 4 cheap ones from eBay and AliExpress. I gave up on calibrating with the library. I finally just solved for slope and offset of a line with a two point calibration. When I get to my computer I will load the code for you. It has worked great for all of them. It is basically y = mx + b.

Anthony

Did you power the HX711 module with 5volt.
Seen several cases here with people trying to connect the HX711 module to a 3.3volt processor.
That will only work with modules that have two power connections.
A 5volt supply to the analogue part of the HX711, and a 3.3volt supply to the digital part of the HX711.
Sparkfun is AFAIK the only retailer with such a board.
Leo..

aiannar974:
I bought 4 cheap ones from eBay and AliExpress. I gave up on calibrating with the library. I finally just solved for slope and offset of a line with a two point calibration. When I get to my computer I will load the code for you. It has worked great for all of them. It is basically y = mx + b.

Anthony

I would do that as well if I felt calibrating was the issue. But regardless of what calibration factor I use, the resulting readings jump all over.

Wawa:
Did you power the HX711 module with 5volt.
Seen several cases here with people trying to connect the HX711 module to a 3.3volt processor.
That will only work with modules that have two power connections.
A 5volt supply to the analogue part of the HX711, and a 3.3volt supply to the digital part of the HX711.
Sparkfun is AFAIK the only retailer with such a board.
Leo..

I'm using an UNO to test with. And powering the HX711 directly from the 5v pin. Come to think of it, I never looked to see what the power draw is on the HX711. Another rookie mistake. But, as I see now, 1.5mA shouldn't be an issue for the UNO with nothing else attached.

I also tried using data pins for the sck and dt lines. As well as analog pins. Neither seemed better than the other.

One other thing I read that "could" help was to add a delay(1) after each reading. Which also resulted in no noticeable difference.

If I were home, I'd attach my sketch. However, it just gets the load cell data and checks if it's higher or lower than what has already been stored as high and low over a specified time. It's so basic I really don't think it's the problem though.

Thinking I may order one of the SparkFun boards. See if it does any better.

Just a follow up.

I ordered a different HX711 module. And, after running the same calibrations and tests, it would appear that the first module was indeed the issue. While waiting for the new module, I ran more (and longer) tests on the first one. Those tests resulted in a nearly 4.5 gram swing.

Once the new module arrived, I put it through the same tests and it did much better. The lowest reading I recorded was -.03g. And the highest was +0.05g. So a swing of 0.08g.

And, while not perfect, it should meet the needs of my project. Time will tell.

Thanks everyone!