news

Perhaps after Ohm’s law, the second most famous law in electronics is Moore’s law: The number of transistors that can be manufactured on an integrated circuit doubles every two years or so. Since the physical size of the chip remains roughly the same, this means that individual transistors will become smaller over time. We have begun to expect a new generation of chips with smaller feature sizes to appear at a normal speed, but what is the point of making things smaller? Does smaller always mean better?
In the past century, electronic engineering has made tremendous progress. In the 1920s, the most advanced AM radios consisted of several vacuum tubes, several huge inductors, capacitors and resistors, dozens of meters of wires used as antennas, and a large set of batteries to power the entire device. Today, you can Listen to more than a dozen music streaming services on the device in your pocket, and you can do more. But miniaturization is not just for portability: it is absolutely necessary to achieve the performance we expect from our devices today.
One obvious benefit of smaller components is that they allow you to include more functionality in the same volume. This is especially important for digital circuits: more components means you can do more processing in the same amount of time. For example, in theory, the amount of information processed by a 64-bit processor is eight times that of an 8-bit CPU running at the same clock frequency. But it also requires eight times as many components: registers, adders, buses, etc. are all eight times larger. So you either need a chip that is eight times larger, or you need a transistor that is eight times smaller.
The same is true for memory chips: By making smaller transistors, you have more storage space in the same volume. The pixels in most displays today are made of thin film transistors, so it makes sense to scale them down and achieve higher resolutions. However, the smaller the transistor, the better, and there is another crucial reason: their performance is greatly improved. But why exactly?
Whenever you make a transistor, it will provide some additional components for free. Each terminal has a resistor in series. Any object that carries current also has self-inductance. Finally, there is a capacitance between any two conductors facing each other. All these effects consume power and slow down the speed of the transistor. Parasitic capacitances are particularly troublesome: transistors need to be charged and discharged each time they are turned on or off, which requires time and current from the power supply.
The capacitance between two conductors is a function of their physical size: a smaller size means a smaller capacitance. And because smaller capacitors mean higher speeds and lower power, smaller transistors can run at higher clock frequencies and dissipate less heat in doing so.
As you shrink the size of transistors, capacitance is not the only effect that changes: there are many strange quantum mechanical effects that are not obvious for larger devices. However, generally speaking, making transistors smaller will make them faster. But electronic products are more than just transistors. When you scale down other components, how do they perform?
Generally speaking, passive components such as resistors, capacitors, and inductors will not get better when they get smaller: in many ways, they will get worse. Therefore, the miniaturization of these components is mainly to be able to compress them into a smaller volume, thereby saving PCB space.
The size of the resistor can be reduced without causing too much loss. The resistance of a piece of material is given by, where l is the length, A is the cross-sectional area, and ρ is the resistivity of the material. You can simply reduce the length and cross-section, and end up with a physically smaller resistor, but still having the same resistance. The only disadvantage is that when dissipating the same power, physically smaller resistors will generate more heat than larger resistors. Therefore, small resistors can only be used in low-power circuits. This table shows how the maximum power rating of SMD resistors decreases as their size decreases.
Today, the smallest resistor you can buy is the metric 03015 size (0.3 mm x 0.15 mm). Their rated power is only 20 mW and are only used for circuits that dissipate very little power and are extremely limited in size. A smaller metric 0201 package (0.2 mm x 0.1 mm) has been released, but has not yet been put into production. But even if they do appear in the manufacturer’s catalog, don’t expect them to be everywhere: most pick and place robots are not accurate enough to handle them, so they may still be niche products.
Capacitors can also be scaled down, but this will reduce their capacitance. The formula for calculating the capacitance of a shunt capacitor is, where A is the area of ​​the board, d is the distance between them, and ε is the dielectric constant (the property of the intermediate material). If the capacitor (basically a flat device) is miniaturized, the area must be reduced, thereby reducing the capacitance. If you still want to pack a lot of nafara in a small volume, the only option is to stack several layers together. Due to advances in materials and manufacturing, which have also made thin films (small d) and special dielectrics (with larger ε) possible, the size of capacitors has shrunk significantly in the past few decades.
The smallest capacitor available today is in an ultra-small metric 0201 package: only 0.25 mm x 0.125 mm. Their capacitance is limited to the still useful 100 nF, and the maximum operating voltage is 6.3 V. Also, these packages are very small and require advanced equipment to handle them, limiting their widespread adoption.
For inductors, the story is a bit tricky. The inductance of a straight coil is given by, where N is the number of turns, A is the cross-sectional area of ​​the coil, l is its length, and μ is the material constant (permeability). If all dimensions are reduced by half, the inductance will also be reduced by half. However, the resistance of the wire remains the same: this is because the length and cross-section of the wire are reduced to a quarter of its original value. This means that you end up with the same resistance in half of the inductance, so you halve the quality (Q) factor of the coil.
The smallest commercially available discrete inductor adopts the inch size 01005 (0.4 mm x 0.2 mm). These are as high as 56 nH and have a resistance of a few ohms. Inductors in an ultra-small metric 0201 package were released in 2014, but apparently they have never been introduced to the market.
The physical limitations of inductors have been solved by using a phenomenon called dynamic inductance, which can be observed in coils made of graphene. But even so, if it can be manufactured in a commercially viable way, it may increase by 50%. Finally, the coil cannot be miniaturized well. However, if your circuit is operating at high frequencies, this is not necessarily a problem. If your signal is in the GHz range, a few nH coils are usually sufficient.
This brings us to another thing that has been miniaturized in the past century but you may not notice immediately: the wavelength we use for communication. Early radio broadcasts used a medium-wave AM frequency of about 1 MHz with a wavelength of about 300 meters. The FM frequency band centered at 100 MHz or 3 meters became popular around the 1960s, and today we mainly use 4G communications around 1 or 2 GHz (about 20 cm). Higher frequencies mean more information transmission capacity. It is because of miniaturization that we have cheap, reliable and energy-saving radios that work on these frequencies.
Shrinking wavelengths can shrink antennas because their size is directly related to the frequency they need to transmit or receive. Today’s mobile phones do not need long protruding antennas, thanks to their dedicated communication at GHz frequencies, for which the antenna only needs to be about one centimeter long. This is why most mobile phones that still contain FM receivers require you to plug in the earphones before use: the radio needs to use the earphone’s wire as an antenna in order to get enough signal strength from those one-meter long waves.
As for the circuits connected to our miniature antennas, when they are smaller, they actually become easier to make. This is not only because transistors have become faster, but also because transmission line effects are no longer an issue. In short, when the length of a wire exceeds one-tenth of the wavelength, you need to consider the phase shift along its length when designing the circuit. At 2.4 GHz, this means that only one centimeter of wire has affected your circuit; if you solder discrete components together, it is a headache, but if you lay out the circuit on a few square millimeters, it is not a problem.
Predicting the demise of Moore’s Law, or showing that these predictions are wrong again and again, has become a recurring theme in the science and technology journalism. The fact remains that Intel, Samsung, and TSMC, the three competitors who are still at the forefront of the game, continue to compress more features per square micrometer, and plan to introduce several generations of improved chips in the future. Even though the progress they have made at each step may not be as great as two decades ago, the miniaturization of transistors continues.
However, for discrete components, we seem to have reached a natural limit: making them smaller does not improve their performance, and the smallest components currently available are smaller than most use cases require. It seems that there is no Moore’s Law for discrete devices, but if there is Moore’s Law, we would love to see how much one person can push the SMD soldering challenge.
I have always wanted to take a picture of a PTH resistor I used in the 1970s, and put an SMD resistor on it, just like I am swapping in/out now. My goal is to make my brothers and sisters (none of them are electronic products) how much change, including I can even see the parts of my work, (as my eyesight is getting worse, my hands are getting worse Trembling).
I like to say, is it together or not. I really hate “improve, get better.” Sometimes your layout works well, but you can no longer get parts. What the hell is that? . A good concept is a good concept, and it is better to keep it as it is, rather than improve it for no reason. Gantt
“The fact remains that the three companies Intel, Samsung and TSMC are still competing at the forefront of this game, constantly squeezing out more features per square micrometer,”
Electronic components are large and expensive. In 1971, the average family had only a few radios, a stereo and a TV. By 1976, computers, calculators, digital clocks and watches had come out, which were small and inexpensive for consumers.
Some miniaturization comes from design. Operational amplifiers allow the use of gyrators, which can replace large inductors in some cases. Active filters also eliminate inductors.
Larger components do promote other things: the minimization of the circuit, that is, trying to use the fewest components to make the circuit work. Today, we don’t care so much. Need something to reverse the signal? Take an operational amplifier. Do you need a state machine? Take an mpu. etc. The components today are really small, but there are actually many components inside. So basically your circuit size increases and power consumption increases. A transistor used to invert a signal uses less power to accomplish the same job than an operational amplifier. But then again, miniaturization will take care of the use of power. It’s just that innovation has gone in a different direction.
You really missed some of the biggest benefits/reasons of reduced size: reduced package parasitics and increased power handling (which seems counterintuitive).
From a practical point of view, once the feature size reaches about 0.25u, you will reach the GHz level, at which time the large SOP package begins to produce the largest* effect. Long bonding wires and those leads will eventually kill you.
At this point, QFN/BGA packages have greatly improved in terms of performance. In addition, when you mount the package flat like this, you end up with *significantly* better thermal performance and exposed pads.
In addition, Intel, Samsung, and TSMC will certainly play an important role, but ASML may be much more important in this list. Of course, this may not apply to the passive voice…
It’s not just about reducing silicon costs through next-generation process nodes. Other things, such as bags. Smaller packages require less materials and wcsp or even less. Smaller packages, smaller PCBs or modules, etc.
I often see some catalog products, where the only driving factor is cost reduction. MHz/memory size is the same, SOC function and pin arrangement are the same. We may use new technologies to reduce power consumption (usually this is not free, so there must be some competitive advantages that customers care about)
One of the advantages of large components is the anti-radiation material. Tiny transistors are more susceptible to the effects of cosmic rays, in this important situation. For example, in space and even high-altitude observatories.
I did not see a major reason for speed increase. The signal speed is approximately 8 inches per nanosecond. So just by reducing the size, faster chips are possible.
You may want to check your own mathematics by calculating the difference in propagation delay due to packaging changes and reduced cycles (1/frequency). That is to reduce the delay/period of factions. You will find that it does not even show up as a rounding factor.
One thing I want to add is that many ICs, especially older designs and analog chips, are not actually downsized, at least internally. Due to improvements in automated manufacturing, packages have become smaller, but that is because DIP packages usually have a lot of remaining space inside, not because transistors etc. have become smaller.
In addition to the problem of making the robot accurate enough to actually handle tiny components in high-speed pick-and-place applications, another issue is reliably welding tiny components. Especially when you still need larger components due to power/capacity requirements. Using special solder paste, special step solder paste templates (apply a small amount of solder paste where needed, but still provide enough solder paste for large components) began to become very expensive. So I think there is a plateau, and further miniaturization at the circuit board level is just a costly and feasible way. At this point, you might as well do more integration at the silicon wafer level and simplify the number of discrete components to an absolute minimum.
You will see this on your phone. Around 1995, I bought some early mobile phones in garage sales for a few dollars each. Most ICs are through-hole. Recognizable CPU and NE570 compander, large reusable IC.
Then I ended up with some updated handheld phones. There are very few components and almost nothing familiar. In a small number of ICs, not only the density is higher, but also a new design (see SDR) is adopted, which eliminates most of the discrete components that were previously indispensable.
> (Apply a small amount of solder paste where needed, but still provide enough solder paste for large components)
Hey, I imagined the “3D/Wave” template to solve this problem: thinner where the smallest components are, and thicker where the power circuit is.
Nowadays, SMT components are very small, you can use real discrete components (not 74xx and other garbage) to design your own CPU and print it on the PCB. Sprinkle it with LED, you can see it working in real time.
Over the years, I certainly appreciate the rapid development of complex and small components. They provide tremendous progress, but at the same time they add a new level of complexity to the iterative process of prototyping.
The adjustment and simulation speed of analog circuits is much faster than what you do in the laboratory. As the frequency of digital circuits rises, the PCB becomes part of the assembly. For example, transmission line effects, propagation delay. Prototyping of any cutting-edge technology is best spent on completing the design correctly, rather than making adjustments in the laboratory.
As for hobby items, evaluation. Circuit boards and modules are a solution to shrinking components and pre-testing modules.
This may make things lose “fun”, but I think getting your project to work for the first time may be more meaningful because of work or hobbies.
I have been converting some designs from through-hole to SMD. Make cheaper products, but it’s not fun to build prototypes by hand. One small mistake: “parallel place” should be read as “parallel plate”.
No. After a system wins, archaeologists will still be confused by its findings. Who knows, maybe in the 23rd century, the Planetary Alliance will adopt a new system…
I could not agree more. What is the size of 0603? Of course, keeping 0603 as the imperial size and “calling” the 0603 metric size 0604 (or 0602) is not that difficult, even if it may be technically incorrect (ie: actual matching size-not that way) anyway. Strict), but at least everyone will know what technology you are talking about (metric/imperial)!
“Generally speaking, passive components such as resistors, capacitors, and inductors won’t get better if you make them smaller.”