WEBVTT

1
00:00:00.040 --> 00:00:02.520
<v Speaker 1>All right, welcome back everyone. Today we're going to be

2
00:00:02.520 --> 00:00:07.000
<v Speaker 1>looking at something super important in electronics design, minimizing and

3
00:00:07.120 --> 00:00:11.160
<v Speaker 1>exploiting leakage in VLSI design. Oh yeah, you know, like,

4
00:00:11.320 --> 00:00:15.400
<v Speaker 1>uh deal, have you ever noticed your phone battery like draining,

5
00:00:15.599 --> 00:00:17.039
<v Speaker 1>even when it's just sitting.

6
00:00:16.719 --> 00:00:18.519
<v Speaker 2>There, happens all the time.

7
00:00:18.800 --> 00:00:22.719
<v Speaker 1>Turns out that's leakage power in action, and it's something

8
00:00:22.800 --> 00:00:25.879
<v Speaker 1>that engineers designing chips have to deal with constantly.

9
00:00:26.039 --> 00:00:27.440
<v Speaker 2>Oh yeah, absolutely, we're going to.

10
00:00:27.440 --> 00:00:30.320
<v Speaker 1>Be taking a deep dive into all that today using

11
00:00:30.760 --> 00:00:35.119
<v Speaker 1>this book by doctor Sunil Pikatre. Okay, he and a

12
00:00:35.200 --> 00:00:38.399
<v Speaker 1>whole team of researchers really dove deep into the world

13
00:00:38.520 --> 00:00:41.640
<v Speaker 1>of low power chip design. Yeah, they really know their stuff,

14
00:00:41.719 --> 00:00:45.520
<v Speaker 1>So get ready to explore how those tiny chips use power,

15
00:00:46.119 --> 00:00:50.000
<v Speaker 1>the clever tricks engineers use to you know, minimize energy waste,

16
00:00:50.280 --> 00:00:52.359
<v Speaker 1>and even some ways that they've learned to actually use

17
00:00:52.439 --> 00:00:55.600
<v Speaker 1>leakage to their advantage. It's pretty amazing, kind of counterintuitive,

18
00:00:55.840 --> 00:00:58.600
<v Speaker 1>yeah for sure. So when we talk about power consumption

19
00:00:58.840 --> 00:01:02.000
<v Speaker 1>in chips, you usually think of dynamic power, right, Yeah,

20
00:01:02.039 --> 00:01:06.000
<v Speaker 1>that's the energy it chip uses when it's actively computing. Yeah, exactly,

21
00:01:06.040 --> 00:01:08.519
<v Speaker 1>it's doing stuff uh huh. But there's this other thing,

22
00:01:08.599 --> 00:01:11.719
<v Speaker 1>this leakage power that we don't always think about.

23
00:01:11.840 --> 00:01:13.680
<v Speaker 2>Yeah, it's like a silent drain on the battery.

24
00:01:14.000 --> 00:01:16.200
<v Speaker 1>Right, even when a transistor is supposed to be off,

25
00:01:16.400 --> 00:01:19.719
<v Speaker 1>that's right, it's still leaking a bit of energy.

26
00:01:19.519 --> 00:01:20.560
<v Speaker 2>Like a leaky faucet.

27
00:01:20.760 --> 00:01:21.959
<v Speaker 1>Yeah, exactly.

28
00:01:22.120 --> 00:01:25.760
<v Speaker 2>And you know what, this problem actually gets worse as

29
00:01:25.799 --> 00:01:27.079
<v Speaker 2>transistors get smaller.

30
00:01:27.519 --> 00:01:28.959
<v Speaker 1>Oh really, why is that?

31
00:01:29.480 --> 00:01:32.200
<v Speaker 2>Well? Think of it like this. As we shrink those

32
00:01:32.239 --> 00:01:35.879
<v Speaker 2>transistors down to pack more and more into our.

33
00:01:35.799 --> 00:01:38.319
<v Speaker 1>Devices, right, trying to get more powerful chips.

34
00:01:38.480 --> 00:01:42.640
<v Speaker 2>Exactly, the gaps between them get tinier and tinier. Okay,

35
00:01:42.840 --> 00:01:45.200
<v Speaker 2>it's like trying to prevent water from seeping through cracks

36
00:01:45.200 --> 00:01:45.719
<v Speaker 2>in a dam.

37
00:01:45.879 --> 00:01:47.719
<v Speaker 1>Right, smaller the gaps, the harder it is.

38
00:01:47.920 --> 00:01:48.359
<v Speaker 2>Exactly.

39
00:01:48.439 --> 00:01:50.879
<v Speaker 1>So it's like the electrons are finding more ways to

40
00:01:50.959 --> 00:01:51.560
<v Speaker 1>sneak through.

41
00:01:51.760 --> 00:01:53.159
<v Speaker 2>You can think of it that way. Yeah, it's a

42
00:01:53.239 --> 00:01:54.719
<v Speaker 2>quantum phenomenon.

43
00:01:54.200 --> 00:01:56.560
<v Speaker 1>Actually quantum wow called tunneling.

44
00:01:57.280 --> 00:02:02.560
<v Speaker 2>Electrons can actually tunnel through barriers that, according to classical physics,

45
00:02:02.879 --> 00:02:05.879
<v Speaker 2>they shouldn't be able to cross, So they're like teleporting.

46
00:02:06.120 --> 00:02:09.319
<v Speaker 2>It's kind of like that. The equation that describes all

47
00:02:09.360 --> 00:02:11.080
<v Speaker 2>this can get pretty complex.

48
00:02:11.280 --> 00:02:11.680
<v Speaker 1>Ill bet.

49
00:02:11.759 --> 00:02:13.840
<v Speaker 2>But the important thing to remember is that the amount

50
00:02:13.879 --> 00:02:17.439
<v Speaker 2>of leakage depends on something called the threshold voltage.

51
00:02:17.520 --> 00:02:19.759
<v Speaker 1>Okay, the threshold voltage.

52
00:02:19.319 --> 00:02:21.199
<v Speaker 2>And it's an exponential relationship.

53
00:02:21.280 --> 00:02:24.919
<v Speaker 1>Exponential, So even a small change in that voltage, yeah,

54
00:02:25.000 --> 00:02:27.800
<v Speaker 1>could mean a big difference in leakage, huge difference.

55
00:02:28.000 --> 00:02:31.360
<v Speaker 2>That's why controlling that threshold voltage is so crucial for

56
00:02:31.439 --> 00:02:32.879
<v Speaker 2>designing low power chips.

57
00:02:33.039 --> 00:02:33.560
<v Speaker 1>Makes sense.

58
00:02:33.800 --> 00:02:37.319
<v Speaker 2>Historically though, we've been focused on making chips faster and

59
00:02:37.400 --> 00:02:41.120
<v Speaker 2>more powerful, right, which often meant lowering that threshold.

60
00:02:40.680 --> 00:02:43.439
<v Speaker 1>Voltage, which then increases leakage unfortunately.

61
00:02:43.520 --> 00:02:47.840
<v Speaker 2>Yes, so there's this constant tug of war between performance

62
00:02:48.280 --> 00:02:49.319
<v Speaker 2>and power efficiency.

63
00:02:49.439 --> 00:02:52.680
<v Speaker 1>So how do engineers tackle this leakage problem?

64
00:02:52.759 --> 00:02:53.560
<v Speaker 2>It's a real challenge.

65
00:02:53.560 --> 00:02:56.759
<v Speaker 1>Do they like build tiny walls to keep the electrons in?

66
00:02:57.000 --> 00:03:01.120
<v Speaker 2>Uh huh? Not quite. But they do have some clever strategies, okay,

67
00:03:01.159 --> 00:03:06.840
<v Speaker 2>like what Well, one common technique is called mtcms mtcmos.

68
00:03:06.919 --> 00:03:07.719
<v Speaker 1>What's that stand for?

69
00:03:07.840 --> 00:03:11.199
<v Speaker 2>Multi threshold cmos. It's basically like putting parts of the

70
00:03:11.319 --> 00:03:13.560
<v Speaker 2>chip to sleep when they're not being used.

71
00:03:13.960 --> 00:03:16.240
<v Speaker 1>Oh, like a power nap for the chip exactly.

72
00:03:16.520 --> 00:03:20.759
<v Speaker 2>They use high threshold voltage transistors as gatekeepers to shut

73
00:03:20.840 --> 00:03:23.159
<v Speaker 2>off power to inactive sections.

74
00:03:23.439 --> 00:03:26.919
<v Speaker 1>So that drastically reduces leakage in those areas. It does,

75
00:03:27.120 --> 00:03:28.960
<v Speaker 1>but I guess there's a trade off there, there is.

76
00:03:29.120 --> 00:03:31.800
<v Speaker 1>Waking those sections back up takes little.

77
00:03:31.560 --> 00:03:34.280
<v Speaker 2>Time, right, so it can slightly impact performance.

78
00:03:34.759 --> 00:03:37.960
<v Speaker 1>Ah, so you lose a bit of responsiveness.

79
00:03:37.159 --> 00:03:39.840
<v Speaker 2>Yeah, but you gain in power savings, gotcha.

80
00:03:39.879 --> 00:03:40.439
<v Speaker 1>So it's always a.

81
00:03:40.439 --> 00:03:41.639
<v Speaker 2>Balance, always a balance. Yeah.

82
00:03:41.719 --> 00:03:43.960
<v Speaker 1>Makes you wonder if there are other ways to reduce

83
00:03:44.039 --> 00:03:46.360
<v Speaker 1>leakage without sacrificing performance. Oh.

84
00:03:46.400 --> 00:03:49.800
<v Speaker 2>Absolutely. Engineers have come up with all sorts of techniques

85
00:03:50.039 --> 00:03:54.840
<v Speaker 2>like what well, things like designing special low leakage cells, okay,

86
00:03:54.879 --> 00:03:57.199
<v Speaker 2>carefully controlling the input signals to the.

87
00:03:57.240 --> 00:03:59.879
<v Speaker 1>Chip right to minimize the leakage.

88
00:03:59.439 --> 00:04:04.080
<v Speaker 2>Exactly, and even adjusting the physical length of those tiny transistors.

89
00:04:04.280 --> 00:04:06.680
<v Speaker 1>Wow, so they're really getting down to the nitty.

90
00:04:06.400 --> 00:04:09.080
<v Speaker 2>Gritty they are. Each approach has its pros and cons.

91
00:04:09.120 --> 00:04:11.800
<v Speaker 2>Of course, it really depends on the specific application and

92
00:04:11.840 --> 00:04:12.919
<v Speaker 2>the design constraints.

93
00:04:13.639 --> 00:04:16.240
<v Speaker 1>I can imagine. It's amazing how much effort goes into

94
00:04:16.279 --> 00:04:18.600
<v Speaker 1>managing something as tiny as electron leakage.

95
00:04:18.680 --> 00:04:19.800
<v Speaker 2>It's really quite impressive.

96
00:04:19.839 --> 00:04:22.439
<v Speaker 1>I mean, it really shows how crucial power efficiency is

97
00:04:22.519 --> 00:04:23.279
<v Speaker 1>in chip design.

98
00:04:23.399 --> 00:04:25.839
<v Speaker 2>It really is, and you know, it makes you wonder,

99
00:04:26.000 --> 00:04:29.240
<v Speaker 2>what's that? What if instead of just trying to minimize

100
00:04:29.279 --> 00:04:33.160
<v Speaker 2>this leakage, we could actually use it to our advantage.

101
00:04:32.800 --> 00:04:35.800
<v Speaker 1>Use it really Yeah, that's.

102
00:04:35.639 --> 00:04:39.000
<v Speaker 2>The idea behind something called sub threshold circuit design.

103
00:04:39.240 --> 00:04:41.360
<v Speaker 1>Sub threshold design, what's that all about?

104
00:04:41.519 --> 00:04:44.519
<v Speaker 2>What's a pretty radical concept. Instead of trying to eliminate

105
00:04:44.600 --> 00:04:48.439
<v Speaker 2>leakage by keeping those transistors firmly off right, sub threshold

106
00:04:48.439 --> 00:04:53.000
<v Speaker 2>design actually operates the transistors below their threshold voltage.

107
00:04:52.800 --> 00:04:56.120
<v Speaker 1>Below the threshold, so they're intentionally leaky exactly.

108
00:04:56.160 --> 00:04:58.720
<v Speaker 2>It might sound crazy, but hear me out. It's a

109
00:04:58.720 --> 00:05:01.959
<v Speaker 2>clever way to achieve all for low power consumption, especially

110
00:05:02.040 --> 00:05:03.680
<v Speaker 2>when speed isn't the main concern.

111
00:05:04.079 --> 00:05:07.120
<v Speaker 1>Okay, I'm intrigued, but why would you want to do that?

112
00:05:07.160 --> 00:05:09.399
<v Speaker 1>I mean, isn't leakage a bad thing traditionally?

113
00:05:09.519 --> 00:05:13.079
<v Speaker 2>Yes, But in the sub threshold region transistors have some

114
00:05:13.199 --> 00:05:17.720
<v Speaker 2>unique characteristics. By operating them at these lower voltages, you

115
00:05:17.759 --> 00:05:20.120
<v Speaker 2>can achieve dramatically lower power consumption.

116
00:05:20.480 --> 00:05:23.519
<v Speaker 1>Interesting, so it's like whispering instead of shouting. You might

117
00:05:23.519 --> 00:05:25.439
<v Speaker 1>not be as fast, but you save a lot of energy.

118
00:05:25.560 --> 00:05:28.240
<v Speaker 2>That's a great analogy, and for things like low power

119
00:05:28.279 --> 00:05:31.959
<v Speaker 2>sensors or medical implants where battery life is crucial. That

120
00:05:32.120 --> 00:05:33.879
<v Speaker 2>energy saving can be a game changer.

121
00:05:34.279 --> 00:05:37.800
<v Speaker 1>So are there real world examples of this sub threshold

122
00:05:37.839 --> 00:05:38.720
<v Speaker 1>design in action.

123
00:05:39.079 --> 00:05:42.480
<v Speaker 2>Absolutely? In fact, the book we're looking at today features

124
00:05:42.480 --> 00:05:43.759
<v Speaker 2>a fascinating case study.

125
00:05:43.839 --> 00:05:44.560
<v Speaker 1>Oh what is it?

126
00:05:44.319 --> 00:05:49.160
<v Speaker 2>It's a sub threshold b FSK transmitter chip DFSK.

127
00:05:49.480 --> 00:05:49.920
<v Speaker 1>What's that?

128
00:05:50.040 --> 00:05:53.040
<v Speaker 2>It stands for a binary frequency shift keying. It's a

129
00:05:53.079 --> 00:05:56.319
<v Speaker 2>method for a wireless communication And they designed this entire

130
00:05:56.360 --> 00:05:59.680
<v Speaker 2>transmitter using sub threshold circuits. Wow, they were able to

131
00:05:59.680 --> 00:06:03.240
<v Speaker 2>achieve significantly lower power consumption compared to traditional designs.

132
00:06:03.399 --> 00:06:07.279
<v Speaker 1>That's impressive a wireless transmitter running on leaky transistors. But

133
00:06:07.560 --> 00:06:10.639
<v Speaker 1>I'm guessing designing in the sub threshold region comes with

134
00:06:10.680 --> 00:06:11.959
<v Speaker 1>its own set of challenges.

135
00:06:12.199 --> 00:06:16.319
<v Speaker 2>You're absolutely right. Transistors in this leaky state are much

136
00:06:16.399 --> 00:06:21.560
<v Speaker 2>more sensitive to variations variations like what things like temperature, voltage,

137
00:06:22.120 --> 00:06:25.519
<v Speaker 2>even tiny manufacturing differences. Oh wow, it's like trying to

138
00:06:25.519 --> 00:06:28.120
<v Speaker 2>build a house of cards on a shaky table. Yeah.

139
00:06:28.160 --> 00:06:31.240
<v Speaker 1>I can see that you need some serious engineering magic

140
00:06:31.319 --> 00:06:33.240
<v Speaker 1>to make that work reliably.

141
00:06:32.959 --> 00:06:36.160
<v Speaker 2>Exactly, And that's what makes these designs so ingenious. They

142
00:06:36.160 --> 00:06:39.040
<v Speaker 2>have to come up with clever tricks to keep everything stable.

143
00:06:39.759 --> 00:06:41.319
<v Speaker 1>So what kind of tricks are we talking about.

144
00:06:41.519 --> 00:06:45.800
<v Speaker 2>Well, one of their secret weapons is something called plaslas

145
00:06:45.800 --> 00:06:51.240
<v Speaker 2>programmable logic arrays. They're like reconfigurable building blocks in the

146
00:06:51.279 --> 00:06:53.920
<v Speaker 2>digital part of the circuit. Okay, they could be programmed

147
00:06:53.959 --> 00:06:55.519
<v Speaker 2>to perform specific.

148
00:06:55.000 --> 00:06:58.600
<v Speaker 1>Functions, so they're like logic legos. You can rearrange them

149
00:06:58.680 --> 00:06:59.959
<v Speaker 1>to create different circuits.

150
00:07:00.160 --> 00:07:01.879
<v Speaker 2>That's a great way to think about it. And what

151
00:07:01.920 --> 00:07:05.279
<v Speaker 2>makes plas particularly useful in sub threshold design is their

152
00:07:05.360 --> 00:07:06.439
<v Speaker 2>predictable structure.

153
00:07:06.720 --> 00:07:09.800
<v Speaker 1>Right, so even though the transistors are leaky, the overall

154
00:07:09.839 --> 00:07:12.199
<v Speaker 1>behavior is still controlled precisely.

155
00:07:12.720 --> 00:07:15.680
<v Speaker 2>Now. On top of that, they use something called adaptive

156
00:07:15.720 --> 00:07:19.120
<v Speaker 2>body biasing to deal with those pesky variations.

157
00:07:19.120 --> 00:07:23.079
<v Speaker 1>We talked about adaptive body biasing. That sounds pretty high tech.

158
00:07:23.160 --> 00:07:26.720
<v Speaker 2>What does it do exactly, Well, it's basically a feedback mechanism.

159
00:07:26.759 --> 00:07:27.439
<v Speaker 1>Feedback.

160
00:07:27.560 --> 00:07:32.560
<v Speaker 2>Yeah. It constantly adjusts the voltage applied to the transistor's substrate.

161
00:07:32.639 --> 00:07:35.199
<v Speaker 2>The substrate it's also called the body of the transistor.

162
00:07:35.439 --> 00:07:36.399
<v Speaker 1>Oh okay, And.

163
00:07:36.360 --> 00:07:40.279
<v Speaker 2>By adjusting that voltage, they can compensate for variations in temperature, voltage,

164
00:07:40.319 --> 00:07:41.279
<v Speaker 2>and manufacturing.

165
00:07:41.560 --> 00:07:43.800
<v Speaker 1>So it's like a self tuning system that keeps everything

166
00:07:43.879 --> 00:07:44.600
<v Speaker 1>running smoothly.

167
00:07:44.720 --> 00:07:47.279
<v Speaker 2>Exactly. It's pretty ingenious, right it is.

168
00:07:47.839 --> 00:07:50.360
<v Speaker 1>This is all fascinating stuff. I'm really starting to see

169
00:07:50.399 --> 00:07:53.319
<v Speaker 1>how this sub threshold design could be a game changer

170
00:07:53.439 --> 00:07:54.279
<v Speaker 1>in electronics.

171
00:07:54.319 --> 00:07:55.519
<v Speaker 2>It really is a new frontier.

172
00:07:55.720 --> 00:07:59.399
<v Speaker 1>It makes you wonder, though, what other applications could benefit

173
00:07:59.399 --> 00:08:03.839
<v Speaker 1>from these techechniques for minimizing and even exploiting leakage.

174
00:08:04.000 --> 00:08:07.759
<v Speaker 2>Well, that's a great question. We've talked about the BFSK transmitter,

175
00:08:08.040 --> 00:08:11.279
<v Speaker 2>but there's so many other possibilities. For example, have you

176
00:08:11.319 --> 00:08:13.360
<v Speaker 2>considered medical implants.

177
00:08:13.160 --> 00:08:18.120
<v Speaker 1>Oh yeah, like pacemakers or insulin pumps. Extending the battery

178
00:08:18.160 --> 00:08:19.720
<v Speaker 1>life of those could be huge.

179
00:08:19.839 --> 00:08:22.639
<v Speaker 2>Exactly, it could be life changing. People wouldn't need risky

180
00:08:22.680 --> 00:08:24.959
<v Speaker 2>surgeries just to replace batteries, that's right.

181
00:08:25.240 --> 00:08:27.759
<v Speaker 1>And what about wearable sensors. Those are getting more and

182
00:08:27.800 --> 00:08:29.079
<v Speaker 1>more popular these days.

183
00:08:29.120 --> 00:08:32.399
<v Speaker 2>Absolutely, imagine a smart watch or fitness tracker that never

184
00:08:32.440 --> 00:08:33.120
<v Speaker 2>needs charging.

185
00:08:33.320 --> 00:08:36.240
<v Speaker 1>Yeah, that would be amazing. No more worrying about my

186
00:08:36.320 --> 00:08:37.440
<v Speaker 1>watch dying in the middle.

187
00:08:37.480 --> 00:08:40.759
<v Speaker 2>Level workout, right, And it's not just consumer devices either.

188
00:08:40.840 --> 00:08:42.240
<v Speaker 2>Remember we talked about data centers.

189
00:08:42.320 --> 00:08:44.879
<v Speaker 1>Oh yeah, those are massive energy hogs they are.

190
00:08:44.919 --> 00:08:48.639
<v Speaker 2>And as our reliance on cloud computing grows, their energy

191
00:08:48.639 --> 00:08:49.759
<v Speaker 2>consumption is only going to.

192
00:08:49.720 --> 00:08:52.759
<v Speaker 1>Increase, so sub threshold design could be a key part

193
00:08:52.759 --> 00:08:53.960
<v Speaker 1>of making them more efficient.

194
00:08:54.159 --> 00:08:57.960
<v Speaker 2>Absolutely, we could reduce their environmental impact and their operating

195
00:08:58.000 --> 00:08:59.519
<v Speaker 2>costs significantly.

196
00:08:59.799 --> 00:09:02.399
<v Speaker 1>It's incredible to think that something as tiny as a

197
00:09:02.519 --> 00:09:06.039
<v Speaker 1>leaky transistor could have such a big impact on a

198
00:09:06.080 --> 00:09:07.080
<v Speaker 1>global scale.

199
00:09:07.320 --> 00:09:09.919
<v Speaker 2>It really highlights the importance of innovation in this field.

200
00:09:10.200 --> 00:09:12.960
<v Speaker 1>This whole deep dive has been a real eye opener

201
00:09:13.000 --> 00:09:16.879
<v Speaker 1>for me. We've gone from this seemingly annoying problem of

202
00:09:17.000 --> 00:09:20.840
<v Speaker 1>leakage power to exploring cutting edge techniques that are pushing

203
00:09:20.879 --> 00:09:22.759
<v Speaker 1>the boundaries of energy efficiency.

204
00:09:23.240 --> 00:09:26.440
<v Speaker 2>It's a great example of how challenging conventional thinking can

205
00:09:26.559 --> 00:09:29.480
<v Speaker 2>lead to some truly groundbreaking solutions.

206
00:09:29.639 --> 00:09:32.279
<v Speaker 1>It makes you wonder what other hidden opportunities might be

207
00:09:32.360 --> 00:09:35.159
<v Speaker 1>lurking out there in the world of electronics, just waiting

208
00:09:35.200 --> 00:09:35.960
<v Speaker 1>to be discovered.

209
00:09:36.159 --> 00:09:38.080
<v Speaker 2>There are definitely more secrets to uncover.

210
00:09:38.240 --> 00:09:41.000
<v Speaker 1>Well. If this episode has sparked your curiosity about chips

211
00:09:41.000 --> 00:09:43.120
<v Speaker 1>and circuits, I encourage you to dive deeper.

212
00:09:43.240 --> 00:09:47.039
<v Speaker 2>Absolutely. Doctor Cottrey's book is a fantastic resource if you

213
00:09:47.039 --> 00:09:48.879
<v Speaker 2>want to explore the technical details in.

214
00:09:48.840 --> 00:09:51.559
<v Speaker 1>More depth, definitely check it out and Remember, the pursuit

215
00:09:51.559 --> 00:09:55.279
<v Speaker 1>of energy efficiency is about more than just extending battery life.

216
00:09:55.360 --> 00:09:58.799
<v Speaker 2>It's about building a more sustainable future for electronics and

217
00:09:58.919 --> 00:10:00.600
<v Speaker 2>reducing our impact on the planet.

218
00:10:00.720 --> 00:10:03.480
<v Speaker 1>That's a great point. So until our next deep dive,

219
00:10:03.759 --> 00:10:07.080
<v Speaker 1>keep those minds curious and those questions flowing. There's always

220
00:10:07.200 --> 00:10:08.720
<v Speaker 1>more to learn, always.

221
00:10:08.879 --> 00:10:11.240
<v Speaker 2>Yeah. So plas they're like, Okay, you can think of

222
00:10:11.279 --> 00:10:15.120
<v Speaker 2>them like a grid, and you connect input signals to

223
00:10:15.240 --> 00:10:20.399
<v Speaker 2>output signals through through a network of transistors, okay, and

224
00:10:20.519 --> 00:10:23.679
<v Speaker 2>by choosing which connections you make, you basically program the PLA.

225
00:10:24.039 --> 00:10:26.559
<v Speaker 1>Oh so it's programmable, yeah, okay, and.

226
00:10:26.519 --> 00:10:28.519
<v Speaker 2>It can do all sorts of different logic functions.

227
00:10:28.639 --> 00:10:30.759
<v Speaker 1>It's kind of like a like a logic lego set, right,

228
00:10:30.799 --> 00:10:32.919
<v Speaker 1>you just rearrange the pieces exactly.

229
00:10:33.080 --> 00:10:34.879
<v Speaker 2>That's a really good way to put it. And what

230
00:10:34.960 --> 00:10:38.000
<v Speaker 2>makes plas really well suited for it for the sub

231
00:10:38.039 --> 00:10:43.039
<v Speaker 2>threshold design is because they're so structured. Okay, they're predictable, right.

232
00:10:42.919 --> 00:10:44.960
<v Speaker 1>So even though the transistors are kind of doing their

233
00:10:44.960 --> 00:10:48.279
<v Speaker 1>own thing, they're leaky leaky, Yeah, but the PLA as

234
00:10:48.320 --> 00:10:51.600
<v Speaker 1>a whole is still like behaving the way it's supposed to.

235
00:10:51.960 --> 00:10:56.039
<v Speaker 2>Exactly. You need that predictability in sub threscial design makes sense.

236
00:10:56.720 --> 00:10:59.919
<v Speaker 1>So what about that adaptive body biasing?

237
00:11:00.240 --> 00:11:02.039
<v Speaker 2>Oh right, yeah, you said it.

238
00:11:01.960 --> 00:11:04.039
<v Speaker 1>Was like a feedback mechanism.

239
00:11:04.440 --> 00:11:04.799
<v Speaker 2>It is.

240
00:11:05.000 --> 00:11:08.080
<v Speaker 1>Does it like have tiny sensors that measure the temperature

241
00:11:08.120 --> 00:11:08.759
<v Speaker 1>and voltage?

242
00:11:08.960 --> 00:11:12.519
<v Speaker 2>Well, not exactly sensors in the traditional sense, okay, but

243
00:11:12.600 --> 00:11:14.639
<v Speaker 2>it's it's really quite clever. How it works.

244
00:11:14.919 --> 00:11:15.759
<v Speaker 1>How does it work? Then?

245
00:11:15.919 --> 00:11:19.320
<v Speaker 2>So basically, there's a special reference.

246
00:11:18.919 --> 00:11:21.159
<v Speaker 1>PLA reference PLA, Yeah.

247
00:11:21.039 --> 00:11:23.200
<v Speaker 2>It's like a representative for all the other plas in

248
00:11:23.200 --> 00:11:28.039
<v Speaker 2>the circuit, okay, And its delay is constantly being.

249
00:11:27.879 --> 00:11:31.919
<v Speaker 1>Measured measured against what a reference clock signal, A clock signal, okay.

250
00:11:32.120 --> 00:11:35.200
<v Speaker 2>And if the delay starts to drift, uh huh, it

251
00:11:35.320 --> 00:11:36.279
<v Speaker 2>means something's off.

252
00:11:36.399 --> 00:11:38.000
<v Speaker 1>So it's like a canary in a coal miner.

253
00:11:38.080 --> 00:11:41.240
<v Speaker 2>Yeah, exactly. And there's this thing called a phase detector

254
00:11:41.320 --> 00:11:45.720
<v Speaker 2>phase detector which compares the delay of that reference PLA

255
00:11:45.960 --> 00:11:47.200
<v Speaker 2>to the clock signal.

256
00:11:47.279 --> 00:11:47.440
<v Speaker 1>Hm.

257
00:11:48.159 --> 00:11:51.360
<v Speaker 2>And if there's a mismatch, yeah, it triggers another circuit called.

258
00:11:51.159 --> 00:11:53.519
<v Speaker 1>A charge pump a charge pump, yeah.

259
00:11:53.559 --> 00:11:58.159
<v Speaker 2>And the charge pump either adds or removes charge from

260
00:11:58.279 --> 00:12:01.120
<v Speaker 2>from the substrate. Of the transition from the substrate, which

261
00:12:01.159 --> 00:12:03.320
<v Speaker 2>is also called the body of the transistort. Oh okay,

262
00:12:03.399 --> 00:12:09.039
<v Speaker 2>and that adjustment to the charge actually tweaks the threshold voltage, so.

263
00:12:08.960 --> 00:12:11.399
<v Speaker 1>They're fine tuning that voltage exactly.

264
00:12:11.720 --> 00:12:14.559
<v Speaker 2>So even with all those variations in temperature and voltage

265
00:12:14.600 --> 00:12:17.320
<v Speaker 2>and manufacturing, right, they can keep the circuit running smoothly.

266
00:12:17.480 --> 00:12:21.639
<v Speaker 1>That's wild. So it's like a self tuning system, it is. Yeah,

267
00:12:21.759 --> 00:12:25.919
<v Speaker 1>that's incredible, it really is. It makes you appreciate the

268
00:12:26.159 --> 00:12:29.600
<v Speaker 1>ingenuity of these engineers. You know, they're working at such

269
00:12:29.600 --> 00:12:32.440
<v Speaker 1>a tiny scale, it's amazing, but they're able to create

270
00:12:32.480 --> 00:12:34.559
<v Speaker 1>these really complex systems.

271
00:12:34.679 --> 00:12:35.519
<v Speaker 2>Yeah, and they work.

272
00:12:35.639 --> 00:12:38.960
<v Speaker 1>They work, that's the amazing part. So we've talked about

273
00:12:39.039 --> 00:12:43.840
<v Speaker 1>the BFSK transmitter, but what other applications could benefit from

274
00:12:43.840 --> 00:12:44.679
<v Speaker 1>these techniques.

275
00:12:45.120 --> 00:12:48.639
<v Speaker 2>Well, let's think about it. What about something like medical implants.

276
00:12:49.440 --> 00:12:52.000
<v Speaker 2>Oh yeah, medical implants, those are a great example.

277
00:12:52.120 --> 00:12:55.080
<v Speaker 1>Yeah, things like pacemakers and insulin pumps.

278
00:12:54.840 --> 00:12:57.919
<v Speaker 2>Right exactly. Yeah, imagine if you could extend the battery

279
00:12:57.960 --> 00:13:01.279
<v Speaker 2>life of those it would be huge huj Yeah, people

280
00:13:01.279 --> 00:13:03.559
<v Speaker 2>wouldn't have to go through those risky surgery so often

281
00:13:03.639 --> 00:13:04.759
<v Speaker 2>just to replace batteries.

282
00:13:05.000 --> 00:13:06.679
<v Speaker 1>Yeah, that's a really good point. And then there's all

283
00:13:06.679 --> 00:13:08.399
<v Speaker 1>those wearable sensors that are becoming so.

284
00:13:08.440 --> 00:13:10.960
<v Speaker 2>Popular, right, fitness trackers, smart watches.

285
00:13:11.000 --> 00:13:12.759
<v Speaker 1>Imagine if those never needed charging.

286
00:13:13.080 --> 00:13:14.840
<v Speaker 2>Oh that would be amazing.

287
00:13:14.879 --> 00:13:16.600
<v Speaker 1>I would never have to take mine on. It would

288
00:13:16.639 --> 00:13:20.120
<v Speaker 1>be so convenient, and wouldn't it be great for you know,

289
00:13:20.399 --> 00:13:22.639
<v Speaker 1>like medical applications too.

290
00:13:22.559 --> 00:13:26.360
<v Speaker 2>Absolutely constant monitoring without worrying about the battery dying.

291
00:13:26.559 --> 00:13:29.279
<v Speaker 1>Yeah, that's a great point. And it's not just limited

292
00:13:29.320 --> 00:13:30.840
<v Speaker 1>to those small devices either.

293
00:13:30.919 --> 00:13:31.080
<v Speaker 2>Right.

294
00:13:31.559 --> 00:13:33.120
<v Speaker 1>Remember we talked about data centers.

295
00:13:33.240 --> 00:13:35.399
<v Speaker 2>Oh yeah, those things use a ton of energy and.

296
00:13:35.360 --> 00:13:37.320
<v Speaker 1>It's only going to get worse as we rely more

297
00:13:37.360 --> 00:13:38.559
<v Speaker 1>and more on cloud computing.

298
00:13:38.720 --> 00:13:42.159
<v Speaker 2>Definitely, sub threshold design could be a key part of

299
00:13:42.200 --> 00:13:43.679
<v Speaker 2>making them more energy efficient.

300
00:13:43.879 --> 00:13:47.320
<v Speaker 1>Yeah, imagine the impact on the environment and they're operating costs.

301
00:13:47.559 --> 00:13:50.159
<v Speaker 1>It's incredible to think about. It is all because of

302
00:13:50.159 --> 00:13:53.080
<v Speaker 1>these tiny, leaky transistors.

303
00:13:53.159 --> 00:13:55.879
<v Speaker 2>It just goes to show that sometimes the biggest solutions

304
00:13:55.919 --> 00:13:57.679
<v Speaker 2>come from thinking outside the box.

305
00:13:58.159 --> 00:14:01.360
<v Speaker 1>Yeah, leakage was always seen as a problem to be solved, right,

306
00:14:01.639 --> 00:14:03.360
<v Speaker 1>but now we're learning to harness.

307
00:14:03.039 --> 00:14:04.480
<v Speaker 2>It and use it to our advantage.

308
00:14:04.519 --> 00:14:07.200
<v Speaker 1>This whole deep dive has been amazing me too. We've

309
00:14:07.240 --> 00:14:12.159
<v Speaker 1>gone from understanding this annoying problem to exploring these incredible

310
00:14:12.200 --> 00:14:15.320
<v Speaker 1>techniques that are changing how we think about energy.

311
00:14:15.080 --> 00:14:16.919
<v Speaker 2>Efficiency is really exciting stuff.

312
00:14:17.159 --> 00:14:20.159
<v Speaker 1>It makes you wonder what other possibilities are out there

313
00:14:20.200 --> 00:14:21.720
<v Speaker 1>just waiting to be discovered.

314
00:14:22.039 --> 00:14:23.279
<v Speaker 2>I think there's a lot more to come.

315
00:14:23.879 --> 00:14:26.639
<v Speaker 1>Can't wait to see what the future holds. While that

316
00:14:26.720 --> 00:14:30.360
<v Speaker 1>about wraps up our deep dive into minimizing and exploiting

317
00:14:30.440 --> 00:14:34.240
<v Speaker 1>leakage in VLSI design, thanks for having me. It was

318
00:14:34.279 --> 00:14:36.679
<v Speaker 1>great having you on. If you want to learn more

319
00:14:36.720 --> 00:14:40.360
<v Speaker 1>about this fascinating topic, I definitely recommend checking out doctor

320
00:14:40.440 --> 00:14:41.240
<v Speaker 1>Kotree's book.

321
00:14:41.360 --> 00:14:42.440
<v Speaker 2>Yeah, it's a great resource.

322
00:14:42.519 --> 00:14:45.600
<v Speaker 1>It's full of insights and technical details. And remember, this

323
00:14:45.679 --> 00:14:48.919
<v Speaker 1>pursuit of energy efficiency is about more than just making

324
00:14:48.919 --> 00:14:52.639
<v Speaker 1>our batteries last longer. It's about creating a more sustainable

325
00:14:52.679 --> 00:14:56.600
<v Speaker 1>future for electronics alutely and minimizing our impact on the planet.

326
00:14:56.679 --> 00:14:58.080
<v Speaker 2>Couldn't have said it better myself.

327
00:14:58.360 --> 00:15:01.639
<v Speaker 1>So until our next deep dive, keep those minds curious,

328
00:15:02.120 --> 00:15:05.240
<v Speaker 1>keep asking questions, and keep learning. See next time.