feedback with diodes?

[ranch varment]

So, I need feedback in my system, or my problem is impossible.

If I make another parallel route from the signal line, I cant guarantee which direction itll cycle,  but I can use a diode, to force the output back into the input.

Does this work -  and is there any way to do it without a diode or transistor?

[ranch varment]

Im going to make you guys so angry again. hehe

I just thought again,  and unless there was 2 feedback points connecting together.  (so there were 2 ways to feedback to the same input) then i dont think i need to bother with a diode because it isnt powered on the feedback.  (because its a different place from the power that instanciated it.)    So  if there were 2 to the same bit, if I wanted to have no diodes, id have to repeat the input, because otherwise im risking shocking backwards into the system, so then id need a diode on both.

If it were only the only feedback position, then i wouldnt need a diode, because it actually hasnt got current there unless it gets here from the output.

Im going to try and make transistorless and diodeless ram,  using this idea.    so im just not allowed to reuse a feedback input,  and then it should work!

[amyk]

You really need to go learn basic electronics if you want to have the slightest chance of being understood, because none of your posts make any sense.

[nitro2k01]

On a serious note. No, just no.

On a more joking note. Just use vacuum tubes or relays, or even something more exotic like acoustic delay lines, and voila you can make transistorless RAM.

[ranch varment]

You just dont understand my idea,  because its hard to explain.

Maybe it wont work,  because without diodes maybe it just 100% conducts and it wouldnt work.     But ill draw it up anyway!  Because its my dream.   

[c4757p]

Yes, please draw a schematic. But this time, don't do it with your feet.

[ranch varment]

Ok I drew it up,  and it didnt look like it was going to work.

I think I need a switch ok,  but -  I want to control the whole computer from just one.   That I want to tick all my ram loops over just on one transistor.   Has this been done before?

[Zero999]

Been smoking more dop ay?

It's possible to make RAM with relays, capacitors and diodes. In fact you can make a whole computer like this.

http://www.northdownfarm.co.uk/rory/tim/tim-8.htm

[tggzzz]

Im going to make you guys so angry again.

Nah. This thread, and especially some of the short responses (e.g. "feet"), are quite amusing. With luck we'll soon approach some of the comments that used to appear every September before 1993.

Those too young to remember can see https://en.wikipedia.org/wiki/Eternal_September

[miguelvp]

Just implement a core memory and be done with it.

[ranch varment]

Thanks for showing me that video,  and I think im bitten by the want to make it completely from scratch bug, too.

Thats why I can seem annoying, I guess.     I guess maybe I did think at first that I was thinking things that other people havent thought about, and im wrong about that.

But, Just because people have worked out 100 different methods to multiply numbers, I dont see why it isnt useful and skillmaking to work out a way that someone else worked out, all by yourself.  -  Even though its reinventing the wheel,  Its my opinion to think its important to.

I want to finish up,  with my from scratch machine,  and I had to ask a million stupid questions before I could finally have it done finished, by my own hand.


So far im stuck with a machine that can react to a sensor or controls,  but it cant react to its previous state.   Im sure I could get that working (the former), but im not sure how difficult its going to be, to be honest.

[ranch varment]

I think im going to drill the instruction line straight into it,  like that guy does,  because it makes the whole thing simpler, and it pretty much handle anything even conditions as long as you turned off sets on a disabling condition,  just runs through the lot the same speed every time, regardless of the forks.

But I dont think making a general purpose computer is where you really win,  its custom logic catered to the task 0 and 1 and burn out your finished transform in a single cycle!

[hamster_nz]

I think im going to drill the instruction line straight into it,  like that guy does,  because it makes the whole thing simpler, and it pretty much handle anything even conditions as long as you turned off sets on a disabling condition,  just runs through the lot the same speed every time, regardless of the forks.

But I dont think making a general purpose computer is where you really win,  its custom logic catered to the task 0 and 1 and burn out your finished transform in a single cycle!

Don't get hung up on doing things in a single cycle.... That will make the whole design as slow as the most complex path through your logic.

On a CPU a 64-bit division can take maybe a hundred cycles. It is possible to design it so it takes one cycle, but that would require a. lot more logic, and the CPU cycle time would need to be much, much longer (50x?) to allow the signals to propagate thorough the additinal logic. There is no point having a CPU where addition is 5000% slower just so division is 2x faster....

Also, have a read of this:

http://american.cs.ucdavis.edu/academic/readings/papers/CRAY-technology.pdf

It may be 'old tech' but It has lots of ideas about the engineering challenges involved to make a really fast computer out of a couple of different types of very simple parts. Power, cooling, electrical noise, cable lengths - and this was for about 80MHz, if I remember correctly!

[tron9000]

So, I need feedback in my system, or my problem is impossible.

If I make another parallel route from the signal line, I cant guarantee which direction itll cycle,  but I can use a diode, to force the output back into the input.

Does this work -  and is there any way to do it without a diode or transistor?

what this should have read was

Hi guy, I'm trying to make some transistorless RAM but have hit a snag. Please see schematic:

(attachment)

I have tried X,Y & Z but it still doesn't work.

blah, blah, blah

many thanks.

Going straight into your problem without tell us what your try to do makes it pretty difficult to help you as most of us don't have crystal balls. just a little advice.

[Fungus]

Please take less drugs before posting....

[ranch varment]

Its got to get the job done however many transistors it takes, so whats the point in pulling the cycle before the jobs done. 

I dont plan on making on making a multiplier, or a summer, or a binary shifter, or other general purpose tools, I actually have planned hardware optical flow  (Comes down to a 3x3 1024 bit descriptor match as ive theorized it.) and if I dont get the "custom" job fully finished, whats the point in another cycle,  why dont I finish on the exact transistor, and get the exact switch speed to the setter,  then itll cycle itself.

On the sample.

All what you said just spoke to me as,  that custom ttl doesnt do shit over just an ordinary cpu because of switch speed?!

So that is a BIG KNOCKDOWN for me!

[tron9000]

Its got to get the job done however many transistors it takes, so whats the point in pulling the cycle before the jobs done. 

I dont plan on making on making a multiplier, or a summer, or a binary shifter, or other general purpose tools, I actually have planned hardware optical flow  (Comes down to a 3x3 1024 bit descriptor match as ive theorized it.) and if I dont get the "custom" job fully finished, whats the point in another cycle,  why dont I finish on the exact transistor, and get the exact switch speed to the setter,  then itll cycle itself.

On the sample.

All what you said just spoke to me as,  that custom ttl doesnt do shit over just an ordinary cpu because of switch speed?!

So that is a BIG KNOCKDOWN for me!

serious, and no offence, is English your first language? cos there looks like a communication barrier here.

Sorry its just there's no flag on your profile.

[Ian.M]

VLSI silicon logic can run fast because the interconnects are very short so have low inductance and capacitance and low propagation delays. Anything you try to build that uses discrete components simply isn't going to be anywhere near as fast for anything more than one or two gates.



Unless you have a budget for this comparable to NASA's, you'd do better to learn VHDL then get the absolute screaming fastest bleeding edge FPGA development board that you can afford/lay your hands on.

[ranch varment]

Oh I have a solution!

what if you can do it switchless after the switch then than run it a picoseconds maximum switch speed running down a nanosecond wire getting noninteractive samples to us but to the robot its developing?

And its brain dead simple, and noone intelligent gave a crap about it... but idiots cared alot.

Lightspeed is no cap on samples collected...  only interactivity.

[tron9000]

And its brain dead simple, and noone intelligent gave a crap about it... but idiots cared alot.

is that an insult? We are trying to help here.

you still haven't shown or told us your setup yet. If your switching signals in order of picoseconds through a rats nest of wires or on a breadboard, your going to hit problems of the transmission line kind.

[ranch varment]

I want you to think about a wire conducting as the signal transmition line.  or even more simply, just the switches in a serial stampede!  (Making good use of your production space.)

You send the switches in waves.

look man you can send send video from one end to the planet to the other, and i think it takes longer than a wire to get there, but i think the signal is still 60hz.

so well now i want you to use your imagination.  And im sending the peak frequency of the switch down preorganized combination of not ands (or switches) converging to a result.  And I can send this peak frequency all the way down the end and collect it at the end (and collect at PEAK frequency of the collection switch!) no matter how long it took to get there.

And then your stuck scratching your head how it works.

Dont worry man, you have to do it for machine learning, otherwise your motor wont evolve fast enough.
If this wasnt done in factories since the 50's wed only have one cheesy ipad a week.


Parallel takes up as much space as serial,  and if you aint stampeding your serial your just wasting space.

[Ian.M]

Actually, signal propagation through wires is slower than light or radio waves in free space, typically only about 2/3 as fast.   You are probably being confused by the extra time lag that relaying a signal via a geostationary satellite adds due to the vast distances involved (geostationary orbit being 35,780 Km above the Equator).

You can certainly transmit information with a bit rate fast enough that more than one bit is in transit down the wire or over the radio link at any one time, the problem starts when you have to synchronise more than one such bitstream that has arrived via a different path so you can do something useful like combine the bitstreams using logic or arithmetic function.   

The reason why asynchronous sequential logic 'hairballs' are no longer used for serious computation is the difficulty of keeping multiple signal paths in step so that you perform the desired operation on the correct pair of bits that originated at the same time, not one with the bit before or after its pair.

[ranch varment]

Looks like we got another head scratcher.

Well I guess you only get the input in the other side at the switch frequency, down to the on off frequency of one switch!, but when it hits the other side it just sets the state flop as it comes,  it sorta just auto detects itself... so i dont think any signal is missing, but i dunno i havent actually done it myself, but beating a womans light speed is piece of cake.

[tron9000]

beat a womans light speed.....buh....wuh...WTF does that mean???

[lewis]

Are we all failing a Turing test?