Forum

Given: When designing for enterprise-wide VoWiFi support, it is important to ensure that your network meets the application requirements specified by the voice vendor’s documentation. Most voice applications have similar design requirements. 

In the answer options:
< — Less than 
> — Greater than
What performance criteria should be met for a VoWiFi application to function optimally?

 A : -75 dBm RSSI, 20 dB SNR, <100 ms latency, <50 ms jitter, 64 kbps throughput, <5% loss

 B : -67 dBm RSSI, 25 dB SNR, <50 ms latency, <5 ms jitter, 64 kbps throughput, < 1% loss

 C : -65 dBm RSSI, 15 dB SNR, <50 ms latency, <50 ms jitter, 64 kbps throughput, <10% loss

 D : -72 dBm RSSI, 20 dB SNR, <5 ms latency, <100 ms jitter, 52 kbps throughput, <5% loss

 E : -67 dBm RSSI, 25 dB SNR, <5 ms latency, <5 ms jitter, 1 Mbps throughput, <1% loss

==========================================================

The solution says the design must meet < 5ms jitter. I am currently working on Voice Enterprise Wifi Cert for my company. The Voice Ent testcase considers 50ms is acceptable.

I would like to know whether to go with 5ms or 50ms.

Thank You.

Repetitive 50ms breaks will be noticeable to anyone listening - although depending on the background noise, and a variety of other conditions, they might not be objectionable.  Especially if they don't happen very often or very regularly.

5ms on the other hand will be nearly impossible for anyone to detect.   With instrumentation yes - with the human ear no.

At about 28ms most people can detect that something is happening, but they may not be able to describe what it is.   Say, for example, a 28ms chunk missing in a piece of music that they know really well.   They'll just realize that something is "off".   

With a visual cue, say an LED pulse in sync with a break, a trained person can recognize breaks down to about 11 msec.   But that takes weeks, or months, of practice.

If I was aiming for what I'd consider to be "crystal clear", I'd shoot for 25 ms.   

However, given that many of todays standards seem to list 50ms, I doubt you'd find a manufacturer that would guarantee better than that.

Totally agree with you Howard. Given the rf environment and other Wifi APs around, its next to impossible to achieve < 5ms of jitter.

The Voice Ent testplan considers jitter less than 50ms is a good value. It would be better if the practice test answers are according to the wifi testplan. Otherwise we will fall into confusion.

Unless you, or your employer, are a member of the Wi-Fi Alliance (WFA), it is sometimes difficult to find out exactly what a particular test entails.  Some of their standards, (e.g. WMM, and Hotspot) you can purchase, but they still do not tell you exactly what, or how, they are tested.   The actual tests are considered proprietary, and are not supposed to be divulged to non-members.

For example, you can download sample copies from the Univeresity of New Hampshire - Interoperabilty Lab (UNH_IOL) of some the tests they perform.  But unless you are a member of the WFA, you can't get samples of their official WFA Pre-Compliance tests.  

The CTIA and WFA have somewhat merged their voice standards.   I think they both now specify 50 ms.

Agreed.

The tests are not proprietary. The same test applies for all the vendors. Since they charge for Wifi certification, they are not distributing the resources. It makes sense. Each vendor has their own way of implemtaion which improves the performance. The WFA only checks for minimum requirement for interop. If the vendor has much more advanced algorithms, by sharing the resources to the competitors they may compromise it.

Howard is there any way to ask questions/ clarify these type of things with the practice test authors?

I still did not get concrete answer for my last question - whether the contention window can be zero or not.

The WFA "certification tests" ARE Proprietary.   It's the IEEE "standards" that are not.
 
WFA specifications are sometimes for sale to the public, but again that doesn't mean that the WFA certification test specifics are available to non-members. 

There are several classes of membership, but the primary one allowing you to have cert tests made against your hardware, is thousands of dollars a year.   Each test is additional thousands paid to one of the "official" test labs that you hired to perform the testing. 

In addition to simple interoperability testing, the WFA and UNH tests also include such things as   "...if option xxx is enabled and the yyy bit is set to one, then fail the test".   These are sometimes included for fields labeled as "reserved, must be zero" in the standard and sometimes for other reasons. 

Or "if a frame of type X is received after the transmission of a type Y frame, then fail the test."    There can be hundreds of these tests, each with many sub-tests.

The publicly available UNH Interoperability Tests samples, will give you an idea of some of their tests.   Take a look here for samples:
    
 https://www.iol.unh.edu/services/testing/wireless/testsuites/