Wednesday, June 24, 2020

GPixel Announces 103MP, 28fps, 12b Global Shutter Sensor

Gpixel announces the GMAX32103, a large format Global Shutter CMOS sensor for industrial applications. The sensor is based on 3.2 µm charge domain GS pixel, provides 11276(H) x 9200(V) resolution (103 MP), and supports up to 28fps with 12bit output. GMAX32103 is aimed to the demanding machine vision applications and aerial imaging.

The 3.2 um pixel achieves a full well capacity of 10k e-, read noise less than 2 e- and maximum DR of 66dB. With the implementation of micro lens and light pipe technologies, the sensor provides a peak QE of 65%, a shutter efficiency of 1/15,000 and excellent angular response. GMAX32103 offers a large FOV to expand single-shot capabilities and a nearly square aspect ratio (1.27:1), which is optimal for inspection applications.

GMAX32103 uses 52 pairs of sub-LVDS channels, each run at a maximum speed of 960MHz. The sensor supports channel multiplexing for lower data rate implementations, and integrates a variety of read-out functions including up to 32 regions of horizontal windowing (region of interest), sub sampling and image flipping. GMAX32103’s is packaged in 209-pin uPGA ceramic package with an outer dimension of 49.5 mm x 48.1 mm.

We are very thrilled with the introduction of GMAX32103. The further expansion of Gpixel’s line up of extremely high-resolution sensors based on an industry proven and widely accepted platform, empowers our customers to tackle demanding applications and to address the industry’s needs for ever increasing image accuracy and throughput. This product is part of our fast growing GMAX product family, which will be further expanded in the very near future with other exciting products,” says Wim Wuyts, CCO of Gpixel.

GMAX32103 engineering samples are expected in November 2020.


  1. 3.2um charge-domain GS - so, probably a 5T arch? So, no true CDS?

    1. Albert Theuwissen - Harvest ImagingJune 25, 2020 at 11:01 PM

      Global shutter with charge-domain storage node can apply CDS, that is not a problem. BTW, if you talk about true CDS, then there must be something like an untrue or false CDS ?

    2. This terminology is used frequently across datasheets.

      So, yes there is a false CDS FYI. And I know that a lot of people talk about CDS when they actually mean Uncorrelated (false) DS. So, better to use the right term.

    3. Albert Theuwissen - Harvest ImagingJune 26, 2020 at 3:03 PM

      100 % agreement : the word CORRELATED refers to the correlated kTC noise in the two samples. So one talks about CDS if the kTC noise is correlated, and one talks about DDS if the kTC noise is not correlated. As easy as that, and not about true- or non-true or false or anything else.

  2. With a read noise less than 2e-, there is probably a true CDS of the sensing node.

  3. Indeed, pixel operates in (true ;-) CDS mode. Noise at nominal gain (max FWC) is ~5 e- which reduces to ~ 2 e- at higher gain (lower FWC too obviously). Max DR is 66.5 dB at max FWC setting.
    Wim Wuyts - Gpixel -


All comments are moderated to avoid spam and personal attacks.