Author: Charles

I got to wondering with ChatGPT after noticing some articles online about the transition to the new paper EZPassMA transponders. I also was curious what info ChatGPT would provide regarding their similarity to the chip embedded credit card technology.

Check out my Gemini conversation here.

I just was dealing with my EZPassMA and discovered that they are now going to a paper replacement for the transponder I have for new customers and replacements. I noticed that there was specific mention of not using batteries in the new stickers that are replacing the transponders. I am interested in knowing how this tecnology works without a battery in the sensor attached to the windshield.

It works by borrowing power from the toll equipment

The new E-ZPass MA sticker is a passive Radio Frequency Identification tag, usually shortened to RFID. It contains:

• A tiny electronic chip
• A very thin metallic antenna
• A paper or plastic backing
• Adhesive for attaching it to the windshield

Massachusetts began issuing these stickers to new customers and as replacements on March 1, 2026. Existing plastic transponders can continue operating. (WBUR)

What happens as you drive underneath a gantry

The overhead toll gantry sends out a powerful radio signal. When your vehicle enters that radio field:

1. The sticker’s antenna catches some radio energy.
2. A microscopic rectifier inside the chip converts that radio-frequency energy into direct-current electricity.
3. That tiny amount of electricity wakes the chip for a fraction of a second.
4. The chip retrieves its identification number from permanent memory.
5. It sends that number back to the gantry.

The identification number remains stored without electricity, much like information stored in flash memory.

But how does it transmit without power?

This is the clever part. The sticker does not generate a conventional radio transmission.

Instead, the chip rapidly changes the electrical load connected to its antenna. That changes how the incoming gantry signal is reflected. The reader detects those changing reflections and translates them into digital information.

This process is called backscatter communication. The National Institute of Standards and Technology describes it as reflecting or backscattering electromagnetic energy received from the reader. (NIST Computer Security Resource Center)

A simplified picture looks like this:

Toll gantry sends radio energy
             ↓
Sticker antenna collects energy
             ↓
Chip wakes and reads its ID
             ↓
Chip changes antenna reflection
             ↓
Gantry receives the reflected ID
             ↓
Computer matches ID to your E-ZPass account

A good electronics analogy

It is somewhat like combining a crystal radio with a radar reflector.

A crystal radio extracts enough energy from a radio signal to operate without a battery. The sticker similarly extracts energy from the gantry. However, instead of producing sound, it electronically changes its reflection to communicate its identification number.

Modern passive RFID tags use printed, etched or deposited conductive antennas. The reader’s radio waves power the chip, and the reader then receives the tag’s response. (Impinj)

Why the windshield position matters

The antenna is carefully designed for its operating frequency and surroundings. The windshield becomes part of that radio-frequency environment. Therefore, incorrect mounting, metallic window coatings or certain heated windshields can weaken the signal.

The sticker should be mounted exactly where the instructions specify. Unlike your removable plastic transponder, a sticker tag generally cannot be transferred between vehicles after installation. Removing it can damage the thin antenna or change its electrical tuning.

Does it transmit continuously?

No. Without a battery, it cannot sit there broadcasting its location. It normally remains electrically inactive until it enters a sufficiently strong field from a compatible reader.

That does not mean it is completely unreadable outside toll roads—a compatible reader placed close enough could potentially energize it. However, it does not independently beacon or transmit while your car is sitting in the driveway.

So the “paper transponder” really is a remarkably thin radio circuit. The expensive transmitter, power supply and sensitive receiver are all located on the toll gantry. The windshield sticker only needs enough circuitry to wake briefly and alter a reflected signal. Pretty ingenious stuff. 📡🚗

Are there any similarities to how the modern credit cards operate that you can “tap” to do your purchase?

Yes—the basic idea is very similar. Both can operate without their own battery by borrowing energy from the reader.

A tap-to-pay credit card contains a tiny chip and antenna. When you hold it near the payment terminal:

1. The terminal creates a radio-frequency field.
2. The card’s antenna captures energy from that field.
3. The energy briefly powers the card’s chip.
4. The chip and terminal exchange payment information.

The main differences are range, radio method, and security.

Tap-to-pay card

A contactless credit card normally uses Near Field Communication, or NFC, at 13.56 megahertz. It works through magnetic coupling—somewhat like a very loosely coupled transformer.

Its range is intentionally short, usually just a few centimeters. The chip also performs cryptographic calculations and generates transaction-specific information. It does not merely transmit your printed card number unchanged.

Toll sticker

A windshield toll sticker generally uses longer-range passive Radio Frequency Identification technology. Instead of transformer-like magnetic coupling, it usually captures energy from a radio wave and responds by changing how that signal reflects back toward the reader.

That allows the toll reader to identify it from several feet away while your vehicle is moving.

The simplest comparison

Tap-to-pay card:
Reader powers chip → chip performs secure transaction → card responds

Toll sticker:
Gantry powers chip → chip supplies tag identity → gantry bills account

So they are close technological cousins. Both are passive radio devices, but the credit card is designed for very close, secure financial communication, while the toll sticker is designed for fast identification at a much greater distance. 💳📡

This chat was about my asking if ChatGPT could help me turn handwriting or printing into a usable font. I chose to ask it about using Linux because there are so many open source programs out there to use on my Linux Debian 13 workstation. Note also that I was ony asking about whether ChatGPT would be able to help me with this type of project – not for it to dive in and start indicating what to do.

The Temporary Chat

I used what is called a “temporary” chat in this one. I don’t think I have covered much about using them in my articles here. ChatGPT has a temporary (unsaved) chat mode available. It goes away entirely once you exit it. I just chose that mode as I wanted a fresh perspective. Creating chats in existing projects or even as new ones that will be saved draw upon all your other materials. Plus, using temporary chats helps to reduce the clutter in my account.

As you can see from the output, one does not have to know Linux in depth in order to use ChatGPT for this type of project. It is much like having a buddy that knows pretty much everything necessary to guide you along. I am fortunate to have an extensive background in much of what I use the AI systems for. That background enables me to have a better sense when developing the questions. Plus, I am confident when it is wrong. Which does happen, just like with our buddies on occasion.

Here was my entire raw chat:

If I print out on white paper the letters of the alphabet and the digits 1 through 9 and some other characters, can you help me to create a font set that is my own printing using my Linux workstation? I am just gathering the idea right now and don’t need deep info yet.

Yes. You can turn your handwriting/printing into a usable font on Linux. 🙂

The usual process is:

1. Print a template sheet with boxes for uppercase, lowercase, digits, punctuation, and symbols.
2. Fill it in neatly with a dark pen or marker.
3. Scan it cleanly as a high-resolution image or PDF.
4. Use font-creation software to trace each character and assign it to its keyboard key.
5. Export it as a TrueType Font (.ttf) or OpenType Font (.otf) file.
6. Install it in your Linux user fonts folder, usually ~/.local/share/fonts/, then refresh the font cache with:

   fc-cache -f -v
   

On Linux, FontForge is the classic powerful tool for this. It can import traced letter shapes, adjust spacing, and generate a real font file. Another route is to use an online handwriting-template service for the initial conversion, then refine the result in FontForge.

A first version does not need to be fancy. Start with:

• A–Z
• a–z
• 0–9
• Period, comma, apostrophe, quotation marks
• Parentheses, hyphen, slash, question mark, exclamation point
• @ # $ % & * + =

The tricky parts are not the letters themselves. They are consistent size, baseline alignment, character spacing, and making similar characters look distinct—such as I, l, 1, O, and 0.

It is a very doable home project, and a printed-template-and-scanner workflow is probably the easiest way to capture your own printing.

Along this line, if I supply some cursive examples of my writing, can I create the same (a usable font) that I can use? I am looking to do this on one of my blog sites so I will need to be able to use it in a web page.

Yes. Cursive can become a usable font too, including on a blog site.

For cursive, you would ideally create several versions of letters because handwriting changes depending on what comes before and after. For example, your lowercase r may connect differently after an o than after a t. A basic font can still look quite good with one form of each character, but advanced cursive fonts use alternate letter shapes and connection rules.

For the web, you would export the font in modern web formats, mainly:

.woff2
.woff

Then upload the font files to your site and load them with CSS like this:

@font-face {
  font-family: "Charles Cursive";
  src: url("/fonts/charles-cursive.woff2") format("woff2");
  font-display: swap;
}

.handwriting {
  font-family: "Charles Cursive", cursive;
}

Then use it in a page or post:

<p class="handwriting">This is written in my own cursive style.</p>

A good first approach would be:

• Write a carefully prepared alphabet sheet in your normal cursive.
• Include lowercase, uppercase, numbers, punctuation, and common symbols.
• Add sample connecting pairs such as th, ch, ll, oo, ing, st, and er.
• Scan it at high resolution with dark ink on plain white paper.
• Build the first font with FontForge on Linux.
• Convert the finished font to Web Open Font Format 2 (.woff2) for the website.

One practical warning: cursive fonts work best for headings, pull quotes, signatures, and short notes. Long blog paragraphs in handwriting can become tiring to read, especially on phones.

Your own cursive signature font for article endings or special callout boxes would be a great fit. ✍️

Would the printed version be better for reading?

Yes—usually printed handwriting is much better for reading, especially on a blog.

Print-style letters:

• Stay clearer at small sizes.
• Work better on phones and tablets.
• Make punctuation and numbers easier to distinguish.
• Need less fancy font programming than cursive connections.
• Can be used for short paragraphs, labels, captions, and callout boxes.

Cursive is nice for a signature, quote, heading, or personal note. But for regular blog text, your printed style would likely be more practical and more accessible.

A nice setup would be:

• Your printed font for headings, side notes, and highlighted text.
• A normal web font for long article paragraphs.
• Your cursive font only for signatures or decorative touches.

That gives the site personality without making readers work too hard. ✍️

Handling ChatGPT Large Files

Here is how ChatGPT responded to my same question I asked Gemini in the previous post about examining a large .zip file containing .jpg items that needed to be OCRed. I have not placed the entire ChatGPT session here because this was a project to check on questions about an annuity policy for a friend:

I have scanned in a large document’s pages as greyscale .jpg page files that I am hoping OCR can be used on them. They are in a .zip file. Are you able to read in a .zip file and answer questions about the overall project from my scanned pages?

Yes, Charles — upload the .zip file and I should be able to work with it. 👍

I can generally:

1. Open/extract the ZIP
2. Review the scanned JPG page images
3. Use image reading and OCR-style extraction where needed
4. Summarize the document
5. Answer questions about the contents
6. Look for names, dates, topics, repeated themes, contradictions, missing pages, etc.
7. Help organize it into sections or create a cleaner typed version

A few practical notes:

• Greyscale JPG scans are usually fine, especially if the text is clear.
• OCR quality depends heavily on scan sharpness, page skew, margins, contrast, handwriting, and font size.
• If the document is very large, I may process it in sections and give you useful partial results rather than pretending everything is perfect.
• If the pages are out of order, filenames like page-001.jpg, page-002.jpg, etc. help a lot.

Go ahead and upload the .zip file when ready.

This is the file. How well are you able to handle it?

I can handle this ZIP pretty well. I inspected it, and it looks very workable. 👍

Here’s what I found:

• The ZIP is about 122 MB.
• It contains 125 JPG page scans, numbered Annuity0000.JPG through Annuity0124.JPG.
• The pages appear complete with no missing page numbers.
• Every image is greyscale JPG.
• Every page is the same size: 2550 × 3299 pixels, which is a good scan size for OCR.
• The folder structure is nicely organized into sections, such as:
  • Preface to contract
  • Actual contract
  • Contract schedule
  • Riders
  • Original annuity application
  • IRA disclosures
  • Privacy notice
  • Withdrawal forms
  • Required minimum distribution form
  • Annuity values / withdrawal benefits
  • Financial professional info

I tested OCR on several sample pages. The results are good to very good on typed contract pages. For example, the OCR clearly picked up contract schedule details, dates, values, headings, and much of the form text.

The weaker spots will likely be:

• handwritten entries
• checkboxes
• small print in dense forms
• pages with logos, shading, or gray background noise
• tables where columns need careful interpretation
• exact dollar amounts, percentages, contract numbers, and dates that must be verified carefully

So, my practical answer is:

Yes, I can read this ZIP and answer questions about the overall project.
For general summaries, explanations, section-by-section review, and finding key items, it should work well. For legally or financially critical details, I would OCR it, then spot-check against the image pages before treating exact numbers as final.

Best next step would be for you to ask something like:

“Give me a plain-English summary of the whole document.”

or

“Find anything related to withdrawals, penalties, surrender charges, death benefits, and required minimum distributions.”