The Parallax Error. That is why old analog meters have the reflective strip behind the needle (indicator). If you see two needles (the actual and the reflected) your point of view is in error. You need to move your angle of view to where the needle (indicator) lines up and covers the reflection. That is where you take your reading from the scale that is behind the needle (indicator).
When copying your LFSR setup I couldn't figure out why my QA and QB (first two outputs) generated a 2-bit sequence and not a 3-bit sequence like yours. I finally found out when you combine the SRCLK and RCLK you get the same result as in your video. I used a MCU to generate 2 clock pulses with a short delay between SRCLK and RCLK. Nice project indeed.
to experimentally check for the all-ones pattern you could clock it as fast as the ICs allow, then add a 41bit and-gate feeding into the set input of an RS latch or similar "can be toggled on" circuit to an led (so you don't have to actually catch the exact clock cycle). then just wait a few hours to months (depending on how fast the ICs can be clocked) for that led to light up.
A nifty trick I just figured out is you can use the 7486 as an inverter for the clock signal. Though I noticed you did just that on another video on LFSRs. But for those who don't know, if you tie input A (or B) of an xor gate high the output is is NOT B. Very useful as I didn't want to muck about with transistors or waste space with a hex inverter I'd only use one gate of. I know my boolean algebra theorems, but it just didn't click until now.
That's approximately the beginning of what was used for encryption devices way back when when I was in the military in the sixties I got to look at some of the encryption devices and they use feedback loops where you change the location of the feedbacks that are being fed back into that shift register. you have brought back some very fine old memories of my military service
Yes, brings back memories for me too. We were using these things in the 70s as easy to implement counters in p-chanel mosfet chips we were designing for various telecoms projects. As Julian demonstrates, they do not have to be maximum length, and the trick was to find the minimum length shift register / tap combination that would cycle through the count you needed. One project in which I was involved used 7 different lsrs to decode the dtmf signals from a push button telephone tone dialer. Note that the key pad is 4 X 3 generating 12 pairs of tones. There were 4 low band tones (1000Hz) if my failing memory serves me correctly after 45 years. Those were the days.
Shifting vertigo I swear they were going backwards when they stopped at one point. You do come up with some strange topics Julian which makes your channel so unique. Cheers 73
When I implemented a 64 bit counter (for clock diagnostics) in an FPGA with a clock of 125 MHz, my boss was surprised when I told him it will not roll over for 4679 years.
Nice minimalist hardware demonstration of the CRC40 computation. I got my first experience with CRC32 with the ZModem transmission protocol. I then discovered that ZIP files use the exact same CRC32. The software that compute the CRC checksum can be made of 8 loops for each bytes that need to be computed. This was slow on the original PC at 4.88 MHz or the 6809 at 4 MHz. It took a few seconds to compute only 64K bytes. There was an algorithm that was 8 times faster. It was using a data table of 256 values of 32 bits each. The checksum was simply computed by doing XOR with the constant in the table using the 8 bit byte as index in the table. I like your careful choice of words as an engineer do. The only detail is your usages of the word "bit" when you describe the total number of value that a given shift register combination can produce. It is not "1 trillion bit..", it is a 40 bits CRC generating a sequence of 1 trillion pseudo random values.
Wow! I just sat down thinking "I'll relax by watching Julian pointlessly make LEDs scroll around" and at first it was amusing, then interesting, then I learned the basis of pseudo-random number generators. Well done!
An interesting thing here, you read the display as a programmer would and not as an electronic tech would. Most of us "el" techs that learned digital electronics in the 70's and 80's read binary output from left to right with the least significant bit on the left. Your circuit is typically laid out from left to right, from power in to display out, and you are even sequencing (shifting) the bits left to right. This is how most technicians and designers would lay this out. Most schematics are designed left to right so we tend to build circuits that reflect that orientation. Consequently, an electronics oriented mind reads your display from left to right / least significant bit on the left. A programmer tends to read numbers in a more traditional number format, with least significant digit on the right. This leads them to read binary the same way - with least significant bit on the right. In the early days where digital numbers were entered into a computer by flipping switches on the front panel, the switches had to be electrically inverted or swapped around to a right to left orientation to accommodate the programmer's binary right to left format. None of this makes any difference in your number sequences except that as an electronic tech I see all your written numbers as inverted or one's compliment plus 1. What you read as "8" I read as "1" or " -7" in one's compliment. The advantage to the tech's orientation is; no matter how many shift registers/bits I add to your cascade to the right, the first bit in the sequence on the left is always bit 1. Using your orientation, the value of the left most bit has to be reassigned depending on how many bits you choose to read. This works on paper and designing programs, but would be a logic nightmare to design as an electronic circuit to alter the output value of each bit.
This video set me off doing some research. I discovered that if an XOR gate has more than two inputs it should not be called an XOR gate, but an odd parity detector. This means that the output is 1 if there are an odd number of 1's on the inputs. This came as a bit of shock to me.
Cool to see these LFSR sequences in "practise" so to speak. An interesting real world application of this pseudo-random property, is it's use in modern CPU memory controllers. They use LFSR's to scramble the memory mapping every boot. This is done partly for security reasons (it is actually possible to recover the state of DRAM sticks after shutdown), and also done for performance reasons, since in many situations, memory is not fully occupied by the OS - and thus the parallellism of the memory architecture is not utilized. By scrambling the memory mapping, you spread the data, no matter how much or how little is currently being used, across all available DRAM channels, and thus increase the throughput. I can't find it now, but I read a paper where a team managed to extract and de-scramble data from DRAM used in a real system, after powerdown, overcoming the LFSR scrambled mapping. Quite impressive.
You never opened a old mechanical touch tone phone, one I had actually had 4x4 coils for tones a b c d. But no buttons. Ham radio operators use all 16 tones.
This brings to light things that we already think are forever random...Pi for example they say goes forever but yet it is determinable and there is a finite limit to the practical precision of it due to planks constant. I suspect Pi, and all things, are related to such long sequences, that they make them appear random. We all may be following some sort of feedback register seeded with our birthdate, etc...
Another interesting video. Would it be possible to modify the circuit to make a shifting logic probe?. Using the base frequency of a circuit & probing a logic out put in the circuit under test would it be possible to gather forty clock steps of data to view. In my working days we used to use signature analyzers combined with an plug-able EPROM to trouble shoot complex PCBs.
A great video. A circuit diagram would be great so that I could recreate it for my students. Dear Julian, please post a circuit diagram for this setup! THX!
Very nice video. I was wondering could you do a video on using an arduino and 8 or 16 multiplexer and opto couplers to read a 7s li-ion battery pack. I am an old hardware engineer not much on programming but trying to learn arduino thanks
I love these videos. I like the sounds you can make with simple LFSRs because they sound just like the noises used in old video games like explosions, thrusters and such. Speaking of LFSR noise, are you ever going to get back to the Vocoder project?
dentakuweb Interestingly, the SID chip in the C64 uses a few LFSR blocks, not just for the noise, but even for the ADSR envelopes and then exponential approximations. A very clever way of saving on logic gates and chip area, as the sequences generated are much longer than the total number of bits in the LFSR itself. It then just used some comparison logic (XOR gates etc.) to trigger other events like the next stage of the ADSR. ;) Yep, you very often heard an LFSR used as noise generators in old computers / consoles / digital synths. hehe Even those cheap "sound effects" key fobs probably used an 8-bit (or lower) LFSR, which is why you often heard the repeating pattern. Great vid, Juilian. Very useful as a refresher / crash-course. ;)
" love these videos. I like the sounds you can make with simple LFSRs because they sound just like the noises used in old video games like explosions, thrusters and such" Yes, this is how these are made, at least for the Gameboy and GBA I know the noise generator is a LSFR
I would think that's pretty common in electronics and sound chips in particular. I know the 'noise' function in the SNES sound chip is an LSFR Then there's Atari's Pokey chip, which is also an audio chip. That has a 5 bit and 9 bit LFSR, and one that is switchable between 7 or 13 bits. You can select any two of these for each audio channel, leading to 8 different possible audio 'noise' profiles, though the pattern isn't quite as random as an actual white noise generator would be. (this is also how the computers using this chip generate random numbers though, if you ask it to create one.) Software implementations of LSFR's are also among the most common seen in computer games to generate random numbers. Especially on older hardware. (on a modern computer you can get away with using a more complex cryptographic PRNG if you really want to, though for a game it rarely matters as long as you use it appropriately.)
So, not random at all, but good enough for many purposes. Thanks Julian. P.S. Thanks for not using a clickbait ‘A 70,000 year Shift Register!!!!! Must Watch!!!!!!!!' title. ;-p
Julian, I really liked this video. However, I have one request. Would you provide a schematic of your circuit so we can use it to experiment with on our own if we want? I would appreciate that. Thank you. Keep up the good videos. :-)
at the MIT museum they have an actual huge year object, not this LFSR but an AC motor driving a gear that 20:1 drives another gear... about a dozen gears long and the last one is welded to an I-beam. they say, falsely, that the energy built up in the billions of years before the weld will break is like a nuclear bomb. as for whitening. the 'best' way to remove bias is as described deep into the following makezine.com/projects/really-really-random-number-generator/ but it tosses away a lot of bits so say 1mbps gets really slowed down say to 10kbps. i think this lfsr injection perhaps preserves full bit rate.
Also your taps actually be wrong if those shift registers have register and output clocks as the output will be 1 clock cycle behind the internal registers if you clock both clock inputs at the same time.
Since the sequence is not actually random, can the position of the output that is all ones be worked out? Then you would know how long you would have to wait to see it.
You could simulate the sequence with some code though, then have it look for that long sequence of 1s. Then you could start the LFSR with a seed value that is only a short way before the 1s appear. Just for fun of course, but at least you'd then know you have so many thousand years to wait before seeing those 1s again. lol
*UPDATE: Found it! Finally!* Apparently, if you call iteration 1 the moment at the beginning of the sequence, when only the leftmost LED lit up, the all ones combination occurs at the 163,096,468,512nd iteration. So it shouldn't take 10,000 years to see this combination, it would actually take "only" 738 years.
I’ve only watched part so far, but curious why you chose to use reverse endian when converting the binary to decimal? ie. Given you are shifting bits in from the left, it would be fair to say the left most register bit is the LSB. So when you decide to mask off the left most 4 bits, they should be read as left most is the LSB. So for the first 4 bits what you called 8 is actually 1, 1 is 8, 4 is 2, 10 is 5 etc.
Ideally, you start with it powered off, turn it on, and use the button to introduce a 1 into the leftmost bit - all the rest will initially be zero. (In his case, where he moved the taps while the previous sequence was still moving through the register, he did effectively start at a random point in the sequence.)
If you don't have a slew of shift registers but you do have a scientific calculator, and you need more than two options (which could more easily be chosen by flipping a coin), you can use the calculator to produce a pseudo-random number. 1. Press the decimal point, then enter some random digits from 0 to 9 (possibly by having friends call out a couple of random digits each); 2. Calculate the natural antilogarithm (E to the X power where X is the displayed value) and then; 3. take the reciprocal (1/X). If you need more than one random number, save the result in the memory for later. 4. Multiply by the number of options you need and add 1 (otherwise you get a number between zero and one less than the number you multiplied by), then ignore the decimal part and use the integer part of the result. If you need another random number, recall the memory and go back to step 2. This will usually be random enough for most purposes since it would take a mathematical savant to predict what number will come up next.
Ravindra Badgujar , I think XNOR gives you a different sequence of a similar set of numbers, (other than all-zeros won't kill it, but all-ones will.) I think NAND leads to uninteresting results because for the four input states to a NAND (00, 01, 10, 11) the two output states don't have equal probability. (1, 1, 1, 0) I read the same Wiki page Julian used about 3 years ago, and it's worth reading.
Woah, I experienced a bit of a optical illusion towards the end there. Looked like the LED display was sliding to the left when you paused the LFSR, anyway good stuff.
"Two trillion..." INCORRECT! It's two hundred and nine *billion* , nine thousand and twenty-three *million* two hundred and fifty-five thousand five hundred and fifty-one.
Prehistoricman We’re majority in the in the world. Plus, lots of mathematicians, like me, think the long scale is superior for a number of things that are explained in Numberphile’s video: ua-cam.com/video/C-52AI_ojyQ/v-deo.html. Ah, and I think he made a mistake even in the short system.
The Parallax Error. That is why old analog meters have the reflective strip behind the needle (indicator). If you see two needles (the actual and the reflected) your point of view is in error. You need to move your angle of view to where the needle (indicator) lines up and covers the reflection. That is where you take your reading from the scale that is behind the needle (indicator).
Great revisit of the LFSR. Much clearer to see whats going on now than in asm :)
Very interesting, Julian! I love it when I learn something new.
There is something oddly relaxing about watching that mega long sequence cycle through.
Your bread boars are always so neat and symmetrical, very pleasing look at!
Glad I stuck it out till the end. Pretty fascinating, think I'm going to need to build one for my desk
When copying your LFSR setup I couldn't figure out why my QA and QB (first two outputs) generated a 2-bit sequence and not a 3-bit sequence like yours. I finally found out when you combine the SRCLK and RCLK you get the same result as in your video. I used a MCU to generate 2 clock pulses with a short delay between SRCLK and RCLK. Nice project indeed.
You know what you need? A dedicated channel that streams it 24/7/365. I'm sure some would sit and watch it!
Like the one broadcasting "flow" of tar drop ...
That was really fascinating. More like this please
thank you, yes more on logic desired!
to experimentally check for the all-ones pattern you could clock it as fast as the ICs allow, then add a 41bit and-gate feeding into the set input of an RS latch or similar "can be toggled on" circuit to an led (so you don't have to actually catch the exact clock cycle). then just wait a few hours to months (depending on how fast the ICs can be clocked) for that led to light up.
Cool. I had a fun couple of hours playing around with this in C++ after this vid. Just what I needed :)
A nifty trick I just figured out is you can use the 7486 as an inverter for the clock signal. Though I noticed you did just that on another video on LFSRs. But for those who don't know, if you tie input A (or B) of an xor gate high the output is is NOT B. Very useful as I didn't want to muck about with transistors or waste space with a hex inverter I'd only use one gate of. I know my boolean algebra theorems, but it just didn't click until now.
Happy New Year Julian, hope the new one brings vibrant video variety.
Quite interesting, both, from the logical and mathematical points of view. Thanks for sharing! Great video.
I am SO going to build this. Thank You again Julian.
AWESOME WORK Julian!....you do know the answer is 42....no need to wait for your bit counter... Great JOB!
That's approximately the beginning of what was used for encryption devices way back when when I was in the military in the sixties I got to look at some of the encryption devices and they use feedback loops where you change the location of the feedbacks that are being fed back into that shift register. you have brought back some very fine old memories of my military service
"service"
Yes, brings back memories for me too. We were using these things in the 70s as easy to implement counters in p-chanel mosfet chips we were designing for various telecoms projects.
As Julian demonstrates, they do not have to be maximum length, and the trick was to find the minimum length shift register / tap combination that would cycle through the count you needed.
One project in which I was involved used 7 different lsrs to decode the dtmf signals from a push button telephone tone dialer. Note that the key pad is 4 X 3 generating 12 pairs of tones. There were 4 low band tones (1000Hz) if my failing memory serves me correctly after 45 years.
Those were the days.
+Peter Jennings those were the days :)
Shifting vertigo I swear they were going backwards when they stopped at one point. You do come up with some strange topics Julian which makes your channel so unique. Cheers 73
That is so unfathomably long, my mind is simply blown!
fantastic work, thanks Julian :)
He needs to sell a shirt. "I only provide XOR feedback" with a sequence pattern underneath and a XOR feedback schematic.
I love it. I would build that and leave it running forever. Imagine your LEDs in a circle. How cool. that looks like fun fun fun
That breadboard looks so clean.
When I implemented a 64 bit counter (for clock diagnostics) in an FPGA with a clock of 125 MHz, my boss was surprised when I told him it will not roll over for 4679 years.
Bosses are idiots :)
A masterpiece! You catch my subscription!
Fascinating, the math is entirely over my head but I'm seriously considering building this as an art project.
Brilliant! Can't wait to try this!
I hope those eneloops are well charged. :-)
I'm having to charge them a lot, this thing does drain them quite quickly :)
Julian Ilett Shame. That’ll start the 10,000 year counter again then. You’ll never see the end!
@@AdamWelchUK Tick Tock Tick Tock. Hey, time is all we got! :-)
Nice minimalist hardware demonstration of the CRC40 computation. I got my first experience with CRC32 with the ZModem transmission protocol. I then discovered that ZIP files use the exact same CRC32.
The software that compute the CRC checksum can be made of 8 loops for each bytes that need to be computed. This was slow on the original PC at 4.88 MHz or the 6809 at 4 MHz. It took a few seconds to compute only 64K bytes. There was an algorithm that was 8 times faster. It was using a data table of 256 values of 32 bits each. The checksum was simply computed by doing XOR with the constant in the table using the 8 bit byte as index in the table.
I like your careful choice of words as an engineer do. The only detail is your usages of the word "bit" when you describe the total number of value that a given shift register combination can produce. It is not "1 trillion bit..", it is a 40 bits CRC generating a sequence of 1 trillion pseudo random values.
In the section where you have pins 2 and 4 feeding back, the pattern seems to be SIX long, not four, which also accounts for the three-length pattern.
Wow! I just sat down thinking "I'll relax by watching Julian pointlessly make LEDs scroll around" and at first it was amusing, then interesting, then I learned the basis of pseudo-random number generators. Well done!
An interesting thing here, you read the display as a programmer would and not as an electronic tech would. Most of us "el" techs that learned digital electronics in the 70's and 80's read binary output from left to right with the least significant bit on the left. Your circuit is typically laid out from left to right, from power in to display out, and you are even sequencing (shifting) the bits left to right. This is how most technicians and designers would lay this out. Most schematics are designed left to right so we tend to build circuits that reflect that orientation. Consequently, an electronics oriented mind reads your display from left to right / least significant bit on the left.
A programmer tends to read numbers in a more traditional number format, with least significant digit on the right. This leads them to read binary the same way - with least significant bit on the right. In the early days where digital numbers were entered into a computer by flipping switches on the front panel, the switches had to be electrically inverted or swapped around to a right to left orientation to accommodate the programmer's binary right to left format.
None of this makes any difference in your number sequences except that as an electronic tech I see all your written numbers as inverted or one's compliment plus 1. What you read as "8" I read as "1" or " -7" in one's compliment. The advantage to the tech's orientation is; no matter how many shift registers/bits I add to your cascade to the right, the first bit in the sequence on the left is always bit 1. Using your orientation, the value of the left most bit has to be reassigned depending on how many bits you choose to read. This works on paper and designing programs, but would be a logic nightmare to design as an electronic circuit to alter the output value of each bit.
This video set me off doing some research. I discovered that if an XOR gate has more than two inputs it should not be called an XOR gate, but an odd parity detector. This means that the output is 1 if there are an odd number of 1's on the inputs. This came as a bit of shock to me.
Cool to see these LFSR sequences in "practise" so to speak. An interesting real world application of this pseudo-random property, is it's use in modern CPU memory controllers. They use LFSR's to scramble the memory mapping every boot. This is done partly for security reasons (it is actually possible to recover the state of DRAM sticks after shutdown), and also done for performance reasons, since in many situations, memory is not fully occupied by the OS - and thus the parallellism of the memory architecture is not utilized. By scrambling the memory mapping, you spread the data, no matter how much or how little is currently being used, across all available DRAM channels, and thus increase the throughput.
I can't find it now, but I read a paper where a team managed to extract and de-scramble data from DRAM used in a real system, after powerdown, overcoming the LFSR scrambled mapping. Quite impressive.
32:24 this remined me of the Scene "How Scared should we be?" from Futurama XD
now you inspired me to build this circuit from my stockpile of 74hc595s and run it at 70kHz with a wrap-around time of one year ;)
or I'll go for the 2^31 version which will give one year of continuously changing patterns at 68Hz ;)
Love this. How about a circuit diagram as it's easier to take in if you can see the layout?
Nice playground. You should add a beep sound or LED that would indicate when the whole nibble was shifted.
Nice build-up!
Yes, digital eliminated this problem a long time ago. Just a piece of history. :-) Julian, you rock.
Thanks :)
Expand this by tying the first four outputs to a 4 bit to 7 segment decoder to drive a seven segment display.............That would look neet
+Luke Curtis that would be excellent :)
If you do that, you would get the cover to The Police - "Ghost in the machine"
remember when I coded in assembler a RND() function, it was 16bits shift, and taking the XOR inputs from bit 7 and 2.
20 years ago.
Now the thing to do would be to put some similar code in a PIC, and then feed it into one of those xmas decorations...
Awesome video like always~~ :D
I didn't like the part following 33:45 :-/
You never opened a old mechanical touch tone phone, one I had actually had 4x4 coils for tones a b c d. But no buttons. Ham radio operators use all 16 tones.
Built and running on my desk at work. It’s driving the night crew insane.
This was about as exciting as watching Shirriff mine bitcoin by hand
This brings to light things that we already think are forever random...Pi for example they say goes forever but yet it is determinable and there is a finite limit to the practical precision of it due to planks constant. I suspect Pi, and all things, are related to such long sequences, that they make them appear random. We all may be following some sort of feedback register seeded with our birthdate, etc...
Now that was an interesting one!
Hi Julian. Great fun video. Im interested in how you dealt with de-bounce on the invert button ?
If the feedback digits are primary numbers, they tend to form more varied permutations
I hadn't thought of using 595's like this
Another interesting video. Would it be possible to modify the circuit to make a shifting logic probe?. Using the base frequency of a circuit & probing a logic out put in the circuit under test would it be possible to gather forty clock steps of data to view. In my working days we used to use signature analyzers combined with an plug-able EPROM to trouble shoot complex PCBs.
Galois LFSR are easily implemented in software. Wiki has pseudo code on the subject. I've used it for a 'random' tune generator circuit.
Gosh, I could keep track of my birthdays so long as the batteries last 10,000 years.
I'd like to see the part where you can hold the button for the input connected to music; I wonder what would happen lol
A great video. A circuit diagram would be great so that I could recreate it for my students. Dear Julian, please post a circuit diagram for this setup! THX!
That's an old video. I don't think I have the circuit made up anymore.
Inevitably, you have all of the right numbers, but not necessarily in the right order.
Very nice video. I was wondering could you do a video on using an arduino and 8 or 16 multiplexer and opto couplers to read a 7s li-ion battery pack. I am an old hardware engineer not much on programming but trying to learn arduino thanks
You could use a spreadsheet with cells with 1 and 0 's in it and do it all in software.
I love these videos. I like the sounds you can make with simple LFSRs because they sound just like the noises used in old video games like explosions, thrusters and such.
Speaking of LFSR noise, are you ever going to get back to the Vocoder project?
dentakuweb
Interestingly, the SID chip in the C64 uses a few LFSR blocks, not just for the noise, but even for the ADSR envelopes and then exponential approximations.
A very clever way of saving on logic gates and chip area, as the sequences generated are much longer than the total number of bits in the LFSR itself.
It then just used some comparison logic (XOR gates etc.) to trigger other events like the next stage of the ADSR. ;)
Yep, you very often heard an LFSR used as noise generators in old computers / consoles / digital synths. hehe
Even those cheap "sound effects" key fobs probably used an 8-bit (or lower) LFSR, which is why you often heard the repeating pattern.
Great vid, Juilian. Very useful as a refresher / crash-course. ;)
+dentakuweb yes, this year it's gonna be circuits, breadboard and all the stuff I enjoy the most
" love these videos. I like the sounds you can make with simple LFSRs because they sound just like the noises used in old video games like explosions, thrusters and such"
Yes, this is how these are made, at least for the Gameboy and GBA I know the noise generator is a LSFR
I would think that's pretty common in electronics and sound chips in particular.
I know the 'noise' function in the SNES sound chip is an LSFR
Then there's Atari's Pokey chip, which is also an audio chip.
That has a 5 bit and 9 bit LFSR, and one that is switchable between 7 or 13 bits.
You can select any two of these for each audio channel, leading to 8 different possible audio 'noise' profiles, though the pattern isn't quite as random as an actual white noise generator would be.
(this is also how the computers using this chip generate random numbers though, if you ask it to create one.)
Software implementations of LSFR's are also among the most common seen in computer games to generate random numbers.
Especially on older hardware.
(on a modern computer you can get away with using a more complex cryptographic PRNG if you really want to, though for a game it rarely matters as long as you use it appropriately.)
If in the late stone age or the early bronze age a man would have build a 10.000 Year Shift Register, it still would be shifting till this very day.
Happy new year all I'm not a fan of the clear bread board It makes it look untidy but your layout is very neat .A nice video though
everything is moving to the right after pausing the video at the end!!!!!! OMG XD
So, not random at all, but good enough for many purposes. Thanks Julian. P.S. Thanks for not using a clickbait ‘A 70,000 year Shift Register!!!!! Must Watch!!!!!!!!' title. ;-p
"You'll totally believe what happens next"
If you speed it up to 1 GHz, you only have to wait about 37 minutes. The only problem is that it might be a bit difficult to see.
i never put such a beautiful resistances and wires on a board :{ how did you do such perfect mount ?
is there a tool to do it ?
Ooh, that's.. pretty cool.
Do you have a Circuit diagram for this?
Because i would love to build something similar to this and btw awesome work 👏
Julian, I really liked this video. However, I have one request. Would you provide a schematic of your circuit so we can use it to experiment with on our own if we want? I would appreciate that. Thank you. Keep up the good videos. :-)
at the MIT museum they have an actual huge year object, not this LFSR but an AC motor driving a gear that 20:1 drives another gear... about a dozen gears long and the last one is welded to an I-beam. they say, falsely, that the energy built up in the billions of years before the weld will break is like a nuclear bomb.
as for whitening. the 'best' way to remove bias is as described deep into the following
makezine.com/projects/really-really-random-number-generator/
but it tosses away a lot of bits so say 1mbps gets really slowed down say to 10kbps.
i think this lfsr injection perhaps preserves full bit rate.
Also your taps actually be wrong if those shift registers have register and output clocks as the output will be 1 clock cycle behind the internal registers if you clock both clock inputs at the same time.
Great video, ut you should share the electric schematic of presented construction.
Very interesting. 21:30 I think you actually can build a Galois FSR with your setup
Awesome...
I feedback therefore I am
I feedback to you, all we need is you to feedback to me and the universe explodes.
infinite loops are dangerous :D
Luckily there are losses.
You must be positive. Being negative is regretful.
+Anvilshock LOL well free energy would be great !
I can only imagine that the display filled with one's while Julian was holding the calculator in front of it
Michel Fugers , ha !
thank you
Good Job, interesting, maybe ufo language stuff
Since the sequence is not actually random, can the position of the output that is all ones be worked out? Then you would know how long you would have to wait to see it.
Simulate it, you could brute-force a solution probably in minutes. Found this thing, btw: github.com/aanunez/LFSR-Simulator
When the total run is 139 laps around the earth, I think the odds are bad to be seeing it soon
You could simulate the sequence with some code though, then have it look for that long sequence of 1s.
Then you could start the LFSR with a seed value that is only a short way before the 1s appear.
Just for fun of course, but at least you'd then know you have so many thousand years to wait before seeing those 1s again. lol
*UPDATE: Found it! Finally!*
Apparently, if you call iteration 1 the moment at the beginning of the sequence, when only the leftmost LED lit up, the all ones combination occurs at the 163,096,468,512nd iteration.
So it shouldn't take 10,000 years to see this combination, it would actually take "only" 738 years.
I’ve only watched part so far, but curious why you chose to use reverse endian when converting the binary to decimal? ie. Given you are shifting bits in from the left, it would be fair to say the left most register bit is the LSB. So when you decide to mask off the left most 4 bits, they should be read as left most is the LSB. So for the first 4 bits what you called 8 is actually 1, 1 is 8, 4 is 2, 10 is 5 etc.
how would you feel if a mathematician figured that it got the 44 1s while you were showing the calculator?
That’s pretty cool. What display are you using?
What if the 41 1's went by while he had the Android tablet held up to the camera?
So what determines when the 10000 year 41 bit sequence goes through? Random emf noise when turned on?
Ideally, you start with it powered off, turn it on, and use the button to introduce a 1 into the leftmost bit - all the rest will initially be zero. (In his case, where he moved the taps while the previous sequence was still moving through the register, he did effectively start at a random point in the sequence.)
You should try filling in a lottery ticket with those numbers!
If you don't have a slew of shift registers but you do have a scientific calculator, and you need more than two options (which could more easily be chosen by flipping a coin), you can use the calculator to produce a pseudo-random number.
1. Press the decimal point, then enter some random digits from 0 to 9 (possibly by having friends call out a couple of random digits each);
2. Calculate the natural antilogarithm (E to the X power where X is the displayed value) and then;
3. take the reciprocal (1/X).
If you need more than one random number, save the result in the memory for later.
4. Multiply by the number of options you need and add 1 (otherwise you get a number between zero and one less than the number you multiplied by), then ignore the decimal part and use the integer part of the result.
If you need another random number, recall the memory and go back to step 2.
This will usually be random enough for most purposes since it would take a mathematical savant to predict what number will come up next.
JasonMasters Yeah, except I have a calculator that has a (pseudo) random number generator function- RND. That's much easier to remember 😜
Why didn't you use resistor arrays?
Would it be possible to provide a schematic?
nice video. what if the feedback is other Logic gate like XNOR or NAND or any other, will the output change?
Ravindra Badgujar , I think XNOR gives you a different sequence of a similar set of numbers, (other than all-zeros won't kill it, but all-ones will.)
I think NAND leads to uninteresting results because for the four input states to a NAND (00, 01, 10, 11) the two output states don't have equal probability. (1, 1, 1, 0)
I read the same Wiki page Julian used about 3 years ago, and it's worth reading.
The chips are learning. ;-)
Woah, I experienced a bit of a optical illusion towards the end there. Looked like the LED display was sliding to the left when you paused the LFSR, anyway good stuff.
RandomSnot , yep.
That feeling when your stomach contents want to exit.
That was fun 😁😂
Julian, could you get back to soldering kits?
Metric system would be nice!
What bargraph display did he use??
Cool!!!
"Two trillion..." INCORRECT! It's two hundred and nine *billion* , nine thousand and twenty-three *million* two hundred and fifty-five thousand five hundred and fifty-one.
You're one of those long form weirdos
Prehistoricman We’re majority in the in the world. Plus, lots of mathematicians, like me, think the long scale is superior for a number of things that are explained in Numberphile’s video: ua-cam.com/video/C-52AI_ojyQ/v-deo.html. Ah, and I think he made a mistake even in the short system.
Then it will be a Time Machine you'll be wanting Julian. Now where's that Flux capicitor?
Ok my head is going to explode 🤣
Very good except for one thing. You're using that transparent breadboard and that makes it difficult to see the pin layouts. Please 86 that thing.