There are two main limits in film - the contrast resolution (depending on noise) and the spatial resolution. For NDT films I will try to see where are the limits. As "measurement instrument" I am using two prototypes of the KowoScan film scanner, one with 27µm pixel size and the second one with 46µm (which is now the KowoScan X50). Additionally I have three different lenses - all with very high quality - which increases the spatial resolution of the scanner.

I start with the highest resolution which I could get - with 8.5µm pixel size at film level. The special macro lens was inserted in the KowoScan XS housing.

The picture shows the captured image, the KowoScanXS prototype and the special lens.

Two films (D2 type) of the famous BAM5 weld were captured with two different duplex wire gauges. The two films are close together that the duplex wire gauges are within one scanned image.
As the contrast in the image here is quite low, I used the filter function of the KowoScan software for this image:


The profile function proves that the smallest duplex wire pair is resolved with more then 30%

which is not only a problem of spatial resolution, also the contrast is quite low as we see 0.02mm on 8mm of steel (0.25% of platinum on steel).

The printed line pair set which was placed above the upper duplex wire gauge proves also a spatial resolution of 20µm - the duplex line pair is resolved with 49%:


If we use this digitizer for a typical 10x24cm film we see this area of the film:


With the KowoScan filter we see much more details


As taken in the contrast limit thread we also use the "nose" for the visibility of details:

Here we see some porosity but the basic weld is very noisy. The film has a density of D=5.2 and the scan with the "standard KowoScanX50" show much less noise.

The scan time was 40 minutes - therefore I do not expect noise from the digitizer.

The next pixel size is 15.3µm on film level. The macro lens is exchanged to a "shorter" lens.

With nearly double the pixel size we get four times the area in one image.
The filtered image shows all the details:


As reference we cut out the nose

and see the same porosity as in the image with 8.5µm image but there is nearly no noise on the weld compared to the version with 8.5µm .

If we compared the resolved spatial resolution from the version with 15.3µm and 8.5µm pixel size

we see the same spatial resolution of both digitizers

The time for scanning with the 15.3µm pixel size was 20 minutes.

When we put the lens from before into the KowoScanX50, we get a pixel size of 30µm on film level.
The scan of the 10x24cm film of the BAM5 shows more than the complete film:


As before we use the filter of the KowoScan software


and cut out the "nose"

We can see some porosity as before and also nice cracks.

The line profile function on the duplex wire gauge shows the 32µm duplex wire resolved with 22%


and the 30µm line pair of the printed film proves this spatial resolution


The scan time for this image was 20 seconds only! (compared to the 20 minutes before just 1.7% of time).
With a little bit longer scan time some more porosities in the nose would be resolved, but the 20s are sufficient to archive scan class DS.

The images of the standard KowoScanX50 with 46.3µm pixel size can be seen in the contrast limit thread.

What digitizer resolution should you select? The old ISO EN14096 requires a pixel size of 15µm for film exposed with 100kV or less and 30µm for films between 100kV and 200kV. Which digitizer has such small pixels?

Ultimately, it's all a question of economic efficiency. Beside the much longer scan time with the smaller pixel also the geometry is important. If you have to digitize 10x24cm films you will get four of them in a KowoScanX50 within one scan:

This is shown with the blue frame.
Putting the longer lens in the KowoScanX50 you get 30µm pixel size and will get in best case two of the 10x24cm films with one scan (if you accept a small overlap between two films in the center). This is marked with the red frame.

Changing to 15µm pixel size you will get only half of the length of the 10x24cm film and the width not completely with one scan (green frame). And the scan time will be more than a few minutes.
The version with the 8.5µm pixel size has no benefit in image quality (spatial resolution and detail visibility) - on the contrary, there is more noise in the image, which could prevent important details from being recognized - and the scanned area of the film is very small (purple frame).


In Table 2 of the updated ISO EN14096 the pixel size is replaced with the basic Spatial Resolution (SRb) measured with the duplex wire gauge or a synthetic line pair set. The minimum there are 30µm SRb - which is resolved with the longer lens in the KowoScanX50.
But in a footnote higher spatial resolutions may be required for films <70kV. Do you think this is necessary?