Old JPEG, BigTIFF, GeoTIFF — what doesn't convert

TIFF's design philosophy is "extensibility by tag". Anyone can register new tags or compression types; readers that don't know them can ignore them and read the basic image. Over 35 years, this has produced a long tail of TIFF variants that work in specialized tools but fail in general-purpose converters. TIFF2PDF handles the common cases; the variants below tend to fail.

Old JPEG — compression 6

An early JPEG-in-TIFF design from 1992 that predates the proper TIFF 6.0 supplement. Multiple TIFF library implementations wrote files that don't conform to any documented spec; readers had to guess what the writer meant.

By 1995 the TIFF Technical Committee deprecated Old JPEG and introduced Compression 7 (modern JPEG-in-TIFF). Old JPEG files still exist in legacy archives (1990s digital library projects, scientific imaging from that era).

Old JPEG TIFFs typically fail to convert. Re-save the file as modern JPEG-in-TIFF using a TIFF utility that supports the legacy decoder, then upload the converted version.

BigTIFF — for files over 4 GB

Standard TIFF uses 32-bit byte offsets, capping the file size at 4 GB. BigTIFF extends to 64-bit offsets. Used for very large GIS rasters, microscopy slides, and astronomical images.

BigTIFF is identifiable by a magic number 43 in the header (vs 42 for standard TIFF). The extension is usually still .tif or .tiff; the difference is in the bytes.

BigTIFF reading is supported. The format itself isn't the issue — it's that BigTIFF is used for content that's typically too large for a PDF page (gigapixel imagery). Conversion fails on memory limits well before reaching billions of pixels, and the resulting PDF page would be impractical to display.

For typical BigTIFFs that exceed the limit, options:

1. Downsample before upload in any image editor (Photoshop, GIMP, Preview on Mac all open large TIFFs and let you resize).
2. Crop a region of interest the same way.
3. Use a tool designed for huge images (a GIS application for geospatial rasters, a slide-imaging tool for microscopy) that produces a pre-rendered preview at PDF-friendly resolution.

GeoTIFF — geographic metadata

A TIFF carrying georeferencing information: which projection the image is in, what coordinate the corners represent in real-world latitude/longitude, what the spatial resolution is on the ground. Used by every GIS tool.

The image data itself is standard TIFF (any compression, color depth). The geo information lives in private tags 33550 (ModelPixelScale), 33922 (ModelTiepoint), 34735 (GeoKeyDirectory), and a few more.

GeoTIFFs convert as regular TIFFs — the resulting PDF page contains the image content, but the geo metadata is dropped, since plain PDF has no mechanism for representing "this page is a map of these coordinates".

If you need the georeferencing preserved in PDF, use a GIS tool that exports to "GeoPDF" (a PDF with geo extensions, supported by some PDF readers and GIS applications). TIFF2PDF is for displaying the image, not preserving its spatial reference.

Floating-point pixel data

TIFF can store pixel values as 32-bit or 64-bit IEEE floats, with SampleFormat = 3. Used for:

• HDR photography (each channel ranges over many orders of magnitude).
• Scientific data (sensor readings calibrated to physical units).
• Computer graphics intermediate buffers.

PDF's image model is integer-valued: 1, 2, 4, 8, or 16 bits per component. There is no PDF native float-pixel format, so floating-point TIFFs are converted to integer-valued pixels by the conversion library before embedding. The exact mapping depends on the library's defaults; expect a generic linear or clipped reduction rather than a perceptual tone-map.

The conversion is destructive — HDR information is lost. For scientific or HDR content where the float values themselves matter, PDF is the wrong delivery format; use the TIFF directly. To control the tone-mapping, run the float TIFF through a dedicated HDR-to-LDR tool first, then upload the resulting 8-bit TIFF.

Tiled images over the size limit

Tiled TIFFs from specialized domains can be enormous:

• Whole-slide microscopy images: 50,000 × 50,000 pixels = 2.5 gigapixels.
• Aerial orthophotos at 5 cm/pixel: 100,000 × 100,000 pixels = 10 gigapixels.

For these, conversion exhausts the per-process memory budget and fails. Pre-process to a smaller version with the source tool's "export overview" or "save as preview" feature, or use a domain-specific tool to render a PDF-friendly preview.

Alpha channels and extra samples

TIFF can have more than 4 channels via the ExtraSamples tag. RGBA is the common case (extra sample = alpha). But some workflows store layers as extra channels: a "selection mask" or a "depth channel" beyond the 3 RGB channels.

TIFF2PDF handles RGBA correctly (the alpha becomes a PDF SMask). For TIFFs with more than 4 samples, only the first 3 (RGB) or 4 (RGBA, if extra sample 1 is marked as alpha) are used. Additional channels are dropped silently.

Byte-order quirks

TIFF supports both little-endian (Intel, "II") and big-endian (Motorola, "MM") byte orders, indicated in the header. Modern tools default to little-endian; some older tools or non-x86 platforms produce big-endian files.

TIFF2PDF handles both. No user action needed.

Checking your TIFF before upload

If TIFF2PDF rejects your file with an "unsupported variant" message and you want to know why, open the TIFF in a full-featured image viewer (Photoshop, GIMP, IrfanView on Windows). The "Image Properties" or "File Info" panel reports the compression type, color mode, and bit depth — the three most common reasons for a reject.

If you see compression labelled "Old JPEG", anything in the "Custom" range (typically vendor-specific), or a non-standard color mode, that's a clue you need to re-save the file with standard settings before uploading.