Assigment 3： Modulation Methosd, FM, AM and Airband, and Digital
Last week you used an integrated program, gqrx or SDR#, to control your SDR and listen to wideband commercial FM, and narrowband FM used in police and fire radio. This week we'll listen to the some of the signals that aircraft use. These use amplitude modulation, or AM. Then we will take a first look at some ways digital data is transmitted over RF channels.
There are many ways to send a voice signal over the air. Last week we listened to FM signals, where the frequency of the carrier wave is varied with the amplitude of the signal. This week we are looking at AM modulation, where the amplitude is varied with the amplitude of the signal. These are the two most common ways of directly transmitting voice. These are known as analog modulation, because the signal itself is transmitted, not it's digital representation.
The basic idea is illustrated below:
The carrier is the pure RF signal when it isn't modulated. It is a constant frequency sinusoid. For last week's signals the frequency was in the 100's of MHz. The sdr's only go down to about 30 MHz, and up to 1700 MHZ.
The message is the second line. In this case it is shown as another sinusoid, which will be in the audio range. The frequencies here will only be a few kHz (the range of frequencies needed to convey voice).
For AM, all we do is add something to the message so that it is always positive, and then multiply it by the carrier. That's what we transmit. Then to receive, all we need to do is follow the envelope of the waveform. A diode and a capacitor will do this. This is all a crystal radio is. Note that the signal goes up and down in amplitude as it goes along.
For FM, all we do is slightly change the frequency of the carrier. It always has the same amplitude. The frequency increases when the signal is high (loud), and decreases when it is low (quiet). The transmitter is a little more complicated, but still pretty simple. The receiver just needs to differentiate the signal, which can be done many ways. When the signal is loud, it varies more quickly, and the derivative is larger. When the signal is low, the signal varies more slowly, and the derivative is smaller.
The first part of the lab is to capture and decode AM signals in the air band, which is right above the commercial FM band we were decoding last week. There is a band from 108-118 MHz that mostly has navigation beacons that identify themselves by Morse code. Then from 118-137 MHz there are several bands used for communication between aircraft and the ground. Communications in these bands use AM modulation. This is because when two users try to talk on the same channel the stronger channel will come through with AM, and this is usually the tower. You will also be able to hear something of the weaker station. With FM you will just hear a buzzing if two users interfere, unless one station is very much stronger, and then you only hear stronger user. AM gives you a better idea of who else is out there.
There are many airports around here. The major ones are shown below.
These all can be heard from here. There are also lots of small airports, and Moffet Field.
The local airport is Palo Alto, which is station KPAO. It transmits on these frequencies
We will also hear traffic from San Francisco (SFO), San Jose (SJC), as well as NORCAL Approach, which coordinates the airspace. The frequencies for SFO are
and San Jose are
and Norcal Approach are
Choose a frequency where you might expect to get a signal. The ATIS frequencies are good initial candidates, because these continuously transmit information about the airport, and how to contact them. The other frequencies, such at the air traffic control frequencies, are more interesting, but are not always in use. Often a transmission lasts just a few seconds, and can be hard to capture.
One busy channel that has a strong signal around here is 310.8 MHz, SFO departures and arrivals from the south. Another active frequency is band around 135.0 MHz, which has two of the approach frequencies from the south. You'll hear all the planes overhead on these.
If you are not in the bay area, you can look your local airport up in
and see what frequencies they use.
Use gqrx or SDR# to see if you can find any activity. If we were to extract just the audio signal from one of these channels, you would see something like this:
When the operator keys up the radio (pushes the Push To Transmit, or PTT button) the radio starts transmitting its carrier frequency. This looks like a constant, because we are tuned to that frequency. When they start to talk, you see the audio waveform added to the carrier.
This is a signal from 310.8 MHz.
Two audio clips are here:
Clip one
Clip two
This is a signal from 135.65
This was collected in Packard EE, where there is a tremendous amount of background noise (the stripes in the waterfall plot). I'd complain, but it could well be my lab!
The wikipedia page on the airband channels is interesting. You can find it here:
Particularly look at the last section about “Unauthorized Use”, and the references cited. It is an interesting story!
To get clear signals, it is helpful to get the antenna length right. For last week it didn't matter too much. FM signals are essentially perfect provided they are above the threshhold where you can determine the frequency of the signal. This happens pretty abruptly, and is known as “full quieting”. This is why FM is so popular. With enough power, it is perfect.
For AM, that is not true. As you get more signal, the sound gets better. Last week we talked about calculating the right length for an antenna. For a half wave dipole (both antenna elements pointed in opposite directions) this should be half a wavelength. For the quarter-wave dipoles you have this year, it should be a quarter of the wavelength.
For AM, you can also tune the antenna length by hand. Set your sdr to the frequency you want, and then adjust the length to maximize the noise. It turns out that will also maximize your signal. Make sure you aren't touching the antenna when you compare levels. You are a lossy dialectric, and that detunes the antenna.
For Part 1 on the assignment, find an airband signal, make a screen shot and audio capture Use the “rec” buttong on gqrx, it leaves it in your top level directory. There is one in SDR# also, which leaves the audio in the same directory as SDR#. Make two slides with
Your screen shot, and a picture of your antenna
Who you are listening to
Upload your slides and audio file to
Airband Signals
Set the sharing for the slides and the audio clip to “Stanford”, so that we can play them in class.
Look for the Morse Code beacons, if you can find them. It is amazing these are still out there!
Next week we are going to start looking
chat gpt 七个盈利模式 chatgpt带来的商业机会
2022年12月，OpenAI发布了对话式AI新模型ChatGPT，一经面世就引发科技界的巨大关注。
ChatGPT可以通过从数百万个网站收集信息，以对话式、人性化的方式生成独特的答案，能在一定程度上替代搜索引擎。
Chat 强调了其为了对话而优化，具有记忆能力，可以完成连续性对话。
它代表了OpenAI最新一代的大型语言模型。
01
据外媒TheInformation 报道显示，微软可能会在 2023 年 3 月之前在 Bing （必应）中应用 ChatGPT，用人工智能 来回答一些搜索查询。若 Bing 引入 ChapGPT，有望以完整句子的形式呈现搜索 结果，这一举措或将改变搜索引擎的运行模式，革新性地提升搜索引擎效率。
02
ChatGPT 基础设施核心环节梳理
ChatGPT上线象征着文本类AI进入新阶段，利好其基础设施及应用。
算法、芯片、存储、数据和算力是AI大模型训练的基础。
1. 云基础设施
国内云基础设施方面，Canalys统计2022年第三季度中国云基础设施服务支出达到78亿美元，同比增长8%。
“中国四朵云”（阿里云、腾讯云、华为云、百度云）格局未变，合计占据80%市场份额。
2.? AI芯片
中国AI芯片市场入局者众多。
海思半导体已成为国内营收规模最大的集成电路设计企业，以N腾系列为主要GPU产品，在AI终端芯片具有优势。云端领域，百度可提供云上GPU服务器；海光信息、天数智心与登临科技均以GPGPU为主打，是国内特定细分领域的主要竞争对手。
芯片领域主要厂商还包括富瀚微、国科微、北京君正、华为海思、星宸科技等；存储器主要厂商主要是长江存储。
3.? 数据
数据集作为数据资源的核心组成部分，是指经过专业化设计、采集、清洗、标注和管理，生产出来的专供人工智能算法模型训练的数据。
目前我国的数据标注与审核业务的参与者主要包括两类，一是人工智能公司内部的标注部门，二是商务流程外包公司。
主要厂商包括海天瑞声（AI训练数据专业提供商）、京东众智、慧听数据、龙猫数据、数据堂、曼孚科技、星辰数据、倍赛、蚂蚁众包、云测、爱数智慧等。
头部数据标注企业概况图：
4.? 对话交互类AI算法
当前国内领先的对话交互类AI算法企业主要包括两类：
以语音识别为核心业务的企业，延生到NLP领域，如科大讯飞，从语音识别延生到语音合成、机器翻译、图文识别、图像理解、阅读理解、机器推理等算法，并在消费者、智慧教育、智慧城市、智慧司法、智能服务、智能汽车、智慧医疗、运营商等领域实现深度应用。
产业链布局相关厂商还包括初灵信息、慧博云通、信雅达、云从科技、天玑科技、科大国创、高伟达、神思电子、高鸿等。
5.? AI 视觉
纵观AI视觉相关产业链，从元器件到集成商，国内厂商均有完善的布局，且在全球竞争中处于优势地位。
人工智能图像识别技术已经进入广泛落地阶段，国内AI视觉领军企业已将技术应用于安防、工业、文旅等领域，为各行各业提出了智慧化解决方案，有效降本增效。
AI视觉产业链：
人工智能产业链基础层相关公司
ChatGPT 玩法大全火了，一键复制就能 get 同款效果：脱口秀张口就来，还能扮演哈利波特
chatgpt是什么意思,chatgpt官网,chatgpt国内能用吗,chatgpt中文正值风头的“网红”ChatGPT 在过去一周算是被网友们玩坏了！
各种有的没的玩法都被网友们发掘出来了……
比如说就有网友被误开了一份停车罚款，又不会写解释信，于是直接“请来”ChatGPT 帮忙。
最终调整出来的解释信是这样的：
一天后就收到了区域市政局回复，罚款直接被取消了！
这可以算通过图灵测试了吧，以后写申请书啥的不就可以解放双手了。（手动狗头）
不过这还都只是 ChatGPT 的冰山一角，在“水面之下”它还有哪些“不为人知”的玩法？
这不，有网友直接整了个大汇总，一一列出 ChatGPT 的各种用处，还贴心地给出了完整的大段提示词，一键复制就能轻松拥有同款效果。
目前在 GitHub 已经标星 600 多次。
具体有哪些，一起来看看～
首先先来讲讲 ChatGPT 的实用性功能。
当然，简单的语言翻译对 ChatGPT 不成问题，不过在翻译润色这块儿，ChatGPT 也是有点东西在身上的。
比如说让它来充当英语翻译者，翻译一句土耳其语，再对其做一些要求：高级词汇、有文学性等等。
最终翻译出来的效果是这样子的：
再用它来试试中文古诗的效果，貌似有点儿意思。
还有，ChatGPT 还能创建 Excel 表格，直接输入所需的表格尺寸，就能分分钟生成。
试着填下数字，也能正确填入对应的位置。
ChatGPT 甚至能被当作 javascript 控制台或者 linux 终端。
又或者让它给出一些具体意见时，它也能讲得头头是道。
除了这些之外，ChatGPT 还有很多待被发掘的功能，这里也就不一一赘述了。
当然，ChatGPT 不只是无情的“生产力工具”，它也是有点文艺细胞在身上的。
大段的脱口秀也是张口就来：
或者直接让 ChatGPT 谱段曲子。
生成的结果是这样子的，中间甚至还加入了合唱部分。
写小说，讲故事等这类 GPT-3 就已经具备的基础功能，ChatGPT 就不用说了，大段大段的文字也是分分钟拿下！
或者无聊时，可以和 ChatGPT 玩玩角色扮演，比如让 ChatGPT 扮演哈利波特。
又或者让它充当免费的旅游指南，想去哪里玩直接问 ChatGPT，有什么要求也可以尽管提。
……
当然，ChatGPT 也还没有进化到真正的人类水平，闹笑话是难免的。
比如说让它“套个娃”：自己生成提示词：
不过用它生成的提示词作为输入时，它自己却不干了。
又或者说让 ChatGPT 讲一段关于“躺平”的脱口秀。
emmmm，一点也不好笑，写到一半还罢工了。
ChatGPT 用法大全还在持续更新中，感兴趣的可以先码住了～
传送门：
https://github.com/f/awesome-chatgpt-prompts
参考链接：
https://notesbylex.com/disputing-a-parking-fine-with-chatgpt.html
本文来自微信公众号：量子位 （ID：QbitAI），作者：Pine
