Axis 20MP Q1659 Camera Tested

Published Jun 13, 2017 17:27 PM

Axis has joined the super high resolution camera trend with their Q1659, a 20MP model equipped with Canon's APS-C sensor.

We tested the Q1659 against Axis' current super low light and integrated IR models:

We tested:

  • Daytime image quality
  • Low light (<1lx) color image quality
  • Dark (<0.1lx) monochrome image quality
  • Performance with external IR
  • CPU load issues
  • VMS support
  • Multi-streaming support
  • Bandwidth/smart codecs

And more. See our full results inside.

Further Testing Coming Up

We plan to test the Q1659 camera vs. competitive large format models from Avigilon, Sony, etc. in upcoming tests.

Summary

In our tests, the Q1659 offered significant low light performance increases over Axis' previous super low light and integrated IR models, with brighter night time images and better subject details and detection at all ranges tested.

Additionally, while the Q1659's bitrates were much higher than 1080p models (~2.5 Mb/s day / ~10 Mb/s night vs. ~1 Mb/s / 5 Mb/s in 1080p), bandwidth usage was not as extreme as in past high resolution 4K+ models. Further, when broken down to Mb/s per megapixel, the Q1659 consumes much less bandwidth per MP than other cameras. 

Users should beware of CPU load when specifying the Q1659, as viewing the camera caused much larger spikes in load than typical 1080p models (~10-20% on both our test machines, versus 2-4% for 1080p). Viewing only 1-2 of these cameras with several other standard cameras in a layout may overload underspecified client machines unless multi-streaming is configured. However, users should check that their VMS supports multi-streaming with this model, as Avigilon Control Center supported no secondary streams, and Exacq supported only one in our tests (versus multiple in Genetec and Milestone).

Q1659 Pricing

Axis Q1659 pricing varies depending on which lens is used, from ~$4,200 USD for the 24mm/F2.8 model through ~$7,200 for the 70-200mm/F2.8 lens. 

This is a large price premium over current super low light and 4K models. Compared to typical "large" imager (~1/2"+) super low light models like Axis' own Q1635 (~$1,200) or their upcoming 4K P1368-E (~$1,099 MSRP), the Q1659 starts at nearly 4x the cost.

Physical Overview

The Q1659 camera body itself is similar to an oversized version of other Q series models. Aside from the notably different size and lens mount, features of the camera are similar to other Q16 models, with I/O, audio, and microSD storage. However, the Q1659 also includes a SFP (Small Form factor Pluggable) slot, allowing users to directly connect fiber to the camera.

Also note that unlike other full frame cameras, such as the Sony VB770 or Avigilon Pro, the Q1659 includes a mechanical IR cut filter, making it IR sensitive. 

Significant CPU Impact

In our tests, viewing the full resolution 20MP stream resulted in increases in CPU load of 8-10% during the day and ~20% at night when bitrate and digital noise increases. These increases were similar on two machines (specs below) and using both ExacqVision and Genetec Security Desk.

Load increase when adding a typical 1080p camera was drastically lower, typically a fraction of a percent.

Client 1 (Laptop):

  • Windows 10 Pro
  • Intel Core i7-3630QM 2.4 GHz
  • 16GB RAM
  • NVIDIA GeForce GT 650M

Client 2 (Desktop):

  • Windows 10 Pro
  • AMD A10-6700 3.70 GHz
  • 8GB RAM
  • NVIDIA GeForce GT 750 Ti

VMS/Multistreaming Support

We were able to connect the Q1659 to Avigilon, Exacq, Genetec, and Milestone without issue. Note that it is currently listed as supported only by Exacq and Genetec.

Multistreaming support or lack thereof varied by VMS:

  • Avigilon Control Center: No secondary stream, full 20MP stream only
  • ExacqVision: Full stream plus one additional stream
  • Genetec Security Desk: Four H.264 and two MJPEG streams (shown below)
  • Milestone XProtect: Four H.264 and two MJPEG streams

 

 

Test Scene

We tested the Q1659 in an open parking lot/field, ~600' long total, shown below. A 50mm lens was used, giving the camera a ~30° angle of view:

The overview below shows the distances used:

Daytime: Similar Details At Over 3X Range

During the day, at 60' distance, details in all cameras are strong, but the Q1659 shows very fine details due to >3x the PPF of 1080p models.

Doubling distance and FOV width, the Q1659 still shows identifying facial details of the subject, while only rough hairstyle/clothing details are visible in the 1080p models. The 1080p models display only line 1 of the test chart while the Q1659 is legible to line 6/7.

At ~250' distance, rough subject details are still visible, as well as 2-3 lines of the test chart, while no details are visible in the 1080p cameras, only detection.

The same is true at 350', the furthest distance tested. The subject's clothing and hair style are discernable, as are 2 lines of the test chart, in the Q1659.

Low Light Color Mode: Better Details / Detection 

Next, we tested at night with exterior lights on to see how cameras performed in color mode at night. Light levels were ~1.25lx close to the building, dropping to ~0.14 at ~250'.

At close range, the test chart is washed out compared to the 1080p models, with only 1-2 lines visible. However, subject details in the Q1659 were much stronger than other cameras, with recognizable facial details.

At ~120' distance, rough subject details are visible in the Q1659, along with lines 1-3 of the test chart, while both 1080p models provide detection only.

Finally, at ~250' distance (~41PPF), the subject is still easily detectable in the Q1659, but obscured by noise in the other cameras.

Dark Monochrome Mode: Q1659 Better Details

We next tested with exterior lights off, dropping to ~0.16 lux with the Q1659 and Q1635 in monochrome mode, and IR enabled in the P3225 Mk II. Note that the moon was 3/4 full at the time of testing, the only source of light in the subject in the scene. 

At ~60', 3-4 lines of the test chart are legible in the Q1659, while only the first line is legible in others. Some facial details are visible in the Q1659, as well as the subject's clothing. The 1/2" Lightfinder Q1635 shows rough facial details of the subject, while the P3225 essentially provides detection only.

Further from the cameras, the Q1659 provides solid detection of the subject, while he is difficult to detect in other cameras.

Finally, at 250', the subject is difficult to detect in the Q1659, and lost in digital noise in others.

Added IR Illumination: Q1659 Superior Details/Detection

Unlike most large format surveillance cameras, such as the Avigilon Pro series or Sony's VB770, the Q1659 includes a motorized IR cut filter, making it IR sensitive at night. Because of this, we tested performance of the camera paired with a Raytec Vario i8 (see our test) to see how external IR impacts performance.

Using its 35° lens, the Vario has a range of 165m/~541'. The clip below shows the subject illuminated at this distance in the Q1659 (to the left of the large utility pole), but no details are available.

At close range, details of the subject are clear in the Q1659, with the chart legible to line 5. The Q1635 is washed out by strong illumination, and the P3225's integrated IR much weaker than the added Raytec.

Doubling distance, details of the subject are still clear in the Q1659, with 3 lines of the test chart legible, while only detection is possible in other cameras.

Finally, at ~250', the subject is still easily detectable in both the Q1659 and Q1635, though without details.

Framerate Limitations

The Q1659 is limited to 8 FPS at its full 20MP resolution, limited compared to many 4K/12MP models which spec 30 or 20 frames as maximum.

As an example of the effects of this framerate, we set up the camera by a busy road. The black SUV entering from the right, traveling ~60 MPH, moves ~30' between each frame. By contrast, in a 30 FPS camera, the vehicle would move only less than 10' per frame, producing many more captured frames over the same area.

With a subject running across the field of view, he moves about 2.5' each frame. In a 30 FPS camera, he would move only a few inches. 

Sample Clips

Interested readers may download day and night clips of the Q1659 at its full 20 megapixel resolution in this ~123 MB .zip file, showing the subject walking from ~500' to the camera location.

Bandwidth Comparison

The Q1659's bitrates were higher than other cameras tested, with bitrates of over 10 Mb/s at night, but not above average compared to 4K and higher resolution cameras in past tests (often 10 Mb/s+). We plan to cover this in upcoming shootouts against competitive large imager models.

Note that all cameras in this test include Zipstream 2. Cameras were set to Zipstream High with Dynamic GOV set to 1200.

When broken down to bitrate per megapixel, the Q1659 is the lowest camera tested by far, coming in 50-80% lower than the 1080p models.

Test Parameters

  • Axis Q1659: 6.55.1
  • Axis Q1635: 6.50.1.1
  • Axis P3225-LVE MKII: 6.55.2
Comments are shown for subscribers only. Login or Join