Evaluating measurements

I programed a PIC 18F to generate pulses going from 120 to 6000 Hz.

Measured the output with two frequencimeters (ME and WA) plus the reading from an analog TRIO scope.

In the Excel file I listed following values:

PIC: frequency expected at the PIC's output.
WA: frequency values obtained with frequencimeter WA.
ME: frequency values obtained with frequencimeter ME.
TRIO: frequency calculated from the readings (which also listed).

Readings approximated to 0.25 of a division. Scope is analogic so I had to actually read the divisions in the mask.

On the right I listed the differences between PIC value and the three instruments.

The x axis of the graphic helds PIC values. The curves, the three differences above.

My conclusions, rather limited, besides that all my instruments are of poor quality:

a) PIC could be not producing the expected frequency. (But MPSIM, with current software, allows to think that I am getting the best of it with values close to the Hz in all cases).

b) I am very bad at reading my scope.

c) Frequencimeter ME is worst than WA. (Both frequencimeters are of the type embeded in garden variety multimeters).

Do you have more useful / educated conclusions to make? I will appreciate that.

Attached file here

**broken link removed**

Looking at your spreadsheet, I dont think the results are all that bad.
How accurate do you want the pulses to be?
If I wanted an accurate frequency, I dont think I would use a PIC but probably a divider from a crystal. But it is late, I am tired and I would have to think a bit more about that!

The scope is at very best only 2% accurate, look at the specifications in the handbook.
You have a couple of rogue values at U18 and U35 in the spreadsheet, but otherwise it is comparable to the ME frequency meter.

I have a DMM with a frequency range, the handbook says 0.2% accuracy. I dont use it to measure frequency as I have a real frequency counter which is far better.

There could be something strange happening with the pic as there is a cyclic variation through the frequency range, which I dont think is due to measurement error.

JimB

I may have misunderstood your plot and am not sure what TRIO actually is. So, the following applies only to WA and ME. Assuming the plot is an error plot versus "actual" frequency, the error for ME looks to be 1% or less of the actual frequency (X-axis). WA has a greater percentage error and an obvious bias as graphed. However, if the bias for WA is corrected (e.g., by calibration), WA may in fact have a smaller average error. I suggest you do some simple statistics on the errors, such as standard deviation.

I would also point out that the data for WA and ME should probably be rounded to whole numbers. It is unlikely the decimals would almost all be "zero."

What level of accuracy were you hoping for? Based on what you describe of the experiment, 1% seems pretty good to me. John