About 10 meters from my home router(half is walled)
X-Y, Y-Z, X-Z direction for 10 times test, average value:

YAGEO:CAN4311153002451K(4dbi), 5.3 x 2.0 -> -70dBm
Supermax: IAN0201(2dbi), 5.2 x 2.0 -> -72dBm
Sunlord: SLDA72-2R470G-S1(2.5dbi) 5.0 x 2.0 -> -68dBm
Rainsun: AN3216-245 3.2×1.6(0.5dbi) -> -72dBm
????: PCB IPX antenna -> -80dBm

So, after that test, remove the IPX for more clearance zone is a better choice.

IT IS SO HARD!!!

I have made 8 VoCore v0.5 but only 2 works, Rest are all NG at BGA part…

The problem is my paste is Sn-Bi-Ag(176℃), but BGA balls are Sn-Ag-Cu(226℃),  I make the hot air to 320℃ the balls are melted down but can not touch and connect to PCB pads.

After serval test, I find a way. Paste under BGA is necessary, put paste on PCB BGA part first, then put RT5350 on that, MUST perfectly on its top. Then use hot air 220℃ until paste is melted about 30 seconds. Rise to 280℃ about 30 seconds, balls and paste will combine to one. After cool down, everything is just fine.

Changed from official U-Boot 1.1.3 for RT5350, remove useless information from the official version. Changed wait time from 5 second to 1 second, now VoCore boots a little faster and the console is more clear. 🙂

I should get my new v0.5 PCB board today, but because of the holiday of May.1, I have to wait three days more. I am patient 🙂

The software part should be stable, has run almost 10 days without a mistake. Unfortunately, the road of Taiyuan is under repairing…my power is cut off this morning, so I can not capture new picture for uptime or continue my stable test.

For the hardware part, I will do three tests.
1. replace 1.2nF to 0.1uF.
90% routers and even raspberrypi are using 0.1uF for the ethernet circuit part, but all MTK/Ralink routers are using 1.2nF or 1nF. Why? That is for ethernet signal, so I think it do not have relation with RT5350 chip. I will test and compare, then the truth will soon be revealed to the world 🙂

RT5350

AR9331

RaspberryPi

2. on board ceramic antenna.
I have tried my best to leave 5.5mm clearance zone for that ceramic antenna, if its performance is too low(<60% 4dBi IPX antenna), I will remove it. Maybe more clearance zone will be better but need to test.

from CAN4311153002451K manual:

3. 22nF to 0.1uF, 4.7uF to 1uF, 1.5pF to 1.2pF
on my board, C8,B5,A5,B3,C9 connected to 22nF, that should be used to make wifi part 1.2V stable, so change it to 0.1uF should be better, but might effect signal, need to test.
on my board I have seven 4.7uF, I do not think that is necessary. This board size is very small, and already have two super big 22uF to stable its power on 3.3V. Replace 4.7uF to 1uF should be acceptable.
on antenna part, 1.5pF(+/-0.25pF) is almost same as 1.2pF(+/-0.25pF), so why use one more capacitance type?
AR2317

MT7620A

VoCore v0.3 is not that easy to test, so I make a dock for v0.5.
Its size is 2.5×2.5×1.6. Once connect it to VoCore v0.5, the total size of the two will be 2.5×2.5×2.0.

This dock contains three parts.

A: USB1(left bottom) is used for power +5.0V, it should be connected to USB power.
B: USB2(top right) is a full function USB.
C: RJ45(right bottom) is for WLAN/LAN connection, connected to EPHY_P0.

It is not the final version of the dock.

From Apr.23, I kept one of my board running. 🙂

I am a newbie on antenna, after study a lot from internet, get some knowledge about it.

PCB board is 1mm, composed by 9 layers:

1. Top Solder Mask      ==> 0.5mil
2. Top Copper           ==> 1.35mil(1oz)
3. PP 7628              ==> 7mil 
4. Inner Copper         ==> 1.35mil
5. Main                 ==> 20mil 
6. Inner Copper         ==> 1.35mil
7. PP 7628              ==> 7mil
8. Bottom Copper        ==> 1.35mil
9. Bottom Solder Mask   ==> 0.5mil

Total: about 40mil

PP7628 DK=4.2~4.7(4.7,<1GHz), use 4.3 for 2.4GHz. width=13mil,  Z=49~50ohm,  ceramics antenna Z=50ohm, perfect matched. So after this calculation, I changed board height from 0.8mm to 1mm and use line width 13mil for antenna.

From Apr.15~Apr.18, I kept it run “top” on the board and every 5 minutes my script sent one test.bin file(about 2MB) to board /tmp/ folder through wifi.
Now after 72 hours, its free memory is 6.5MB compared to 9.9MB at startup. Its temperature is about 65C at end(house about 24C) without heat sink. Looks good for me. 🙂

看来最近开源硬件确实挺火，深圳的创客们搞出来个WRTnode，基本理念和我不谋而合。我们用的都是OpenWrt系统，这样软件都可以共用。
其实我开始时也想用MT7620做设计，但是我最后由于以下几点原因否决了。
1.成本高，芯片是RT5350的一倍多，内存芯片价格也略高。
我希望做开源的东西，能以最低的门槛，做最大众化的东西。记得十年前，我上大学时从淘宝买到一块60块钱的二手8051实验板时的快乐，希望这样的快乐能更多的传递下去。
2.用的是DDR2，线很难布。
DDR2的频率比较高(>200MHz)，MT7620在内存和芯片间需要匹配阻抗。这样就导致了板子不能做的小,而且由于阻抗的存在，成本也会难以控制。WRTnode做到了40×30，虽然没有VoCore小，但我相信这基本是MT7620的极限了，而VoCore则不然，如果舍得成本，换成BGA封装的SDRAM内存，体积和功耗还可以进一步减小。
3.多余的性能性价比不高。
多余的CPU性能比较鸡肋，360MHz足够处理应用，580MHz实际性能比它快不了多少，基本可以说360MHz干不了的580MHz也干不了。
DDR2速度没有用处，瓶颈在总线速度上，不能完全发挥DDR2的速度。

WRTnode的优势：
1. 功耗低，MT7620运行电流是140mA左右，RT5350是200mA，同时带来的好处是发热大大减小。我粗略计算一下，考虑到多种因素，MT7620比RT5350节能20%～30%.
2. 速度快，MT7620的CPU可以运行在580MHz， RT5350是360MHz。
3. 支持内存大，MT7620的DDR2单片最小都是64MB，而SDRAM最大只有32MB，64MB还得接两片。
4. 双天线设计，这个真的很好，可以发挥出802n的最大威力，必须赞。

VoCore的优势:
1. 价格便宜，成本只有MT7620解决方案的1/2到1/3.
2. 体积小，体积只有WRTnode的三分之一，WRTnode核心板5×4.5=22.5平方厘米，VoCore是2.4×2.45=5.88平方厘米.
3. 容易扩展，GPIO多，导出了RT5350所有的GPIO，一共22个，还有JTAG，USB，I2C，I2S，PCM，网线，串口，一个都不少.
4. 支持多种电源电压，用3.3V~6V供电都可以，可以方便的使用锂电池供电不需要转换.

WRTnode设计上值得我学习的：
1. 导出了SPI的接口。
这个设计我没有考虑到，有这个接口确实很方便，省了60块买flash刷写夹具的钱。
这个是夹具链接，给他们做个广告http://item.taobao.com/item.htm?spm=a230r.1.14.1.6HxkOH&id=36799806985&_u=f40cgkm9c63
2. 接口都是2mm的插孔
我做出了24mm的半孔版后，觉得半孔虽然减小了很大的体积(0.5平方厘米)，但是焊接，接插都很不方便，很容易把半孔铜皮弄坏。2mm的插孔只要插上标准的杜邦线就行了，适合做实验。在下一版我会改进，使用完整插孔，由于板子面积有限，使用标准间距1.27mm，做下来的板子应该是990milx990mil(25.146mm)＝6.33平方厘米

WRTnode我要吐槽的：
1. 导出的网线太多了。
这是做智能设备控制又不是做路由，把5个网线接口都导出来，至少两个口基本一辈子都不会用到。平空增加布线复杂度。这样可能会不得不在中间电源层布线，而这样的布线会对信号层有很大干扰，特别是DDR2本来就是高频，网线口的速度也是100MHz以上的，这样的板子可能会不稳定。我在做5350的时候虽然最高只有166MHz也是小心翼翼的，绝不敢在电源层走线（别看VoCore小，只在top和bottom走线），高频尽量3W实在不行2W。串扰模拟时按1GHz，3.3V得出最大串扰电压<100mV，应该还算稳定。
2. 买不到……
这板子想搞个研究下，可惜哪都没卖的，是卖的太好么……郁闷，不过淘宝上其他的MT7620的开发板倒是很多。

最后，好久没有写中文博客，多说一句：
我坚信，不远的未来我们只需要通用的硬件，而功能都由软件实现。VoCore就是为了实践此而诞生的。