ADC Accuracy

TomS

After reviewing AN001 I'm amazed with what the Propeller can do. I want to convert a dc level to digital with 8 bit accuracy and·I see on page 17 of an001 the resolution can be just about any value. What will the accuracy be, ie, vs temp and VDD changes?· Will it be monotonic? I would imagine the change in the digital value with change in VDD would·track well.· That means I could do a ratio of input to VDD and it would eliminate VDD from the equation. I suppose you could correct a DAC output in the same manner.· Has anybody actually measured the performance?

Also, would running without a crystal osc have any effect on accuracy as long as the clk remained stable for the duration of on measurement?

Thanks
Tom

Paul Baker

If accuracy is the important criteria, use an external ADC. The sigma delta capabilities of the Propeller are good for applications where low-cost and speed are the predominate criteria. Higher accuracy can be generated by making the sample aperature larger and throwing away the bits below the noise floor, but all of the other points you mention are factors.

The onboard oscillators are accurate enough in the measurment window, however repeatability·is an issue especially with any change in the ambient temperature.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Paul Baker
Propeller Applications Engineer

Parallax, Inc.

Post Edited (Paul Baker (Parallax)) : 2/15/2007 10:58:01 PM GMT

TomS

So, you don't think I can achieve 8 bit accuracy? I would very much like to keep the parts count down.

Tom

Newzed

Tom, 8-bit accuracy is pretty coarse.· If you are trying to digitize 0 to 3 volts, each ADC count is almost 12mv.· With my TLC2543 ADC, reading
0 to 5 volts,·I get 1.21mv per ADC count.· Reading 0 to 3 volts, I get about .74mv per ADC count.

Sid

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Yesterday is history, tomorrow is a mystery, and today is a gift.

That is why they call it the present.

Don't have VGA?
Newzed@aol.com
·

Paul Baker

Yes 8 bit accuracy is possble, you will likely have to do a summation greater than 256 clock cycles to achieve the required accuracy but it is possible since you are measuring a DC value. It will be suceptible to changes in power supply, temperature (if you are using the onboard oscillator), etc.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Paul Baker
Propeller Applications Engineer

Parallax, Inc.

TomS

I can do a summation of any number of clock cycles as long as I get 2 readings per second. Anything faster could be averaged. As I said in my first posting, I think I can compensate for VDD changes and the temperature range is room temp +/- 10 deg F. Not much really. Can one cog do two ADC's? I guess the only way to find out if it will work is to spend the money and build a prototype.

Tom

Paul Baker

Then it should be no problem, yes each cog can perform 2 ADC functions for a total of 16 per chip (works out nicely that this consumes all 32 I/O pins)

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Paul Baker
Propeller Applications Engineer

Parallax, Inc.

TomS

When I get it up and running I'll let you know how it works.

Tom

pjv

Hi Paul & Tom;

In regard to oscillator changes due to temperature.... it has negligible effect. A typical conversion (especially only 8 bits) will be completed in such a short time that the oscillator can be considered stable. Temperature affecting the switching threshold is another story that I have yet to ferret out.

The ratiometric concept definitely works.

I'm presently pursuing a 16 bit model using an SX (quite noisy), and we'll see how it goes.... so far the signs are encouraging.

Cheers

Peter (pjv)