I am a real newbie on hardware 🙂

VoCore 0.2 has failed, when I connect it to power, some time the current cost is 0.20A but some time the current cost is 0.23A and RXD2/TXD2 is always high, it should change to low once the chip on boot process. So I guess the problem is before bootstrap, and the problem might be I floated PORST_N which is used to reset the chip.

For v0.2, I just simply followed the datasheet of RT5350F and trust its “internal pull up”. But in fact, reset can not work that easy way. it will be pull high once it connect to 3.3V then it will be pull low for a few ms(140ms~400ms), so if I just pull it to high will cause problem.

OK, then I jump a wire under BGA as following. Plan to connect a chip named MAX809S, it is a reset circuit. It is used to make reset stable.

Then solid all parts on. Then do a few test…

Connect to power, now it current consume is fixed to 0.20A, not float between 0.20A ~ 0.23A anymore. So I think at least the reset part is normal now.

Connect with TXD2 to a PL2302 USB-TTL device, I find TXD2 changed to low after boot, so boot strap is normal.

Then I try to read from the chip, no luck, nothing output. I tried many times, but no success. Only one time it output some REAL char when I press the memory hard onto the board(I am so happy when I get that output, it is close to success). I think there must be something wrong on the memory. Then I use ohmmeter to check between memory solid pad and memory chip foot, find at least 5 feet are not really connected to the pad. For fix that, I tried to use the high skill called “drag-soldering”, but just make my board into mess… and I lose one memory chip. 🙂

Another try is to solid another board, but this time BGA has some problem because of the jump wire. I have to stop here, or I will lose more RT5350F. :'(

For next step, I’d better to make another board, then use the “print-solid” way, much easier for a newbie and that will have high rate to success.

Show my current work.

VoCore v0.1 is a test version, it is used to make sure I am able to route all the nets though 2.7mm x 2.7mm board. VoCore v0.2 is a version to make the real PCB. I must be very careful to check the board.

I need about one month to make it done.

http://zh.wikipedia.org/wiki/5G

5G information from wifi. 5G is using 28GHz, that means the wave length is only 10mm. How could that be possible to transfer 2000m?Unless there is no block. And how to save power, it will cost more power absolutely.

Very very funny, looking forward for this tech and more information.

http://www.alternatezone.com/electronics/files/PCBDesignTutorialRevA.pdf

无线通讯技术的极限

目前wifi使用的是2.4GHz的通讯频率，波长可以简单的算出为3e11mm / 2.4e9 = 125mm 大概一分米多一点，所以wifi的信号穿墙是衰减很大的。2.4GHz理论的极限的通讯速度是2.4G / 8 = 300MB/s, 但是无线信号被干扰的很厉害，需要加大量的检错码，一般实际的极限只有1/6的速度，就是50MB/s ＝ 400Mb/s。大家购买的路由器300Mb/s就很快了。同时，速度和距离也有很大关系，距离越远由于干扰和信号衰减，速度会以非常快的速度下降。

未来的无限技术

我认为可以使用多频技术并且提高信号频率

多频的意思就是2.41G, 2.42G, 2.43G …同时工作 -> 每多一路，功率加一，处理的芯片加一。

提高信号频率就是用5G, 6G等更高频来传输数据 -> 但是频率越高功率越大，且越容易被遮挡。

而我的期待是可以通过无线传输两路HDMI，这样的速度大概是30GHz，通过占用10个3G通道应该可以实现，但是这样的设备小型化就比较难了，功耗也很难控制，相当等于12个2.4G的wifi芯片同时工作。理论上是可以实现的，我的无线路由做完后着手做一个这样模型，希望可以实现HDMI数据的传输。 一般wifi发射功率是100mW，距离也就是在80平米的房间里勉强覆盖，如果家里都是混凝土浇筑的估计信号衰减的会很厉害。 现在的电力线传输技术倒是可以解决一部分，但是电力线只有一根，像这样的超高频不使用差分信号且电力线都很长，就是一根接干扰的天线，现在的HiFi基本速度就极限了。 而这样的速度还是无法实现4K的传输的，那需要4个HDMI的速度。

在低频下，那就不用说了，只要不长到绕地球一周，几乎没有影响。高频则不同。走线和过孔都存在等效的电容和电感。电感会阻碍高频信号，所以对高频信号来说，导线的电阻也许微不足道，但产生的阻抗却是无法忽略的。

假如1ohm是阻抗的分界线，那么根据电感阻抗计算公式

X = 2 * PI * F * L

假设F=166MHz，标准的SDRAM的工作频率，这样可以算出1ohm的阻抗恰好是1nH

若F=2.4GHz，那么1ohm的阻抗是66pH

网上找的经验公式L=5.08h[ln(4h/d)+1]，一个4层1.6mm的板子上打10mil的过孔，那么L=1.35nH, 这样10mil的过孔就是1.4ohm的等效电阻。在2.4G的高频下，电阻相当于21ohm

计算过孔的寄生电容

C=1.41εTD1/(D2-D1)

这样我的过孔是10mil孔焊盘16mil,厚度T是1.6mm

C=1.41 * 4.4 * 64 * 10 / 6 = 0.66pF

这个在2.4GHz频率下等效电阻为X=1 / (2 * PI * f * C)=0.01,很小，可以忽略

在166MHz下等效电阻为0.15,也不大，过孔的寄生电容对高频影响并不大，只是在信号上下沿会导致延时，因为信号保持时过孔电压一样，就不存在电容了。

根据电容引起的上升时间变化通过公式计算大概在30ps左右，2.4GHz上升时间为416ps / 10 = 41.6ps，这个电容还是有一定影响的，所以高频到2.4GHz，导线必须尽量等长，这样大家都一样延时这么多时间就和谐了。

再算一下导线的寄生电容，

C=εS/4πkd=εLw/4πkd

走线厚度，w一般是1oz 35um。L走线长度假设1000mil=0.0254mm*1000=26mm=0.026m, ε基板介电常数，常用的是FR-4，大概4.4, d是间距，5mil布线间距=0.13mm=0.00013m,k是静电力常数9e9Nm^2/C^2,都换算成标准单位带进去算一下

C=4.4*0.026*35e-6/4/3.14/9e9/0.00013=2.72e-13F=0.27pF

这样对应的阻抗就是

X = 1 / (2 * PI * f * C)

在166MHz的高频下，等于这两根导线间接了一个X= 3.5k ohm的电阻.在2.4G的高频下等于接了一个240ohm的电阻。

继续计算导线的寄生电感

L = 0.01 * D * N * N / (L / D + 0.44) 网上的无敌经验公式

L单位是mH,D线圈直径cm,N线圈数，我们的是导线，此为1，L线圈长度，单位cm。

这样1000mil的导线假设围成1圈，电感为

L = 0.01 * 0.9cm / (2 * 3.14 + 0.44) = 1.3nH

假设我们的PCB板打了两个个过孔，在166MHz下相当于增加了3ohm的电阻，导线长度1000mil(25.4mm)相当于3000ohm，那么很容易算出我们有1/1000的分压，就是说3.3V的电压会感应出3.3mV，貌似不是很多，若旁边有10条平行线的话，大概是0.01V~0.02V的感应电压，这样的噪声在1%以内，理论上可以接受。

但是2.4GHz下情况就不同了，我们等于线之间只有240ohm阻抗而过孔电感和导线的电感产生的阻抗却有40～60ohm，3.3V产生的分压达到了1/5，如果边上还有高频导线干扰甚至可以达到1/2，如果考虑到环境电磁噪声，这样的线路板必然无法正常工作。

VoCore.MT3410

[Capacitance]
|1   |17  | C9,C16,C43,C92,C98,C104,  | 0402, 0.1uF, 10V, X5R, +/-10%
|    |    | C108,C113,C114,C387,C388, |
|    |    | C389,C390,C391,C414,C422, |
|    |    | C110                      |
|2   |4   | C48,C50,C77,C79           | 0402, 1.2nF, 50V, X7R, +/-10%
|3   |3   | C36,C40,C409              | 0402, 1.2pF, 50V, NPO, +/-20%
|4   |1   | C105                      | 0402, 1.5nF, 50V, X7R, +/-10%
|5   |2   | C7,C10                    | 0402, 100pF, 50V, NPO, +/-5%
|6   |1   | C4                        | 0402, 10nF, 16V, X7R, +/-10%
|7   |4   | C6,C17,C278,C443          | 0402, 10pF, 50V, NPO, +/-5%
|8   |3   | C8,C14,C111               | 0402, 1uF, 6.3V, X5R, +/-10%
|9   |2   | C35,C38                   | 0402, 2.7pF, 50V, NPO, +/-10%
|10  |1   | C18                       | 0402, 22nF, 25V, X7R, +/-10%
|11  |1   | C15                       | 0402, 27pF, 50V, NPO, +/-5%
|12  |7   | C2,C42,C84,C93,C109,C112, | 0402, 4.7uF, 6.3V, X5R, +/-20%
|    |    | C251                      |
|13  |1   | C161                      | 0402, 470pF, 50V, NPO, +/-5%
|14  |2   | C105,C106                 | 0603, 22uF, 10V, X7R, +/-20%
[Inductance]
|1   |1   | L11                       | 0402, 2.2nH, +/-10%
|2   |1   | L12                       | 0402, 2.7nH, +/-10%
|3   |1   | L444                      | 0402, 3.3nH, +/-10%
|4   |1   | L14                       | 0805, 4.7uH, +/-20%, 1100mA
[Resistance]
|1   |1   | R1                        | 0402, 12k, +/-1%
|3   |1   | SR1                       | 0402, 22, +/-1%
|5   |1   | R212                      | 0402, 300, +/-5%
|4   |1   | R67                       | 0402, 330k, +/-1%
|7   |1   | R68                       | 0402, 680k, +/-1%
|6   |15  | R71,R94,R95,R107,R118,    | 0402, 4.7k, +/-5%
|    |    | R119,R142,R144,R150,R154, |
|    |    | R192,R293,R308,R318,SR2   |
|8   |8   | R26,R27,R28,R29,R55,R56,  | 0402, 49.9, +/-1%
|    |    | R57,R58                   | 
|9   |1   | R300                      | 0402, 51, +/-5%
|10  |1   | R8                        | 0402, 8.2k, +/-1%
|11  |1   | R50                       | 0402, 9.1k, +/-1%
[Chip]
|1   |1   | MT3410L                   | SOT23-5
|2   |1   | RT5350F                   | TFBGA-196B
|3   |1   | EM63A165TS                | TSOP-54
|4   |1   | W25Q64FV                  | SOIC 208mil
[Other]
|1   |1   | CRYSTAL 20MHz             | 3225
|2   |1   | U.FL IPEX/IPX, SMT        | 3225

I made a simple application to show me the pin, so it is easier for me to location every pin on RT5350F and get the best result of every compose layout.
The app run like this. Source code is attached.

RT5350F PinMap Source Code Here