nm

12/13 Exit?

一半的出口，代表著我有超過一半的機會......

清晨的米子下著雨，重頭戲正要登場，
繼昨天的歡迎宴之後，今天是開幕儀式與發表會。

也許是移動日消耗太多體力，加上天氣偏冷，有點想要睡回籠覺，
強迫自己一定要起床，並且收拾東西去吃早餐，
拿著前一天發的資料袋，喔不，是相機包(照相=回憶，比論文集重要多了)，
準備去9樓銀座吃早餐，
華盛頓的早餐有分和定食與洋定食，
照字面上就是日式和西式的不同，
第一天早餐選擇了日式，有米飯，魚等....(如圖所示...細節忘了)


吃飽後趕著九點前到了會場
會場人也不多，

咦?日本人不是都不遲到的，
結果等了半天9:15分才開始，
然後就開始用日式發英文的演講講了一個半小時，都聽的快睡著了。

會場大廳
 
抬頭一看，圍幕玻璃
 
 
會場，演講人遲到


邊聽報告邊準備


因為自己的session在下午，完全沒急迫性，
趁演講時就順便準備了一下報告內容，
monorning tea也順便跟人交流了一下，
趁機多發幾張名片，要不然印了兩盒完全用不完，
有時遇到台灣人不免要寒暄幾句，
畢竟他鄉遇故知真是有一種特別的感覺
中間還跟一位當地媽媽級的工作人員聊天，
她很訝異我是台灣人，還說我長的像比較像日本人，
能被當地人給認同，真是天大的榮幸阿。
中午，就在超長排隊人潮下草草結束，自助式，不難吃。

巧遇台灣大學的博士生，很和善，人不錯
 
就是這位媽媽工作人員說我像日本人，趕緊合照一張
 

當我走進了UWB session的講堂，我失望了，
2006.4曾經參加過台灣的VLSI研討會，是"大"講堂，
滿心期待結果ISPACS只是間小教室，我愣了好久.....，
Session主持人是黑人，整個過程都蠻奇怪。

第一位上台報告的是瀋陽人，在日本千葉大學唸書，
很專業，英文說的很流利，表現的不錯，
而且名字很特別喔，叫做"張曉維"，
是台灣人就應該知道這是什麼意思吧，
當我還沒出發的時候就知道有這樣有趣的事，
不過我沒想到是女人，後來還一起去玩，真是奇妙。

第二位?沒來，聽說沒到現場發表會取消刊登權利(所以我才出現在這)。

第三位是個日本小胖，答問時間被問倒，答不出來，
最扯的就是他的老師還跳出來幫他回答，

小胖的老師:大概這樣子使用了，數據我忘記了，小胖你還記得嗎?

小胖:我也忘了

這....還一搭一唱

我還看到一個session的主持人主持到一半說

Chaitrman : Next paper,oh It's mine.

(這位主持人請了台下一位教授代理，並且上台報告)

真是個鳥研討會阿，什麼鳥事都有。

第四位就是我啦，等了好久，
不過因為這幾天沒辦法準備，才剛上臺就忘詞了，
報完之後不知所云，我想台下也聽不懂吧，因為我是做電路，其他論文是系統。
問答時間，有兩個人一直找我碴....

<問題一>

馬來西亞?泰國?來的教授 : $%$%^$^.
(媽呀，聽都聽不懂，發音超怪)
事後我翻譯了一下，

馬?泰?國教授: 你的雜訊太高了吧，你的模擬有包含其他的溫度嗎?

我 : 有阿，模擬結果在0~70度都在接受範圍底下，而且我的功率是別人的1/3....

我不斷重複這句話，因為那位發音奇怪的教授一直重複，所以我也一直重複

<問題二>

小胖教授 : 你的線性度似乎不太好，要用在MB-OFDM不太合乎需求
(我也是聽了很久才聽出來的，你也知道日本人發音不好)

我 : 若你要將mixer以及lna設計在一起，可以把lna的buffer拿掉，這樣就會變好了

小胖教授 : (&%&%&^%&$#@$#@$
(接下來的我聽不太懂)


很奇怪的是，他最後還給我建議，而且説我其實做的不錯，
很莫名奇妙的把我當他學生一樣說教，給我個台階下。
問答時間結束，黑人主持人替我解圍，一切終於結束了，
簽到之後(好像點名)，我具有畢業資格了!!!!!!
而且第一次用英文報告，真是難忘的經驗。

以下是我的講稿



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

<Title Page>

Good afternoon everyone. My name is Sheng-Hsiang Yen, I'm from Taiwan. It's my presure to present my paper here .The title is "A Low-Power and Low-Noise Amplifier for 3-5GHz UWB Applications".Later, i'd like to describe a common gate LNA for Direct series-UWB Low-band and Multiband-OFDM Group-A system. The amplifier is implemented in TSMC 0.18um RF CMOS process.

In this circuit, We concerned more about a brand new design of low power. Unfortuntely,these issues low power LNA always cause some problums,for instance, Higer NF , Lower Gain etc. In order to solve these problums, I'll introduce some good method. OK, Let's begin to talk more about that.

<Introduction>

In this slide I'd like to describe UWB system. Since February 2002 , UWB system is approved by the Federal Communications Commission, but up to now the stander has not yet been defined .there are two major architectures, DS-UWB and MB-OFDM .In DS-UWB, the band groups are separated into

two parts, High band (6.2 to 9.7GHz) and Low band (3.1 to 4.9 GHz). In MB-OFDM, some major Band Groups were defined, such as Group A (3.1 to 4.9GHz), C (6 to 8.1 GHz),D (8.1 to 10.6 GHz).The Group B (4.9 to 6 GHz) is not considered because it covers some other systems .For UWB application, designer should avoid the band of 5~6 GHz.

<Motivation>

In previous silde ,i introduced the UWB system. Designer should avoid the band of 5~6 GHz. In our design , we chooses the low band and the Group A. just working at 3.1~5GHz. We have some challenges.

Wideband matching , and how to design a very low power LNA with good power gain and NF performance.

So we have to chose one topology either CGLNA or CSLNA.CSLNA provides high gain and good noise performance.CGLNA has advantages of lower power consumption, better linearity et cetira. Finaly we chose CGLNA because it is a better choice for the design of a low-power LNA.

<Topology of LNA>

Compared with CGLNA, CSLNA provides higher gain and better NF. In NF equations ,the minimum NF of CSLNA is dependent on ω and ω T, which are operating and unity current gain frequencies. but the NF of CGLNA is approximately constant when input matching condition gm=Rs . In order to decrease NF of CGLNA,We should increase gm. The thing is " increasing gm cause bad matching". Due to this reason, So We need to insert a new network to fix this problems.

<Effective Gm>

Gm is the effective transconductance of the amplifier stage. It affects the gain and NF. The figure shows the input stage with inductor-terminated common gate LNA. The inductor Lg is introduced to get higher Gm, increase the gain and reduce input–referred noise. We derive the equation which directly shows the relation of Effective Gm and Lg , It is a function of operating frequency

and product of Lg and Cgs. According to the equation , When operating frequency increases, the effective Gm goes higher than the original value.

<Input Matching>

Reactive components are placed in circuits skillfully to get wideband matching and enhance bandwidth.

In this case a double-tuned interstage coupling network is used as the input matching circuit. The networks have been widely used for realizing maximally flat amplitude responses in band-pass. To avoid using transformer in our physical design, We used the equivalent circuit as shown

in right part. The coupling coefficient, k, depends on the inductor L. By

changing the value of L, the impedance response can be categorized into three major types, under-coupling, criticalcoupling and over-coupling. For the case of critical-coupling, the result is a broadband impedance matching that is suitable for UWB application.

<Input network>

For the proposed UWB LNA, the double-tuned interstage coupling networks are connected in parallel with the input stage of CGLNA. To achieve better noise performance, Cp is parallel with Cgs for higher Gm, as shown in Figure.


<Complete LNA Circuit>

M1 is connected to input networks in parallel for 50Ω matching
Bias current of M1 is 1.8mA
M2 provides higher gain and better isolation
M3 is a buffer for output matching

<Simulation Gain and NF>

The power gain (S21)
shown in Fig. 8 ranges from 15.4 dB to 14.2dB and the noise
figure decreases from 4.3dB at 3.1GHz to 3.8dB at 5GHz.

<Simulation input and output return loss>

the input return loss (S11) and output return
loss (S22) are below -10.6dB and -13.3dB, respectively.

<Simulation input-referred 1dB>

The input-referred 1dB as shown in Fig. 10 is -21.8dBm at 4GHz.

<Performance Summary>

The LNA circuit shows power consumption of 3.24mW
without output buffer. The summary of the proposed LNA is
listed in Table I.

<Conclusion>

According to the result i've talked b 4 .To sum up, at first,CGLNA is considered a better choice for the design of a low-power LNA. if you want to design a low power amplifier , it's a good decision.NEXT, The inductor-terminated common-gate LNA goes higher gm value.Higer gm could increase the power gain and decrease the NF . The third, Double-tuned interstage coupling network realizes maximally flat responses.For wideband matching design, it's very useful.

<Acknowledgment>

I want to thank the chip implementation center (CIC) for chip fabrication and technical support


<Performance Comparison>

This work has advantage of very low power design.

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


這次報告完回去檢討了一下，有以下幾點問題

1.經驗不足

沒有實際的英文報告經驗，遇到一些狀況無法立刻反應，
而上台時就應該打定主意選擇要看稿或是不看稿，而不要三心二意，
一下看稿一下又不看，而讓節奏亂掉。

2.準備不足

因為一開始就打定主意是個鳥演討會，很晚稿才完成，
而每天晚上都因為某老兄生活習慣不好而無法準備，所以準備不足忘詞

3.地點不佳

教室型態的場地，連講台都沒有，對於報告的人來說真的不知道要怎麼看稿
這三點我會記在心裡，若以後有機會，千萬要避免。

論文發表，小研討會，亂報好了


竟然忘詞.....