[sane-devel] Monolta Dimage Scan Dual II (avision backend)

[Bernd Jöpen]

Hi Alex, hi everyone,

I came across your August discussion about the Minolta DiMAGE Scan Dual II and the gamma-table handling in the SANE Avision backend.

Over the past few weeks, I have been investigating very similar issues with the same scanner (USB ID 0638:026a) using SANE 1.4.0 on macOS.

Interestingly, our independent experiments led to a similar finding regarding the gamma-table transfer.

Here is a short summary of our results:

Film-holder detection: We implemented a bounded retry mechanism for BUSY responses during film-holder detection.

Gamma-table handling: We investigated several approaches to the 4096-byte gamma transfer, including padding, repeating the existing table, and interpolation.

Final gamma experiment: Our Research-009 implementation calculates and transfers 4096 gamma values per channel. The hardware logs confirm successful gamma upload and scan completion.

Scanning: We successfully completed RGB8 scans at 600 dpi and 2820 dpi, including a sequential four-frame batch at full optical resolution.

Scanner initialization: We found that opening the SANE device before inserting the film holder is important for reliable initialization in our setup.

The final 2820 dpi scans produced 2664 × 3996 RGB8 images, with 31,936,032 bytes transferred per frame.

We have documented the experiments, patch versions, hardware results, and known limitations in a public GitHub repository:

https://github.com/bjoepen/minolta-sane-patch

Our findings are based on actual hardware tests, but we do not yet claim calibrated color accuracy or complete validation of the patches for upstream inclusion.

I was particularly interested in your observation that the scanner expects 4096 gamma bytes, as this closely matches our findings.

I would be interested in your thoughts on a few points:

Have you investigated whether the 4096-byte gamma-table behavior is specific to the Scan Dual II, or shared by other Minolta devices using the Avision backend?

Did you observe any image-quality differences between the original 512-value handling and your modified implementation?

Would it make sense to investigate the film-holder BUSY handling and gamma-table correction as separate, device-specific changes for possible upstream inclusion?

I would be happy to share additional technical details and compare our findings.

Thanks for your work on this scanner and for sharing your investigation!

Best regards,
Bernd

GitHub: https://github.com/bjoepen/minolta-sane-patch
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://alioth-lists.debian.net/pipermail/sane-devel/attachments/20261016/e1b01310/attachment.htm>