About 4G in USA

The current lack of a clear 4G definition has left many consumers in the dark about what they're really buying.

Sprint Nextel (S, Fortune 500) was the first to launch a network called 4G, going live with it in earlier 2010. It is based on a network standard called WiMAX, which is capable of speeds ranging from 3 Mbps to more than 10 Mbps.

Time for Verizon to buy Sprint?

T-Mobile soon followed suit, claiming at the time to be "America's largest 4G network." But T-Mobile's 4G network is actually an expansion of its existing 3G network, which operates on a standard known as HSPA+. Despite not being true "4G," T-Mobile's network can reach speeds of 5 Mbps to 12 Mbps -- faster than Sprint's.

Verizon (VZ, Fortune 500) launched its 4G network late last year, and it now claims to be the nation's largest and the fastest. The company's 4G network operates on the Long Term Evolution (LTE) standard, which will average 5 Mbps to 12 Mbps speeds.

AT&T (T, Fortune 500), which is expected to unveil its 4G-LTE network in the coming months, decided not to be the lone major carrier without a 4G network. So earlier this year, the company also began calling its 3G-HSPA+ network "4G." That means AT&T will soon have two 4G-branded networks, each using different technologies -- and its current 4G offering could, in some instances, be faster than its soon-to-be-released, new 4G network.

http://money.cnn.com/2011/06/22/technology/4g_bill/index.htm

Peer to Peer video call is working

Finally we have reached 480p end-to-end RTP demo on i.MX53 tablet. Here is the video clip (the file is too big, so I put it on Youtube, click the link below) showing the video between two  iMX53 tablets. We’ve added I-frame re-generation so that it is more robust during packet loss. In the meantime, we are working on adding SIP signaling.

http://www.youtube.com/watch?v=ioWNzSEazok

Thanks to the great efforts of Alex Lien, and the great support of Chris and Zubair.

As for 720p, the CPU MHz is still high during simultaneous encode/decode RTP loop.

As for Skype,  the Skypekit itself takes 500MHz at audio-only mode.  If a customer really wants Skype, we need to work closely with Skype to drive this MHz further lower.   

5/5

Marco released one version, I felt it is not bad. Waiting for MSI's feedback.

Machine gun noise

昨天Marco加了vPort出來到audio fringer的buffer size, 從50ms to 750ms, 似乎是有解，但是偶爾還是會發生，不過Roger說這樣的話應該就是Android來不及處理，跟我們無關。

Thanks Olivers suggestion for machine gun noise

1.      Check if using G.729 has higher such noise chance than G.711.

2.      When use G.711, put pockets by using iperf or smartbit to check if machine gun noise is more serious.

3.      What is your WiFi driver implement method, work queue or tasklet?

4.      Would it be possible to use one version that is only can make call, without all other drivers to verify?

In parallel, we will implement a code to check MHz utilization and release to you tomorrow.

4/15/2011

1. MSI在WiFi lab內測試，發現還是有機關槍聲，所以他們認為在傳送過程中掉封包是可以排除
2. Jon認為是Jitter buffer來不及處理，也就是CPU不夠力，不夠力的原因可能是我們的D2C耗掉過多資源，所以才要用ZB已經解決的版本來試試。
 It’s more like our jitter buffer is underrun.  I was guessing that it is due to high CPU utilization of D2C (same problem as Johor and FSL), so that it causes the late arrival of incoming audio packets, and thus got dropped by the jitter buffer.  
3. RY認為還要檢查PLC, Adaptive Jitter Buffer是否開啟。另外還建議要錄PCM
4.