X-raying banknotes

X-raying banknotes is an obvious hobby for me, since I am a radiographer. 

You cannot X-ray a banknote with a standard X-ray machine because those units have a minimum “penetrating ability” which exceeds the resistance offered by a single sheet of paper or polymer.

Without going into the technical details, this is how it is done:

1. The note is X-rayed on a pathology specimen X-ray unit which can image small items of low density

2. The field of view of this unit is not very large, so a banknote might have to be X-rayed in two portions to cover the whole note

If Wilhelm Roentgen appeared on a banknote, I would have made sure to X-ray that note as homage to him. Second best is Marie Curie who appears on at least three banknotes.

My “control” note is a 20 Złotych commemorative note from 2011:

20 Zlotys (Marie Skłodowska-Curie) - Poland – Numista

This is the radiograph:

Changes in the paper density can be detected, such as the watermark and electrotyping. Also the iridescent ink shows up as an area of relatively higher density. On a normal radiograph, structures that attenuate the X-ray beam appear lighter and those that offer less attenuation appear darker.

That note was X-rayed in two pieces and I tried four different ways to stitch it:

1. Hand stitched in Photoshop. This is very time-consuming and is prone to distortion if the two pieces cannot be rotated the exact same angle before joining.

2. Automated stitching using the “Photomerge” feature in Photoshop. In half the cases I tried, there was an obvious step where the two images had been joined.

3. PTGUI. This is not free, I used a trial version which has all the features but watermarks the image. This software was developed for stitching several photos into one panorama and therefore relies on a lot of camera EXIF data such as focal lengths and lens types. However the company says you can trick the software into working with scanned images by specifying a focal length of 1000mm and doing a few other steps also. I tried this, and it does in fact work, very well. It works also on backlight scans I have, where the note was too large to scan in one piece. The software will even show you where the join is before exporting the image. The downside is, this is £155 for the standard version and £315 for the pro.

4. Canva Affinity V3.x which is free software but you have to register an email address to use it. It does not show the join but it does work:
 

I tried that also on two backlight images and it worked on those also. There has been one case where it could not work with the two images I provided unless I cropped and rotated one of them first. The PTGUI had no issues with those same images, but Affinity is free.

I have X-rayed almost 100 notes so far and it hasn't been simple. There are settings on the X-ray machine that have to be tweaked to get a decent image.

The big thing that I found is hidden features located in security threads. I haven't decided whether I will publish everything I found on LinkedIn, but one thing that is simple enough is the X-ray appearances of Tyvek, Bradvek, polymer and paper. I'll be writing an article on that at some point but I need to do some other imaging on those notes first.

Here's another note, a $10 from 2017 which I also X-rayed in two pieces but then stitched manually in Photoshop:

Lastly, a very pretty note when X-rayed. The 200 Rand note from 2004 which was X-rayed in two pieces and stitched using Affinity:

Interesting, as always.

I wonder if X-ray imaging might assist in revealing watermarks on old well-worn banknotes and on banknotes where the ink density is heavy and makes it difficult to see watermarks.

I believe it will help, as long as there is sufficient density difference where the watermark is.

That is interesting. 

I wonder if X-ray imaging might assist in revealing watermarks on old well-worn banknotes

I suspect it would be difficult to get a good image of a well-worn banknote since the density of the paper would likely be so uneven (less where for every crease/fold) that the creases/low density regions might mask other regions one wishes to examine.  The wear & tear might make it tough to see the watermark. It would be interesting just to see a slightly circulated EF or VF note to see if the creases do show up.

Some research would be in order, I think, on a specific note in grades VG to VF.

A design with a watermark under the printing.

Some interesting images coming up. This is a damaged well folded 5 Roubles note with hand-written 62 on it.

Photo:

Backlight image (flatbed scanner with transparency backlight):

Plain infrared image:

Transmitted infrared (I have to move the note to one side because there is a cross embedded in the platform of my device):

Oblique infrared (this is the best for demonstrating folds):

Another oblique, from the other direction:

Now the radiograph:

It seems to me that X-ray could be useful for a damaged note, because it isn't affected by folds (well not to a great extent on this note)

Another one that I did, this time a polymer note that has been folded. Malaysia 1 Ringgit.

Photo:

Backlight image (flatbed scanner with transparency backlight):

Plain infrared image:

Transmitted infrared (I have to move the note to one side because there is a cross embedded in the platform of my device):

Oblique infrared (this is the best for demonstrating folds):

The radiograph:

I don't know what that horizontal line is across the middle of the note. Nothing shows up under IR or UV.

Some UV ink has been picked up on X-ray, here's an uncirculated note under UV:

All of the ink drops out in the X-ray image of the circulated Russian paper note revealing the extensive watermark very effectively. 

The IR images are less good at showing the watermark. 

Is it possible to vary the wavelength in IR in an attempt to drop out more of the ‘noise’ from the ink on the note to make the watermark stand out more clearly?

The machine I use for the transmitted IR image has a fixed wavelength, I can't change it. That is 850nm.

850 is close to visible (relatively) - it would be interesting to see what can be seen by increasing the wavelength incrementally. I ought to be able to get a chance to do that when I take a look at some of my notes under IR, soon.

I'm going to post one of my radiographs of this note:

500 Piso (Seal type 7; without tactile marks) - Philippines – Numista

Do you see what I see?

odd job

Do you see what I see?

Vertical
 

B

S

P

?

That you own that note and could replace the low resolution banknote museum ones?

Yes, I could replace the catalogue images, but I am delaying that because of a problem with the UV images. I don't replace the main images by themselves if I have UV images also (which is true most of the time).

@bumagomarazzi is correct, the item of interest is the security thread. Although the vertical BSP can be seen on the reverse of the note, the strange thing is how dense that lettering is on  X-ray. I would expect it to be seen under backlight but it is invisible.

I suspect it is magnetic ink, but I would need access to a suitable device to prove it though.

One more interesting, is the 2020 Bulgaria 5 Leva:

5 Leva - Bulgaria – Numista

It has a solid security thread windowed into four pieces on the reverse. First the obverse UV:

Now the radiograph:

Must be a magnetic thread, has opaque segments on X-ray. Also the thick tactile stripes are opaque on the sides and so are the UV fibres (which is not the case for all banknotes)

Bran, сould you take a radiograph of a modern 20 euro note? I'm interested the security strip in the center. It has strong magnetic properties, as far as I can see. It would be interesting to compare...

This is a picture from my mag device; the geometry is still far from perfect, but it's something.

That's quite nice! How did you make that device, me wants

I do have that radiograph here it is:

Here's a bonus for you Vlad, a 2002 note:

Note how nicely the coding watermark bars show up on the left

Both the 2002 and 2015-2019 notes have UV fibres but the fibres on the 2002 note are not obvious on X-ray. 

Bran, thank you for your radiographs. I see a familiar pattern on the security strip.

About my device - in short, it's a Magnetic Image Sensor and the mechanism for moving it toward the banknote. Basically, it works like a regular scanner, only instead of a row of optical pixels, it has a magnetic image sensor (MIS).

A MIS is essentially many magnetic heads (like on a tape recorder), only assembled in a single housing. Here's an example of such assemblies from different companies:
- https://www.hec-itochu.com/product/mis/01.html 
- https://www.dowaytech.com/en/TMR6318C-TMR-Magnetic-Image-Sensors.html 

I don't know if someone can buy it directly from them, the prices are very high. I got one on the second-hand marketplace four years ago, and I've been building my own magnetic scanner ever since.

Like a regular optical scanner, the MIS provides magnetic field information along a single line, and to scan a flat object (a banknote in my case), I need to move the sensor evenly. As you can see from my image, the mechanical part is still struggling, but I'm persistent. This is generally sufficient for a quick analysis, but I want to get beautiful images. The main problem is synchronizing the MIS signal with the moving part. It also took me a lot of time to write the software to convert the MIS data into an image.

Before I spent a lot of time researching magnetic field visualization, but in the end, only two options seemed viable:

1 - Magneto-optical crystals with a Faraday effect (used in the devices whose images you showed somewhere earlier, Regula use this). Pros: high image resolution. Cons: expensive, small workspace. If I were to go this route, I could move a small sensor like a CNC machine to scan the entire surface (like Regula 7701M).

2 - Magneto-resistive assemblies, like the one I use. Pros: the working size of the assemblies is quite large (up to 200 mm), allowing for processing large documents (100 mm is sufficient for banknotes; there are only a few giant banknotes that won't fit). Cons: expensive, low resolution (I haven't seen anything higher than 100 dpi). For this implementation, it's enough to move the sensor in one direction, like a regular scanner.

In the end, my choice was determined by chance – I found a magnetoresistive sensor at a reasonable price, and I decided to build my device around it. I would have included photos, but right now it's just a prototype, like an "engineer's heart attack" 🤣

Well, you have done a good job, the resolution is certainly enough, in my opinion. 

Another interesting one, the 2019 vs 2024 versions of the Kenya 500 Shillings:

They have different security threads.

As a side issue, there are enough design differences that I think these should be on their own pages.