ADC selection for weigh scale

Hi, thanks for the reply and sorry for the confusing question. The 16 bit adc solution was just to tell you that i was able to get the same resolution with the amplifier and the 16 bit adc with same as ads1232.

Here are the requirements below:

1. Application is for weighing automation for packaging and weighind industries.

2. Input signal will be from 0mv - 30mv max but in practical condition may be the span is as less as 0 - 10mv.

3. Resolution: i need minimum 19-20 bits noise free bits.

4. Communication is reliable spi , serial , i2c anyone will be OK.

5. If there is internal reference then will be more preferable.

6.Cost : maximum 200-250 INR.

The document attached was very helpful. 

Also what are the design consideration i need to check for using a 24 bit adc to get the maximum noise free bits.

Hi Ankit,

One key factor is the amount of noise of your system relative to the full-scale range of the ADC.

A couple of things to consider.  First, the integrated PGA will perform better than an external amplifier as although the resolution in bits decreases, the noise also decreases.  Also, with an integrated amplifier you can make use of the ratiometric measurement, whereas with the external amplifier you add the amplifier noise/drift which will not cancel as a part of the measurement.

Instead of looking at noise-free bits, it may be better to look at noise-free counts for the desired input range relative to full-scale.  Usually the desire is to have higher noise-free counts than the system is able to deliver.  If you want to use a 10kg load cell and want to get 1 gram resolution, you would need 10,000 noise-free counts (no flicker on the weigh scale).  If you want 0.05 grams resolution than that would be 200,000 noise-free counts. The noise-free counts is equal to the ratio of the load cell voltage output divided by the full-scale range times the noise-free bits. So the number of noise-free bits does affect the overall resolution and assumes a 1 to 1 ratio where the load cell output equals the full-scale range.  The number of noise-free counts of scale resolution decreases depending on the ratio of the load cell output to the full-scale range.  So it is usually best to work backwards from the desired scale resolution to see which ADCs will perform the desired task.

Another thing to remember is the noise for the ADC is computed when the noise is the lowest (shorted input case).  This is a best case scenario and does not include any system noise, or the gaining of noise.  As you don't measure a shorted input, the actual measurement will most likely have higher noise content.  What the shorted input case will do is give a point of reference.  For example, if you short the inputs and are able to achieve the datasheet performance, then you have designed a good system with good layout.  If you are not able to achieve the shorted-input numbers, then you will have decreased performance where some correction will be required.

Best regards,

Bob B

Hi, 

Thanks a lot Bob B.  That explanation really helped me and clear my knowlegde about the noise free counts.

I need 1gram resolution in 50kg loadcell at least and 1 gram in 100kg loadcell will be a great achievement .In that sense in first phase i need to get 50,000 noise free counts and if possible 100,000 noise free counts in the next phase.

Thanks again for detailed description.

Hi Ankit,

When comparing ADCs, you often will see a noise table relative to a certain reference voltage.  These tables will not always match up with the reference you want to use.  For example, a 5V reference/excitation voltage is desired.  The ADS1232 has specific noise tables for a 5V reference and you can use the noise tables directly to calculate noise-free counts.  The ADS1220 has noise tables for 2.048V reference, so you cannot directly use the noise-free bits from this table as the scaling is incorrect.  However, you can use the peak to peak noise from the tables.  At 20sps for the ADS1220 the noise is 410nV peak to peak when using gain of 128.

We can use the noise relative to full-scale to determine the noise-free bits (equation 2 in the ADS1220 datasheet), or we can use a less complicated formula and determine the smallest measurable weight relative to full-scale capacity of the load cell if we also know the load cell sensitivity (mV/V) and the excitation voltage.  To explain further for our example, we will want to use a load cell where the sensitivity is 2mV/V and the load cell capacity of 50kg.  With 5V excitation this gives an output of 10mV at full-scale for 50kg.  So the simplified formula is (capacity/full-scale voltage output) *noise peak to peak.

50kg/10mV*410nV = 2g which also happens to equal the capacity of the load cell divided by the noise-free counts.

So you may ask, why I didn't give you the simpler formula earlier?  The main reason is so you can understand that number of bits of the converter (16-, 18-, 20-, 24-bit, etc.) doesn't tell the whole story.  You need to understand how noise affects the ability to get the weight resolution you require.  Not all converters are equal for the same data resolution (i.e. 24-bit converters) as there is a great dependency on the reference, PGA, etc..

Best regards,

Bob B

Thanks BOB B. ,

So until this i understand that as per the example you explained below:

For a 50kg loadcell , excitation 5v and a loadcell of 2mv/V. 

case 1:

for ads1220

Minimum resolution will be 2 grams.  

50kg/10mV*410nV = 2g which also happens to equal the capacity of the load cell divided by the noise-free counts which are 18.4

case 2 :

For ads1232

peak to peak noise for 10sps at 128 gain is 110nv

50kg/10mV*110nv = 0.5 gm which is not as per the noise free counts which are same as ads1220 which is 18.4 at 128 gain and 20sps.

so which one to consider as better the adc with minimum peak to peak or not?

Also ads1220 has internal reference which will be a plus point as external reference also will also add some noise and there is less design analysis require if there is internal reference.

Best regards ,

Ankit Patel