Mirrorless


Panasonic G9 Camera Review

February 15, 2018

By Greg Scoblete

PDN is a member of the Technical Image Press Association which has contracted with Image Engineering to perform detailed lab tests of digital cameras. See here for a full methodological rundown of how Image Engineering puts cameras through their paces. Full res files of every visual in this review are available to download for your pixel-peeping pleasure here. A hands-on review of the Panasonic G9 will appear in the April issue of PDN.

Introduction

While Panasonic’s GH5 is no slouch in the stills department, it’s so packed with video features that you could be forgiven for treating it as a video camera. The G9 looks to rectify that. If the GH5 is the hybrid flagship, the G9 is Panasonic’s flagship mirrorless still camera. It’s modeled closely on the GH5, but optimized for still photography.

It features a 20-megapixel CMOS sensor with no low-pass filter and a third of a stop more dynamic range than the GH5.

It can crank at up to 20 fps using an electronic shutter with autofocusing or up to 60 fps with focus fixed on the first frame for up to 50 RAW files. If you’re using a mechanical shutter, you’ll hit burst rates of 9 fps in AF-C and 12 fps with focus fixed on the first frame.

Panasonic has improved its dual image stabilization system, which combines the body stabilizer with select image-stabilized Panasonic lenses to deliver up to 6.5 stops of image stabilization. Importantly, Panasonic has improved the system so that it can stabilize effectively at both wide and telephoto focal lengths (up to the equivalent of 280mm). Previously, dual stabilization was effective at wide focal lengths but less so out at telephoto ranges.

It can also deliver the 6.5 stops of correction even if you’re not using a stabilized lens,

While the G9 isn’t pitched as heavily toward filmmakers as the GH5, it still boasts 4K/60p video recording (8-bit 4:2:0 internally and 8-bit, 4:2:2 externally via HDMI). Full HD recording is available up to 180 fps. It offers 6K and 4K photo modes to isolate high-res images from 30fps video bursts.

Panasonic has borrowed a page from Olympus and Ricoh by incorporating a high-resolution mode that moves the image sensor by half a pixel for a total of eight images that are compiled in camera to create a single high res image equivalent to about 80-megapixels.

You’ll find a new AF Point Scope feature that enlarges the area in focus, plus the camera will remember the focus point as you switch between a vertical and horizontal orientation. As you move focus points horizontally they will now loop around to the other side of the frame, Pac-Man style, rather than get stuck at the edge.

Additional features include:

  • weather proof build
  • EVF with .83x view and 120fps refresh rate (EVF magnification can be adjusted to .7x and .77x)
  • a status LCD on the top of the camera
  • 225 AF points
  • a joystick to select AF points
  • dual SD card slots
  • 200K shutter life
  • 3-inch free angle LCD display
  • Wi-Fi and Bluetooth
  • mechanical shutter speeds of 60-1/8,000 sec.
  • electronic shutter speeds up to 1/32,000 sec.

The G9 will is available now for $1,700. Panasonic will also offer a battery grip for the camera for $350.

Image Engineering Test Results

Resolution   

  • At ISO 200, 1860 line pairs per picture height (LP/PH; 96 percent of the theoretical maximum of the 20.3 megapixel sensor), and at ISO 400, 867 LP/PH, 96 percent of maximum.
  • Resolution declines very slightly as ISO increases: at ISO 3200 and ISO 6400, 1686 and 1684 LP/PH are recorded, both 87 percent of the theoretical maximum.
  • At the highest ISO of ISO 25600, 1474 LP/PH are recorded, 76 percent of the theoretical maximum.
  • Slightly better resolution than the DMC-G81, a predecessor with a 16 megapixel sensor, which produced resolution measurements at ISO 200 of 1787 LP/PH (104 percent of its theoretical maximum).

This graph shows the loss of contrast (y-axis) as a function of the spatial frequency in line pairs per picture height (x-axis) for different ISO-sensitivities (colored lines). The further to the right a curve stretches before descending, the better the resolution at that ISO. The limiting resolution for each ISO can be found by identifying to the highest spatial frequency which results in a contrast of 0.1, or where the ISO curve crosses the thicker horizontal thicker black line marking 0.1. The vertical pink line is a reference representing half the number of pixels in the sensor height (the Nyquist frequency).

Texture loss

  • At ISO 200, the G9 produces a MTF50 value of 1364 LP/PH in high contrast scenes, and 1237 LP/PH in low-contrast parts of the scene.
  • At ISO 800, MTF50 is 1308 LP/PH (high contrast) and 1009 (low contrast); at ISO 6400, MTF50 is 658 LP/PH (high contrast) and 460 LP/PH in low contrast.
  • Somewhat better than the G81, which at ISO 800 produced MTF50 values of 928 LP/PH in high-contrast scenes and 804 in low contrast.
  • Relatively few artifacts are produced by the G9: 12.0 percent at ISO 200 (high contrast) and 19.7 percent low contrast.
  • At ISO 6400, the G9 produces 40.3 percent artifacts in recording high-contrast scenes and 52.8 percent at low contrast scenes.
  • The DMC-G81 produced rather more artifacts at lower ISO when recording both high and low-contrast textured patterns: at ISO 200, 27 percent (high contrast) and 25.2 percent (low contrast). However, at higher ISO, results were similar: 43.1 percent at ISO 6400 (high contrast) and 53.3 percent (low contrast).

An artifact is an alteration in a digital image due to technology or technique of processing. Artifacts stem from noise, compression, and sharpening. This graph plots the calculated difference in digital signal between two methods (DeadLeavesCross & DeadLeavesDirect). The colored lines represent response at different ISOs and in reference to a high-contrast target and a low-contrast target. Values plotted are the Dead Leaves SFR difference against the spatial frequency. The larger the area under the curve, the more artifacts are present. 

Edge contrast / sharpening   

  • Moderate overshoot (for example, 4.9 percent at ISO 200) in the sharpening along high-contrast edges at nearly all ISOs tested, and zero undershoot.
  • The DMC-G81 produced stronger overshoot and moderate undershoot along the high contrast edges (e.g. 12.7 percent overshoot and 5.5 percent undershoot at ISO 200).
  • Along low-contrast edges, the G9 produces a small degree of sharpening (6.3 percent overshoot and 3.4 percent undershoot at ISO 200; 3.7 percent overshoot and 1.5 percent undershoot at ISO 800). At the higher ISOs of 12800 and ISO 25600, both undershoot and overshoot at 0.0 percent.

This graph shows the degree of sharpening in the image by representing an over- and undershoot along contrasted edges. The colored lines represent measurements at different ISOs and in high- and low-contrast situations. The size of the dip before the edge (in both depth and breadth) indicates the degree of undershoot; similarly, the amount overshoot is indicated by the height and breadth of the peak. Thus, larger dips and/or peaks indicate that a sharpening effect is visible.

OECF VN / visual noise      

  • In Viewing Condition 1 (100 percent), noise is not noticeable (0.9) at ISO 200 and barely visible (1.0) at ISO 400. However, at ISO 6400 and above, the visual noise is very obvious (2.1 ranging up to 3.1).
  • In Viewing Condition 2, visual noise will only become obvious at ISO 12800 (1.3) and above.
  • In Viewing Condition 3, visual noise would be well below the threshold of being noticeable in images shot at ISOs between ISO 200 (0.5) and ISO 1600 (0.9). Visual noise would be barely discernable at ISO 3200 (1.0) and 6400 (1.1), thereafter increasing in visibility with higher ISO.
  • Noise is most visible in the darkest tones.

This chart shows the noise behavior at various ISO-sensitivities (colored lines) as a function of the brightness of the target image, which is indicated by the relative darkness of the circle on the outer edge of the diagram (noise in shadowed areas are above, and in highlights below). The larger the area inside a curve, the stronger the noise. The degree to which noise disturbs the appreciation of an image, depends on the image size and the viewing condition. The right-hand side of the chart shows the visibility of the noise in an image that is displayed 100% on a monitor (VN1). The left-hand half shows the visibility of noise in a 40-cm tall print (VN3).

This chart shows the noise behavior at various ISO-sensitivities (colored lines) as a function of the brightness of the target image. The perception of noise is represented by the area that is encircled by the curve. The larger the area, the stronger the noise. How much the noise disturbs the viewing of an image, depends on the image size and the viewing distance. This chart shows the noise visibility for an image that is displayed 100% on a monitor (VN1).

This chart shows the noise behavior at various ISO-sensitivities (colored lines) as a function of the brightness of the target image. The perception of noise is represented by the area that is encircled by the curve. The larger the area, the stronger the noise. How much the noise disturbs the viewing of an image, depends on the image size and the viewing distance. The chart shows the noise visibility for an image that is about postcard size (scaled to a height of 10cm) viewed at a distance of 25cm.

Dynamic Range       

  • The dynamic range exhibited by the G9 is good, with over 10 f-stops at many ISOs tested (10.6 at ISO 200, 10.1 at ISO 400, and 10.3 and 10.2 at ISOs 1600 and 3200, respectively.)
  • The dynamic range is slightly less wide at ISO 800 (9.5 f-stops) ad ISO 6400 and above (ranging from 8.4 to 9.0).

Color Reproduction

  • Color reproduction is good, albeit with seven colors showing strong deviation from the original color. Of these seven, six were bright reds.

Color reproduction is shown here in two ways. The upper figure is a chart comparing a reference color (right-hand half of each color patch) directly with the color reproduced by the camera (left-hand half of the color patch). Below is a table that lists the DeltaE of each color patch. Red cells indicated strong color deviations, light green cells represent colors with noticeable deviations, and a dark green field represents a moderate deviation.

Automatic white balance  

  • The automatic white balance is adequate at most ISO speeds. It is the best at ISO 25600 (1.4) and 1600 (1.5). At ISO 200 and ISO 800, automatic white balance is 1.9.
  • The poorest performance of the automatic white balance was measured at ISO 6400 (2.5)

Video                                                                

  • Resolution in frames grabbed from videos shot at low ISO was 1050 (LP/PH; 97 percent of the theoretical maximum), and nearly the same (1047 LP/PH) in frames shot at high ISO.
  • At low ISO, MTF50 of a high-contrast textured pattern was measured at 826 LP/PH, and 812 at high ISO.
  • The MTF50 was measured at 800 LP/PH at low ISO when recording a low-contrast textured scene, and 774 LP/PH at high ISO.
  • Video frames showed greater sharpening than in stills: 15.8 percent overshoot and 6.9 percent undershoot were measured in frames shot at low ISO. At high ISO, the results were nearly identical to those measured from low ISO: 14.6 percent overshoot and 7.1 percent undershoot.
  • Visual noise in video frames is moderate at 100 percent (Viewing Condition 1: 1.3 at low ISO and 1.2 at high ISO), but not noticeable in the two other viewing conditions, with values ranging from 0.5 to 0.8.
  • Dynamic range in video is very good, 10.3 f-stops at low ISO and 9.0 at high ISO.
  • White balance is acceptable: 1.8 at low ISO and 0.2 at high ISO.

This chart shows the noise behavior at two ISO-sensitivities (ISO100 and ISO1600) as a function of the brightness of the target image. The amount of noise perceived is reflected in the size of the area encircled by the curves. The larger the area, the stronger the noise and its perception. The degree to which the noise disturbs the viewer, depends on the image size and the viewing distance. This chart shows the noise visibility for a video frame that is displayed 100% on a monitor (VN1).

This graph shows the loss of contrast (y-axis) as a function of the spatial frequency in line pairs per picture height (x-axis) for two ISO-sensitivities in video mode (colored lines). The further to the right a curve stretches before descending, the better the resolution at that ISO.
The limiting resolution for each ISO can be found by identifying to the highest spatial frequency which results in a contrast of 0.1, or where the ISO curve crosses the thicker horizontal thicker black line marking 0.1. The vertical pink line is a reference representing half the number of pixels in the sensor height (the Nyquist frequency).

This graph shows the sharpening in the image due to an over- and undershoot along edges. Depending on the size (based on width and height) of the additional emerging area, a lower (shallower additional area) or stronger (higher and narrower additional area) sharpening effect is visible.

Speed                                                                                            

Start-up time           

  • Starts up in 0.8 seconds, faster than the 1.0 seconds needed for the G81.

Continuous shooting         

  • The G9 shoots 60.0 frames per second in JPEG format, up to 50 images in a row, much faster than the 9.4 frames per second of the G81.
  • Burst shooting in RAW format is the same: 60.0 frames per second up to a total of 50.

Autofocus (300lx)   and Autofocus (30lx)         

  • Only measured in live view.

Autofocus (300lx) Live View       

  • Autofocus time in bright light took 0.06 seconds, a little longer than claimed by the manufacturer, for a total shooting time of 0.13 seconds.
  • This is still faster than the 0.19 seconds measured in the G81.

Autofocus (30lx) Live View          

  • Autofocus in low light took longer, 0.12 second, for a total shooting time of 0.2 seconds.
  • Nearly twice as fast of the 0.37 seconds needed for the G81.


In Depth

The Panasonic DC-G9 is a mirrorless camera with a 20.3 megapixel sensor. It produces still images of very acceptable quality, and good quality video, although its real strength is in a variety of in-camera processing features that will appeal to some users.

The resolution in images produced by the G9 is decent, considering that many competitors have a good deal more megapixels in their sensors. At the lowest native ISO, ISO 200, the G9 records 1860 line pairs per picture height (LP/PH; 96 percent of the theoretical maximum). At ISO 400, the resolution is similar (1867 LP/PH, 96 percent of maximum). Subjective visual inspection of the test images captured by the G9, indicate that images captured at ISO 1600 appear nearly as high in resolution as images made at ISO 200. The resolution declines very slightly as ISO increases: at ISO 3200 and ISO 6400, 1686 and 1684 LP/PH are recorded, both 87 percent of the theoretical maximum. At the highest ISO available in this camera, ISO 25600, 1474 LP/PH are recorded, 76 percent of the theoretical maximum.

The resolution in images produced by the G9 is not quite as good as its sister camera, the Panasonic DC-GH5, which also has a 20.3 megapixel sensor. Images made in the laboratory by the GH5 at ISO 200 recorded 1905 LP/PH (98 percent of the theoretical maximum at ISO 200). However, the G9 does show slightly better resolution than one of its predecessors, the Panasonic DMC-G81. The G81 had a 16 megapixel sensor and captured 1787 LP/PH at ISO 200 (104 percent of the theoretical maximum).

Texture reproduction by the G9 is fairly good. At ISO 200, the G9 produces a MTF50 value of 1364 line pairs per picture height (LP/PH) in high-contrast scenes, and 1237 LP/PH in low-contrast parts of the scene. At ISO 800, MTF50 is 1308 LP/PH (high contrast) and 1009 (low contrast). Smoothing is obvious at higher ISOs: at ISO 6400, the MTF50 is only 658 LP/PH (high contrast) and 460 LP/PH in low contrast. Subjective inspection of images captured also shows a loss of the finest details at ISO 1600 compared to ISO 200; at ISO 6400 and higher, smoothing is very obvious. The G9 also produced relatively many artifacts at ISO 6400: 40.3 percent in high-contrast parts of the image scenes and 52.8 percent in low contrast areas.

Texture smoothing in the G9 is improved over its predecessor G81: for example, at ISO800 where the G9 produced an MTF50 of 1308 LP/PH in high-contrast areas, the G81 produced an MTF50 928 LP/PH. In low contrast parts of a scene, the G9 produced 1009 LP/PH, in comparison to 804 LP/PH produced by the G81. Compared to the G9, the G81 produced rather more artifacts at lower ISO when recording both high and low-contrast textured patterns: at ISO 200, the G9 produced 12.0 percent artifacts in high contrast areas and 19.7 percent in scenes with low contrast; while the G81 produced 27 percent (high contrast) and 25.2 percent (low contrast). However, at higher ISO, the G81’s results were similar to those cited above for the G9: 43.1 percent at ISO 6400 (high contrast) and 53.3 percent (low contrast).

The G9 produces very moderate sharpening. A mild overshoot can be measured (for example, 4.9 percent at ISO 200) in the sharpening along high-contrast edges at nearly all ISOs tested, and zero undershoot. In contrast, the DMC-G81 produced stronger overshoot and moderate undershoot along the high contrast edges (e.g. 12.7 percent overshoot and 5.5 percent undershoot at ISO 200). Along low-contrast edges, the G9 produces a small degree of sharpening (6.3 percent overshoot and 3.4 percent undershoot at ISO 200; 3.7 percent overshoot and 1.5 percent undershoot at ISO 800). At the higher ISOs of 12800 and ISO 25600, both undershoot and overshoot are 0.0 percent.

Visual noise is in general not very noticeable. In Viewing Condition 1 (simulating viewing at 100 percent), noise is not noticeable (0.9) at ISO 200 and barely visible (1.0) at ISO 400. Visual noise would become very obvious at ISO 6400 and higher (2.1, ranging up to 3.1). In Viewing Condition 2, a small screen or postcard-sized print, the visual noise would only pass the threshold of being obvious at ISO of 12800 (1.3) and above. In Viewing Condition 3, simulating a large size print, the noise would be well below the threshold of being noticeable, in images shot at ISOs between ISO 200 (0.5) and ISO 1600 (0.9). The noise would be barely discernable at ISO 3200 (1.0) and 6400 (1.1), thereafter increasing in visibility with higher ISO. Noise is most visible in the darkest tones.

The dynamic range exhibited by the G9 is good, with over 10 f-stops at many of the ISOs tested (10.6 at ISO 200, 10.1 at ISO 400, and 10.3 and 10.2 at ISOs 1600 and 3200, respectively.) The dynamic range is slightly less wide at ISO 800 (9.5 f-stops), or at ISO 6400 and higher (ranging from 8.4 to 9.0).

Color reproduction is good, albeit with seven colors showing strong deviation from the original color. Of these seven, six were bright reds, and one was a turquoise swatch. The color fidelity, combined with the broad dynamic range and the high-resolution pixel-shift option, may indicate that this camera is suitable for landscape and possibly architecture photography.

The automatic white balance is adequate at most ISO speeds. It performs best at ISO 25600 (1.4) and ISO 1600 (1.5). At ISO 200 and ISO 800, automatic white balance is 1.9. The poorest performance of the automatic white balance was measured at ISO 6400 (2.5)

The G9 has excellent speed results. It starts up in 0.8 seconds, faster than the 1.0 seconds needed for the G81. The burst shooting is amazing: the G9 shoots 60.0 frames per second in JPEG format, up to 50 images in a row, much faster than the 9.4 frames per second of the G81. (These results indicate that burst shooting would last less than a second in total – hopefully one has chosen the right 5/6 of a second to record!) Burst shooting in RAW format is just as fast: 60.0 frames per second up to a total of 50.

The Panasonic G9 is described by the manufacturer as the ‘world’s fastest autofocus’, with the Panasonic.com website claiming “0.04 sec. (approx.) lock focus.” The Image Engineering speed test protocol produced measurements of autofocus time in bright light a little slower than claimed, 0.06 seconds for a total shooting time of 0.13 seconds. This is faster than the 0.19 seconds measured in the G81.

Autofocus in low light took longer, 0.12 seconds, for a total shooting time of 0.2 seconds, nearly twice as fast as the 0.37 seconds needed for the G81.

Video                                                                

Resolution in frames grabbed from videos shot at low ISO was 1050 (LP/PH; 97 percent of the theoretical maximum). Videos shot at high ISO showed nearly the same resolution (1047 LP/PH).

Texture was fairly well reproduced in video frames: at low ISO, MTF50 of a high-contrast scene was measured at 826 LP/PH, and 800 LP/PH in low-contrast scenes. MTF50 was 812 LP/PH in frames shot of high-contrast scenes at high ISO, and 774 LP/PH in low contrast scenes.

Video frames showed greater sharpening than in stills: 15.8 percent overshoot and 6.9 percent undershoot were measured in frames shot at low ISO. At high ISO, the results were nearly identical to those measured from low ISO: 14.6 percent overshoot and 7.1 percent undershoot.

Visual noise in video frames is moderate at 100 percent (Viewing Condition 1: 1.3 at low ISO and 1.2 at high ISO), but not noticeable in the two other viewing conditions, with values ranging from 0.5 to 0.8.

Dynamic range in video is very good, 10.3 f-stops at low ISO and 9.0 at high ISO. White balance is acceptable: 1.8 at low ISO and 0.2 at high ISO.