Thio.

Revisit the subject off how computer memory works.

If you remember from a few previous computer files several years ago, we looked at how we could store information, a single bit of information in an electronic circuit.

So we took the classic nor gate.

We took two of them in fact, Phegley and or gauge on.

We wired them up.

So the output of one fed back to the input of the other on Bryce of Bertha.

And then we had to import scene that could be used to reset or set the output.

And we had to outputs of with the opposite of each other on this circuit.

If you build, it could go watch.

Previous episode would be to be set to be one.

If this picture was taken one and he would remember that had been set or you could reset it.

If this bit was taken to be one, which point it would then go to be false or zero until you changed it and that circuit I would remember a single bit of information, and then we saw how we could build that up and use multiple off them to store lots of information.

We gave each one and address, and then we arrange them in a grid, and we split that address up to refer to the row and columns so we could access each individual bit now at stores, single bits.

And we could have multiple ones that we can address individually by giving it a binary address.

And if we wanted to store multiple bits, all we do is have multiple one of them want to store each bit and run them in Paris.

We'll concentrate on using a single bitch.

But if we look at a normal memory trip is a SIM from long ago but is actually 12345678 different chips on here that he used to store each of the individual bits that make up a bite.

We can build this circuit its doors a single bit of information, but we're using four transistors minimum to store each bit, which seems a lot.

And is there a way we can store it using less Now, this type of memory is what's called static graham, because once you've stored the bit, it will remember that until the power is turned off.

As we said before, we could arrange these into acreage.

And so they say, We're going to get store four bits a story one bit that's during another, that storing another that's storing another, and then we can address them from the row.

So that should be rosy.

Rose to be robe Ron.

This would be column zero this would be column born and then we can refer to each individual cell on.

We could expand that out.

We could go to four by fourth on DSO on, and now we can refer to any of the 16 things by giving it address those 0001110011011111 over here.

But as we said, we need to use four transistors for each of these cells, and that's going to sort of set however much silicon we've got, that's gonna be the limit of how much around we can get in there unless we can get the size of this down.

So is there a way we can build a memory circuit, which uses less than four transistors?

The answer is yes.

but it comes with a problem.

The advantage of building static memory like we saw before is as soon as it's turned on and set.

It remembers that until he turned the power off.

If you want to reduce the number of transistors used, we have to compromise and say, OK, we're gonna let the memory forget we're gonna build a memory circuit that inherently is built to forget that the information is stored and this is what we use inside all computers these days, the memory that you're building will forget what it's stored after about 64 milliseconds.

How does that work?

Well, the way we do that is by using not just the transistor, but a transistor and a capacitor.

So we build each memory cell rather than from 4 to 6 transistors.

We build them from one transistor on one capacity and the ones that work what I've got here wide up is an led resisted Just stop it exploding all the current getting too grateful to be more accurate on a switch so I can turn it on and off.

And so the circuit is relatively straightforward.

We've got a power source.

I have a switch.

I have an early D giving off light, and I have a resistor, Probably 3 30 homes.

That classic thing.

When I close that switch, that light comes on.

But that's also modify that circuits line.

He said that we've got a capacitor in there.

What we're gonna do is we're gonna wire it up.

So we've got zero votes I returned down here.

We've got plus five volts here, and we're gonna put the capacity in parallel with the Lodz.

Signora, Switch on.

We're gonna have the l E d.

On the resistor would alongside that will have the capacitor as well.

So that's why that up and we should see a difference.

So when I turn it on now, when I press this election light comes on when I let go, it goes off for those interested at home.

I'm using a 470 micro Farid capacitor here.

So we put that in the circuit and now when I press the button, the light comes on.

But when I let go, it gradually dies down.

Click it and the light guys down there.

What's going on here?

Well, the capacitor is a component that stores charge So as I close this switch, current flows down for the other day, but it also flows into the capacitor and its stores.

The charge charge accumulates between these two plates, so the current flows down here in lights he led when I close the switch.

But it also flows down, starts to charge the top plate of the capacitor so that when I let go the switch, the charge can then flow around this to former circuit.

But the charges on your small amount on it, as we see it, decays over time and it starts to run out.

But that capacitor Ashley stores whether it's five volts or zero.

But if we don't charge it, then it zero vaults for its five volts and install that for a short period of time in that period of time is dependent on the type of capacity.

If I change this for 47 micro foward capacitor, sure, I get it the right way round the wise that might be an explosion.

The same thing is happening as a slave.

My show is changing.

The capacitance of the capacitor changes the amount of time that the information is stored for a larger capacitor stores it for a longer period of time.

But we've got a problem.

We've got a circuit here which can store information for a short period of time on the capacity is typically used in dynamic RAM circuits in the computer.

The recommendation is it'll last for about 64 milliseconds.

We're a bit longer, but that's no good.

We want our computer to remember things.

So how do we get around it?

Well, let's look at the circuit again.

As I press it, we can see its decaying.

And if I want you to store that, I could say that I was the on.

I'll press it again and I'll press it again and I'll press it again.

And every time I press it, it starts the decay process from four brightness again, and as long as I can refreshing what's stored in there and keep pressing the button, then it stays on and it stays on and it stays on.

And that's exactly what the computer does.

Every 64 milliseconds.

It looks at that a bit of information that's stored on their works out, whether it's zero or one that's been stored and stores it again haven't.

64 minutes.

Seconds later it looks at it and stores again and on and on, on and on and on and so on.

So we've got a circuit now that considerable information for a short period of time, but we need to refresh it.

Otherwise, we lose the information in that.

Actually, this is not that much different from the first computer core memories that we use way back when when When you read the information, you destroyed it.

So you have to store it back in again.

So it has president.

But we can use this as the basis other thing because we can build a capacitor in a single transistor in much less space on our silicon chip.

Then we use for transistors to do it.

We store in exactly the same arrangement.

And rather than having a physical switch like I had, that we use a transistor is a switch.

So each of ourselves will become a capacity which is connected to ground, and we have a transistor and then we can put them into a grid arrangement, as we had before and again we can address them as they rose 0011011 for the road and 00011011 of the column.

But we have another problem with the static ram chips that we had before they store five volts or zero vaults with a capacity, as we saw its decaying as it's going on.

So rather just having the values coming straight out the capacitor.

You need to have some extra circuits here, which we call the sense amplifiers, which take the value from the bit on dhe converted to being a zero or one at 05 Also, whatever system computers using.

So we can't just take the value directly like we could before when he had extra bit of circuitry in there to amplify it to the right levels on the way.

These arranged is that you select a particular row off the dynamic Ram circuit.

Let's say we want to select Rose 01 and the sense amplifies and then switched to be connected to that and whatever values are stored in there, let's say we got 1010 then sort of read from there, and it produced out.

It's the output at the end, and this leads to an interesting property because whenever we select a value from memory in Dynamic Ram, we have to copy this through the sentence amplifiers.

And then it's buffered using afraid to be some sort, static Graham type stuff after that, which means that it takes longer to select a new row than to switch between a different problem.

So once we've selected a row, so be very quick to switch from reading on this road column.

002 columns They were one, but we'd have to reload the things if you wanted to go and read from bro 00 column 00 So if we can speed things up slightly by reading across the row each time and actually, if you use a CPU cash, you can pre load your cash lines by using that.

But what we can also do is well, as we've read, they're saying we can sort of feed the values back in to refresh our circuit.

And this is called dynamic Graham.

Refresh on dhe.

You need to build that in.

If you're gonna use dynamic ram in your computer.

What most computers do these days.

You need to build that in to your actual system on these various ways.

You could do it.

You could do it literally.

In software, the original son one micro workstation actually had code that refresh the dynamic ram every 64 milliseconds.

It was an intricate was fired, and it just went through reading the values from each of those locations which caused the ram chips to update it.

A problem with that is, if you probably crashes, then you can't debug it because your whole member gets wiped.

S o these days is actually usually built into harder.

And actually, what will happen is after a certain amount of time, 64 milliseconds, the hardware will go through and refresh each of the bro's of the dynamic ram in turn to refresh each of the memory cells is the data that's stored that now.

Actually, in reality, you probably won't do it all in Mongo.

You stop it regularly and do a row at a time and go through.

Then you start again each of the way through, and there's lots of support built into various chips to do.

This is an interesting thing about this, though.

When we build static Graham, the data is stored until the power goes out.

As soon as the power goes off, then the value is stored.

We're using capacities here, and so it's an interesting side effect in that perhaps can show it with this.

I haven't wind this up, but I'm gonna turn my light on in, charge it.

And rather than let him go the switch, I am going to pull the power supply out.

And even though there's no power into the circuit because that capacitors built a charge as I remove it, then it's still stores.

A bit of data on this actually happens in your dynamic ram chips.

If you pull the power where a static ram would forget everything, the dynamic Ram chips store the data for a short period of time.

And interestingly, while the dynamic ram chips manufacturers say you need to refresh it after six before 64 milliseconds is up, in reality, it can sometimes be a bit longer.

I've seen computers where this has managed to stay for over 10 seconds after the Paris put for machines and actually one of the security risk for computers that the capacitance changes depending on temperature.

So if you get hold of someone's computer and make it very, very cold and reboot in a particular way.

You have a good chance of being able to recover data that was in that memory so we can start off by putting an and gate in here connected to that.

So now we have when a in contain zero we're going to do is try and find the peak okay of these, these bunnies.

So, actually, we're starting off in focus here, and we've got a value of about five million.