What's the Minimum Velocity That PIV can Measure ?

[ABHIMANYU DUBEY]

My doubt is not related to PIV-lab per say, but with PIV in general. 

I've attached the raw image and the Pre processed Image for reference here. I'm doing experiments in a rectangular channel, and the images cover a bit more than half of the channel in the cross stream direction. The bright line in the images near the bottom edge of the image is actually the lower edge of the channel which was removed after background subtraction and contrast adjustment.

 Now this bright spot creates a problem in the determination of the position of wall. The bright spot spans approximately 30 pixels in height. So what I tried here is that one I've eliminated the wall completely, I process the complete Image as shown in the vector file Image(That's actually an Instantaneous Frame) , do spatial and frame averaging over say 3000 to 5000 frames and then based on the velocity profile I determine the wall position. 

Now the distance b/t two successive vectors here is 4 pixels along cross stream direction, So I'm able to measure velocity statistics at a distance of 4 to 8 pixels from the wall, which translates to 24 to 30 microns, ( assuming a calibration factor of 8 micron/pixel, for most of the cases it's actually less than 8 ). Quite surprisingly even the standard error for these very close to wall measurements is quite less. 

I've attached a MAT file here which contains 3 columns, Y_in_Pixels , U_in_Pixel_per_Frame , and Standard_Error_In_U_Pixel_per_Frame. Last 8 rows   of this matrix confirms the same. A quick error plot of this data show how good or bad the measurement was. 

I'm a bit skeptical about these results because they seem too good to be true. So I have 2 questions regarding this situation.

1. How can I calculate what's the minimum velocity that a PIV system can measure. I can get a measure of uncertainty by evaluating peak to noise peak ( for example) ratio from the correlation plane, but uncertainty in measurement is not the same as  error in measurement. 
2. Is this work around of estimating the position of wall by using high resolution PIV data like this a good alternative ? Because if wall position is changed then the number of pixels lying between the wall and the mid Plane of the channel will change which in turn would change the micron to pixel conversion factor. 

I also take an image of a scale for the purpose of calibration but since the camera the moved parallel to the focal plane for imaging upper and lower half of the channel, irrespective of however careful I'm during the experiments there is a slight change in the calibration factor. For this case for example the calib came out to be 7.88 microns/pixel from the scale, but upon re-calculating it based on the number of pixels lying b/t the wall and the mid plane of the channel it came out to be 7.022 micron/pixel. 

If you are still reading, then I sincerely thank you for kind efforts. 

Thanks Very Much,

Regards,

Abhimanyu

Google Drive Link For Same Images

[ABHIMANYU DUBEY]

My apologies,

I had attached the wrong google drive folder in the last post. 

 Kindly please find the relevant link to google drive and the data attached here. 

 

The data (velocity_data.mat in the attached folder)contains three columns:

1. Y coordinates in pixels,

2. U in Pixel/Frame

3. Corresponding Standard Error in U ( Pixel/Frame )

The Origin has been shifted to the lower left part of the image, and the date is w.r.t to this origin. 

Calibration Factor = 7.022 microns/pixel

Straddle time = 20 microsec : 

Thanks,

Apologies once again. 

 

Regards,

Abhimanyu

 

Link For Google Drive