As an electrical engineer who has been designing / building analog and digital circuits for work since the late 80's, I'd have to say that the main reason for going binary boils down to power - at least with transistors. The MOSFETs in modern computers consume very little power when they are "ON" (1) or "OFF" (0). When ON the voltage is essentially at the positive power supply rail, and when off they are at ground - a voltage swing that is nearly 100% of the power supply rail. Any time they are between those 2 voltages, (when switching between states for example,) they consume many times more power. It is entirely possible to build transistor circuits with multiple logic levels, but you will suffer a large increase in power when doing so. However, switching between those states will be faster - turning fully ON (saturating) or OFF takes extra time. CRAY computers took advantage of ECL (Emitter-Coupled Logic) and their faster speed by using bipolar transistors that were always "on," i.e. conducting current. The 1 and 0 states were still differentiated by high and low voltages, but their swing was only about 16% of the power rail. This led to very fast computers for Cray, but also required those machines to use exotic and expensive cooling systems to keep running. Over time however, MOSFETs have gotten much smaller and faster - so much so that the ability to use MANY more of them for the same amount of power greatly overcomes the speed advantage of using ECL or other non-saturating logic. Arguably there is another low-power state MOSFETs could use - the Tri-state output - it is neither at the high or low voltage, but rather disconnected from the line. This however would take more transistors at each input and output to decode and encode the signal. And this system is already in use in computers - but it is used for allowing multiple devices to share the same bus (RAM, for example). As far as I know, no one has ever found it advantageous for performing logic or arithmetic as part of the data processing tasks.
as someone who understands this stuff why binary tho wouldn't it be more efficient to use more numbers based on different voltages. while it would uproot an entire system everything is built on wouldnt it create a faster computer while more complicated wouldnt it allow faster information travel. instead o 10010 youd just say 4 or 9 or even a 2 digit number that would relay the same information im just asking.
@@learn905he just explained that. He said you CAN do this stuff and it IS faster, but at a much greater power requirement. He also said that since mosfets have become smaller and faster, you’re really not saving much time or efficiency by switching to a non-binary system. It’s much more effective, at this time anyway, to remain in a binary system.
This series of videos is truly great. I absolutely love listening to Professor Brailsford. I'm a unix and storage guy. I got into this because i loved the technology, but after a while it gets to be a bit of a drag. Watching these videos helped me reignite the passion that made me get into this field. I love it!
What doesn't come out here is that the process of doing digital electronic mathematics with anything other than two states requires far more complex electronics.
+chrisofnottingham I agree. It appears to be a minimalization issue whereby you reduce the number of state-elements (0, 1, etc.) but also maximizing state-space. Clearly, as one UA-cam comment noted by suggesting a 1-based system of zeroes, the next best solution is zeroes and ones. Any extra state-elements added to the system have a progressive lower effect on the "usefulness" of their existence.
Well yes and no. The complexity from having more than two states will be figuring out how to make transistors relay more than two signals. The complexity of the circuit will overall stay the same.
+TJ.Lewis I'm not convinced the complexity does remain the same. Doing addition with more than two states pretty much turns into analogue computing plus multi level quantizing, which is very much more complex, or some kind of multi level logic that is processed using binary logic anyway. Transistors and valves are just naturally binary or continuous. So we can do binary or continuous mathematics fairly easily but it just isn't easy to impose another fixed number of states. Whereas by contrast, gears can in principle work naturally in any base. It is just the nature of the medium.
chrisofnottingham I do not think it will become completely analogue computing for n states if n > 1, only because it is not a continuous sinusoidal signal. For instance binary signals graphed out will have pits and hills because of having two possible states making rectangles. With three states it starts taking a shape of a triangle. To play devils advocate, at some point it will become somewhat sinusoidal, and graphing will have to be done using integration. In the case of complexity, the schematics of a processing chip in relations to logic gates; even tho it is inherently binary by nature does not mean the chip as a whole using n > 2 states cannot function. With logic gates, any signal not a 0 or 1 will be lost or ignored.
+TJ.Lewis By using ternary the logic gates would be much more complex. With binary a logic gate is a very simple circuit with just 2 transistors. With ternary those circuits would be much more complex than what you gain.
Another advantage of representing numbers in a binary format is it greatly simplifies error correction. Find the location of the error - and you automatically know the correct data - it's simply the inverse of the error.
The end of that video reminds me of Jevons Paradox. The basic idea is that by increasing efficiency of something (in an attempt to conserve that resource) you can drop the price which stimulates demand to a degree that more than makes up for the increase in efficiency.
That "bi-quinary" system reminds me of those crazy "Diamond Edge 3D" cards that came out in the 90s - they rendered only in quadrilaterals and not triangles like modern graphics cards. Speaking of which, it'd be cool to see more videos related to GPUs! (Pardon the non sequitur-ish nature of my request.)
+derbuchholzer If there's anything I can do to ease the pain I'll try.... >Sean EDIT: Although, perhaps the fact that they're not involute contributes to the slip...
+messianicrogue tl;dr: Binary is not necessary, and the alternative might even be easier to build for some cases. But it would be a power-hog lick no other, and be quite costly. Is that enough?
+109Rage Is it still a power hog with current technology? I just heard him say it was back in the day, and in the same exact vein as to why we use decimal in standard use, we still use binary.
This guy is fantastic! The way he talks about concepts that are so foreign to most - like it's nothing - is great. I'd love to chat with him even though I'd be lost.
+Matt Maloney Nowadays chip designers for arithmetic units will cheerfully go fully binary and accept the factor of 3.3 for the number of binary digits compared to decimal This is because each binary logic element is simple and is a low-power transistor or capacitor. But transistors weren't invented until the 1950s. Hence, in Tommy Flowers' day in 1943, each logic element was a power-hungry valve. So, if he'd "gone binary" in his counters as well as for his logic elements, the extra power consumption was non-negligible. On the other hand he couldn't "go fully decimal" because he couldn't keep 10 voltages stable and differentiated. Hence the "bi-qui" compromise. For more on this look inside Jack Copeland's "Colossus" book on page 123 (see EXTRA BITS video, linked off this one, for more about this book)
I dont know if I would call it ironic. Although it got a lot of media attention that defect was on one samsung design and only happened to a very small number of phones. I think it would be ironic if samsung continuously produced things that caught on fire to the point of being known as the phones that always catch on fire.
it always boils down to "on" or "off" of an atom. decimal or any other number systems are just interpretations. but if you are able to control the eigenstates of an electron, each atom can represent one trillion-cimal, meet the future computer.
Nice vid, but doesn't actually explain the actual question asked. The reason, short version, is that actually building the electronics to perform mathematical operations, becomes AMAZINGLY more complex, when you have to use more than 2 possible states. Everything else, like keeping voltages apart, can be solved by improving the technology involved, but the complexity of the circuits required, can't.
But now the question becomes this: if you're using multiple analog voltage separations to encode the values 0-4, how do you *store* those values? Currently we can store data because we can turn things on/off, or reorient magnets north/south, etc. How would you do that with 5 possible states, or worse, 10?
***** That works for large-scale machines, but how could you do it fully electronically so it can be miniaturized enough for laptops, tablets, phones, or even just desktop PCs?
+Joe Mills Multiples of 2 is what makes sense given the computer systems we have that uses the storage. If you build a 5 level system, you're of course free to use only 5 of the 8 levels in TLC flash (or build specific 5-level flash). What I replied to was the question of how to store multiple levels in memory, nothing about existing such memory not being binary based.
Michael Tempsch But the entire point of the video was "why don't we use something other than binary?" To say "you could do it by using parts of binary" is redundant.
+IceMetalPunk As stated, you don't have to use 'parts of binary.' You could design a specific 5-level flash memory - the tech is there, currently up to 8 levels. Given this question in your original post: "Currently we can store data because we can turn things on/off, or reorient magnets north/south, etc. How would you do that with 5 possible states, or worse, 10?", I pointed to a current technique that actually does this. I fail to see how the basic technique must be disqualified because it in current implementations uses a number of levels that is a power of 2.
There were some attempts at ternary computers in the Soviet Union, but circumstances of the Cold War led them to be scrapped in favour of stealing binary systems from the West to save resources and research time.
Ternary would be an even better option. About 64% more efficient than binary for storing a random data set. (Base e is perfectly efficient but not much better than 3 and much much harder to apply)
But I do want my guitar amp made with valves :) interestingly it's rare to see the bias in the grid in guitar amps, it's mostly in the cathode and the signal is in the grid. I guess you get better amplification when the signal is the control voltage in the grid, you can make cathode have rather negative voltage.
Storage is one thing, but what about the logic itself, which is inherently binary? That would all have to be converted to base 5 or base 10 or whatever. I feel like that would be incredibly difficult, but maybe I'm missing something.
He just wanted to show of! But on a serious note...to store one bit of data you need one basic memory component, namely flip flop and becuase we need to store, say 99, we need atleast 7 such basic memory blocks. But if somehow we have designed a basic memory block capable of having 10 states for representing 10 binary digits instead of current capability of 2 for binary then we would need only 2 basic memory blocks for storing 99; one for each digit. 7/2 = 3.5 which is slightly greater than 3.22 cause we always have natural number for counting. So our storing capacity and in fact entire binary based digital system would be atleast 3.22 times more bulky than its decimal based counterpart
As mentioned by him later in the same video it is to calculate the maximum number of bits required to represent any 'n' digit number in its binary equivalent. Log 10 base 2 is 3.22. The professor mentions clearly that if you multiply this value 3.22 with the number of digits 'n' of your decimal number and them take the ceiling value you would get the maximum number of bits required for its representation. For Example if you have a 2 digit number say 35 or 48 or 99(the greatest 2 digit number) then you require 2×3.22=6.44 and take its ceiling ie 7. So 7 bits or a 7 bit length binary number can represent all 2 digit decimal numbers. Similarly for 3 digit numbers say 999 the max number of bits for binary representation is 3×3.22=9.66 and then take ceiling of 9.66 ie 10. So a 10 bit binary number can represent all 3 digit decimals. Same for 4 digits and so on. Binary numbers reduce complexity of logic but as you can see increase the circuitry by a lot. For a 2 digit decimal number you have to use 7 times the circuitry for doing the same.
In summary: decimal is only 3.3x more efficient than binary, while being significantly less reliable and harder to implement. It's simply not worth it.
Well you probably wold not get those 3 degrre burns anywhy, because you phone , let alone the batery to run the thing fir more than 2 secoonds, wokd not fin in your pocet if the phone was made with thermionic valves but tge comparison made me smile anyway, thanks for another great video
I wonder... with our current manufacturing and fabricating abilities, is making a decimal computer system still THAT inefficient compared to binary anymore? I mean yeah it might be a little bit, but considering how small we can make things, how efficient on power they are, it has to be somewhat plausible. I'd love to see that as an exploration of our computing abilities to see if perhaps there is a better way to, well, computer, from the ground up.
Why Binary? Because all electronic computers , from the Colossus at Bletchley Park to today's smartphones and tablets, at their most basic level are collections of many on/off switches, there are only two possible states for every switch. To use Base 10, every switch would have to have ten possible states, and the whole system would get horrendously complicated very quickly.
Babbage should have invented the CNC mill. The textile industry had already began to embrace that kind of automation, this would not have been a foreign concept to him.
You could argue that binary is not in fact used for DSL transmission where there can be as many as 15 bits per symbol. They use a combination of voltage levels and phase modulation to encode more bits into each symbol.
In Soviet Russia they made trinary computer, it was much more efficient compared to binary. But unfortunately because money issues they didn't make it better, they started copying west. But the interesting 1 and 0 was they used 1, 0 and -1, which made negative numbers easier to express than in binary. also some calculating stuff was easier
There is *so* much Soviet tech with incredible potentials that we just straight up lost because of the pressures and constraints of the Cold War. I'm not mad the Soviets lost. I'm mad at the science that never happened because they were forced to fight a war instead of spending time and resources on scientific pursuits for the sake of scientific advancement.
Let's say we build a hexadecimal (or any other base>2) computer. For every value we then have to differentiate between 16 distinct 'positions'. If we can build technology that precise then we can also (in most cases) build technology where those hexadecimal numbers are replaced with 4 binary bits 1/4th the 'size' each (or whatever measurement is relevant). Since bits are simpler and easier to read they are readable even at smaller sizes. Since a lot of operations (like logic gates) are naturally done with binary it is easier to build a binary computer. And that is why binary is the favoured number system for electric computers.
Well, we've had analog computers before where precise voltages represent numbers and various circuits combine those voltages in different ways. At some point you have to take a measurement and record a number which will be accurate to whatever precision the machine allows. Logic would work along those lines, but the advantage of binary is the complete lack of ambiguity and the ability to determine the state of any bit with a single voltage threshold.
Can someone explain. When he was talking about Log to the base 2 of 10 = 3.322, does that just give you the maximum number of bits needed for a two digit or any digit number because 10 being a two digit number definitely doesn't need 7 bits.
It's how many bits you need to represent all numbers up to the highest 2 digit number. For the highest 4 digit number (9999) you need 4 * 3.322 = 13.288 = 14 bits.
This needed an example of 0-4 as far as voltage goes. would they still be using +5 -5 and just detecting outputs @ 5,3,0,-3,+5? MUCH more info needed on how they decided it was "stable" i can see computers splitting logical commands from calculator commands but.. i'm really concerned with the latency of transferring data between base 5 and base 2. Also concerned that data storage wasn't mentioned. Please do a follow up.
Why not mention anything else, like balanced ternary? The question "why binary" is not answered in this vid. It's seems more like a trailer for a vid that would actually answer the question, with a couple interesting tangental facts tossed in.
i always thought that binary was arbitrary due to the 2-way logic gates. its AND using both 1, OR if 1 is an actual 1 and 1 is an actual 0. i thought thats why base 2 was being used, instead of base 5 or base 10
I don't claim to know too much about computers, so please tell me if this is a stupid question. But if we were to switch to using decimal rather than binary, could we in theory have a CPU that is 2-3 times faster with the same size chip?
+teehee1604 No, because even though it would take less digits to represent the same numbers, what slows down computers isn't the number of bits - it's not too hard to turn a 32bit cp into a 64bit cpu, and if there were any reason to build 128bit or 256bit computers, we would be doing it. The limiting factor for speed is how fast the gates can switch and how fast signals can go through the chip (ie the speed of light). So even if you replaced your 64bit computer with a 19 digit computer (about equivalent), it still wouldn't be faster.
+Robin Powell There is only one operation where 10 is easiest to compute, and that is raising to powers. You don't have to multiply or figure anything, just bang on another zero.
+Rob Kinney The reason it seems the base ten system makes exponents easier to compute is simply because of the figures we use. In a number system devised for a different base, it would likely appear just as easy
Penny Lane No I just think it sounds cool :/ It _does_ have the lowest radix economy though, which is loosely related to the topic of the video (which is why I chose that specific number to ask about)
+saharahgaiht Why does it look like it takes 3 binary digits to store the number 3? (which binary stores in 2 ). And so far all 3 of those numbers look the same to a circuit. Off, off, still off.
Simon WoodburyForget "0.0" and ".0" mean the same thing. Not just in "unary", but also in actual programming. float zero= .0f; is the exact same as float zero= 0.0f; The value of 0.0 in unary would be zero and (I think) one half. 0.00 would be one third, etc. At this point the really serious flaw becomes apparent, and I've put too much thought into a non functional numbering system.
+Joshua Pearce The most interesting non-standard number system is the factorial one, but it has its difficulties as well. Unary, as is used in tally marks, can't represent 0 or fractions, as it is bijective and every digit (the same symbol) just represents a 1, such as 13(dec.) = 1111111111111(unary) = 1111*111+1 in unary, if 1 is used as the symbol. Even using a lot of symbols for a large base like 60 is more useful.
After hours of discussion with my best friend (we both know programming and electronic on transistor level), we agree that Duodecimal (base 12) is the best for people and for computer Trinary is better than Binary, Trinary have sadly been incorrectly done and everyone is focus on a poor working prototype to say that Trinary don't work while it is superior to Binary, this is the future !
+Demongornot Maybe in the very far future It is possible for a computer run off of an n number of states. But for the technology we have right now, along with the hypothetical Q-bits, I think it is much more feasible to obtain a process of Quaternary system. Instead of using a quantum process, it would have transistors that has 4 levels of state instead of two.
+TJ.Lewis Trinary is already possible, its not limited by technology but by mainstream, people have started with Binary because it was for simple current hard to manage other than 2 electrical state (voltage) and no one really take serious or deep look on trinary while it is one of the most efficient system, more than trinary is less efficient and binary is not efficient enough, floating point precision is actually a mess because binary offer really poor precision for fractions. Quaternary is starting to be too much, exactly as binary is not enough.
+Demongornot We'll what I said was that having a state of existence up to something such as a 4th state system was feasible. But if it was all possible to have any state of existence the higher that state, the better the floating points will be. I say that because of the problem of the recursion and bit limit problem. For simplicity purposes, if a decimal system was possible, that would be the ideal system. I say that because 1x10^n is easier than 1xN^n for N >=1 and n >= 0
It's a shame that all modern computers rely on binary, there are logical advantages to trinary, and many different ways you can implement it in terms of electrical components.
Thank you for this video. A colleague of mine asked me that a while ago. I tried to explain it but I wasn't able to put it into simple enough words. I'll send him this video.
This felt like the first half of a video. Seems like he got to the point of saying "bi-quinary is better than binary" and then didn't deliver a punchline as to why we wound up with binary anyway.
Left as an exercise for the reader. Anyway, it's because modern bits have lower voltage differences, so it's not as feasible to do multiple voltage levels. You can also see that with reduction in efficiency in TLC/QLC memory in SSDs.
Yevgeniy Gorbachev Didn’t he give that as a reason for not using base 10? I thought he was saying Flowers reckoned he could make biquinary work...Perhaps he was just mistaken in that and quickly discovered as much
@@yevgeniygorbachev5152 Voltage has nothing to do with it. It's a matter of efficiently maximizing space because binary requires less transistors to count. 8 transistors can represent up to 256 in binary, because it uses exponential counting, decimal needs 10 transistors to get to 55. Binary Maximum Count: 2+4+8+16+32+64+128=256. Decimal Maximum Count: 1+2+3+4+5+6+7+8+9+10=55
@@yevgeniygorbachev5152 I think your misunderstanding on voltage is you're thinking of electricity circuits vs computing systems. In electrical circuits, 0 = zero power, but 1 can equal different kinds of voltage. In computing a lower voltage to a transistor = 0, any higher voltage = 1 and no voltage the computer is off. It's about counting in powers. The binary counting system is in base-2, which means you reach a new base every two digits or in mathematical terms a new "power" every 2 digits. In decimal, it is done base-10, so you need 10 digits to reach a new power.
Did you notice that they used +5V and -5V? So why not 0V as well? Well, there was actually an experimental computer using balanced ternary (as this system is called) but it was more of emulating it than using it. Transistors are binary only and that's the main reason why we use binary nowadays. However as transistors are very close to hit their physical limits, new methods are developed and these methods (optical, Josephson junction) are in fact ternary. Donald Knuth (father of the analysis of algorithms and author of The Art of Computer Programming, among bazillion other achievements in computer science) predicted that balanced ternary would be the system of the future. (I hope this will be covered in a future Computerfile video.)
I would love to see a balanced ternary video. The logic behind it is just so nice and not all that more difficult really. Equally something I wish was experimented with was ternary or quaternary on hard drives. But it would matter much now since HDD are bssically on there way out. The only really larger density increase recently was SMR, which is horribly slow and only aimed at archival markets. SSD are much cheaper now and sloely catching up on price and density.
+Herve Shango For anyone too lazy to find a place to convert it, this is what it converts to: Did you notice that they used +5V and -5V? So why not 0V as well? Well, there was actually an experimental computer using balanced ternary (as this system is called) but it was more of emulating it than using it. Transistors are binary only and that's the main reason why we use binary nowadays. However as transistors are very close to hit their physical limits, new methods are developed and these methods (optical, Josephson junction) are in fact ternary. Donald Knuth (father of the analysis of algorithms and author of The Art of Computer Programming, among bazillion other achievements in computer science) predicted that balanced ternary would be the system of the future. (I hope this will be covered in a future Computerfile video.
+Jan Sten Adámek Not only did balanced ternary blow my mind, but Knuth also describes Fibonacci counting (very useful in information theory) and factorial counting (as a curiosity).
I want to know more about the non-binary attempts at making computers. There isn't much data online about Dekatron vacuum tubes and the Setun balanced ternary computer.
You should look at the lore yourself, it even explicitly states it's an electronical device ("The RobCo Pip-Boy (Personal Information Processor) is an electronic device manufactured by RobCo Industries."). Lack of transistors doesn't make it mechanical.
How about doing a video about modern vacuum tubes? They are used extensively in communication satellites (traveling wave tubes are pretty much the only amplifier that can reliably be used in a small while offering a small and compact package) and micro-tubes are being developed to be used in cellphones and other microwave communication gear such as Wifi access points so he might get tubes in his mobile sooner than he thinks ::)
+DOGMA1138 They don't really have that much to do with computers, so I don't think they fit here (or people watching computerphile would be interested/have physics under their belt). They are analog amplifiers.
Why has this channel not done a video or even talked about memristors yet? And the computer technological significance of this equipment. I am sure many of your viewers have not even heard about memristors yet. ;-)
This was actually quite interesting. I had heard in my computing course that it was because of the voltage variation, that they couldn't guarantee that say, 4 volts would always actually be 4 volts and wouldn't drift as he says and become 5 or 3 and mess things up but they could easily guarantee it with binary by making it simply any voltage or no voltage. This video showed that there was a lot more to it than that though, I never really thought about the debate while the technology was still being developed, only about why we use it in terms of modern computing.
Ah, another lovely story by the brilliant Professor Brailsford. What voice he has, perfect for story telling. But maybe I'm biased by the fact that I'm American and any proper English accent sounds perfect for story telling.
+Franklin Cerpico What, pray, is a 'proper' English accent? (speaking as someone from the north east of England) -- And I agree, his voice is perfect for story-telling.
Gammel Prutte Well if I had to narrow it down, there's the accent which the Professor has, which I for lack of a better word label as 'proper' in order to distinguish it from a 'cockney' accent. Not to say a hint of 'cockney' isn't nice too, take Michael Cain for example.
The real reason? Computers took a lot of space in the beginning some were entire buildings. You only need 8 transistors to count to 256 in binary, you need 10 to count to 55 in decimal. Binary Maximum Count: 2+4+8+16+32+64+128=256. Decimal Maximum Count: 1+2+3+4+5+6+7+8+9+10=55
This didn't really explain why binary is used now. So there was an alternative base 10 system on the valve computers at Bletchley Park. Why aren't we still using that system then? What happened to it?
Transistor logic happenned. The big difference between transistor logic and valve logic is that the spike in power consumption is only when switching not all the time. Valves are power hogs all the time, transistors only when switching. So simply having a register to hold a value would draw power, so it made sense to have the smallest amount of circuitery. In transistor logic, you can have a lot of static circuitery which will not consume that much power when nothing happens on them. So representing numbers in binary allows for simpler circuits, even if there are more of them in the computer. It's the same thing with relays. Most relay computers also used binary representation (Zuse Z3 even in 1941) because in relays it's also the transition that is costly.
A circuit that can natively store and retrieve one of 10 states would be complicated to build and would include a lot of analog components that aren't efficient to miniaturize on ICs. Two state logic requires so few parts per bit that it makes sense to keep using it, at least with current technology.
Far fetched ...Nonsense...We use 2 states only (binary ones and zeros) because transistors use far less electricity deciding over this two states...If we use 8 different voltages??? than to make processor fast as mobile phone-we'll need 60 KW ..Nonsense...Even as a test this is unachievable cause the voltage threshold will vary too much and thus to much errors...Not possible...
"3 times more circuitry" - not true. To make 10 levels just as reliable as binary will require more hardware anyway. OK, that still _might_ be less but not 3 times.
As an electrical engineer who has been designing / building analog and digital circuits for work since the late 80's, I'd have to say that the main reason for going binary boils down to power - at least with transistors. The MOSFETs in modern computers consume very little power when they are "ON" (1) or "OFF" (0). When ON the voltage is essentially at the positive power supply rail, and when off they are at ground - a voltage swing that is nearly 100% of the power supply rail. Any time they are between those 2 voltages, (when switching between states for example,) they consume many times more power.
It is entirely possible to build transistor circuits with multiple logic levels, but you will suffer a large increase in power when doing so. However, switching between those states will be faster - turning fully ON (saturating) or OFF takes extra time. CRAY computers took advantage of ECL (Emitter-Coupled Logic) and their faster speed by using bipolar transistors that were always "on," i.e. conducting current. The 1 and 0 states were still differentiated by high and low voltages, but their swing was only about 16% of the power rail. This led to very fast computers for Cray, but also required those machines to use exotic and expensive cooling systems to keep running.
Over time however, MOSFETs have gotten much smaller and faster - so much so that the ability to use MANY more of them for the same amount of power greatly overcomes the speed advantage of using ECL or other non-saturating logic.
Arguably there is another low-power state MOSFETs could use - the Tri-state output - it is neither at the high or low voltage, but rather disconnected from the line. This however would take more transistors at each input and output to decode and encode the signal. And this system is already in use in computers - but it is used for allowing multiple devices to share the same bus (RAM, for example). As far as I know, no one has ever found it advantageous for performing logic or arithmetic as part of the data processing tasks.
as someone who understands this stuff why binary tho wouldn't it be more efficient to use more numbers based on different voltages. while it would uproot an entire system everything is built on wouldnt it create a faster computer while more complicated wouldnt it allow faster information travel. instead o 10010 youd just say 4 or 9 or even a 2 digit number that would relay the same information im just asking.
@@learn905he just explained that. He said you CAN do this stuff and it IS faster, but at a much greater power requirement. He also said that since mosfets have become smaller and faster, you’re really not saving much time or efficiency by switching to a non-binary system. It’s much more effective, at this time anyway, to remain in a binary system.
This series of videos is truly great.
I absolutely love listening to Professor Brailsford.
I'm a unix and storage guy. I got into this because i loved the technology, but after a while it gets to be a bit of a drag.
Watching these videos helped me reignite the passion that made me get into this field.
I love it!
Every idiot can count to one
-Bob Widlar
+AvZ „Astatine“ NaV Every idiot can count to 10.
tomlxyz Not with a single bit, you can't
AvZ NaV If you start with 1 you can.
+AvZ “Astatine” NaV "A" in hexadecimal, tadah.
You asked for it
WXVwLCBJIHdpbg==
What doesn't come out here is that the process of doing digital electronic mathematics with anything other than two states requires far more complex electronics.
+chrisofnottingham I agree. It appears to be a minimalization issue whereby you reduce the number of state-elements (0, 1, etc.) but also maximizing state-space. Clearly, as one UA-cam comment noted by suggesting a 1-based system of zeroes, the next best solution is zeroes and ones. Any extra state-elements added to the system have a progressive lower effect on the "usefulness" of their existence.
Well yes and no. The complexity from having more than two states will be figuring out how to make transistors relay more than two signals. The complexity of the circuit will overall stay the same.
+TJ.Lewis I'm not convinced the complexity does remain the same. Doing addition with more than two states pretty much turns into analogue computing plus multi level quantizing, which is very much more complex, or some kind of multi level logic that is processed using binary logic anyway.
Transistors and valves are just naturally binary or continuous. So we can do binary or continuous mathematics fairly easily but it just isn't easy to impose another fixed number of states. Whereas by contrast, gears can in principle work naturally in any base. It is just the nature of the medium.
chrisofnottingham I do not think it will become completely analogue computing for n states if n > 1, only because it is not a continuous sinusoidal signal. For instance binary signals graphed out will have pits and hills because of having two possible states making rectangles. With three states it starts taking a shape of a triangle. To play devils advocate, at some point it will become somewhat sinusoidal, and graphing will have to be done using integration. In the case of complexity, the schematics of a processing chip in relations to logic gates; even tho it is inherently binary by nature does not mean the chip as a whole using n > 2 states cannot function. With logic gates, any signal not a 0 or 1 will be lost or ignored.
+TJ.Lewis By using ternary the logic gates would be much more complex. With binary a logic gate is a very simple circuit with just 2 transistors. With ternary those circuits would be much more complex than what you gain.
Another advantage of representing numbers in a binary format is it greatly simplifies error correction. Find the location of the error - and you automatically know the correct data - it's simply the inverse of the error.
Hamming codes you mean?
The end of that video reminds me of Jevons Paradox. The basic idea is that by increasing efficiency of something (in an attempt to conserve that resource) you can drop the price which stimulates demand to a degree that more than makes up for the increase in efficiency.
That "bi-quinary" system reminds me of those crazy "Diamond Edge 3D" cards that came out in the 90s - they rendered only in quadrilaterals and not triangles like modern graphics cards.
Speaking of which, it'd be cool to see more videos related to GPUs! (Pardon the non sequitur-ish nature of my request.)
Professor Brailsford is awesome! Thank you for introducing him; I only wish I'd had more profs like him.
Interesting video.
But the lack of using involute gears in that animation at 0:53 was a bit painful to watch.
+derbuchholzer If there's anything I can do to ease the pain I'll try.... >Sean
EDIT: Although, perhaps the fact that they're not involute contributes to the slip...
+Computerphile _Anything?_ ( ͡° ͜ʖ ͡°)
+derbuchholzer wow ur so clever plz marry me, bst cmmnt on utube, 10/10, point score > 9000.
+derbuchholzer can you explain to those of us who are wondering why you are getting likes?
+derbuchholzer does it grind your gears?
No idea what was being said for 90% of this video, I understood every word being said, I just don't comprehend any of it.
+messianicrogue
tl;dr: Binary is not necessary, and the alternative might even be easier to build for some cases. But it would be a power-hog lick no other, and be quite costly.
Is that enough?
+109Rage Great summary, thanks!
+messianicrogue Mine was the opposite. His voice is too raspy for me to hear what he's saying.
+messianicrogue
Thats computerphile and numberphile for you
+109Rage Is it still a power hog with current technology? I just heard him say it was back in the day, and in the same exact vein as to why we use decimal in standard use, we still use binary.
This guy is fantastic! The way he talks about concepts that are so foreign to most - like it's nothing - is great. I'd love to chat with him even though I'd be lost.
i only understand like 3% of what this man is saying but i would watch him explain anything
As usual, this is wonderfully insightful. But, does it definitively answer the question, "Why Binary?" -- I must be missing something.
+Matt Maloney
Nowadays chip designers for arithmetic units will cheerfully go fully binary and accept the factor of 3.3 for the number of binary digits compared to decimal This is because each binary logic element is simple and is a low-power transistor or capacitor. But transistors weren't invented until the 1950s. Hence, in Tommy Flowers' day in 1943, each logic element was a power-hungry valve. So, if he'd "gone binary" in his counters as well as for his logic elements, the extra power consumption was non-negligible. On the other hand he couldn't "go fully decimal" because he couldn't keep 10 voltages stable and differentiated. Hence the "bi-qui" compromise. For more on this look inside Jack Copeland's "Colossus" book on page 123 (see EXTRA BITS video, linked off this one, for more about this book)
Professor Brailsford demonstrating a Samsung phone while talking about not wanting third degree burns from his computer is suddenly rather ironic.
that's not binary's fault LOL
feanenatreides xD
I dont know if I would call it ironic. Although it got a lot of media attention that defect was on one samsung design and only happened to a very small number of phones. I think it would be ironic if samsung continuously produced things that caught on fire to the point of being known as the phones that always catch on fire.
+Cole Knapek And it wasn't even a Samsung phone, it was a vendor-supplied battery.
feanenatreides lol.
it always boils down to "on" or "off" of an atom. decimal or any other number systems are just interpretations. but if you are able to control the eigenstates of an electron, each atom can represent one trillion-cimal, meet the future computer.
Professor Brailsford is a treasure. I enjoy his videos so much!
Computerphile. Would you recommend that colossus book? It seems interesting.
+Torrey Braman Professor Brailsford would heartily recommend that book, in fact see the 'extra bits' video for his personal recommendation! >Sean
Awesome! I think ill look into it!
I could listen to this man the whole day… And since I love to binge-watch that, I think I kinda do that.
Mobile device with thermionic valves = Pip Boy
Nice vid, but doesn't actually explain the actual question asked.
The reason, short version, is that actually building the electronics to perform mathematical operations, becomes AMAZINGLY more complex, when you have to use more than 2 possible states. Everything else, like keeping voltages apart, can be solved by improving the technology involved, but the complexity of the circuits required, can't.
But now the question becomes this: if you're using multiple analog voltage separations to encode the values 0-4, how do you *store* those values? Currently we can store data because we can turn things on/off, or reorient magnets north/south, etc. How would you do that with 5 possible states, or worse, 10?
*****
That works for large-scale machines, but how could you do it fully electronically so it can be miniaturized enough for laptops, tablets, phones, or even just desktop PCs?
*****
That's what I figured XD I just wondered why that wasn't addressed in the video, since it's a very important reason to use binary.
+Joe Mills Multiples of 2 is what makes sense given the computer systems we have that uses the storage.
If you build a 5 level system, you're of course free to use only 5 of the 8 levels in TLC flash (or build specific 5-level flash).
What I replied to was the question of how to store multiple levels in memory, nothing about existing such memory not being binary based.
Michael Tempsch
But the entire point of the video was "why don't we use something other than binary?" To say "you could do it by using parts of binary" is redundant.
+IceMetalPunk As stated, you don't have to use 'parts of binary.'
You could design a specific 5-level flash memory - the tech is there, currently up to 8 levels.
Given this question in your original post: "Currently we can store data because we can turn things on/off, or reorient magnets north/south, etc. How would you do that with 5 possible states, or worse, 10?", I pointed to a current technique that actually does this. I fail to see how the basic technique must be disqualified because it in current implementations uses a number of levels that is a power of 2.
I understand the issue with decimal, but why not hexadecimal, or heck, even base 4?
They're powers of 2.
11 x 17" greenbar paper! Nostalgia washes over me.
Did someone find a better alternative so far? or even tried to ? just out of curiosity
There were some attempts at ternary computers in the Soviet Union, but circumstances of the Cold War led them to be scrapped in favour of stealing binary systems from the West to save resources and research time.
Ternary would be an even better option. About 64% more efficient than binary for storing a random data set. (Base e is perfectly efficient but not much better than 3 and much much harder to apply)
But I do want my guitar amp made with valves :) interestingly it's rare to see the bias in the grid in guitar amps, it's mostly in the cathode and the signal is in the grid. I guess you get better amplification when the signal is the control voltage in the grid, you can make cathode have rather negative voltage.
Storage is one thing, but what about the logic itself, which is inherently binary? That would all have to be converted to base 5 or base 10 or whatever. I feel like that would be incredibly difficult, but maybe I'm missing something.
Why did he calculate log10 base 2 .
He just wanted to show of!
But on a serious note...to store one bit of data you need one basic memory component, namely flip flop and becuase we need to store, say 99, we need atleast 7 such basic memory blocks. But if somehow we have designed a basic memory block capable of having 10 states for representing 10 binary digits instead of current capability of 2 for binary then we would need only 2 basic memory blocks for storing 99; one for each digit. 7/2 = 3.5 which is slightly greater than 3.22 cause we always have natural number for counting. So our storing capacity and in fact entire binary based digital system would be atleast 3.22 times more bulky than its decimal based counterpart
As mentioned by him later in the same video it is to calculate the maximum number of bits required to represent any 'n' digit number in its binary equivalent. Log 10 base 2 is 3.22. The professor mentions clearly that if you multiply this value 3.22 with the number of digits 'n' of your decimal number and them take the ceiling value you would get the maximum number of bits required for its representation. For Example if you have a 2 digit number say 35 or 48 or 99(the greatest 2 digit number) then you require 2×3.22=6.44 and take its ceiling ie 7. So 7 bits or a 7 bit length binary number can represent all 2 digit decimal numbers. Similarly for 3 digit numbers say 999 the max number of bits for binary representation is 3×3.22=9.66 and then take ceiling of 9.66 ie 10. So a 10 bit binary number can represent all 3 digit decimals. Same for 4 digits and so on. Binary numbers reduce complexity of logic but as you can see increase the circuitry by a lot. For a 2 digit decimal number you have to use 7 times the circuitry for doing the same.
In summary: decimal is only 3.3x more efficient than binary, while being significantly less reliable and harder to implement. It's simply not worth it.
Well you probably wold not get those 3 degrre burns anywhy, because you phone , let alone the batery to run the thing fir more than 2 secoonds, wokd not fin in your pocet if the phone was made with thermionic valves but tge comparison made me smile anyway, thanks for another great video
one of the best story tellers ever
It would be interesting to see a modern base ten computer.
Mother nature uses base4 for the DNA code. Why not use that?
I wonder... with our current manufacturing and fabricating abilities, is making a decimal computer system still THAT inefficient compared to binary anymore? I mean yeah it might be a little bit, but considering how small we can make things, how efficient on power they are, it has to be somewhat plausible. I'd love to see that as an exploration of our computing abilities to see if perhaps there is a better way to, well, computer, from the ground up.
Why Binary? Because all electronic computers , from the Colossus at Bletchley Park to today's smartphones and tablets, at their most basic level are collections of many on/off switches, there are only two possible states for every switch.
To use Base 10, every switch would have to have ten possible states, and the whole system would get horrendously complicated very quickly.
Babbage should have invented the CNC mill. The textile industry had already began to embrace that kind of automation, this would not have been a foreign concept to him.
You could argue that binary is not in fact used for DSL transmission where there can be as many as 15 bits per symbol. They use a combination of voltage levels and phase modulation to encode more bits into each symbol.
In Soviet Russia they made trinary computer, it was much more efficient compared to binary. But unfortunately because money issues they didn't make it better, they started copying west. But the interesting 1 and 0 was they used 1, 0 and -1, which made negative numbers easier to express than in binary. also some calculating stuff was easier
There is *so* much Soviet tech with incredible potentials that we just straight up lost because of the pressures and constraints of the Cold War.
I'm not mad the Soviets lost. I'm mad at the science that never happened because they were forced to fight a war instead of spending time and resources on scientific pursuits for the sake of scientific advancement.
3:24 I'd also buy one... +Computerphile - Any chance of you branching into merch??
Let's say we build a hexadecimal (or any other base>2) computer. For every value we then have to differentiate between 16 distinct 'positions'. If we can build technology that precise then we can also (in most cases) build technology where those hexadecimal numbers are replaced with 4 binary bits 1/4th the 'size' each (or whatever measurement is relevant). Since bits are simpler and easier to read they are readable even at smaller sizes. Since a lot of operations (like logic gates) are naturally done with binary it is easier to build a binary computer. And that is why binary is the favoured number system for electric computers.
But how would decimal logic gates even function?
Well, we've had analog computers before where precise voltages represent numbers and various circuits combine those voltages in different ways. At some point you have to take a measurement and record a number which will be accurate to whatever precision the machine allows. Logic would work along those lines, but the advantage of binary is the complete lack of ambiguity and the ability to determine the state of any bit with a single voltage threshold.
just move to dozenal logic
Who else came expecting him to talk about ternary computers?
Can someone explain. When he was talking about Log to the base 2 of 10 = 3.322, does that just give you the maximum number of bits needed for a two digit or any digit number because 10 being a two digit number definitely doesn't need 7 bits.
It's how many bits you need to represent all numbers up to the highest 2 digit number. For the highest 4 digit number (9999) you need 4 * 3.322 = 13.288 = 14 bits.
This needed an example of 0-4 as far as voltage goes. would they still be using +5 -5 and just detecting outputs @ 5,3,0,-3,+5? MUCH more info needed on how they decided it was "stable"
i can see computers splitting logical commands from calculator commands but.. i'm really concerned with the latency of transferring data between base 5 and base 2.
Also concerned that data storage wasn't mentioned. Please do a follow up.
Because it is easier than quaternary when constructing logic gates and latches. But solid state storage uses multiple voltage levels.
Correct me if I'm wrong,
But the MLC, TLC and QLC technology is doing kind of the same thing...
Why not mention anything else, like balanced ternary? The question "why binary" is not answered in this vid. It's seems more like a trailer for a vid that would actually answer the question, with a couple interesting tangental facts tossed in.
But what about trinary?
Have you guys done a video on analog computers already?
i always thought that binary was arbitrary due to the 2-way logic gates.
its AND using both 1, OR if 1 is an actual 1 and 1 is an actual 0.
i thought thats why base 2 was being used, instead of base 5 or base 10
I didn't get the end of the video. So why isn't bi-quinary used in modern computers?
Can you guys now do one about excess notation ?! pls ?!?
Binary is easy to build rather than other number systems. is it?
I don't claim to know too much about computers, so please tell me if this is a stupid question. But if we were to switch to using decimal rather than binary, could we in theory have a CPU that is 2-3 times faster with the same size chip?
+teehee1604 No, because even though it would take less digits to represent the same numbers, what slows down computers isn't the number of bits - it's not too hard to turn a 32bit cp into a 64bit cpu, and if there were any reason to build 128bit or 256bit computers, we would be doing it. The limiting factor for speed is how fast the gates can switch and how fast signals can go through the chip (ie the speed of light). So even if you replaced your 64bit computer with a 19 digit computer (about equivalent), it still wouldn't be faster.
Fascinating video. thanks
This whole interview could have been boiled down to one word: transistors.
10 is terrible number to base a system on, it has nothing special about it neither in regards to divisibility nor to how easy it is to compute.
+Robin Powell There is only one operation where 10 is easiest to compute, and that is raising to powers. You don't have to multiply or figure anything, just bang on another zero.
+Rob Kinney The reason it seems the base ten system makes exponents easier to compute is simply because of the figures we use. In a number system devised for a different base, it would likely appear just as easy
I suppose, going up by powers of 2 in base 2 also involves just adding zeros.
+Ryan Cobourn I thought your binary code response is binaryle awesome!
Well why not use hexadecimal it’s shorter than decimal
This guy is so interesting to listen too!
I guess I can call my Marshall tube amp a Marshall thermionic valve guitar amp now.
so is it actually worth it to possibly switch from binary to that 0-9?
How about a complex base number system?
What about base e?
+MMMIK13 using an irrational base doesn't make much sense, at least to my knowledge..
+Neue Ära Using base e would definitely be irrational.
+Neue Ära Of course it doesn't. +MMMIK13 just wanted to show off.
Penny Lane No I just think it sounds cool :/
It _does_ have the lowest radix economy though, which is loosely related to the topic of the video (which is why I chose that specific number to ask about)
+MMMIK13 Please explain how having the "lowest radix economy" makes it an ideal choice.
Behold my invention: Unary. 0 = 1, 00 = 2 and 000 = 3... It consumes half the energy and performs twice as fast as binary. Where's my phd?
+saharahgaiht Why does it look like it takes 3 binary digits to store the number 3? (which binary stores in 2 ). And so far all 3 of those numbers look the same to a circuit. Off, off, still off.
+saharahgaiht So 47 = 00000000000000000000000000000000000000000000000
You forgot to account for zero. Zero should be 0, one is 00.
Simon WoodburyForget "0.0" and ".0" mean the same thing. Not just in "unary", but also in actual programming. float zero= .0f; is the exact same as float zero= 0.0f;
The value of 0.0 in unary would be zero and (I think) one half. 0.00 would be one third, etc. At this point the really serious flaw becomes apparent, and I've put too much thought into a non functional numbering system.
+Joshua Pearce The most interesting non-standard number system is the factorial one, but it has its difficulties as well. Unary, as is used in tally marks, can't represent 0 or fractions, as it is bijective and every digit (the same symbol) just represents a 1, such as 13(dec.) = 1111111111111(unary) = 1111*111+1 in unary, if 1 is used as the symbol. Even using a lot of symbols for a large base like 60 is more useful.
bloody interesting video, thanks!
So he just trailer off at the end there and didn't answer why they didn't use the base 5
big respect prof
Binary can be represented within the machine as off and on. It seems logical to me to use binary
I guess the bottom line is: it`s the most basic form of information there is.
After hours of discussion with my best friend (we both know programming and electronic on transistor level), we agree that Duodecimal (base 12) is the best for people and for computer Trinary is better than Binary, Trinary have sadly been incorrectly done and everyone is focus on a poor working prototype to say that Trinary don't work while it is superior to Binary, this is the future !
+Demongornot Maybe in the very far future It is possible for a computer run off of an n number of states. But for the technology we have right now, along with the hypothetical Q-bits, I think it is much more feasible to obtain a process of Quaternary system. Instead of using a quantum process, it would have transistors that has 4 levels of state instead of two.
+Demongornot why would duodecimal be better for humans, we have 10 fingers...
+TJ.Lewis Trinary is already possible, its not limited by technology but by mainstream, people have started with Binary because it was for simple current hard to manage other than 2 electrical state (voltage) and no one really take serious or deep look on trinary while it is one of the most efficient system, more than trinary is less efficient and binary is not efficient enough, floating point precision is actually a mess because binary offer really poor precision for fractions.
Quaternary is starting to be too much, exactly as binary is not enough.
so can humans do faster mental arithmetic with base 12? Also I think counting fingers in binary mode (up to 1023) is too error prone.
+Demongornot We'll what I said was that having a state of existence up to something such as a 4th state system was feasible. But if it was all possible to have any state of existence the higher that state, the better the floating points will be. I say that because of the problem of the recursion and bit limit problem. For simplicity purposes, if a decimal system was possible, that would be the ideal system. I say that because 1x10^n is easier than 1xN^n for N >=1 and n >= 0
I heart this channel.
It's a shame that all modern computers rely on binary, there are logical advantages to trinary, and many different ways you can implement it in terms of electrical components.
I'm watching this on the exact same phone model as him. lol
Because of you, I made a computer! Check out the video on my page. I made it control servers, leds, sounds and more! I just had to say thank you.
so?
cool
@C:/Users/ntwede.exe please dude. It's hard enough to study CS in binary..
At 4:30 I have given up 🤣😁
Thank you for this video. A colleague of mine asked me that a while ago. I tried to explain it but I wasn't able to put it into simple enough words. I'll send him this video.
I have so much respect for this man, absolutely fascinating!
He is a lovely man too! He taught me at university!
@@mrs-m He's a pleasure to listen to and quite informative.
This felt like the first half of a video. Seems like he got to the point of saying "bi-quinary is better than binary" and then didn't deliver a punchline as to why we wound up with binary anyway.
Left as an exercise for the reader. Anyway, it's because modern bits have lower voltage differences, so it's not as feasible to do multiple voltage levels. You can also see that with reduction in efficiency in TLC/QLC memory in SSDs.
Yevgeniy Gorbachev Didn’t he give that as a reason for not using base 10? I thought he was saying Flowers reckoned he could make biquinary work...Perhaps he was just mistaken in that and quickly discovered as much
@@yevgeniygorbachev5152 Voltage has nothing to do with it. It's a matter of efficiently maximizing space because binary requires less transistors to count. 8 transistors can represent up to 256 in binary, because it uses exponential counting, decimal needs 10 transistors to get to 55.
Binary Maximum Count: 2+4+8+16+32+64+128=256.
Decimal Maximum Count: 1+2+3+4+5+6+7+8+9+10=55
@@mikeef747 Why are you multiplying by two in one expression and adding one in the other? I was under the impression that decimal was 1 + 10 + ...
@@yevgeniygorbachev5152 I think your misunderstanding on voltage is you're thinking of electricity circuits vs computing systems. In electrical circuits, 0 = zero power, but 1 can equal different kinds of voltage. In computing a lower voltage to a transistor = 0, any higher voltage = 1 and no voltage the computer is off.
It's about counting in powers. The binary counting system is in base-2, which means you reach a new base every two digits or in mathematical terms a new "power" every 2 digits. In decimal, it is done base-10, so you need 10 digits to reach a new power.
The Note 7 gives you 3rd degree burns even without those valves ;)
Did you notice that they used +5V and -5V? So why not 0V as well? Well, there was actually an experimental computer using balanced ternary (as this system is called) but it was more of emulating it than using it. Transistors are binary only and that's the main reason why we use binary nowadays. However as transistors are very close to hit their physical limits, new methods are developed and these methods (optical, Josephson junction) are in fact ternary. Donald Knuth (father of the analysis of algorithms and author of The Art of Computer Programming, among bazillion other achievements in computer science) predicted that balanced ternary would be the system of the future.
(I hope this will be covered in a future Computerfile video.)
I would love to see a balanced ternary video. The logic behind it is just so nice and not all that more difficult really.
Equally something I wish was experimented with was ternary or quaternary on hard drives. But it would matter much now since HDD are bssically on there way out.
The only really larger density increase recently was SMR, which is horribly slow and only aimed at archival markets.
SSD are much cheaper now and sloely catching up on price and density.
+Jan Sten Adámek 01000100 01101001 01100100 00100000 01111001 01101111 01110101 00100000 01101110 01101111 01110100 01101001 01100011 01100101 00100000 01110100 01101000 01100001 01110100 00100000 01110100 01101000 01100101 01111001 00100000 01110101 01110011 01100101 01100100 00100000 00101011 00110101 01010110 00100000 01100001 01101110 01100100 00100000 00101101 00110101 01010110 00111111 00100000 01010011 01101111 00100000 01110111 01101000 01111001 00100000 01101110 01101111 01110100 00100000 00110000 01010110 00100000 01100001 01110011 00100000 01110111 01100101 01101100 01101100 00111111 00100000 01010111 01100101 01101100 01101100 00101100 00100000 01110100 01101000 01100101 01110010 01100101 00100000 01110111 01100001 01110011 00100000 01100001 01100011 01110100 01110101 01100001 01101100 01101100 01111001 00100000 01100001 01101110 00100000 01100101 01111000 01110000 01100101 01110010 01101001 01101101 01100101 01101110 01110100 01100001 01101100 00100000 01100011 01101111 01101101 01110000 01110101 01110100 01100101 01110010 00100000 01110101 01110011 01101001 01101110 01100111 00100000 01100010 01100001 01101100 01100001 01101110 01100011 01100101 01100100 00100000 01110100 01100101 01110010 01101110 01100001 01110010 01111001 00100000 00101000 01100001 01110011 00100000 01110100 01101000 01101001 01110011 00100000 01110011 01111001 01110011 01110100 01100101 01101101 00100000 01101001 01110011 00100000 01100011 01100001 01101100 01101100 01100101 01100100 00101001 00100000 01100010 01110101 01110100 00100000 01101001 01110100 00100000 01110111 01100001 01110011 00100000 01101101 01101111 01110010 01100101 00100000 01101111 01100110 00100000 01100101 01101101 01110101 01101100 01100001 01110100 01101001 01101110 01100111 00100000 01101001 01110100 00100000 01110100 01101000 01100001 01101110 00100000 01110101 01110011 01101001 01101110 01100111 00100000 01101001 01110100 00101110 00100000 01010100 01110010 01100001 01101110 01110011 01101001 01110011 01110100 01101111 01110010 01110011 00100000 01100001 01110010 01100101 00100000 01100010 01101001 01101110 01100001 01110010 01111001 00100000 01101111 01101110 01101100 01111001 00100000 01100001 01101110 01100100 00100000 01110100 01101000 01100001 01110100 00100111 01110011 00100000 01110100 01101000 01100101 00100000 01101101 01100001 01101001 01101110 00100000 01110010 01100101 01100001 01110011 01101111 01101110 00100000 01110111 01101000 01111001 00100000 01110111 01100101 00100000 01110101 01110011 01100101 00100000 01100010 01101001 01101110 01100001 01110010 01111001 00100000 01101110 01101111 01110111 01100001 01100100 01100001 01111001 01110011 00101110 00100000 01001000 01101111 01110111 01100101 01110110 01100101 01110010 00100000 01100001 01110011 00100000 01110100 01110010 01100001 01101110 01110011 01101001 01110011 01110100 01101111 01110010 01110011 00100000 01100001 01110010 01100101 00100000 01110110 01100101 01110010 01111001 00100000 01100011 01101100 01101111 01110011 01100101 00100000 01110100 01101111 00100000 01101000 01101001 01110100 00100000 01110100 01101000 01100101 01101001 01110010 00100000 01110000 01101000 01111001 01110011 01101001 01100011 01100001 01101100 00100000 01101100 01101001 01101101 01101001 01110100 01110011 00101100 00100000 01101110 01100101 01110111 00100000 01101101 01100101 01110100 01101000 01101111 01100100 01110011 00100000 01100001 01110010 01100101 00100000 01100100 01100101 01110110 01100101 01101100 01101111 01110000 01100101 01100100 00100000 01100001 01101110 01100100 00100000 01110100 01101000 01100101 01110011 01100101 00100000 01101101 01100101 01110100 01101000 01101111 01100100 01110011 00100000 00101000 01101111 01110000 01110100 01101001 01100011 01100001 01101100 00101100 00100000 01001010 01101111 01110011 01100101 01110000 01101000 01110011 01101111 01101110 00100000 01101010 01110101 01101110 01100011 01110100 01101001 01101111 01101110 00101001 00100000 01100001 01110010 01100101 00100000 01101001 01101110 00100000 01100110 01100001 01100011 01110100 00100000 01110100 01100101 01110010 01101110 01100001 01110010 01111001 00101110 00100000 01000100 01101111 01101110 01100001 01101100 01100100 00100000 01001011 01101110 01110101 01110100 01101000 00100000 00101000 01100110 01100001 01110100 01101000 01100101 01110010 00100000 01101111 01100110 00100000 01110100 01101000 01100101 00100000 01100001 01101110 01100001 01101100 01111001 01110011 01101001 01110011 00100000 01101111 01100110 00100000 01100001 01101100 01100111 01101111 01110010 01101001 01110100 01101000 01101101 01110011 00100000 01100001 01101110 01100100 00100000 01100001 01110101 01110100 01101000 01101111 01110010 00100000 01101111 01100110 00100000 01010100 01101000 01100101 00100000 01000001 01110010 01110100 00100000 01101111 01100110 00100000 01000011 01101111 01101101 01110000 01110101 01110100 01100101 01110010 00100000 01010000 01110010 01101111 01100111 01110010 01100001 01101101 01101101 01101001 01101110 01100111 00101100 00100000 01100001 01101101 01101111 01101110 01100111 00100000 01100010 01100001 01111010 01101001 01101100 01101100 01101001 01101111 01101110 00100000 01101111 01110100 01101000 01100101 01110010 00100000 01100001 01100011 01101000 01101001 01100101 01110110 01100101 01101101 01100101 01101110 01110100 01110011 00100000 01101001 01101110 00100000 01100011 01101111 01101101 01110000 01110101 01110100 01100101 01110010 00100000 01110011 01100011 01101001 01100101 01101110 01100011 01100101 00101001 00100000 01110000 01110010 01100101 01100100 01101001 01100011 01110100 01100101 01100100 00100000 01110100 01101000 01100001 01110100 00100000 01100010 01100001 01101100 01100001 01101110 01100011 01100101 01100100 00100000 01110100 01100101 01110010 01101110 01100001 01110010 01111001 00100000 01110111 01101111 01110101 01101100 01100100 00100000 01100010 01100101 00100000 01110100 01101000 01100101 00100000 01110011 01111001 01110011 01110100 01100101 01101101 00100000 01101111 01100110 00100000 01110100 01101000 01100101 00100000 01100110 01110101 01110100 01110101 01110010 01100101 00101110 00001101 00001010 00001101 00001010 00101000 01001001 00100000 01101000 01101111 01110000 01100101 00100000 01110100 01101000 01101001 01110011 00100000 01110111 01101001 01101100 01101100 00100000 01100010 01100101 00100000 01100011 01101111 01110110 01100101 01110010 01100101 01100100 00100000 01101001 01101110 00100000 01100001 00100000 01100110 01110101 01110100 01110101 01110010 01100101 00100000 01000011 01101111 01101101 01110000 01110101 01110100 01100101 01110010 01100110 01101001 01101100 01100101 00100000 01110110 01101001 01100100 01100101 01101111 00101110
+Jan Sten Adámek Why not use 0V? Easy: fault detection. Especially in the early days, these valves burned out a lot.
+Herve Shango For anyone too lazy to find a place to convert it, this is what it converts to:
Did you notice that they used +5V and -5V? So why not 0V as well? Well, there was actually an experimental computer using balanced ternary (as this system is called) but it was more of emulating it than using it. Transistors are binary only and that's the main reason why we use binary nowadays. However as transistors are very close to hit their physical limits, new methods are developed and these methods (optical, Josephson junction) are in fact ternary. Donald Knuth (father of the analysis of algorithms and author of The Art of Computer Programming, among bazillion other achievements in computer science) predicted that balanced ternary would be the system of the future.
(I hope this will be covered in a future Computerfile video.
+Jan Sten Adámek Not only did balanced ternary blow my mind, but Knuth also describes Fibonacci counting (very useful in information theory) and factorial counting (as a curiosity).
I want to know more about the non-binary attempts at making computers. There isn't much data online about Dekatron vacuum tubes and the Setun balanced ternary computer.
This man is a gifted communicator. His life's work has definitely advanced mankind. Respect +1
Imagine how big and power hungry a non-electronic smartphone-equivalent would be.
+C0deH0wler Probably something like this : vignette3.wikia.nocookie.net/fallout/images/7/76/Pip-Boy_3000.jpg/revision/latest?cb=20110712154420
+András Bíró Well that's still electronic.
*****
Electronic**. You don't call your speakers electromechanical devices because they have an analog knob that can turn, do you?
You should look at the lore yourself, it even explicitly states it's an electronical device ("The RobCo Pip-Boy (Personal Information Processor) is an electronic device manufactured by RobCo Industries."). Lack of transistors doesn't make it mechanical.
You don't know what electromechanical devices are.
A better question might be, why when your base building block consists of transistors, would you want to use any other base?
Balanced ternary.
How about doing a video about modern vacuum tubes? They are used extensively in communication satellites (traveling wave tubes are pretty much the only amplifier that can reliably be used in a small while offering a small and compact package) and micro-tubes are being developed to be used in cellphones and other microwave communication gear such as Wifi access points so he might get tubes in his mobile sooner than he thinks ::)
+DOGMA1138
They don't really have that much to do with computers, so I don't think they fit here (or people watching computerphile would be interested/have physics under their belt). They are analog amplifiers.
The prof is due for a cellphone upgrade I think.
Why has this channel not done a video or even talked about memristors yet? And the computer technological significance of this equipment. I am sure many of your viewers have not even heard about memristors yet. ;-)
I wish this man was my Grandfather.
I would talk to him all the time.
This was actually quite interesting. I had heard in my computing course that it was because of the voltage variation, that they couldn't guarantee that say, 4 volts would always actually be 4 volts and wouldn't drift as he says and become 5 or 3 and mess things up but they could easily guarantee it with binary by making it simply any voltage or no voltage. This video showed that there was a lot more to it than that though, I never really thought about the debate while the technology was still being developed, only about why we use it in terms of modern computing.
Ah, another lovely story by the brilliant Professor Brailsford. What voice he has, perfect for story telling. But maybe I'm biased by the fact that I'm American and any proper English accent sounds perfect for story telling.
+Franklin Cerpico
What, pray, is a 'proper' English accent? (speaking as someone from the north east of England) -- And I agree, his voice is perfect for story-telling.
Gammel Prutte Well if I had to narrow it down, there's the accent which the Professor has, which I for lack of a better word label as 'proper' in order to distinguish it from a 'cockney' accent. Not to say a hint of 'cockney' isn't nice too, take Michael Cain for example.
I keep watching these videos and completely don't understand what they are talking about.
The real reason? Computers took a lot of space in the beginning some were entire buildings. You only need 8 transistors to count to 256 in binary, you need 10 to count to 55 in decimal.
Binary Maximum Count: 2+4+8+16+32+64+128=256.
Decimal Maximum Count: 1+2+3+4+5+6+7+8+9+10=55
"There are 10 kinds of people, those who understand binary, and those who don't" ... a joke probably older than computers.
This didn't really explain why binary is used now. So there was an alternative base 10 system on the valve computers at Bletchley Park. Why aren't we still using that system then? What happened to it?
Transistor logic happenned. The big difference between transistor logic and valve logic is that the spike in power consumption is only when switching not all the time. Valves are power hogs all the time, transistors only when switching. So simply having a register to hold a value would draw power, so it made sense to have the smallest amount of circuitery. In transistor logic, you can have a lot of static circuitery which will not consume that much power when nothing happens on them. So representing numbers in binary allows for simpler circuits, even if there are more of them in the computer. It's the same thing with relays. Most relay computers also used binary representation (Zuse Z3 even in 1941) because in relays it's also the transition that is costly.
We figured out that physics doesn't have ten fingers.
A circuit that can natively store and retrieve one of 10 states would be complicated to build and would include a lot of analog components that aren't efficient to miniaturize on ICs. Two state logic requires so few parts per bit that it makes sense to keep using it, at least with current technology.
Far fetched ...Nonsense...We use 2 states only (binary ones and zeros) because transistors use far less electricity deciding over this two states...If we use 8 different voltages??? than to make processor fast as mobile phone-we'll need 60 KW ..Nonsense...Even as a test this is unachievable cause the voltage threshold will vary too much and thus to much errors...Not possible...
Technically, the most effective base for arithmetic is e, but that's silly. 3 would still be better than two though
I have heard that before! Why is that and can you give a source so i can read more about it please?
"3 times more circuitry" - not true. To make 10 levels just as reliable as binary will require more hardware anyway. OK, that still _might_ be less but not 3 times.