HackerTube English?
表示、直觉与反馈 →

同一台机器,换个地方量,为什么结论变了?

在第一轮排水时打开洗碗机,再把面包从烤面包机中拿走。两个小实验迫使我们说清楚测量从哪里开始、到哪里结束,以及仪器究竟感知了什么。

本文的问题 · 10
  1. 少了压缩机,热去了哪里?
  2. 换了涂层,想要的颜色还在吗?
  3. 洗完一轮,究竟是哪轮水洗的?
  4. 热泵吹出热风,能说明它此刻有多省电吗?
  5. 站在路边哪里,看起来这条路会很快?
  6. 说「伦敦人口」前先画出边界
  7. 要在晶圆上印一个 L,掩模该画什么?
  8. 采样装置开始工作时,响应计时该停了吗?
  9. 聚变增益的分母,画到了电站的哪一层?
  10. 计数从关卡入口开始,还是从整段流程开始?

少了压缩机,热去了哪里?

把温度计贴在热电模块冷面上,它的确能变冷。把视野扩大到整台设备,热端还需要散热片、风扇和足够的气流。地源热泵的讲解把比较边界再向外推:地下回路交换热量的对象也参与决定效率。两台机器若要放上同一张表,应写明是在比一个部件能有多冷,还是在比从一个空间移走同样多的热共用了多少电。

原文与时间戳 · 4
you need a finned heat sink
Thermoelectric cooling: it's not great. · ▶ 4:59–5:06
can move more heat energy out of the fridge's interior than it takes to run the heat pump
Thermoelectric cooling: it's not great. · ▶ 16:58–17:05
the temperature of the fluid returning from the ground loop
Ground Source / Geothermal Heat Pumps and Other Info · ▶ 10:34–10:37
the system simply puts heat back into the Earth
Ground Source / Geothermal Heat Pumps and Other Info · ▶ 12:15–12:19

换了涂层,想要的颜色还在吗?

给灯罩着色时,染料的颜色只是一个变量。不透明的覆层还能把尖亮的光点扩散开。对同一盏灯分别看光源近处、灯罩表面与整串灯,从哪个位置判断「像不像记忆里的灯」,会改变下一次应修改的部件。

原文与时间戳 · 2
the opaque white kind of frosting
LED Christmas lights still annoy me. Let's try some more fixes. · ▶ 10:24–10:29
water, white craft paint, and PVA glue
LED Christmas lights still annoy me. Let's try some more fixes. · ▶ 10:29–10:37

洗完一轮,究竟是哪轮水洗的?

设想两次洗碗,结束时盘子都干净。要检查预洗剂的作用,可以像节目里那样在第一轮排水就停机观察;等主洗结束再评判,就让另一种洗涤剂也参与了。拿同样的办法想一台烤面包机:把面包拿走,以为只删掉了「被烤的东西」,其实还删掉了双金属片接收热辐射的对象。一次改动究竟删去了哪个输入,要把传感器也画进去才看得见。

原文与时间戳 · 4
release detergent immediately for the first fill
Detergent packs are kinda wishy-washy (Dishwashers Explained) · ▶ 12:17–12:20
stopping it after it drains the first fill
Detergent packs are kinda wishy-washy (Dishwashers Explained) · ▶ 19:00–19:06
warmed mainly by the heat radiating from the bread itself.
The Antique Toaster that's Better than Yours · ▶ 4:19–4:24
the toaster barely runs before shutting off.
The Antique Toaster that's Better than Yours · ▶ 3:56–3:59

热泵吹出热风,能说明它此刻有多省电吗?

热泵的性能系数要用送进屋里的热量除以耗电量。节目里那台设备在严寒中仍吹出暖风,讲者却把当时的系数明确说成估计。一只温度计无法同时给出这两项读数。再把观察从一分钟拉到电池的整个使用期:制造电池的材料在多轮充电中使用,燃料却每次都要重新供给。此时分母若仍只写「一件产品」,两者的服务量就对不上了。

原文与时间戳 · 4
how much heat energy the heat pump moves compared to how much it consumes.
Heat Pumps: the Future of Home Heating · ▶ 21:10–21:18
its coefficient of performance in this scenario may have been just 1
Heat Pumps: the Future of Home Heating · ▶ 26:23–26:28
extracting and purifying resources
You are being misled about renewable energy technology. · ▶ 54:40–54:45
make a durable good
You are being misled about renewable energy technology. · ▶ 54:54–54:58

站在路边哪里,看起来这条路会很快?

在两个路口之间测速,很容易看见宽直车道能让车开多快。把起点移到车道入口,把终点移到街对面的商店,转弯车辆、冲突点与红灯就进入了计时范围。Not Just Bikes 用这种街道解释为何看似高速的设施仍会让人等候。选一个你熟悉的行程,两端各画一道线:换一个终点后,哪段原本被略过的等待回到了账上?

原文与时间戳 · 2
a high-speed connection between two places.
Stroads are Ugly, Expensive, and Dangerous (and they're everywhere) [ST05] · ▶ 0:55–1:01
many points of conflict that require drivers to constantly slow down
Stroads are Ugly, Expensive, and Dangerous (and they're everywhere) [ST05] · ▶ 4:55–5:03

说「伦敦人口」前先画出边界

Grey 用人口数字玩了个看似荒唐的开场。他先说的是 City of London 那片小区域,再把周围更大的伦敦算进来。边界一动,即使逐人计数完全准确,答案也会跟着变。想比较两个地方的拥堵程度时,可以先请两个人各画一张图:中心区、行政区域与日常通勤区域,哪一种才和读者想解决的问题相符?

原文与时间戳 · 2
a one-square mile hole near the middle
London Swallowed a City but Couldn't Digest It. · ▶ 0:28–0:32
the two Londons have separate city halls
London Swallowed a City but Couldn't Digest It. · ▶ 0:33–0:37

要在晶圆上印一个 L,掩模该画什么?

QueueTard 先让观众猜:要得到一个 L 形导线拐角,在掩模上也画同样的 L 就行吗?演示里成像把拐角印走了样;工程师于是反过来计算,让掩模预先带上一些形变。用于核查的图纸和实物应该分开摆:若拿掩模照片去和最终设计逐像素比,最有效的修正反而会被当作缺陷。再看锡滴光源的两次脉冲,第一枪打出的形状也得等到第二枪才能检验用途。两段说的工序不同,比较时先选定你关心哪一层的输出。

原文与时间戳 · 4
you'd think I'd put that L shape on a mask.
Indistinguishable From Magic: Manufacturing Modern Computer Chips · ▶ 13:49–13:55
That shape is numerically determined, basically calculated in a supercomputer.
Indistinguishable From Magic: Manufacturing Modern Computer Chips · ▶ 14:11–14:19
reshape the tin droplet from a sphere into a concave sheet.
The Extreme Engineering of ASML’s EUV Light Source · ▶ 11:03–11:11
the concave sheet tilts in the wrong direction and messes up the second shot.
The Extreme Engineering of ASML’s EUV Light Source · ▶ 11:15–11:21

采样装置开始工作时,响应计时该停了吗?

Sotos 讲到会吸入空气的监测车。若只统计采样次数,仪表盘可能看起来一直正常运行;要知道多久可以采取行动,还得跟着样本走:谁取走,何时送到实验室,何时完成分析?这与在路口之间量车速的问题相通。一次采样完成、或一段道路走完,都是合理的停止点;但如果读者问的是端到端耗时,就要把下一段等待也写进答案。

原文与时间戳 · 3
an infrastructure that will be able to detect new agents
DEF CON 25 - John Sotos - Genetic Diseases to Guide Digital Hacks of the Human Genome · ▶ 27:01–27:08
a human comes by, picks up the samples
DEF CON 25 - John Sotos - Genetic Diseases to Guide Digital Hacks of the Human Genome · ▶ 48:18–48:24
many points of conflict that require drivers to constantly slow down
Stroads are Ugly, Expensive, and Dangerous (and they're everywhere) [ST05] · ▶ 4:55–5:03

聚变增益的分母,画到了电站的哪一层?

面对聚变装置的增益数字,先沿能量流画两条边界:一条围住等离子体,另一条围住设施及其供电系统。Hossenfelder 强调,送入等离子体的能量没有包含设施全部耗能;即使产生了热量,也不能把它直接算成同样多的可输出电力。这里增补的是核算问题:分子是聚变热量还是电力,分母计了加热器输出还是整套设施输入?两种比值回答不同层次的问题。视频中的历史数字与转换效率估算不作为本站的当前性能数据;这也不是说局部实验增益没有价值,而是要在据此判断净供电前补齐系统边界。

原文与时间戳 · 2
it’s the Q_total that matters, not the Q_plasma.
How close is nuclear fusion power? · ▶ 4:36–4:43
Fusion reactors take a lot of energy to run, and *most* of that energy never goes into the plasma.
How close is nuclear fusion power? · ▶ 4:43–4:55

计数从关卡入口开始,还是从整段流程开始?

「半次 A 键」首先是一个记账边界问题。按讲者的解释,某段操作只要求 A 处于按住状态;若按下事件早已为前一段发生,这一段就未必新增一次按下。把同一段拿出来独立开始,则要把建立这个状态的动作算回来。讨论某颗星需要几次按键前,可以先写下入口状态:按钮是松开的,还是沿用前段的按住状态?这里比较的是分段需求与完整流程的新增成本,不是把一次物理按键切成一半;也不能从这个计数约定推导后半段游戏路线必然成功。

原文与时间戳 · 2
only required a to be held not actually pressed
SM64 - Watch for Rolling Rocks - 0.5x A Presses (Commentated) [OUTDATED] · ▶ 2:45–2:56
on a single star basis you'd round that up to three
SM64 - Watch for Rolling Rocks - 0.5x A Presses (Commentated) [OUTDATED] · ▶ 3:03–3:15

读完再想一步

为身边一件机器画出三样东西:使用者要的结果,仪器能读到的量,开始与停止读数的时刻。移动其中一个,写下原来能回答的问题还剩哪一半。