Load Cell Questions - capacity of load cell

Hi all,

I have been trying to find some resources on how to properly select a load cell for a test stand.

The main questions I am curious about right now revolve around load cell capacity.

For the time being we will be primarily focusing on 38 mm motors. In the range of 200-300 lbf of thrust max, however someday I am sure we will go larger.
A while back, before I started looking into the specifics, My dad and I found and purchased a used 2000 lbf capacity Revere SSB type load cell.

Other than the fact that it is used, and may be damaged, would this load cell be effective in the 0-300 lbf range. Or do load cells have a "butter zone" where they are most accurate?
I have read through the "Load Cell Primer" on the Tripoli Gerlach website, but the datasheet doesn't seem to list a verification interval for the 1 ton model that we have.

As I most load cells' load-strain graphs are very linear, and we should be able to calibrate the load cell with known values of mass.

Thanks so much for your input. Let me know if I can clarify any of my thoughts or question.

Dave

Yes it will work, but it will be less sensitive than a load cell with a lower maximum range.

This load cell has a sensitivity of 2 mv / volt of excitation. You would typically apply a 10 volt excitation voltage and if you did, you would have a 20 mv signal change when you put a 2000 pound weight on the load cell. This would make the response equal to 2000 pounds / 20 mv = 100 pounds per mv.

This number is low so you really need a load cell amp to increase the signal from the load cell to match the input of the digitizer your are using to record the data.

If you have a 0-10 volt or a -5 to +5 volt digitizer, the load cell amp ideally needs a gain of 10 volts /0.020 volts = 500 to match the full scale range of the load cell to the full scale range of the digitizer. If you had a 12-bit digitizer, the sensitivity of your measurement is 2000 pound / (4096-1) bit steps = 0.488 pounds / bit step. That's ~ 400 steps for a 200 pound max thrust curve.

You need the load cell amplifier to provide gain to amplify the load cell signal to a reasonable set size so you have good resolution in your thrust curve. For example if you set the gain to 5000, the resolution is 4000 bit steps with matches the resolution of the ADC. The problem can be that you need to adjust the output offset or you will go off scale which is the reason for the load cell amplifier.

A load cell amplifier supplies the bias voltage for the load cell, amplifies the signal from a submillivolt signal level to a volt signal level, puts in a voltage offset to allow you to adjust the zero signal level voltage to match the minimum value of the digitizer, and a low pass filter to eliminate high frequency noise. A 400 Hz low pass filter is optimum of a test stand.

Most load cells have a full scale sensitivity between 0.5 to 5 mv per volt of excitation, and can have an excitation voltage between 5 to 10 volts.

A good load cell amplifier will supply excitation voltage of 5 or 10 volts, have continuous gain adjustment form 100 to 5000, have -10 to +10 volt offset capability, and may also have a 1 to 15 scaling amplifier with a 400 hz low pass filter. The best use a TI IN125 chip as the load cell amp, and if it is built to the recommended circuit, the gain, excitation voltage, offset, scaling and frequency are independently adjustable. Sadly many load cell amps are not designed this way and are a pain to calibrate.

Bob

Bob,
Thank you for your response!
I really appreciate your help.
Sorry for the slow response, I have been trying to digest this as much as possible - as well as trying to figure out exactly what I don't understand and doing more online research...
I am rather green when it comes to load cells, amplifiers and how to set them up. However I am very much willing to learn. If you would please bear with me, as I may have some silly questions.

bobkrech said:

Yes it will work, but it will be less sensitive than a load cell with a lower maximum range.

This load cell has a sensitivity of 2 mv / volt of excitation. You would typically apply a 10 volt excitation voltage and if you did, you would have a 20 mv signal change when you put a 2000 pound weight on the load cell. This would make the response equal to 2000 pounds / 20 mv = 100 pounds per mv.
Bob

Click to expand...

I think I understand for the most part up to this point and maybe a little further.
Basically the excitation voltage is a reference voltage and determines your full range.
The excitaion voltage and "sensitivity" are both from the load cell datasheet:
https://www.intertechnology.com/Revere_Transducers/pdfs/SSB.pdf

Let me see if I really get this...
At 0 lbf load the response in mv from the load cell = 0 mv
At 2000 lbf (Full Capacity of load cell) the response in mv from the load cell = 20 mv [2mv/1V = Xmv/10V, solve for X = 20 mv]

Correct?

bobkrech said:

This number is low so you really need a load cell amp to increase the signal from the load cell to match the input of the digitizer your are using to record the data.

Bob

Click to expand...

Correct me if I am wrong, but this statement means that to be able to accurately measure 50 lbs, we would some how need to be able to measure 0.5 mv. And to be able to measure 1 pound, I'd need to be able to decipher 0.01 mv? To be able to obtain the submilivolt signals and read them we need some sort of amplifier, which you go into in your next section...

bobkrech said:

If you have a 0-10 volt or a -5 to +5 volt digitizer, the load cell amp ideally needs a gain of 10 volts /0.020 volts = 500 to match the full scale range of the load cell to the full scale range of the digitizer. If you had a 12-bit digitizer, the sensitivity of your measurement is 2000 pound / (4096-1) bit steps = 0.488 pounds / bit step. That's ~ 400 steps for a 200 pound max thrust curve.

You need the load cell amplifier to provide gain to amplify the load cell signal to a reasonable set size so you have good resolution in your thrust curve. For example if you set the gain to 5000, the resolution is 4000 bit steps with matches the resolution of the ADC. The problem can be that you need to adjust the output offset or you will go off scale which is the reason for the load cell amplifier.
Bob

Click to expand...

When you say "digitizer", are you referring to some sort of analog to digital converter (ADC) Like a DATAQ unit? Or is it part of the amplifier circuit?
Could you please explain the concept of a gain and how you came up with the value of 500?

Let me see if I understand the overarching principles though...
With the gain of 500, we have a "resolution" of ~0.5 lbs over the full range of the load cell up to 2000 lbs. However, if we are only testing motors up to 200 lbs, our data would very coarse.
Using a gain of 5000, we can obtain a "resolution" of ~0.05 lbs, but only up to about 200 pounds. If our motor produces 250 lbs of thrust, that data will be truncated?

Either way, if a load cell with a capacity closer to the intended max thrust of the data we want to capture is selected, a much smaller gain is required and the better the data is, if I am starting to wrap my head around this correctly?

bobkrech said:

A load cell amplifier supplies the bias voltage for the load cell, amplifies the signal from a submillivolt signal level to a volt signal level, puts in a voltage offset to allow you to adjust the zero signal level voltage to match the minimum value of the digitizer, and a low pass filter to eliminate high frequency noise. A 400 Hz low pass filter is optimum of a test stand.

Most load cells have a full scale sensitivity between 0.5 to 5 mv per volt of excitation, and can have an excitation voltage between 5 to 10 volts.

A good load cell amplifier will supply excitation voltage of 5 or 10 volts, have continuous gain adjustment form 100 to 5000, have -10 to +10 volt offset capability, and may also have a 1 to 15 scaling amplifier with a 400 hz low pass filter. The best use a TI IN125 chip as the load cell amp, and if it is built to the recommended circuit, the gain, excitation voltage, offset, scaling and frequency are independently adjustable. Sadly many load cell amps are not designed this way and are a pain to calibrate.

Bob

Click to expand...

Is the low pass filter applied real time as the data is being captured or can it be applied after the data has been gathered?


Please let me know if I can clarify any of my statements. Also, please let me know if I am understanding the basic principles so far. Like I said this is all relatively new to me. Do you know of any other resources you could point me to to help with my understanding as well?

I hope I didn't complicate matters with all of the quotes...
Thanks so much for your help.

Dave

dvdsnyd said:

Bob,
Thank you for your response!
I really appreciate your help.
Sorry for the slow response, I have been trying to digest this as much as possible - as well as trying to figure out exactly what I don't understand and doing more online research...
I am rather green when it comes to load cells, amplifiers and how to set them up. However I am very much willing to learn. If you would please bear with me, as I may have some silly questions.

Click to expand...

Your questions are good questions.

I think I understand for the most part up to this point and maybe a little further.
Basically the excitation voltage is a reference voltage and determines your full range.
The excitaion voltage and "sensitivity" are both from the load cell datasheet:
https://www.intertechnology.com/Revere_Transducers/pdfs/SSB.pdf

Click to expand...

Yes. You should understand what the transducer needs to operate and what signal levels it output. It is also important to know how far you can push the load cell in case your estimate of thrust is low, or if the motor catos. Most load cells can be "overloaded" to 150% of their rating and if you hit 300% of the load rating you have permanently deformed the transducer.

Let me see if I really get this...
At 0 lbf load the response in mv from the load cell = 0 mv
At 2000 lbf (Full Capacity of load cell) the response in mv from the load cell = 20 mv [2mv/1V = Xmv/10V, solve for X = 20 mv]

Correct?

Click to expand...

The response with no load is nominally 0 mv but there can be a DC offset of 1% of full scale with your transducer. When you multiply the offset voltage by the amplifier gain, it could be up to a volt or so with a gain of 5000, but it is constant and does not change unless you have overstressed the transducer.

Correct me if I am wrong, but this statement means that to be able to accurately measure 50 lbs, we would some how need to be able to measure 0.5 mv. And to be able to measure 1 pound, I'd need to be able to decipher 0.01 mv? To be able to obtain the submilivolt signals and read them we need some sort of amplifier, which you go into in your next section...

Click to expand...

Not only do you have to measure 0.5 mv or 0.01 mv, but you need to have some resolution capability as well. The accuracy of your measurement is determine by the number of bits you can resolve for any delta volts you observe. If you want to resolve 0.5 pound thrust changes with your transducer, you have to measure the value to (o.5#/2000#) x 20 mv = (20 mv/4000) = 5 uV! If you want to know that value to +/-20%, you need a system accuracy to measure 1 uV! This is why you need a load cell amplifier.

When you say "digitizer", are you referring to some sort of analog to digital converter (ADC) Like a DATAQ unit? Or is it part of the amplifier circuit?

Click to expand...

You are correct. IMO the DATAQ is the most cost effective and simplest ADC you can buy for the job. Anyone can make the hardware, but the DATAQ software is what makes it simple to use.

Could you please explain the concept of a gain and how you came up with the value of 500?

Click to expand...

The reason why you need an amplifier with gain is to raise the low level load cell output voltage to match the nominal 10 V input level of your ADC. 10 V FS ADC / 0.020 V FS LC = 500 gain. If your ADC range is different, then the gain would be different. Your load cell has a full scale range of 2000#, If you were testing a motor that had a peak thrust of only 150#, you might want to add an addition gain of 10 to change the effective load cell full scale output to 200#. This can be accomplished in one step if you can raise the gain to 5000, or in 2 step by used a 2 stage amplifier where the first stage has a gain of 500 which serves as the input to the second stage with a gain of 10.

Let me see if I understand the overarching principles though...
With the gain of 500, we have a "resolution" of ~0.5 lbs over the full range of the load cell up to 2000 lbs. However, if we are only testing motors up to 200 lbs, our data would very coarse.
Using a gain of 5000, we can obtain a "resolution" of ~0.05 lbs, but only up to about 200 pounds. If our motor produces 250 lbs of thrust, that data will be truncated?

Click to expand...

Course is a relative term. If you resolve a single point thrust value to 1% (you have 100 counts of resolution) the mathematical standard deviation is the sqrt(100)/100=10%. If you have 100 time steps in your thrust curve, the product of the thrust curve and the delta time is the total impulse. The standard deviation of the total impulse could be as high as sqrt(100x100)/(100x100)=1% if the thrust curve was a square pulse. The standard deviation is the sqrt(sum of the volt values/(counts of the peak x the number of time elements).

Either way, if a load cell with a capacity closer to the intended max thrust of the data we want to capture is selected, a much smaller gain is required and the better the data is, if I am starting to wrap my head around this correctly?

Click to expand...

That is correct, but the typical load cell is linear to about 0.001% so if you have good low noise electronics, you can still get good data from a slightly larger load cell, up to a factor of 10 or so more than your peak reading.

Is the low pass filter applied real time as the data is being captured or can it be applied after the data has been gathered?

Click to expand...

It is best to low pass the analog signal so you do not have high frequency noise in the data, however you can use the FFT/Inverse FFT function in the DATAQ software to remove high frequency and resonant frequency signals.

I really like this LCA from PICOM. https://picom2.com/Phillips_Instruments.html It uses the TI INA 125 chip and all amplifiers are independent so IMO it is done right and inexpensive. The only thing is that you should ask for a 400 Hz low pass instead of the standard lower value, or simply change one component in the circuit to set the low pass band frequency.

Please let me know if I can clarify any of my statements. Also, please let me know if I am understanding the basic principles so far. Like I said this is all relatively new to me. Do you know of any other resources you could point me to to help with my understanding as well?

I hope I didn't complicate matters with all of the quotes...
Thanks so much for your help.

Dave

Click to expand...

I'd also recommend taking a look at Aerocon Systems. https://aeroconsystems.com/cart/load-cells/

They have a bunch of different weight and type cells, and their prices are pretty reasonable across the board. In addition to the ~$100 cells up through 1000 lbf, they also sell a single channel amplifier, so that if you do decide to go with a larger cell, you can at least boost the signal voltage.

I recently used a test stand for my school's rocket club, and we ran into the same issue. We had a 10000N load cell with motors peaking around 300N. We were pushing the mv excitation range, and unfortunately that was within the MyDaq's error %. So a couple guys got together and built an op-amp to boost the signal by I think 100 times. We were able to get usable data off it for at least the largest motor we fired at 861N average thrust over 1.5s.

Jared

The instrumentation and procedures we use to generate thrust curves when we conduct static tests for NAR S&T motor certifications is found here. https://www.nar.org/SandT/docs/ST-MotorTestingManual.pdf

After we obtain the thrust curve, we look at the raw data, remove any test stand resonance using the FFT/Inverse FFT method on the data file. and then run the data through John DeMar's excellent Thrust Curve Tool program to obtain the wRasp.eng file for the certification certificate. https://www.thrustgear.com/software.html

Bob

Bob, Jared -
Thank you for your help and replies!

bobkrech said:

The response with no load is nominally 0 mv but there can be a DC offset of 1% of full scale with your transducer. When you multiply the offset voltage by the amplifier gain, it could be up to a volt or so with a gain of 5000, but it is constant and does not change unless you have overstressed the transducer.
Bob

Click to expand...

With the Dataq reading in the signal from the load cell attached to the amplifier, and with a gain of 5000, I could see a reading of upwards of 1 volt, even though nothing is on the load cell. Calculating this:
FS = 20mv
Gain = 5000
offset = up to 1%

(.01*20mv)*5000 = 1000mv = 1v
I can take note of this within the Dataq software correct, just set this reading to be the zero, right?

bobkrech said:

Not only do you have to measure 0.5 mv or 0.01 mv, but you need to have some resolution capability as well. The accuracy of your measurement is determine by the number of bits you can resolve for any delta volts you observe. If you want to resolve 0.5 pound thrust changes with your transducer, you have to measure the value to (o.5#/2000#) x 20 mv = (20 mv/4000) = 5 uV! If you want to know that value to +/-20%, you need a system accuracy to measure 1 uV! This is why you need a load cell amplifier.

Bob

Click to expand...

For testing motors, generally, what type of resolution capability do we want to shoot for? Obviously higher is better. But at what minimum would you want to stay under, and what generally is good?
Is 0.5 lb thrust change resolution acceptable? Or is it best to go lower yet?

How many bits of resolution would be required to measure down to 1uv?

Looking at the Dataq website. There are really 2 units for sale:
DI-145 which has 10 bit ADC resolution. They claim to be able to measure down to 19.5 mV resolution.
DI-155 which has 13 bit ADC resolution. Which claims to be able to measure down to 610 uV resolution.

So, if I try to solve for how much change in thrust I can resolve with the DI-155 unamplified I would get:
X = 0.61*2000/20 = 61 pounds.
This tells me with this Dataq unit measuring full scale, I can only resolve thrust changes of 61 pounds..

How does adding amplification/gain help be able to bring the 61 pounds down to 1 pound or less?
I guess there are two scenarios I don't understand, the gain when applied to the full scale 2000 pounds, and the gain when applied to only wanting to measure something around 200 pounds.

bobkrech said:

The reason why you need an amplifier with gain is to raise the low level load cell output voltage to match the nominal 10 V input level of your ADC. 10 V FS ADC / 0.020 V FS LC = 500 gain. If your ADC range is different, then the gain would be different. Your load cell has a full scale range of 2000#, If you were testing a motor that had a peak thrust of only 150#, you might want to add an addition gain of 10 to change the effective load cell full scale output to 200#. This can be accomplished in one step if you can raise the gain to 5000, or in 2 step by used a 2 stage amplifier where the first stage has a gain of 500 which serves as the input to the second stage with a gain of 10.

Bob

Click to expand...

Maybe I just don't understand this statement?

Which Dataq would you recommend the 155 or 145?


Thanks for the NAR resource Bob, I will have a look at that. Looks like some good procedural stuff too.

Thanks again for all your help! Look forward to any and all help..

Dave

Start with the sensor. It is 2mV/V full scale. You appear to be assuming a 10V excitation to get 20mV full scale.

The DI-145 has an input range of +/-10V. Assuming your amplifier can reach 10V, then you require a gain of 10V/20mV = 500. Resolution is 2000lbf/512 or about 4lbf. The DI-145 has 1024 counts for a 20V range but you are using only half or 512 counts. In other words, it is effectively a 9 bit ADC in this application. (If it is like the DI-194 then by making some simple changes you could alter the range to 0-5V. That way you only need a gain of 250 and you get the full ADC resolution.)

If you use a gain of 5,000 then the range is only 200lbf and the resolution is 0.4lbf. The downside is that the noise gets amplified as well.


The DI-145 is not a good choice. Not only is the ADC resolution lacking but the sample rate is pathetic. The DI-155 is the better choice by far. But if I was doing this I would use something like the AD7705. It can be connected directly to the load cell without mucking about with instrumentation amplifiers and such.

https://www.analog.com/static/imported-files/circuit_notes/CN0119.pdf is a circuit note from Analog Devices that describes how to make load cell measurements using an IC with an integrated 24 bit Sigma-Delta ADC and a PGA (Programmable Gain Amplifier). For probably 15 years or more Analog Devices has been making this series of ICs for digital scale applications. I agree that this would be the modern way to go, and only you, and a few dozen other TRF members are capable of designing, building and programing a device with one. I am unaware of anyone who makes a $100 or less DLCA (Digital Load Cell Amplifier) employing these ICs, and after you built it, it needs acquisition software to acquire and record the data.

For the rest of us who don't do digital design, I think the LCA/inexpensive ADC combo is the only viable option in the $100 price range.

Bob

dvdsnyd said:

Bob, Jared -
Thank you for your help and replies!

With the Dataq reading in the signal from the load cell attached to the amplifier, and with a gain of 5000, I could see a reading of upwards of 1 volt, even though nothing is on the load cell. Calculating this:
FS = 20mv
Gain = 5000
offset = up to 1%
(.01*20mv)*5000 = 1000mv = 1v
I can take note of this within the Dataq software correct, just set this reading to be the zero, right?

Click to expand...

You would use that value a Baseline zero in your analysis. The Thrust is proportional to delta volts = volts reading - volts baseline.

ThrustCurveTool does the math for you....

For testing motors, generally, what type of resolution capability do we want to shoot for? Obviously higher is better. But at what minimum would you want to stay under, and what generally is good?
Is 0.5 lb thrust change resolution acceptable? Or is it best to go lower yet?
How many bits of resolution would be required to measure down to 1uv?

Click to expand...

It really depends on what you are trying to do. Anything more than 100 counts at the peak gives you a 1% estimate of the peak thrust.
That depends on the motor. You could get a useful thrust curve for a motor with a peak thrust over 50 pounds (~220N) with that setting.
You are asking the wrong question. Your bit resolution is the ADC step size divided by the LCA gain. If you have a 20 mv ADC step and a gain of 500, your measurement resolution is 20/500 = 0.04 mv - 40 uv. If your ADC step size is 0.6 mv and you use a gain of 500, your resolution in 0.6/500 = 0.0012 mv = 1.2 uv.[/quote0

Looking at the Dataq website. There are really 2 units for sale:
DI-145 which has 10 bit ADC resolution. They claim to be able to measure down to 19.5 mV resolution.
DI-155 which has 13 bit ADC resolution. Which claims to be able to measure down to 610 uV resolution.

So, if I try to solve for how much change in thrust I can resolve with the DI-155 unamplified I would get:
X = 0.61*2000/20 = 61 pounds.
This tells me with this Dataq unit measuring full scale, I can only resolve thrust changes of 61 pounds..

Click to expand...

don't was your time or mine by not using a load cell amp.

How does adding amplification/gain help be able to bring the 61 pounds down to 1 pound or less?
I guess there are two scenarios I don't understand, the gain when applied to the full scale 2000 pounds, and the gain when applied to only wanting to measure something around 200 pounds.

Maybe I just don't understand this statement?

Which Dataq would you recommend the 155 or 145?

Thanks for the NAR resource Bob, I will have a look at that. Looks like some good procedural stuff too.

Thanks again for all your help! Look forward to any and all help..

Dave

Click to expand...

UhClem answered that question.

BTW Aerocon has great prices on loadcells. You can get a lower rated load cell for $30.......

Bob

bobkrech said:

I agree that this would be the modern way to go,and only you, and a few dozen other TRF members are capable of designing, building and programing a device with one.

Click to expand...

Except that these days any idiot can buy and program an Arduino.

While checking the price of the AD7192 I found a great deal: A $55 evaluation board (pictured in the application note you cited) that includes everything you need. AD7190, power supply (battery, USB, or external), micro-controller, LCD display, and USB interface.

https://www.analog.com/en/evaluation/eval-ad7192/eb.html

You can't beat that. (with a stick)

UhClem said:

Start with the sensor. It is 2mV/V full scale. You appear to be assuming a 10V excitation to get 20mV full scale.

The DI-145 has an input range of +/-10V. Assuming your amplifier can reach 10V, then you require a gain of 10V/20mV = 500. Resolution is 2000lbf/512 or about 4lbf. The DI-145 has 1024 counts for a 20V range but you are using only half or 512 counts. In other words, it is effectively a 9 bit ADC in this application.

Click to expand...

Dave,
Thanks for your help.

Could you please explain in a little more detail your above statement?

Specifically:
What is meant by the input range for the Dataq?
Where does this voltage come from? The load cell/amplifier combination?
Is the range related to the excitation voltage of the load cell?
What is the limiting factor to being able to supply the necessary 20 volt range for the DATAQ in the above scenario?

UhClem said:

The DI-145 is not a good choice. Not only is the ADC resolution lacking but the sample rate is pathetic. The DI-155 is the better choice by far.

Click to expand...

From the looks of it, the DI-155 has programmable input voltage range. The manual also mentions something about gain settings. Are these the same type of gains like an amplifier would supply? Or is it only for the input voltage to be able to utilize the full scale of the ADC over multiple input ranges?


Thank you very much for your help!

Dave

Input range. You can apply a voltage in the range of +/-10V to the input. -10V will result in a count of 0 while +10V gives 1023. Or perhaps it is two's complement with a range of -512 to +511. You will have to read the documentation to find out. If you don't know what two's complement is, then you have more reading to do.


The voltage comes from whatever source you connect to it. In this case the output of your load cell amplifier/filter.

The output of the load cell varies with the excitation voltage. This is what is meant by the 2mV/V specification. Multiply by the excitation to get full scale output voltage.

What limiting factor? Your circuit must be able to provide that. You will have to consult the design detail to see if they can do it. A key hint is the supply voltages.


The gain setting on the DI-155 sets the input voltage range. Pick the one that matches the output range of your source. Note that these ranges are all far greater than the full scale output voltage of the load cell so amplification is required.