Testing Bandwidth vs. Frame Rate

By IPVM Team, Published Jan 23, 2019, 03:10pm EST

Selecting frame rate has a major impact on surveillance bandwidth and storage consumption. But with smart codecs now common and cameras more efficient than ever, what is the "right" frame rate?

In this report, we tested 10 cameras from 3 manufacturers, testing framerates ranging from 1-30 FPS, with smart codecs on and off, using H.265 and H.264, in multiple varying complexity scenes.

Inside, we give our key findings and two key recommendations for optimizing surveillance framerates.

free image

Key ********

***** *** ******* *** takeaways **** *** *****:

  • **% ******* ******** **** low ***** **** (*-* FPS) ** ******* (**-** FPS):** *******, ****** **** low ***** **** ** 10-15 *** (******* ** ************) ******** ** * ~15% ******** **** ***** smart ******, *** ~**% without, **** **** ********** discussed *****.
  • ****** ******** (~**%) **** average ***** **** ** high:******** ********* **** ************* when ****** **** **-** FPS ** ** ***, about **% **** ***** codecs *******, *** ******** higher (~**%) **** **** disabled.
  • ***** ****** ************* ****** bitrate ** *** ***** rates:***** ****** ******* ******** significantly **** *** ** night ** *** *** levels ******, ************* ~**% in *** ******** ******* scenes *** **-**% ** more ******* **** ******** night *****, ********** **** our **** *****.
  • ***** ****** ****** ********* from ****** ***** *****:** **** *** *** night ******, ** *** resolutions, ***** ****** ******** the ******* ********* **** when ********** ***** ****.
  • *.*** ****** ********* ** H.264:** *** *****, *.*** camera ******** ****** ************* similarly ** *.***, **** smart ****** **** ** and ***, ****** ******** were ******* ***** ** 20-30%.
  • ** ******** ***** ** high ***** *****:**/*** ****** ******** ********* more ******* **** ***** and *** ******* ** high ******** ******, **** than ******** ******* **** 1 ** ** ***, while ***** *********** ********* by ~**-**%.

***** ** ***** ********, we ******* *** *** recommendations:

Recommendation *: *** ***** ******

*** ** ***** *********** benefits ** *** ***** rates ******, ** ********* that ***** ********* ****** frame ***** *** ***** codecs ******** ********. *** only ** **** ******* overall ***** ******** ****** all ***** ***** (~**-**% day/low ****** *** ~**% night/high ******), *** **** reduce *** ****** ** increased ********, ** ****. For *******, ********** ***** rate **** * *** to **-** ******* ** a ~**% ******** ***** smart ******, *** **** 30% *******.

Recommendation *: **-** *** ******* *** **** ************

***** ** ***** *****, 10-15 *** ** *********** the "***** ****" ** the ******** ******* ***** rate *** *********. ****** to **-** *** **** 3-7 *** ********* ******** by ***** **-**% **** using ***** ****** (**-**% without) ******** ** *-* FPS. ** ********, ****** from **-** *** ** 30 *** ** * much **** ******* ********, ~30-35%, ****** ** * much **** ********* ******** for *** ******* ***** additional ****** ******* ** most ************.

30 ***: +**% ******* *** +***% ******

*******, **** **** ***** 30 *** ******** ~**% more ********* **** **-** FPS, **** ***** *** find **** * ********** trade *** *** ***% more ******, ** ***% vs. ** ***. **** may ** **** ****** in ************ **** **** movement, **** ** **** intersections ** **** ******** areas ***** **** ** or ** *** *** miss ****** ********* ** for ***** **** ****** storage ******* *** ****** wish ** ******* *** most ****** ****** ********.

Note: *** * ********

**** **** ******* ** this **** **** **** to ****** ****** ** frame **** ***************. ******** ***************** ************ ************, ****** ** view *** ***** ******** varied ***** ******. ******* of ****, ******* ****** not *** ***** ************ to ******* ** ******* between ******** ******* ** resolutions. ** **** ** perform ************ ******* ** more ******* ** ******* scenarios ** *** ****** months ** ****** **** ********.

Low ****** *******

** ******* *** ***** in * ****** ******* interior *****, **** *** motion, *** ****** ***** rate **** * ** 30 *** ** **** camera.

**** ***** ***** ******, bitrates ********* ** ******* of ~**% **** ******** from * *** ** 10-15 ***, **** ******* for ************. ** ********, increasing **** ** *** to ** *** ********* bitrates ** ~**%, * more *********** ******.

**** **** ***** *** and ***/** *********** ****** fairly ********* ** **** scene, ***** ******* ****** to ******** **** ******** from *-** ***, **** a **** ******* **** at ** ***.

low activity day comparison smart codec

********* ******* ***** ****** enabled **** ******, ~**% when ******** **** * to ** ***. ** 1080p *******, ******* ********* faster ** ****** *** settings, **** ******* ********* when ****** **** * to * ** ** FPS, ******* ** **** smart ****** **** *******. Also *******, ** ****** resolutions, ******** ********* ****** at ***** ********, **** 1 ** * *** the ******* ********** ********.

low activity day comparison no smart codec

Note: **** ***** ****** ***** **** ** ****

**** **** ** **** cameras, ***** ****** ******** to ***** ****/*** **** when ***** * *** frame *****, **** ******** much ****** **** ***** settings. *** *******, ** the********* **-**********-*, ******** **** ******** highest ** * *** with ***** ****** *******.

smart codec bitrate higher at 1 fps

*******, ******* ** * stream ********, ** ******* that ******* *** *** disabled **** *** ****** was *** ** * FPS **** *.***+ ****** on, ***** ** ***** frame ***** ** ****** to ~** *******.

High ********** *********

** *** *** ******** increased ** **** ******* scenes, ** ****** ** a **** ***** (~*.* Lux) **** ** **, with ***** ****** ****** throughout *** *****.

***** ****** **** ******* less ********* ** *****, with ******** ~**% ***** than ***-***** ***** *********, less **** *** ~**% reductions **** ** *** daytime ***** *****.

**** ******* ** **** scene, ***/** ****** ******** increased **** **** ******* than *** ** ***** cameras **** ********** *******. While ***** ***** ***** and *** ******* ********* only ~**% ** ******* when ****** **** * to ** ******** ** ** ***, average ******** ** ***/** cameras **** **** ****** when ******** **** * to ** ***, *** increase ** **** **% when ****** **** ** to ** ***.

high activity night comparison

*** **** *** **** with ***** ****** ********. 1080p *** *** ******* both ********* ********, *** less **** ******, ***** 8MP/4K ****** ******** **** than ******. **** **** 1080p/4MP ********* **** **** significant **** **** ***** codecs *******, ~**-**% **** moving **** *-** *** (vs. ~**%) *** **-**% when ******** **** ** to ** *** (**** vs. ~**%)

high activity night comparison 0 1lx

H.265 ****** ********* ** *.***

****** *** ****** ******, H.265 ****** ********* ** H.264, **** ********** ***** overall ********. ***** *.*** models **** ******, ****** because ***** ****** *** available, *** ****** ******** overall *******.

*** *******, ** *** nighttime **** ********** *****, both *.*** *** *.*** trended ******* ** ******* rates. *.*** ******** ******** about **% ***** ** this *****.

h 265 scales similary to h 264

**** *** **** ** all ****** ******, *******, nighttime, **** *** ******* smart ******.

Cameras ******

*** ********* ******* **** used ** **** ****:

  • **** *****-*** ** **
  • **** *****-***
  • **** *****-***
  • **** *****-*** ** **
  • ********* **-**********-*
  • ********* **-**********-*
  • ********* **-*********-***
  • ****** ***-*****
  • ****** ***-*****
  • ****** ***-*****

 

Comments (9)

Avatar

Sergey Bystrov

NetworkOptix
01/23/19 09:02pm

MPEG(H.264, H.265) compression is based(among other things) on image correlation between frames. More noise => less correlation between frames => less compression.

Big sensors => small pixel size => more noise(especially at night) => less correlation between frames => less compression.

But still surprised 8MP did so much different at night. 
Looks like such cameras frequently should be used with a lower resolution during the night. If let's say 4 pixels downscaled to 1(which I believe camera simply does if the smaller image size is requested, it's denoise spatial filter by itself ). 

I wonder how your 8mp cameras would behave at night if used as 2MP camera. And are the results consistent across cameras in this case?

Avatar

Ethan Ace

IPVM | In reply to Sergey Bystrov | 01/24/19 05:17am

The results were pretty consistent across 8MP cameras. They all differed moderately in bitrate, but the trends were similar, with higher framerates spiking faster.

Avatar

Sergey Bystrov

NetworkOptix
In reply to Ethan Ace | 01/24/19 10:23am

Ethan, I'm sorry if I did not make myself clear...

I'm interested to see how 8MP cameras behave if used as 2 MP cameras and if results are consistent across cameras under this scenario.

Why is this important?

 

I assume that since 8MP cameras exist, somebody needs it. 
Apparently (based on your tests) it's expensive at night form storage point...


The thought is to use 8MP cameras as 2MP at night.

However, it's would be nice to know if it makes sense. 

Avatar

Jon Dillabaugh

Pro Focus LLC
In reply to Sergey Bystrov | 01/27/19 04:20pm

I guess this would depend on whether the sensor crops to 2MP or if it down samples to 2MP. I have seen both happen.

Avatar

Sergey Bystrov

NetworkOptix
In reply to Jon Dillabaugh | 01/28/19 11:08am

Jon,

I believe cropping is done by a minority of manufactures. ( You know, we do dual streaming and field of view typically match.

Undisclosed #1

IPVMU Certified | In reply to Sergey Bystrov | 01/24/19 09:23am

Big sensors => small pixel size => more noise(especially at night) => less correlation between frames => less compression.

Big sensors => large pixel size, no?

Do you mean

High res sensors => small pixel size ?

I wonder how your 8mp cameras would behave at night if used as 2MP camera.

IMHO, the 8MP camera “downscaled” to 2MP could be expected to have half the noise of the unscaled 8MP, but still would have 2x the noise of a native 2MP sensor, assuming a CMOS sensor.

Avatar

Sergey Bystrov

NetworkOptix
In reply to Undisclosed #1 | 01/24/19 10:10am

High res sensors => small pixel size ?

Yes, this is what I meant, thanks. 

Not sure about the numbers. I just do not know physics of sensors enough, maybe you are right. I think it also would depend on which noise reduction filters are are currently used on cameras. For this reason, I interested IPVM to test it. 

Donald Gordon

IPVMU Certified | 01/24/19 12:06am

Great article.  Your findings correlate well with what we have identified during our day and night testing with 1080p cameras and 8MP cameras.  In a site with several hundred cameras, we always see a handful of 8MP cameras that increase their nighttime bandwidth well beyond the other cameras (20mbps to 100mbps).  These cameras typically have much more digital noise caused by a number of factors including dirty lenses, IR illuminator energy reflecting off of adjacent structures, IR illuminator energy from other adjacent cameras illuminating the problem camera, street and building lights shining into the cameras, floating debris in the air (dust and dirt), and insects flying around the cameras.

Vin Imbesi

01/24/19 05:01pm

Great article!  It would be nice to add a section regarding the impacts that these compressions have on server and workstation system requirements.  

 

Minimal improvements in throughput, for example, may not be worth the additional horsepower that would be necessary in these scenarios.

Read this IPVM report for free.

This article is part of IPVM's 6,736 reports, 909 tests and is only available to members. To get a one-time preview of our work, enter your work email to access the full article.

Already a member? Login here | Join now
Loading Related Reports