Load Cell to Arduino

Here's an idea I was curious about to increase A/D resolution.

Lets say for example the LSB represented 1 millivolt. Could I use a resistor network (or DAC and voltage divider) to drop 1/4mV, 1/2mV, and

3/4mV. The idea would be to add these voltages to the load cell signal, read with 0mV offset, next read with 1/4mV offset, then 1/2mV offset... It would seem that averaging with offset voltages less than the LSB could get a higher resolution and hopefully with linear accuracy. Just wondering if that would work? The A/D I intend on using will probably be 24 bit so I'm not going to higher than that, but sure would be nice for the lower resolution load cells sometimes. I thought about using a DAC with a voltage divider to make a programmable offset voltage to divide the LSB. Offset 1/2 LSB, compare, offset 1/4 LSB, compare... till I find the fraction of the LSB where noise toggles it.

So if I were to do this with the 10 bit Arduino ADC, offsetting 1/64LSB at a time over 64 readings, would the sum of the readings be (almost) as good as a 16 bit ADC? Don't know if it would be practical but sounds interesting and educational :-)

RogerN

Google delta-sigma converter.

--
Failure does not prove something is impossible, failure simply
indicates you are not using the right tools...
nico@nctdevpuntnl (punt=.)
--------------------------------------------------------------

Show me how to get 8 bits resolution with ONE comparator...

--
Failure does not prove something is impossible, failure simply
indicates you are not using the right tools...
nico@nctdevpuntnl (punt=.)
--------------------------------------------------------------

As long as the noise is uncorrelated with the signal, it's very easy.

You've never seen a delta-sigma ADC?

A successive approximation A/D only has one comparator. ;-)

--
You can't have a sense of humor, if you have no sense.

Not to mention dual-slope ADCs.

--
My liberal friends think I'm a conservative kook.
My conservative friends think I'm a liberal kook.
Why am I not happy that they have found common ground?

Tim Wescott, Communications, Control, Circuits & Software
http://www.wescottdesign.com

I'll see your two slopes and raise you multiple slopes.

The HP3458A, and indeed most high-end meters, uses a multislope technique to get accuracy equivalent to that of a 28 bit ADC.

From a single comparator of course! :)

--

John Devereux

another approach is just to add noise, and average it out.

24 bits is over 7 significant figures, seems like bucketloads.

yeah, if done right, probably. a 5 stage R-2r ladder should be linear enough.

sigma-delta is a another A/D scheme with fewer "moving parts", a single comparitor, a capacitor, a resistor, a d-type latch, a clock and a couple of counters. if you want to play with ADCs it's probably a good place to start.

--
?? 100% natural

Thanks all for the info, I'm learning interesting things!

I have an AD7730 on my solderless breadboard right now planning to test out after I come up with a 4.9152Mhz clock source. I expect a lot of noise on the breadboard but it's a start.

Another Idea I had is using auto ranging like a voltmeter and windowing like an expanded scale voltmeter. I guess this would be something like having a lower resolution ADC set for example a 1mV range and changing that 1mV window perhaps using an DAC. Or what about a digital pot to set the gain of the instrumentation amp for auto ranging?

RogerN

I don't know what you call it but years ago, I made a frequency to analog converter to drive a Joy stick port on a C-64 computer for a fax program from low band radio.

I had a comparator in the input to convert the audio tone from the receiver to square waves. Used the edges on the slopes for an edge switch where I would control a 4066 analog switch. This switch would take care of updating the sample and hold op-amp which was a jfet input and also discharge the current controlled C time constant used to generate a linear ramp time reference where it would be then translated over to the sample and hold.. The current source was a jfet and worked excellent.

THis circuit would update to the rate of what the incoming cycle was.

I know you can get charge pump chips that do the same thing however, I found the one I tried from LM, did work, but did not produce a very linear output and seem to be narrow on the range of its abilities. They did work well for basic F to V converters if you didn't mind that issue.

Jamie

The 50/60 Hz setting is related to the sampling frequency such that powerline frequency is greatly attenuated. This should give you an idea

If you can, put the ADC in the scale so you don't have to deal with noise pickup in the cabling. Much better to run SPI signals in a cable (with suitable drivers/receivers and signaling rates) than it is load cell signals.

If you must have the load cell remote to the ADC, use twisted pairs in a shield. Excitation as one pair and bridge output as the other pair.

Signal is only as significant as your reference is stable. ADC reference should also be the excitation signal.

Depending on your resolution needs, some folks forgo the instrumentation amp and go directly into a differential input channel of the ADC. I generally prefer to use an IA for the high CMRR but with every stage of amplification you also get noise.

There are a lot of apnotes out there under "strain gauge application" (strain gauge being the basic sensing element in the load cell)

noise=20

=20

=20

amp=20

generally=20

amplification=20

The signal coming out of a load cell is only a few mV/V full scale. Keep the IA.

?-)

for.=20

signals=20

=20

=20

It=20

get a=20

=20

I'm=20

voltage=20

1/2=20

LSB=20

at=20

good=20

interesting=20

A thought to keep track of: the load cell has only about 0.01% accuracy / linearity / resolution in the first place. No matter what you do with = the signal afterwards that is all you can get from them.

?-)

I put an AD7730 chip on the breadboard, got it interfaced to the Arduino, and am getting numbers from a load cell. This took some playing with to get working almost correctly but I'm getting readings that look good, 5 grams on a 30 Kg load cell is producing about 25 counts from the A/D value / 128. Should be able to do a little better with actual calibration instead of internal reference calibrations.

Next I'll write some calibration and tare functions plus conversions for lbs, oz, grams, Kg's, grains, etc.

One concern I'm having so far is sometimes the weight reading go down with the addition of a little weight, not sure if this is because of a bad load cell or if there is some misreading from the AD7730 chip. I tried using all SPI modes from the arduino library, I couldn't get mode 0 to work at all, mode 1 works most reliable, mode 2 and 3 both work but the chip locks up after a few readings.

RogerN

That part is based around a sigma-delta DAC and so has no reason not to be monotonic. I'd try substituting a different source that you know to be monotonic - eg: a pot in series with a big resistor - and repeating your measurements. bracketing the discontinuity may give some information as to what the cause is.

--
?? 100% natural

--- Posted via news://freenews.netfront.net/ - Complaints to news@netfront.net