Adding Load Cells/Strain Gauges to Kinetis

[myke_predko]

Gentlebeings,

Does anybody have experience adding Load Cells/Strain Gauges to Kinetis? 

It seems to have been done on Arduino using ADS1100 or INA125 (as well as INA122) chips very easily, but I was wondering if anybody had any experience using the built in dual-input ADCs on the Kinetis or with cheaper instrumentation amplifiers? 

I am planning on using the K22 for the product and I am hoping to avoid the expensive interface chips above. 

Thanx,

myke

[xiangjun_rong]

Hi, Myke,

As you know that the Strain Gauges output is a differential signal, generally, it's output signal is very small, if you want to amplify it, you have to use instrumental AMP(triple AMP), which is expensive.

I suggest you use Kinetis KM family, the chip has two ADC module, one is SAR ADC module, another is AFE module, both of them support differential analog input and some channels of both ADC modules have PGA which can amplify the differential signal with programmable gain.

The AFE module supports PGA function and can test differential signal. It's rating resolution is 24 bits, but the actual resolution ENOB is defined in data sheet. the AFE module uses signal-delta technology, which modulate the tested signal to a high frequency signal, from theory, the technology can improve the time/temperature drift, the feature is important when you test a DC type signal.

This is the Kinetis KM family website:

ARM Cortex-M0+ Cores|Kinetis M MCUs|NXP

Hope it can help you.

BR

Xiangjun Rong

[myke_predko]

Thanx for the pointer - I was wondering if there was a Kinetis part that does the job.

Unfortunately, it doesn't look like the KM1x and KM3x have USB cores in them (which is another requirement).

Any other suggestions?

myke

[xiangjun_rong]

Hi, Myke,

Unfortunately, only the KM family has AFE module, the other family does not have AFE module, but the KM   family does not have USB, as an alternative, you can use an external USB controller with SPI or the other interface.

You can consider the SAR ADC, which also support differential input.

Hope it can help you.

BR

Xiangjun Rong

[travis_l]

Hi Myke,

I have lots of experience in talking to strain gauge load cells with Kinetis MCU's since I work for a major weigh scale manufacturing company.  The first obstacle you will need to overcome is resolution.  SAR A2D's are great but speed is often your enemy when reading a strain gauge, so I would definitely look into a Sigma-Delta type A2D.  There are some Kinetis MCU's that have this capability, but we haven't used any due to the fact that the programmable gain was not enough for our purpose.  The other thing that you are going to run into is drift due to temperature.  From what I've seen, lots of Sigma Delta A2D's experience significant drift over temperature.  We generally rely on an external A2D that has good temperature stability, along with an analog front end that uses a little higher quality components for better noise rejection.  I would highly recommend the Cirrus Logic 553x family of A2D's, which have great hardware filtering and temperature response.

Travis

[myke_predko]

Hi Travis,

Thank you for taking the time to reply - that helps me to confirm I'm on the right track. 

The Cirrus parts look interesting but I'm IO limited which is why I am considering the TI ADS1100 which has an I2C bus interface (which I'm already using) - do you have any experience with that part? 

Also how noisy are the readings (assuming a well filtered excitation voltage to the load cell and used as Vdd/Vref for the ADC) can you expect from a load cell?  I want to capture the change in force on the load cell.

Thanx again,

myke

[travis_l]

Myke,

I've never used the ADS1100, but I know TI makes some really good A2D's.  I have an ADS1262 32 bit sigma-delta eval board sitting on my desk right now.  If you are limited on your IO, then I2C makes a good choice if you don't already have an SPI bus and an extra pin for a chip select.  In looking at the data sheet for the ADS1100, it appears to have good linearity and temperature specs, especially at the slower sample rates. 

The noise of the readings can be into the microvolts, especially at higher sample rates.  It looks like TI compensates for this by giving you less bits of resolution for the higher sample rates.  This is an important factor when considering an A2D, so keep an eye on the ENOB (effective number of bits) which should tell you  how many bits are 'noise free'.  You will still see some jitter in your readings which we usually overcome using some type of digital filtering, typically we use a finite impulse response (FIR) method to filter as that gives us good responsiveness when the force is changing quickly and good filtering when the force is static.  Typically, integrating just a few samples can eliminate most of the noise that you would see.

When looking at a load cell in a weigh scale, we can typically see 10000 - 20000 stable, discrete divisions and sometimes more depending on the setup and strain gauge.  So, for instance, using a load cell rated for 10 pounds, we can accurately measure a thousandth of a pound.  In order to get this kind of resolution out of a strain gauge, we typically have a 24 bit sigma-delta A2D with a gain of at least 32 or more.

Travis