Class 1 Assignment - Measure And Report Video Bandwidth

This Mornings first test camera set to 5FPS, bit rate was 2048:-

Sony DH110 - Scene is a office Corridor well lit, H264.

1). I frame 1 per second 1280x720- 266,705 B/sec

2). I frame 1 per second 800x480 - 256,148 B/sec

3). I Frame 1 per Second 320x192 - 172,647 B/sec

What I noticed above is the jump in resolution from 2 to 1 wasnt that significant. I guess this proves resolution isnt everything. Clearly as well there has to be a "setup cost" to the stream, the connection itself and the video being sent across that.

1). I frame 1 per second 1280x720- 262,632 B/sec

2). I frame 3 per second 1280x 720 - 270,813 B/sec

3). I Frame 5 per second 1280 x 720 - 271,925 B/sec

This is odd result, I changed the I frame to send 5 I's per second and its not that much different to 3. Maybe I am missing something but the 3 I frames per second again creates a "overhead / setup cost" which after that the extra I frame is not much different?

I wondered if the setup costs was in fact the bitrate on the camera throttling thing back, so I set the camera to use adaptive rate control

Adaptive rate control

On Off

But setting it on doesnt actually seem to make any difference or does it..

1). I frame 1 per second 1280x720- 256,709 B/sec

2). I frame 3 per second 1280x720- 264,949 B/sec

3). I frame 5 per second 1280x720- 265,964 B/sec

Just a small drop in bandwidth use?

What is clear on the images that the 5 I frame per second had the least artifacts in line which the IPVM article perdicted and actually that didnt come at a massive cost to bandwidth, but that of course depends on the size of the system, the network design and the operational requirements from the customer. I have a naggling feeling that my test is being changed by the bitrate setting, compared to what Matt is seeing I wonder if its making my test screwy?

I noticed that refreshing the brower reset the bandwidth use, I think the activeX was stuck on the previous test resolution, slightly odd, but a refresh sorted it. It might be a good idea for others to refresh the page between tests.

I have a question what do people see the impact based on addtional video metadata ( DEPA etc), how much does this add to the mix?

Bandwidth Consumption without video streaming:

Bandwidth Consumption:

URLS: How Apple Watch will boost earnings: Analyst

Twitch

Streaming at medium quality

Streaming at High Quality

GameTrailers - YouTube

GAME TRAILERS STREAMING HD

Note: There were times in all different streaming scenarios that the bandwidth consumption was idle as shown below:

This bandwith was capture before live video.

.This bandwith was capture after live video

Before opening any video, the bandwidth consumption was: (I close every application).

For this assignment I use Fox 5 News: (Fox 5 Washington DC on Livestream).

720p – This was the best video quality, but I experienced some lags and the bandwidth consumption was high comparing with others.

432p

270p – This was the worst video quality, I barely could identify faces in video. But the bandwidth consumption was minimal. I didn’t experience problems loading the video.

I ran some videos on CNET started with no video at 0 Mbps and throughout the video had fluctuation between 2-4 Mbps.

First test – changing image size/resolution

Second test – adjusting I Frame or I Picture setting

Sony IPELA SNC-DF70N – Keep in mind, at least a 4-5 year old IP camera with MPEG4 and JPEG as only two Video Modes

Default?, but current settings

CODEC - MPEG4

Image size/resolution – 640x480

Frame rate – 6 FPS

Bit rate – 384 KBPS

I-Picture interval – 3 Sec

=62,067 B/Sec = .47 Mbps

Changed image size/resolution – 320x240

I-Picture interval – 3 Sec

= 59,324 B/Sec = .45 Mbps

Changed image size/resolution – 160x120

I-Picture interval – 3 Sec

=51,796 B/Sec = .39 Mbps

Image size/resolution – 640x480

I-Picture interval – 1 sec

= 64,726 B/Sec = .49 Mbps

I-Picture interval – 2 Sec

=59,650 B/Sec = .45 Mbps

I-Picture Interval – 5 Sec

=11,968 B/Sec = .09 Mbps

I noticed that if I tried to change the bit rate setting in the camera to 1024 kbps it would save for a bit or until I changed the resolution, then it would automatically change back to 385 kbps. Not sure why.

So with the results I received – there was a substantial decrease in bandwidth when I-Picture setting was changed to 5 from 1 and even from 2, contrary to the article, but with that came along severe negative image quality as the article stated.

This is assuming I did this exercise correctly…… J

I used an Axis M3005-V as the only camera connected to an Exacq HVR (thus the bandwidth is measured in kilobytes per second). I started the camera at 1080p, 30FPS, default quality. I then dropped the resolution down to 720P, VGA, and QVGA. I left all the other settings the same. I used a very simple scene just looking at a white ceiling and wall. I captured the bandwidth usage from the Exacq Client instead of the Task Manager t

Here are the bandwidth results:

1080p - 111.10 kB/s

720p - 24.82 kB/s

VGA - 10.02 kB/s

QVGA - 5.41 kB/s

Bandwidth consumption no streaming (youtub

Bandwidth consumption with youtube streaming

Notice a large spike from streaming youtube (Send/receive for the first: 248/720 vs Send/receive for the second: 360/156.326). would try to an ip camera later.

Hello Mamadou,

When using YouTube to playback a video file, YouTube will begin downloading a portion of the video until there is enough data to display to you. It will then periodically download additional portions of the video until it has fully downloaded.

This would explain the peaks and valleys you are seeing from the YouTube stream.

I suspect you will get more consistent results when you connect an IP camera and retest.

I ran a live feed via web browser on an Axis P3364 looking at an exterior parkinglot for a warehouse in daylight, with 720p resolution and very little movement. I was getting ~240 KB/s (~1920 Kb/s) on task manager. I also had my exacqVision client open to the same scene and was getting nearly the exact same numbers.

As I was watching the feed, a large truck moved thru the FOV and I expected to see a bandwidth spike. Instead, I saw a larger valley and a lower spike... Now I'm confused.

I'm using Windows 7, so the attached is the best I can do. I did a video feed from our local news station. The first screen shot is activity loading the web site. The second shot is activity once the video started running, on that particular news story. The attachment is a Word document, created in landscape. You may have to increase magnification to read the document.

I'm not sure the attachment worked.... But what it shows is the following:

Sending B/sec. b/sec. Receiving B/sec. b/sec.

3,388 27,104 35,386 283,088

1,007 8,056 384,429 3,075,432

I think I did the conversions correctly. I am not connected to any IP video cameras.

Here are 8 different profiles and the corresponding views and BW via Bosch 7000 HD 720 Starlight Camera:

Profile 1 BW and View

Profile 1 settings

Profile 2 BW and View

Profile 2 settings

Profile 3 BW and View

Profile 3 settings

Profile 4 BW and View

Profile 4 settings

Profile 5 BW and View

Baseline, no video:

Viewing four IP cameras:

And the same four cameras but at a lower resolution and fps:

I noticed a big drop initially when I lowered the resolution and fps (~1.6Mbs) and then bandwidth usage went up to about 6.Mbs.

I connected wirelessly to one of our IP cameras looking out of our office and changed the compression, resolution and codec settings. Results as follows:

1. No Connection - 0.14 Mbps

2. VBR/Highest Quality Compression at 1080P - 8.23 Mbps

3. VBR/Lowest Quality Compression at 1080P - 0.39 Mbps

4. VBR/Highest Quality Compression at 704 x 480 - 1.02 Mbps

5. VBR/Highest Quality Compression at 704 x 480 (MJPEG) - 11.51 Mbps

6. VBR/Highest Quality Compression at 1080P (MJPEG) - 23.75 Mbps

Summary:

1. Changing the Codec from H.264 (1) to MJPEG (6) tripled the bandwidth used.

2. Reducing the resolution from 1080P (1) to 704x480 (4) reduced the bandwidth from 8Mbps to 1Mbps.

Reducing compression quality to lowest (3) from highest (2) reduced bandwidth by a factor of 20.

5MP ACTi E54 camera on 1Gbps LAN

none: 0.13% of 1Gbps
480p at 15fps: 0.55%
720p at 15fps: 0.49%
1080p at 15fps: 0.69%
3MP at 15fps: 0.64%
5MP at 5fps: 0.90%

I checked the bandwidth on eight 1080p cameras set to record at 3fps. Six of the cameras were outdoors and 2 were indoors. The bandwidth of all cameras was surprisingly similar during periods of no motion in the daytime. No lights were used indoors during the test:

Cam 1 outdoors - 1742 kb/s

Cam 2 outdoors - 1400 kb/s

Cam 3 outdoors - 1396 kb/s

Cam 4 outdoors - 1542 kb/s

Cam 5 outdoors - 1394 kb/s

Cam 6 outdoors - 1465 kb/s

Cam 7 indoors - 1558 kb/s

Cam 8 indoors - 1381 kb/s

Periodically, all outdoor cameras would increase by about 200 kb/s. This was explained by light breezes causing some movement in the surrounding vegetation.

Occasionally camera 1 would increase to over 5000 kb/s before settling down to normal levels. There is a very leafy bush in the lower foreground of this camera and it is exposed to more wind than any of the other locations monitored by the other cameras. So when there were stronger gusts of wind, the bush thrashed around and caused a bandwidth spike on camera 1. The other outdoor cameras are in much more sheltered positions and/or show less vegetation. They only increased in bandwidth by about 200 kb/s during strong wind gusts.

I streamed some CCTV camera streams from around the Earth. Unfortunately I didn't learn the settings of the camera.

Went to Times square, some activity, it's nightime. I get between 300-800 Kb/s depending how many people walking into view.

Next went to a Florida Beach. Very dark. Steady around 500 Kb/s

Finally went to Syndey where it is the next day. Watching a bridge. I had 600-900 kb/s zoomed out and as I zoomed in on the bridge, active with cars, it went up to over 1,000 at times.

Hi Brian,

I've made a few tests this morning (in Europe) with an Axis Camera M1114 :

OS Ubuntu 14.04

Browser Firefox 36.0.1

Axis Firmware 5.50.3

Change of resolution

Test 1

H. 264

Resolution : 1280x720 (16:9)

Compression : 30

30 fps

Gov 1 (= 30 x I. Frame per second)

252 KB/sec

Test 2

H. 264

Resolution : 800x450 (16:9)

Compression : 30

30 fps

Gov 1 (= 30 x I. Frame per second)

105 KB/sec

Test 3

H. 264

Resolution : 640x360 (16:9)

Compression : 30

30 fps

Gov 1 (= 30 x I. Frame per second)

80 KB/sec

-----------------------------------------------------------------------------------------

Change of frame rate

Test A

H. 264

Resolution : 1280x720 (16:9)

Compression : 30

30 fps

Gov 1 (= 30 x I. Frame per second)

252 KB/sec

Test B

H. 264

Resolution : 1280x720 (16:9)

Compression : 30

15 fps

Gov 1 (= 30 x I. Frame per second)

205 KB/sec

Test C

H. 264

Resolution : 1280x720 (16:9)

Compression : 30

5 fps

Gov 1 (= 30 x I. Frame per second)

67 KB/sec

-----------------------------------------------------------------------------------------

Change of GOV length

Test Alpha

H. 264

Resolution : 1280x720 (16:9)

Compression : 30

30 fps

Gov 1 (= 30 x I. Frame per second)

252 KB/sec

Test Beta

H. 264

Resolution : 1280x720 (16:9)

Compression : 30

30 fps

Gov 15 (= 2 x I. Frame per second)

40 KB/sec

Test Charlie

H. 264

Resolution : 1280x720 (16:9)

Compression : 30

30 fps

Gov 30 (= 1 x I. Frame per second)

32 KB/sec

This tests show the influence of some parameters on the bandwith. It's interesting to notice that GOV length in this case is the best solution that allows to strongly reduce the bandwith without losing quality.

I've connect to to an IP Camera Samsung SNB3002 and few testing.

This is the result as below:

Test 1 - 950Kbps:

MJPEG

704x576

5fp

Test 2 - 800Kbps :

H.264

704x576

25fps

GOV length: 30

Test 3 - 300Kbps:

H.264

704x576

5fps

Test 4 - 2.2Mbps :

H.264

704x576

25fps

GOV Length: 1

I noticed that

1. The bandwidth will drop to about 1/3 when switch from MJPEG to H.264.

2. The bandwidth increase almost triple when change the GOV from 30 to 1. it is not much change if adjust the GOV value to greater than 30.

3. The bandwidth decrease more than half when changing from 25fps to 5fps

Hey Matt,

I can't stream video off the internet so I decided to run tests through our March DVRs and report the differences in bandwidth. Our stores are cookie cutter and therefore the scene complexity is the same. I waited till there where no customers so it would be even accross the board. I was looking to see what different camera models would yield. All cameras are the same resolution and the same model DVR except one.

Axis 3304 March 4000 series DVR

Default settings 4fps - 525Kbps

the picture was terrible so I made some changes to see what that would do

Decreased color by 5 and increased contrast by 5 - 450Kbps. The picture improved some so I'm guessing the reduction in noise/distortion helped. Still not very good pictures

Reduced compression by 10 - 350Kbps, not sure why the bandwidth dropped, but only marginal improvements in quality.

Next was an Axis 3214 March 4000 series DVR

Default settings 4fps - 125Kbps, interestingly the bandwidth with is much lower but the quality is much higher.

Dropped the compression by 10 - 200Kbps, pretty much expected reduced compression increased bandwidth, though the quality improvement wasn't great.

Next up Axis 3384 and March 4000 series DVR

Default settings and 4fps- 150Kbps - best video of all of them with, really outstanding.

Drop the compression by 10 - 162Kbps, not what I was expecting, very little increase in bandwidth.

I then checked the bandwidth on an 8000 series March DVR and Axis 3384 at default settings and 4fps - 130Kbps. Interesting, lower bandwidth and the best picture quality of all of them. Of course both of these come at a premium.

Goes to show the differences in models when all else is equal.

Hi Brian,

Here are the Test result with an Axis Camera P3354:

OS Windows 7 Professional

Browser: IE 11

Axis Firmware 5.40.17

Change of resolution

--------------------------------------------------

Test 1

• H. 264
• Resolution : 1280x960 (4:3)
• Compression : 30
• 25 fps
• Gov 32
• 2.20Mb/sec

------------------------------------------------------------------

Test 2

• H. 264
• Resolution : 480x360 (4:3)
• Compression : 30
• 25 fps
• Gov 32
• 4.44 Mb/Sec.

------------------------------------------------------------------------------

Test 3

• H. 264
• Resolution : 160x120 (4:3)
• Compression : 30
• 25 fps
• Gov 32
• 0.96 Mb/Sec

-----------------------------------------------------------------------------------------

Change of frame rate

Test A

• H. 264
• Resolution : 1280x960 (4:3)
• Compression : 30
• 20 fps
• Gov 32
• 2.0Mb/Sec

-----------------------------------------------------------------------------

Test B

• H. 264
• Resolution : 1280x960 (4:3)
• Compression : 30
• 15 fps
• Gov 32
• 1.30Mb/Sec.

--------------------------------------------------------------------------

Test C

• H. 264
• Resolution : 1280x960 (4:3)
• Compression : 30
• 5 fps
• Gov 32
• 0.60Mb/Sec.

-----------------------------------------------------------------------------------------

Change of GOV length

Test 1

• H. 264
• Resolution : 1280x960 (4:3)
• Compression : 30
• 25 fps
• Gov 25
• 2.3 Mb/Sec.

---------------------------------------------------------------------------------

Test 2

• H. 264
• Resolution : 1280x960 (4:3)
• Compression : 30
• 25 fps
• Gov 15
• 3.35Mb/Sec.

-----------------------------------------------------------------------------

Test 3

• H. 264
• Resolution : 1280x960 (4:3)
• Compression : 30
• 25 fps
• Gov 5
• 7.77Mb/Sec.

Best regards

Markus

I am testing Hikvision DS-2DF5286-AEL.Camera is mounted on the intersection and doing constant pattern.

1080p resolution at 15 fps

While on pattern bandwidth is 10Mbps to 16Mbps depending on the camera's position.

When I park the camera looking at a parking lot with minimal movement bandwidth is 3Mbps to 6Mbps

Now 720p

Pattern 1.8Mbps to 8Mbps

Parked 1Mbps to 3Mbps

Mpeg4

720p

Pattern 2.3Mbps to 9Mbps

Parked 1.2Mbps to 3Mbps

1080p

Pattern 10Mbps to 24 Mbps

Parked 3Mbps to 7Mbps

Conclusion HD PTZ cameras require a lot more bandwidth than stationary cameras. H264 still better compression but not as noticable when camera is moving.

Also found camera control much smoother (less lag) on Mpeg4

Markus,

Those are some incredibly low bandwidths, what are the cameras looking at? My bandwidth for a 3384 which is almost the same as the 3354, just more bells and whistles, isn't even close.

Hi Markus,

The results you see for Test 2 seems intersting to me.

Test 2/H.264/Resolution : 480x360 (4:3)/Compression : 30/25 fps/Gov 32/4.44 Mbps.

4.44 Mbps is very high for such a small resolution. Can you confirm that this was definitely using H.264?

In Test 2 you dropped the resolution substantially, but more than doubled the bandwidth.

Can you double check your figures?

Or let me know how you got to those end results?

Thanks again Markus!

Markus,

Oops, I glossed over you're using Mbps and I was using Kbps.

Using a Pelco Camera

1. Frame rate: 6 fps

Resolution: 1280 x 720

Format: H.264

Bytes Per Interval = 245,033

2. Frame rate: 30 fps

Resolution: 640 x 350

Format: H.264

Bytes Per Interval = 166,914

3. Frame rate: 15 fps

Resolution: 800 x 605

Format: H.264

Bytes Per Interval = 358,255

The one with the highest resolution but lowest fps is in the middle as far as bandwidth use.

Even though 1 is a higher resolution than 3, the fps is less than half.

Honeywell HD-45 IP Camera

Resolution 850X 450

Network Utilization: .03%

Increased Resolution to 1280 x 720

Network utilzation: .06% - Just about double

Same resolution but added motion to the shot

Network Utilization: 0.11%

Definitely see the link between resolution and bandwidth and scene complexity and resolution

I did my test with a Avigilon 3 MP WDR Dome Camera viewing a seating area in the lobby without any activity. Changing the compression (Image Quality) I got the following results:

Trying to change any of the other settings caused the camera to reboot and disconnect from the system so not convenient for testing. Also, I cannot adjust the image rate (fps). It won't go any higher than 20 fps on this particular camera and when I try to select a lower setting it automatically jumps back to 20. It appears the fps are linked to ...the bitrate? The bitrate also cannot be set any higher than what it is defaulted to.

Hello,

I did the test using a grandeye camera through a laptop.

first test was 640x480 fisheye used 4000 to 6000 bytes.

second was 1024x1024 fisheye and used 8000 to 10000 bytes.

I connected to C/Net for this example.

The first connection was at 720P the video quality was good being HD, I did experience spikes but I would expect this, the bandwidth consumption was generally high.

This demonstrates the importance of compression.

The Second connection was streamed at 432p, the video quality was watchable and the bandwidth was average, again there were peaks but nothing significant.

The third and final connection was at 270p, this was by far the worst video quality in my option it is unwatchable, the bandwidth is very low but the video quality is not worth the trade-off.

It is clear why when using HD CCTV compression is a must.

Hi Matt,

I went to the NASA real-time video stream of Earth from the ISS. There's a few cameras on board that are cycled though and in constant movement. The site is WATCH: NASA unveils real-time video stream of Earth from ISS | Globalnews.ca (for those interested in checking it out.) You can easily pick between image resolutions; and below are my findings related to bandwidth consumption. (I went to CNN and C|Net, and could find how to change the resolution...?) Anyhow, using the task manager, I saw ...

@720p = 10 Mbps up to 17 Mbps

@480p = 4-8 Mbps

@240p = 1 - 3.7 Mbps

Rafael M.

I used an AXIS M1054 for the tests. 

In deed, when increasing resolution, the B/W increased. 

Also, I left resolution unchanged, and changed other parameters, with these results: 

1) Changing FPS: In fact, increasing FPS increases B/W

2) Increasing Compression from 30 to 100: The B/W decreased, but the change was not significant. I would say it's because the camera had a quiet scene. 

3) Changing I Frame interval: Increasing the interval decreases the B/W. Again, it was a quiet scene, so the change was not very significant. 

In general, the results were as expected. 

I did the test using a Brickcom VD-202Ap camera.

I tested the frame rates and quality settings.

One of biggest things that I noticed was the increase in bandwith when changing the quality from a medium setting to high.

1. Camera set at 1920X1080 frame rate 30, quality 3, bandwith 576kbps

2. Camera set at 1920X1080, frame rate 7, quality 6, bandwith 4.4mbps

3. Camera set at 1920X1080, frame rate 7, quality 3 bandwith 552kbps

4. Camera set at 192X1080, frame rate 30, quality 6 bandwith 15.9mbps

Thanks to everyone responding so far. Just a quick note on image size: Your uploads look best if they are no bigger than 600 pixels horizontal. Huge images are just a pain to scroll past.

Make sure your uploads are sized with courtesy to other readers in mind!

Thanks everyone. Great findings so far!

I tested a Digital Watchdog Megapix 2.1MP camera (DWC-MPTZ5X)

The scene was a standard lighting living room with an outside window in the background. Very little movement.

Setup and Results:

480X272

H264

7fps

.19Mbps

------------

1920X1080

H264

15fps

WDR ON: 2Mbps

WDR Off: no change

-----------

1920X1080

H264

7fps

WDR ON: .7Mbps

WDR Of: .92Mbps

-----------

320X176

H264

20fps

WDR ON: 2.9Mbps

WDR Off: 2.6Mbps

-----------

320X176

H264

10fps

WDR ON: 1.35Mbps

WDR Off: No change

*It was interesting that the camera is working harder to process the scene with WDR on yet it had very little impact on the bandwidth. Lower resolution with higher frame rates had the most impact on bandwidth.

*I was using DW Spectrum VMS to measure/show the bandwidth at the different frame rates/resolutions.

Jason

Using an Avigilon 2MP mico dome camera I selected a scene in a control room with one person at a computer. Used the same Image rate (9.0) and Image Quality (3) but changed the resolution from 768 x 432, 1280 x 720, and 1920 x 1080. The bandwidth had a big jump from changing to the highest resolution setting from 1.0 Mbps to 9.6 Mbps. Below are the screen shots.

Hello Matt,

I'm monitoring my network/bandwidth via Nettop (cmd line tool for OS X).

OS - OS X 10.9.5 / Browser - Chrome Version 39.0.2171.95 (64-bit)

Source: An open IP Camera - Princeton University, NJ
(I did not have access to the camera for this test)

hoyt-gen-cam.princeton.edu
http://128.112.232.19/appletvid.html

Test 1:

IMG width=640 height=480 aspect ratio=4:3

bytes in bytes out rtt avg
306 MB 8515 KB 82.97 ms

IMG width=2560 height=1920 aspect ratio=4:3

bytes in bytes out rtt avg
323 MB 9001 KB 151.25 ms

The largest deltas were in the Round Trip Time (RTT) averages compared to the "bytes in". The RTT averages got me thinking of latency, whether its my specific machine, wireless connection, network, environmental conditions, etc. So, I just listed some variables that I thought also may be factors for a system deployment:

-The data transfer rates of the source's Internet / network connection(s)

-The nature of the transmission medium (copper, fiber, wireless)

-The physical distance between the source and the destination

-The number of nodes between the source and the destination

-The amount of traffic on the LAN / VLAN to which the cameras/edge devices are connected

-The number of other requests being handled by intermediate nodes, remote and virtual servers

-The presence of interference in the circuit

Me and my co

workers used a March CamPx Microdome camera to make the following measurements.

encoder:H.264

resolution:Full D1

Ips :25

Bitrate:variable

Quality:medium

Results:minimal motion/steady light=at camera 1.3-1.5Mbps/task manager 2.2-2.4Mbps

no motion/low light=at camera 2.5-2.7/task manager 2.3-2.7Mbps

lots of motion/low light=at camera 7.6-8.2 Mbps/task manager 8.1-10.3 Mbps

lots motion/steady light=at camera 8.2-8.4Mbps/task manager 8.0-9.3Mbps

Then left all settings the same except we changed compression to maximum setting and the results are as follows:

minimal motion/steady light=at camera76-79Kbps/task manager104-120Kbps

no motion/low light=at camera 96-100 Kbps/task manager 120-136Kbps

lots of motion /low light=at camera 127-383Kbps/task manager 416-456Kbps

lots of motion/steady light= at camera 118-350/task manager 392-424 Kbps

We noticed a huge difference in bandwidth from medium to maximum compression, also the difference at the camera verses

the difference at task manager and also big difference between low light and steady lighting. One thing that seems to be constant,

is scenes with lots of motion in good light or low light the bandwidth was very high,obviously scene complexity will always play a

role !

,

I completed the assignment at home and did not have access to an IP camera, so I used my local new channels streaming video for my test. I changed resolutions based on the options in the player (270p, 480p and 720p) and measured the bandwidth consumption for each option using a free tool called NetWorx.

My results are as follows:

• No video = 0 Kbps

• 270p = 232 Kbps

• 480p = 1.68 Mbps

• 720p = 2.54Mbps

The 270p video was noticeably degraded and at times pixelated. The picture quality significantly improved when moving to 480p, however bandwidth consumption increased 7x. When moving to 720p, bandwidth only increased 1.5x when compared to 480p and the video quality was marginally improved.

I tested my bandwidth consumption by streaming live video and adjusting the resolution. My results were pretty consistant and predictible. They are as follows:

NO VIDEO

Sent: 0Kb/s

Recieved: 0Kb/s

VIDEO @ 270P

Sent: 48Kb/s

Recieved: 1.6Mb/s

VIDEO @ 480P

Sent: 120Kb/s

Recieved: 5.5Mb/s

VIDEO @ 720P

Sent: 1Mb/s

Recieved: 12 Mb/s

I noticed that my bandwith consumption was pretty regular and increased pretty regularly as well as I jumped between resolutions. Watching the consumption, I noticed larger fluctuations with lower resolution whereas when the resolution was higher, the bandwith was more consistently high. Video quality at 270P was very bad and 480P was much better and 720P even better yet. I did notice the most noticable quality change when changing from 270 to 480.

Hi

I am doing the assignment over Remote Desktop to my office machine, and using Windows 7 - does not make it easy to run OR to measure, but here goes:

I followed this with a number of similar scenarios, using two different cameras attached to my VMS, and changing resolution, compression and frame rate.

I did several more comparisons refining the results. I still need to work on the 'Received Data' in the Windows 7 Task Manager graphs, so that I can isolate bandwidth of the camera from the other data traffic going on in the application.

Rob

In answer to the question about Analytics in the class - 'Does running Analytics affect the bandwidth?' I have a response.

I tested an ioimage camera, running the same kind of monitoring that I used for my Bandwidth examples - Task Manager in Windows 7, but running over a Remote Desktop, which makes it a bit tricky!

I could NOT see any significant consistent difference in bandwidth with the Analytics on or off.

I checked with ourt R&D team who write the software. They told me that the Analytics information in the picture (i.e. the 'boxes' and 'trails'), are sent as an additional TCP stream which is part of the overall communications with the camera. these are combined when the camera streams are rendered on the display. While there IS a separate stream, and therefore there IS added data, the amount of data for the additional pictures is very small and not significant in the overall result - i.e. you can size for an Analytics camera without concern for whether the Analytics is switched on or not.

(The 'wavyness' of the graph when Analytics is ON seems to coincide with the movement of the cars, not with the addition of the Analytics metadata)

Using a Interlogix TVD-M3245E-2 3MPX Camera H.264 WDR

Bitrate Type – Variable

Max Bitrate 2048

Video Quality – High, I Frame 15, 15FPS

Resolution

All 2048 – 1920 & 720 are 2Mbps

Video Quality – Medium, I Frame 15, 30FPS except 2048 is still maxed out at 15FPS

Resolution

All again 2048 – 1920 & 720 are 2Mbps

Video Quality – High, I Frame 15, 15FPS, 30FPS except 2048 is still maxed out at 15FPS

Resolution

All 2048 – 1920 & 720 are 2Mbps

Why after all changes are using 2Mbps?

I have measured bandwidth using an Avigilon 5.0-H3-BO2-IR at 1FPS, 7FPS and 13FPS. The scene was exterior of a building and parking area with no vehicles present, for the most part the scene was quite while doing the test so a good representation of how FPS affects bandwidth.

1FPS gave a network utilization of approximately 2.4%

7FPS gave a network utilization of approximately 24%

13FPS gave a network utilization of approximately 40%

Dramatic difference in bandwidth and network utilization when FPS was increased any amount.

Logged into Vivotek IP8332 Camera with VMS and changed resolution as follows with Bandwidth readings at different resolutions.

At 640 x 360 Resolution, the bitrate was 470-560 Kbps

At 1280 x 720 Resolution, the bitrate was 1300-1500 Kbps

At 1920 x 1080 Resolution, the bitrate was 2600-2900 Kbps.

All readings were at 5fps on outdoor scene with daylight.

With all else being equal, the Bandwidth has a direct relationship to Resolution.

The above screen shot is an Axis 3004 set to 240p 30fps

The above screen shot is an Axis 3004 set to 480p 30fps

The above screen shot is an Axis 3004 set to 720p 30fps

The above screen shot is an Axis 3004 set to 1MP 30fps

My test was done using an Axis m3004-v 30fps MJPEG 30% compresstion

My results showed as follows

24Kbps up and 1.5 Mbps down - 240p

56Kbps up and 5.2 Mbps down - 480p

104Kbps up and 16.1 Mbps down - 720p

136Kbps up and 22.4 Mbps down - 1MP

small jump from 240p to 480p, a huge jump from 480p to 720p, and a small jump from 720p to 1MP. I understand more pixels/ better resolution more bandwidth useage down but why is my upload going up too?

Using a Topline BC-2188 IP camera @ 5 fps, H264 Base Profile, with a baseline bandwidth of 0 bits/second.

This is the result from a SANYO vdc-hd3300 5mp camera.

I forgot to take a screen shot with no camera attached for referance.

Jim

Hi all,

I have tested using Windows 7, Networks and a Axis P1354 (5.60.1).

TEST 1: MJPEG

We can see that dividing the framerate per 2 aprox. divides the bandwith per 2.

Same for the resolution.

TEST 2: H264 with GOV = 1

As GOV = 1, bandwith is very close to MJPEG. It doesn't makes sense to make H264 with GOV = 1.

Same comments than test 1 for linear relationship between framerate or resolution and bandwith.

TEST 3: H264 with GOV = 25

Here, with GOV = 25 (1 I frame each 25 frames), it makes sense to do H264. We can see cleary that the bandwith is here aprox. 20 times lower.

Same comments than test 1 for linear relationship between framerate or resolution and bandwith.

GOV is not the only one setting important for H264, compression is also one. Let's look at it in test 4.

TEST 4: H264 with different compression levels

Here we can see the same 720p 25fps GOV25 H264 stream with different compression levels.

As it's a Axis camera, the compression can be set from 0 to 100. For other suppliers, the scale can be different, for example for Dahua it's from 1 to 6. What is common for all suppliers is called quantisation level. The lower it is, the less compressed the video is (so the quality is better).

What is interesting here is that there's 10x from medium quality to high quality, and only 3x from low to medium.

TEST 5: Light conditions can have an impact on bandwith

The WDR test is not significant because the scene was not complex and light conditions were easy.

What is interesting is that in very low lightn with no aditional IR leds, the bandwith increases by 5x.

This means it needs to take care when dimensioning network speed and storage.

TEST 6: Image settings can have an impact on bandwith

Here we can see that depending on the contrast level settings the bandwith can vary a lot. For example, there's 3x more bandwith from contrast 50 compared to contrast 90.

I find it !! now this is my bandwidth with a live video.

Okay I’ve done some more configuration and came up with interesting results for me.

Interlogix TVD-M3245E-2 3MPX Camera

Bitrate / Variable, Video Quality / Highest, Frame Rate / 15FPS, Max Bitrate / 2048, Encoding / H.264

I Frame = 1;

2048* 1536P = 25Mbps, 1920* 1080P = 29Mbps, 1280* 720P = 6Mbps

I Frame = 5;

2048* 1536P = 8Mbps, 1920* 1080P = 4Mbps, 1280* 720P = 2Mbps

I Frame = 10;

2048* 1536P = 3-5Mbps, 1920* 1080P = 2-3Mbps, 1280* 720P = 1-3Mbps

I Frame = 15;

2048* 1536P = 3Mbps, 1920* 1080P = 2Mbps, 1280* 720P = 2Mbps

Hello everyone,

I did my testing using a Bosch Dinion 1080p, factory defaulted then pulled into a Geutebruck VMS. I checked my bandwidth at three resolutions (704x480, 1280x720, and 1920x1080) and two frame rates (10 and 30). Without adjustment I found my usage to be a fair bit higher than most figures posted on here. My scene was fairly complex viewing 70% outdoors with moving bushes and trees, I tried to avoid taking measurements when there was large amounts of activity, such as the wind blowing or people passing by the window. I was surprised to see how how much the bandwidth fluxuated (100kb-300kb) with activity or changes in the scene not readily visible to me.

As demonstrated in class, the Megabits per second increased with the increase in images per second.

A much larger increase was seen when we increased resolution, but not so much with the framerate increase in this case. I attribute this to my taking the measurement at a lower point in the fluxuations I was seeing.

As before, another increase with resolution, although not quote as drastic this time. This could be due to my own error or changes in the scene I didn't notice.

If time allows I may redo this experiment in more tightly controlled conditions to reduce or eliminate the range of numbers I was seeing.

Kind regards,

Chris Wahl

Hey guys,

Used the Axis P3343 set at 4 frames per second. MJPeg

800 X 600

1124 kbps

640 x 480

651 kbps

489x360

528 kbps

Noted the obvious lower network usage changes, unable to do a print screen very well with our video wall set up but did get a screen shot of the camera location. No activity on this camera throughout the test period

Hello,
thanks to the class. The performed test are well understood and the band
width-intensive parameters are clear.
Thanks for the input of many different cameras.

I used MileStone XProtect, and a Brickcom model CB-500Ap-01 megapixel camera, PoE, H.264. My scene was just around my desk area in the office, well lit, although moving the camera around didn't seem to have much effect on bandwidth. I was able to see the bandwidth usage in XProtect, so task manager was unnecessary for this exercise. Here are the results (hopefully the columns will line up so it's readable)

FPS Res Bit rate Bandwidth

• 30 1920X1080 CBR 2.5Mb/s
• 30 1920X1080 VBR 1.4Mb/s
• 30 1280X720 CBR 4.0Mb/s
• 30 1280X720 VBR 0.6Mb/s
• 10 1920X1080 CBR 6.7Mb/s
• 10 1920X1080 VBR 1.0Mb/s
• 10 1280X720 CBR 6.0Mb/s
• 10 1280X720 VBR 0.2Mb/s

I noticed that lowering the resolution doesn't necessarily reduce bandwidth, but going from constant bit rate to variable bit rate had a profound effect on lowering the bandwidth. In any event, this is a great technique for evaluating real-life bandwidth usage.

Third Test However Camera Position had been moved towards widow 10’ away with no movement in window, a quite view as opposed to another quite view facing an empty wall 3’ away

Not sure why those readings were so high because these are not

Okay I’ve done some more configuration and came up with these results.

Interlogix TVD-M3245E-2 3MPX Camera

Bitrate / Variable, Video Quality / Highest, Frame Rate (measured as Its second between) / 15FPS, Max Bitrate / 2048, Encoding / H.264

I Frame = 1;

2048* 1536P = 16Mbps, 1920* 1080P = 9Mbps, 1280* 720P = 3Mbps

I Frame = 5;

2048* 1536P = 4Mbps, 1920* 1080P = 2Mbps, 1280* 720P = 2Mbps

I Frame = 10;

2048* 1536P = 2Mbps, 1920* 1080P = 2Mbps, 1280* 720P = 2Mbps

I Frame = 15;

2048* 1536P = 2Mbps, 1920* 1080P = 2Mbps, 1280* 720P = 2Mbps

Bitrate / Variable, Video Quality / Lowest, Frame Rate / 15FPS, Max Bitrate / 2048, Encoding / H.264

I Frame = 1;

2048* 1536P = 12Mbps, 1920* 1080P = 6Mbps, 1280* 720P = 2Mbps

I Frame = 5;

2048* 1536P = 2Mbps, 1920* 1080P = 1Mbps, 1280* 720P = 850Kbps

I Frame = 10;

2048* 1536P = 2Mbps, 1920* 1080P = 2Mbps, 1280* 720P = 504-618Kbps

I Frame = 15;

2048* 1536P = 2Mbps, 1920* 1080P = 1-2Mbps, 1280* 720P = 450-502Kbps

It looks like you can get high quality video at 5 I Frame and not eat Bandwidth as you would 1 and there is not enough differance between 10 and 15 I Frame setting to use 15.

However it lookks like at 10 I F. you would not use much bandwidth for a larger system what would be the trade off using 10 I F. instead of a lower I Frame Setting.

I used a Pelco Spectra IV PTZ camera looking at a blank office wall. Lighting was daylight in first runs and darkened room for second. Bandwidth use was measured using Windows Task Manager looking at percentages of a possible 100 Mbps data flow.

30 fps/MPEG 4 Day: 2.61% Dark: 4.25%

15 fps/MPEG 4 Day 2.26% Dark: 4.01%

30 fps/H264 Day: 2.58% Dark: 3.98%

15 fps/H264 Day: 2.1% Dark: 3.3%

Tim Lawson

An Axis Q1755 was used.

Before opening

480x270:

720p

1080p

I used a Axis P1346. Variations in bandwith traffic were (using MJPEG):

1) 1024 x 768

2) 1280 x 960

3) 1600 x 1200

Using 1280 x 960

1) Compression 10/100

2) Compression 20/100

3) Compression 30/100

Using 1280 x 960

1) 30 fps

2) 20 fps

3) 10 fps