Load cell strange problem

[imanjl]

ad7715 user

Hi

I use a load cell to measure weight in range 0-50Kg ,It is connected to a 16bit ADC and PIC mcu to display the result on LCD in grams . the load cell is from Bongshin Company :

DBBP-50 Kg(S type)
**broken link removed**

The problem is when I touch and move the end side of the load cell cable (which is connected to loadcell case) it comes out of calibration and while there is no weight on it, it displays random numbers in range 20-600 gr! This is fixed when I locate the cable in a particular position or I manually push the load cell in opposite direction of weight force in order to compensate the unwanted weight. Sometimes even without moving the cable the problem happens and there is random numbers on display .
I believe this is not about noise or any other electrical interface because it completely fixes with a physical action on load cell.

Does the load cell itself have a problem? Or it is a common behavior of a load cell in this range (50KG) .

Anyone has any idea about this ?

Thanks

 

[laktronics]

load cell noise

Hi,
Mostly cut wire/ solder joint making loose contact on and off.

Regards,
Laktronics

 

[imanjl]

adc loadcell

Sure ,,,I replaced the loadcell with a new one and all the signal path contacts
seems ok , But the problem still exists .

 

[laktronics]

use a load cell

Hi,
Did you change the cable also?
Regards,
Laktronics

 

[FvM]

what does the inside of a load cell look like

I think, the cable is attached to the load cell.

If mechanical cable movements (e. g. bending) near the load cell causes the effect, the problem may come from a design flaw of the load cell that results in a high mechanical sensitivity at this point. Fixing the cable to the construction thorougly would be the only solution, apart from checking the issue with the manufacturer of course.

I expect, that the bridge terminals are connected by approriate techniques inside the load cell and also at your equipment, if so, there should be no room for loose contacts.

A special point may be the cable screen. If it's connected to the load cell case and the load cell grounded at the construction, you should check if the screen connection at your equipment side allows current flow affecting the input circuit, e. g. through a common ground. Also AC interferences (either mains frequency or RF) may be picked up by the input circuit and cause an offset. In this case, also the cable position could have an effect. These electrical effects wouldn't show (or show different) with an isolated load cell not mounted to the construction.

 

[gebepic]

ad7715 to load cell

Is the system earthed in any way?
1/
It looks like that a high amount of 50 cycles hum is picked up.
Possible double earth connection on BOTH sides of the shielding.
The shielding MUST ONLY be earthed at ONE side ONLY.Preferably side closest to input circuit.
(Only my opinon)

2/

Faulty design of input circuit ?




UPdate
MISSED quote !!!

I believe this is not about noise or any other electrical interface because it completely fixes with a physical action on load cell.

Sorry about it but there are so many things like that.....

It looks like FvM is close to the truth
A way to "fix" that would be to secure the cable tighly to the chassis
A further way to ascertain that it is the problem would be to calibrate and zero it

with the cable FXED in one position.

Then without touching the calibration at all
Move and fix the cable at another place and when fixed ,,, see if there is any difference in zero and calibration

 

[imanjl]

mulitile load cell accuracy

I think, the cable is attached to the load cell.

Yes , It cannot be replaced . I tried to cut the cable and make it smaller in length , now the results are a little more stable , but the cable position problem still exists .

If mechanical cable movements (e. g. bending) near the load cell causes the effect, the problem may come from a design flaw of the load cell that results in a high mechanical sensitivity at this point. Fixing the cable to the construction thorougly would be the only solution, apart from checking the issue with the manufacturer of course.

Im also pretty sure about this and going to contact the manufacturer tomorrow .

A special point may be the cable screen. If it's connected to the load cell case and the load cell grounded at the construction

The Cable screen is not connected to Load cell case , and I disconnected the loadcell from construction in order to reject any interfere . The cable screen (shield) is connected just from one side to analog ground on the PCB .

By the way thanks for you helpful info ... .

 

[imanjl]

ad7715 loadcell

The cable position problem was solved with inversing the load cell "S" in the construction . I had wrongly mounted the load cell in the construction and the cable weight itself caused the extra random weight !

But still I have 5 digit LSB noise ! which is a pretty large amount of error in this application . I use AD7715-5(16bit Delta-Sigma ADC) from analog device and it limits me to drive the load cell with +5V excitation while the recommended Ext voltage is mentioned +10v in loadcell datasheet .

Is it true that with increasing the Exc voltage we get less LSB noise ?

Do you suggest to use 24 bit ADC ? if yes which part number is well suited for my application regarding measuring weight (Range=0-50Kg Sensitivity=0.01 Kg) .

 

[gebepic]

In my opinion, doubling the excitation will double the usable output resolution(like adding one bit resolution) but it would not improve the noise except divide it by 2 which considering the "5 digits" noise would be little
It could also be any vibrations around the site or low frequency noise that creates a resonnance in the frame .
That seems far fetched but .....I 've seen plenty of funny things..

To be sure put an oscilloscope on the signal and see if it is random or a specific frequency like 100 cycles or 50 cycles (hum)



Also usually a sensitive weighing system should use 3 loadcells of which the outputs are summed
You seem to speek about one load cell only...

quote
Do you suggest to use 24 bit ADC ? if yes which part number is well suited for my application regarding measuring weight (Range=0-50Kg Sensitivity=0.01 Kg) .

24 bits will only show more of the lower part of the signal and therefore "increase"
the perception of noise
(My opinion only)I stand to be corrected on all points...

 

[FvM]

You didn't tell the ADC operating parameters. They are important, cause they affect AD generated noise considerably. I would expect, that you use 2.5V Ref, Gain 128, minimal sampling rate. You should calculate ADC noise with your operation parameters and check, if the observed noise can be expected to be sourced mainly by the ADC itself.

Shorting the ADC input would remove possible load cell generated or mechanical noise.

Using the AD internal gain block is convenient, but doesn't achieve low noise. A precision bipolar instrumentation amplifier as preamplifier should give lower noise. Comparing 16- and 24-Bit ADC from the same family (Analog has some types) shows basically identical noise voltages. Thus, if 16 bit resolution is sufficient, a 24-Bit converter wouldn't give advantages.

If the ADC generated noise is dominant, doubling the exitation voltage would help. A -2.5/+7.5 load cell supply could be used for instance. It could be divided to a 2.5 Vref as well.

Other ADC families may have lower noise than AD7715. Also averaging rspectively digital filtering is an option to reduce displayed noise. Other ADC may allow higher conversion speed without increasing noise as much as with AD7715. When averaging the readings, overall noise can be lower then.

 

[imanjl]

Also usually a sensitive weighing system should use 3 loadcells of which the outputs are summed You seem to speek about one load cell only...

Due to some limitations I can use only one load cell ,,But Im sure somethings is wrong about the ADC or interfacing technic.

You didn't tell the ADC operating parameters

Many thanks Fvm , you really pinpoint the problem , I configured AD7715 like this way :

Internal gain block (PGA) : gain=128
Sample rate : 500Hz->filter cut off=131hz
Vref=1.92 and it is not 2.5V

Because my load cell Rated output(R.O.) is 3mV/V and with +5V Excitation voltage we have : 3mv*5=15mV full scale voltage so 15mv*128(gain)=1.92 VRef
I think I must use negative voltage for increasing Vref as you mentioned . With Shorting the ADC input still I have 3-4 digit lsb noise !! but it happens less frequently than the time load cell is connected .

 

[FvM]

According to the datasheet, AD7715 has uV RMS noise with 500 Hz sampling frequency, corresponding to 15 kg RMS or rouglhy 75 kg peak-peak. Reducing the sampling frequency to 50 Hz should reduce the noise also by a factor of 10 (Table V, unbuffered mode). Basically, the AD7715 doesn't achieve full resolution at higher data rates, table VI shows this in an umpressive way.

To achieve a similar effect with averaging would need 100 samples or the tenfold measuring time. The intended measurimg speed of your application of course sets a limit a averging or filtering.

Also usually a sensitive weighing system should use 3 loadcells of which the outputs are summed
You seem to speek about one load cell only...

I can't confirm this opinion. 2 or 3 load cells are used, if the construction doesn't allow acessing the weight force at a single point, but e. g. general purpose scales have mostly a single load cell. Using multiple load cells is neither a matter of accuray nor noise, to my opinion.

 

[gebepic]

All the industrial weighing system I worked on did use 3 load cells.
I might have missed something here...NOT INDUSTRIAL PROCESS SCALE ?


for a few reasons(3 points scale)

The main ones:

Vibrations of the mounting frame.
Movement of the mass to mesure(i.e.liquid sloshing (partial cure with baffles))
Irregular repartition of the mass semi solid(i.e.coal)

What is lost in weight in some cells is gained in the others(side to side motion(as minute as it is))

With one point only , side thrust (vibrations)and any irregularity in the positionning affects the reading .

I have seen one scale with one point and it did not work very well , at the end it was changed to 3 points(not by me)

Of course there are some systems with one point only, but it usually is for relatively course weighing.




Also:Not knowing the weights to mesure nor the accuracy

Resolution:
Here speaking of going to 24 bits AD means a formidable resolution

From 16 bits to 24 bits , depending of the weights
Resolution per Ton Span:
1000 kg in 16 bits 1000/65535===>>resolution of 15.25 grams
Possible Display: 999.99 Kgs
1000 kg in 24 bits 1000/ ffffff==>> resolution of 6 milligrams
Possible Display: 999.99999 Kgs
If I am not mistaken

which in my opinion would be an overkill if there is already an instability of the 5 lowest digits in 16 bits

of course it can be formatted to hide them

You seem to have the problem well in control ,so I' ll leave it to you chaps.Good luck.

 

[imanjl]

Reducing the sampling frequency to 50 Hz should reduce the noise also by a factor of 10

Sample rate was changed to 50Hz now I acheive 2 digit LSB noise !! The result is pretty stable on LCD and I think it is acceptable for AD7715 and also my project requirments .

I like to focus on PCB design and grounding if I get the better result I'll keep you updated guys .

Many thanks ... .

 

[mandar_mahajan]

Hello, it seems to be a ground related problem. check the voltages bet'n ground,&if it is there ,then give ur adc a seperate ground path.

I have faced this type of problem(24 bit adc:cs5460) but that was different case.
Now I want to work on ADC interfacing with load cells. Can u suggest me which adc should I use & how to giv my load cell mv to adc.(I want to use 4 load cells for my system, In what fashion should i giv these mv to adc. Either all 4 o/p's on different channels of adc or multiplexing them & giving them to single channel.)
PLZ HELP!!!