Ethan, how are you bud!? Looking forward to our next cab ride in Vegas at ISC. That was a wicked cool article. Very creative test. I am getting some Gen 6 Sony cameras tomorrow for testing and I will take a crack at doing this test myself. Keep up the good work.
Advanced Camera Bandwidth Test Results
Published Aug 12, 2013 04:00 AM
[*******: ** *** ***** ************ ** all ***** '**********' *** '************',***** ***. **** **-***** ********* ****** ******* never ******** ******** ************* ********* ********* and *******.]
*****'* **** **** ***** ** **** might ******. ****** *** *** ***** of *** ************ ******* ********** *** bandwidth. *********, *** **** ********** ********* ** *********** ********** **** *** **** **********. **** of *** **** **** **** ********* can **** ********* ** *** ***** (think ********* **. ************).
*** ******* ******* *** **** **** experiments **** *** **** ***** *** different***.
*** ********, ******* *** ****** *****. Which ** ***** ****** ******** **** bandwidth?
*** ***** *****?
** ********* **** **** ***********. *** about *********?
** **** ******* ******* *** ********* impact ** *******:
- ********
- **********
- *****
- *****
- ***
**** ** **** **** ****** ********, but **** ******** ******* ****** ** bandwidth ***********. **** **** ***********, *** results ***** ****** ****** ****** ***** tested ********* ****, ********, *****, ***** and ****.
Key ********
**** *** *** *** ******** **** this ****** ** *****:
- *********** ********* ******* *************, *** **** models ****** * ****** ****, **** with *** ***** ******* ********** (**********, framrerate, *********** ******, ***.), *********** ** 2.5 ** * **/* ******** ****** cameras ****** *** *** *****.
- ********* * *****'* ***** ** **** resulted ** *********** ********* ********** (** much ** +/-**%) **** ****** *** basic ***** ******** *** ****.
- ********** ***: ****** **** **** ******'* ********* to ****** ***** ** **** ** a **** ************ *** **** ****** track ******** ** * **%+ ******** in *********.
- *************: ******** ** ***** ******, ********* **** wide ** ****** ** * **** resulted ** * ******** ** ~**% overall. *******, *** **** *****'* ****** default ********* *** *** *********** ******** in ********* *** **** ****** *****.
- ********** *********: ****** **** ******* ******** ** minimum ********* ** **** ****** ******** in ~**-**% ******** ** *********, ****** details **** ******* ******* ** **** scenes.
- ********** *********: ********** ********* **** ******* ** maximum ******** ** ~**% ******** ** bandwidth ** *******, **** ** ********* improvement ** ***** *******. ****** **** numerous ***** **** **** ******** (*****, blowing ******, *******, ******, ***.) *** the ******** **** ** *********.
- ********* **********: ********* ******** ****** ******* ******* *************, with **** ******** * **** ******* adjustment *****. *** **** ***** *** example, *** ******* ** ******* ******* than **** ************'* ********.
- *** ****** ***** ******* ** ******* were ***** ******** ** *********** ** sharpness.
- ** **** ******, ********* ********* **** WDR *** ****** **, ****** *** difference ****** ************* **** ***** ** scene.
- **** ***** ******* *********: *** **** ****** **** *** more *** ** ***** * ****** was, *** ***** *** ********* *********** was. **** ***** ******** *** ** focus ******* ********* *********** **%+.
- ******** *** ***** ********* ******: ***** ********* ** ******** *** color ******** ******** ** ** ** 50% ********* ** ********* ******** ** default ********. ********** ***** ******** ** maximum ********* ********* ** ***% ** more.
- **********: ****** ******** *** *****, ********** and ********** ********** **** ********* ********* due ** **** ** ******* ** either ******** ****** ** ************.
***************
**** *** *** *** ***************:
- **** **** ******** ****** ** *** intended ***** *** ***. **** ** **** **** ********* **** historically ******* (*** * ******** **** setup) ******* ********** ** *** ** the **** ***** *** ***** *********** differences *** ***** *********** **** ************* across *************.
- ** ******* ***** ******* **************** ** ********* ******* ** ********* spikes.
- ******** ******** *********** * *** ** ****** *********, but ** **** ** **** ** in ***** ** ****** *** ******* tradeoffs.
******* ********** ** ********* *** ** gained ** ******** ***** ********, **** as **********, ********, *** *****. *******, users ****** ** ******* **** ********* these ** ***** *********** ****** ***** quality.
Wide **. ****** ***
** *****, ***** ****** ** **** seem **** ******* **** ****** ****** of ****, ****** *** ** *** greater ****** ** ******* ** *** scene. *******, ***** ******* **** ******** pixels *** ****, ******* **** ****** of ***** *******, ** ***** **** bandwidth *** ******** ** **** *****. Additionally, ******* ** ****** ******* * much ****** **** ** *** ***** when ***** ******** ****** ** ****, further ********** *********.
** ****** ** **** ** *** following ********* **** *** *******. *** comparison ****** *** * ********* ** these ******* ***** **** ********** **** scene.
************
** **** *****, ** *** ** the ******* ****** ** ****** ************ around **** ****. ******** ***** ******* the ***** ***** ** ***** **-** miles *** ****. ** ***** ********* increased ** *** *******, ** ******* of **-**%. *** ***** ***-**** *** an ******* ** **** ****, ********** only **********.
*****
** **** *****, ** *** ** the ******* ******* * **** ****** running *****, ** *** ** *** public *** ******* *** *******. ********* increased ** ***** **% ** **** scene, *** ** *** ********* ******* of ********** ******** **** ** **** objects ** *** *****, **** ** the *******, *****, ***** **** *****, bleachers, ***., *** ** ***** ********** to ******* ***** *** ***********. ** *** widest ***, ***** ******* *** *** appear ** ****** ********, ** *********** and ********* **** ******* ***********.
****/******* ***
*******, ** **** ***** ** *** up *** ******* ******* * ****** park, ** *** ******* ***. ** this ****, ********* ** **** ******* decreased, *** ** ***** **** ** removal ** ***** (***** **** ****** in *** ****) **** *** ********** of *** *****. ***** ******* **** in *** ****** *** *** *** provide *** **** ***** ** ****** as *** ****** ***** ** *** wide ***. **** **** ** **** case *** ***** ******** *********, ****** other *******. **** ** *** ** large **** ** *** ****** ******* sharpness, ***** ***** *** **** ****** and **** **** *********** ** *** grass, ********, *** ****** ******* *** than ***** ******* ** *******.
*********
********* *** ********* ** ********* ******** in ***** ********** ** *********. *** effects ** **** **** **** ********** with ********** ******** ** *** *****, including *****, ******, ********, ******, ******, and ******. ******* ***** *** ********* are ********* *** ** *** ********** of ***** ********, ********* ** ******* higher *********. ** ****** ********* ** multiple ******, ** ******* ******* *********** in ***** *******. ******* ****** ******, depending ** ****** ************, *****, *** lighting ****.
*** *******, ** *** ****** *****, while *** ********** *** ******* ** all *******, *** ********** ** ******* details ** *** ******* *** ******** lines ** *** ***** *** ***** relative ** ********* ** *** ********, Axis *****, *** ***** *******. ****** saw * ***** ********* ** ****** (click ***** *** **** **** *******).
*******, ** ******* ******, **** ** this ******* ****, *********** ** ******* quality **** *********** ** **** *******, though *** ***** ******* ********** ********** (click ***** *** **** **** *******).
*****, **** **** ***** ***** ******* parameters, ***** *** ** ********* *** sharpness ******* *******, ******* ********, ********, maximums, *** ********** ****** **** ************ between ******, **** **** *** **** manufacturer. *** *******, *** **** ***** was ***** **** *** ******** ****** in *** **** ** *** ********. At *** ******* ********* (* ** a ***** ** ***), ** *** sharper **** *** ******** **'* ******* (100 ** * ***** ** ***), as **** ** **** *****, ******* more **** ****** ** *** ***** and ****** **********:
*** ***** *******, ***** ***** ****** ****** of *** *** ******* ** ****** sharp *****, ** ** **** ********** to *********** ******* *******, ** **** must ** ********* **********.
Focus *******
**** **** *********, ***** ******* ** focus *** ************ ****** *********. *** to ****** ********** ** ******* *********, bandwidth ** ******* **** ******* *** even ******** *** ** *****, ******* differences ***** *** *** ** *********** noticeable ******* ***** **********. **** ********** shows **** ********* ** **** ****, using ** **** ****** ** ** example:
WDR ** **. ***
** **** ** *** **** ******, bandwidth ********* **** ******* *** ** compared ** ***. ** *** ************ scene, **** *** *** ** ********** details ********* ** ******** ****** *****, such ** ***** *** ********* *** awning ****** *** ******. ******* *** on **** ********* ******** ** *** asphalt ** *** ******* *** ** the **** *****, ********** ***********.
** *** **** *****, *** ***** increased **** ******* ** *** *****, such ** *** *****, *****, *** stone ******** ** *** **** *****, increasing *********, ****** *** ** ************* as *** ************.
****** *** ***** *** **** *********, the *****'* ********* ********* ** *** track *****. **** ** *** ** the ******'* *** ********, ***** ********* some **** *******, *** *** ****** to ***** ********** *** *** ****** amount ** *** ** *** *****. This *** ** **** ******** ***** of *** *****, **** ** *** light ****** **** ** *** **** field, *** *** ********* ***** ****** out.
Other ***** ***********
** **** ******** ***** ***** **********, such ** ********, *****, *** **********. Like *********, *** ***** ** ********** and ******* ******** ****** ****** ******* cameras, ** ******* ** ********* *** image ******* **** ***********. *******, ** all *****, ********* ***** ******* ********* results ** ** ***** ****** ********* savings. ***** *** **** **** ***** from ********* ***** ******* ********* ********, but *** ******** ******* ** ***** quality *** ********* **** ** *** up **** ****** *** ******** ***** small *****.
********
****** ******** ******** ** **-**% ********* in ********* ** *** *** ******* we ******. *******, ***** ********** ****** the ******* ** *** **** ******* and ******, ********* ** * ********* dull ******* *****. ******** ********* ******** in *** ******* ********* ****** ***** these **********, **** ** ***** ****** benefit.
**********
********* ********** ********* ******** ** ******* bandwidth ********* ******* **-**% ******** ** default ********. *******, **** ***** ********* may **** *********** ******** ****** ** performance, ********** ** *** ***** ********. Unlike ******** *** *****, ********* ** brightness **** ** ********* ** *********, since *** ***** ******* ****** ***, with **** ** ** ********* ***** white.
*****
******** ***** *** *** ******** ****** on ********* ** ***** ******** ******, an ******* ** ***** **% ******** to ********. *******, ********** ***** *** much ******* ******, ********** ********* ** about **%. **** **** ** ******* setting, *** ****** ******* **********, **** no ***** *********** **********, ***** ** the **** *** ******* **** ** oversaturation *** ****.
Comments (27)
Jon Klieman
Aug 13, 2013MI
Matt Ion
Aug 13, 2013"Much like sharpness, small changes in focus may dramatically affect focus." - well I should hope so! ;)
UI
Undisclosed Integrator #1
Aug 13, 2013Ethan, nice report. You could maybe include the use of in-camera privacy zones to reduce bandwidth, eg block out trees or areas of no interest. I think this should be the same for all manufacturers cameras who have privacy zones?
[IPVM Response: See our test Reducing Bandwidth Through Privacy Masks?]
JC
Jason Clement
Aug 13, 2013Great report!
Mike Dotson
Aug 13, 2013Ethan,
Is there any relationship between the Sharpness settings and image compression settings that we can take advantage of?
For example, if we increase the Compression....can we use Sharpness to keep some of the lost detail?
My guess is that it depends on the order of these operations in the individual cameras. ie RAW >>>> Compression >>> Sharpness or RAW >>>> Sharpness >>>Compression.
SH
Stephen Holtzhausen
Aug 13, 2013Nice report, maybe for some other additional tests:
SVC vs H.264 on FullHD+ resolutions.
And perhaps the "smart-stream" feature that Vivotek provides on their Supreme series cameras. What they do on their 5Mpix cameras and soon to be launched 10Mpix cameras is pushing 720p or 1080p streams on a static image, whenever there's motion, only the parts that create motion are displayed and recorded in the 5/10Mpix image quality the "background" remains at the set 720p or 1080p resolution.
RB
Rob Beck
Aug 13, 2013This was a fantastic learning experience for me. Another plus of membership right here.
MM
Massimiliano Moretto
Aug 13, 2013another great piece of an educational path. Thank you Ethan!
Max Shumeyko
Aug 13, 2013Quite interesting tests. Thank you for your work.
I wonder Ethan, do you see any ways, how can we improve algorythms of bandwidth & storage calculators with such test results?
The best option to calculate required bandwith & storage space is to make real field tests with cameras on the end-user facility and measure the real cameras bandwidth.
But it is not always possible to visit the facility. In many cases all the video surveillance designer has is a site map.
SH
Stephen Holtzhausen
Aug 13, 2013@ Max Shumeyko:
Personally I calculate using specific CBR values for specific resolutions.
This way I get to the worse-case scenario without undercutting bandwidth, throughput and/or storage requirements.
Example (Based average results over different brands using H.264 Baseline Profile):
4CIF/D1 @ 25fps = +/- 2.5Mpbs Max
720p @ 25fps = +/- 4Mbps Max
1080p @ 25fps = +/- 8Mbps Max
So in theory if the client requires 32x Full HD cameras recording at 12.5fps, I can calculate using 32 x 4Mbps = (128Mbps Throughput and Bandwidth requirement for recordin)
From here I can continue with the whole storge calculation based on hours per day, motion factor within those hours and of course retention time.
But that's just me, as I've burnt my fingers way back in the day with VBR. Certain low-light scenarios can push a Full HD VBR stream (H.264) right up to 16+Mbps per camera resulting in either too little storage or even in worse case scenarios a NVR that's not capable of the throughput that crashes.
GA
Greg Austin
Aug 13, 2013We have found as a rule of thumb, average CBR balance between quality/quantity to be this:
H.264 & 7fps
1 Mbps per 1 megapixel
SH
Stephen Holtzhausen
Aug 13, 2013@ Greg, that does calculate about right.
Only problem with where I'm from (South Africa) our clients INSIST on 25fps. Doesn't matter how much you try and explain to them the "if's and the what not's"...
So 7fps is difficult to sell here and would help them with storage requirements as majority of our end-users require 180+ days of storage for legislation purposes...
GA
Greg Austin
Aug 13, 2013Stephen,
It's a work in progress! Educating the customer is sometimes easier than getting the A&E's to stop specifying full 30fps requirements...
We use 7 or 8 fps for most cameras & 12 to 15 fps for license plate cameras. This covers the majority of situations.
MN
Morten Nielsen
Aug 13, 2013Stephen Holtzhausen : Vivoteks solution sounds interesting, it's very similar to a suggestion I made a while ago (I know that ideas are like a....., but I am thrilled to see Vivotek making some strides in this area - the others should take note I think).
I'd like to know more about the camera's compression pipeline; using an out of focus lens to reduce bw sounds like a terrible idea, but so does using a sharpening kernel on the raw frame that is fed to the compressor. The sharpening kernel will cause image noise to be amplified, and noise just doesn't compress well (also the reason AGC can be a BW hog). The client can easily run a sharpening kernel over the frame once it has been received - or you can export and do it in Photoshop (this will amplify encoding artifacts too, but if you are not compressing too much, they should be negligible).
Ideally, I'd want image stabilization prior to the encoding process. I must admit I haven't checked the cost of BW on a camera that is shaking in the wind vs. one that is stabilized - but it would also make any motion detection a lot more robust.
I suppose there's not much you can do - if the customer requires a certain FOV and quality, then you just need to adjust the storage requirements accordingly, or do customers generally accept changing the FOV to lower bandwidth? How big a part of the budget is storage generally? I would imagine it was a fairly cheap part of the system.
JH
John Honovich
Aug 13, 2013Morten,
So clearly differences in the compression pipeline are a factor. Good point to raise. This is a little bit of a black box so we can only speculate about what is happening inside. However, we confirmed with one chip manufacturer that this is to be expected.
As for your question: "How big a part of the budget is storage generally? I would imagine it was a fairly cheap part of the system."
Storage is a big part. It's not as big as the storage vendors claim it is, but ~20% of an entire system is not unrealistic for bigger deployments (though it obviously depends on frame rate and storage duration, where preferences vary dramatically - see: frame rate stats, storage duration stats).
The big issue with storage is that estimating is such a shot in the dark (being off by 50% is common) and that it comes as a nasty surprise at the end of the project ("Yeah we can do it but we need 10TB more"), which the integrator and end user need to debate who eats it.
Ethan Ace
Aug 13, 2013Undisclosed, we actually did a test of reducing bandwidth through privacy masks. Thanks for reminding me; I'll add a link to the report.
Mike, good question re: sharpness/compression. The answer is: I don't think so, though I haven't tested it. Sharpness is a big subject, honestly, and small questions like this could go on and on.
Stephen, thanks for the tip on Vivotek. I know of a couple manufacturers that will encode regions in different qualities, but this is the first I remember hearing of someone changing those regions dynamically. We may look into that. As far as SVC, it would be difficult to test right now because so few people support it. We'd definitely like to, but it needs more mainstream support.
JH
John Honovich
Aug 13, 2013To those recommending rules of thumbs, "X Mb/s per Y resolution and Z frame rate". Sure, you can do that but you are likely going to be off regularly unless you consistently use the same models in the same conditions (i.e., you specialize in bank branches and use ABC cam model 123).
To Max's question: "do you see any ways, how can we improve algorithms of bandwidth & storage calculators with such test results?"
I don't see any easy way. Clearly the wide vs narrow FoV test shows that the size of detailed objects in a FoV is critical regardless of the scene type but unfortunately that can come from the patterns of carpeting, roads, etc. Plus, since it frequently varies significantly by camera model, it's hard to imagine a way to model it without carefully checking and tracking low level issues.
Dennis Dyer
Aug 13, 2013John - great work, thank you very much.
JH
John Honovich
Aug 13, 2013AG
Alexander Gutierrez
Aug 13, 2013This report is a great explanation on how much Manufacturers BW and Storage calculators vary on when using their calculators (Axis, Samsung, Vivotek,..) using the "same" parameters. Great work!
CE
Cynthia Ezell
Feb 17, 2014This statement under "Key Findings":
In most scenes, bandwidth decreased when WDR was turned on, though the difference varied substantially from scene to scene.
is contradicted in another statement under "WDR On vs Off":
In most of our test scenes, bandwidth increased when turning WDR on compared to off.
Am I missing something?
Ethan Ace
Feb 17, 2014Hi Cynthia, you aren't missing something. You caught a typo. It should read increases in both cases. I've updated the report.
WA
Wassim Abu-Zent
Feb 22, 2014There is a mistake in "Increasing FoV", under the key findings. It should read moving from narrowest to widest.
Markus Segginger
Oct 10, 2016Hello Ethan. Thank you for this interesting Article.
Concerning the FoV: coming to the conclusion, narrow FoV does not gives you the Guarantee of having more Bandwidth. It depends the content of the Scene. How big are the Details within a Scene and the Speed of Movement.
U
Undisclosed #2
Oct 10, 2016True, but in this case narrowing the FOV while the road is still kept in frame increases bandwidth because of the movement of the cars, which now takes up more of the image.
KB
Keith Beatty
Oct 16, 2016OH Man. That video clip at the beginning was FUNNY.
I don't even know what anybody was saying, or how anybody could get that hype.
CR
Chad Ray
Dec 17, 2016I am from West Virginia so I am always behind, at least this time only 3 years!
Great Article Ethan.
UE
Undisclosed End User #3
May 28, 2018Clear explanation on focus & BW.