Author Topic: Experimenting with TTL Cpu, 74LS chips, old vs New? Retro style switches?  (Read 7250 times)

0 Members and 1 Guest are viewing this topic.

Offline rwgast_lowlevellogicdesinTopic starter

  • Frequent Contributor
  • **
  • Posts: 659
  • Country: us
    • LowLevel-LogicDesign
I am interested in building Ben Eaters 8 bit TTL breadboard computer, I have never actually used 74 series logic for anything, with the functionality of all the chips out there these days and rarity of parallel buses these days, i have just never had a need. I have always been in to computer design and over the years I have wanted to build an old school style computer with a blinken lights and switch panel, along with a punch card reader and serial terminal. I decided to go TTL instead of z80 or 65c02, is because for some reason I am a real bottom up type of learner. A lot of this project is also about writing a ground up software stack too.

I plan to start out just following ben's design (and reading Petzolds "Code:Hidden language of computer hardare and software") and soldering through hole on veriboard, instead highly capacitive breadboard design. But once i know what im doing i would like to take it in the direction of the Magic-1, and make it do something useful while learning more. At that point I would probably want to use a proper PCB and surface mount 74AHC components to get a better clock speed.

Right now im on digikey making an order for parts and I have realized I have about 90% of the parts on the projects BOM. A while back I got a box full of 74 and 4000 series logic chips, But being fairly uneducated about the 74LS series im wondering if there speed/rise time difference between 74LS from the 70s & 80s compared to the newer TI ones sold on digikey today? If there is a difference what effect would having chips with different rise/propagation times have? I'm thinking I should just order all new chips but if it is just a waste on 30 bucks that wont give any better speeds/noise/power consumption then I would probably be wise to just use what I have sitting around in case I need it one day. Lastly what are the names of the type of switches used on PDP an IMASI8080 front panels? If its possible I would like to pick up the "duck bill" style switches on digi key as well.

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Y'mean comparing new 74LS?  Wouldn't think there's a difference...

74HC and HCT are compatible, HCT being fully compatible and HC being HC-output to TTL-input compatible (and TTL-output to HC-input compatible with a pull-up resistor, if you don't mind it being slow; but HCT is better to just drop in).  HC and HCT have identical outputs.

Also 74F, 74AC(T) and 74ABT are faster, if you start running out of propagation delays in critical paths.  Mind the signal quality, and provide good grounding and bypassing!

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline jfiresto

  • Frequent Contributor
  • **
  • Posts: 896
  • Country: de
I always feel some cognitive dissonance when people refer to 7[A]HC[T]... parts as TTL, but I have been a CMOS fan for the last 45 years. Could that misnaming be an issue?
-John
 

Offline greenpossum

  • Frequent Contributor
  • **
  • Posts: 408
  • Country: au
I always feel some cognitive dissonance when people refer to 7[A]HC[T]... parts as TTL, but I have been a CMOS fan for the last 45 years. Could that misnaming be an issue?

Well strictly speaking CMOS also uses transistors, they don't have to be BJT. And TTL is used to distinguish from RTL, DTL, ECL (which really would have to be renamed if ever there could be a CMOS equivalent), etc. But I understand your feeling of dissonance.
 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 17428
  • Country: us
  • DavidH
But being fairly uneducated about the 74LS series im wondering if there speed/rise time difference between 74LS from the 70s & 80s compared to the newer TI ones sold on digikey today?

There is no difference; new low power schottky has the same performance as old low power schottky.

Quote
If there is a difference what effect would having chips with different rise/propagation times have? I'm thinking I should just order all new chips but if it is just a waste on 30 bucks that wont give any better speeds/noise/power consumption then I would probably be wise to just use what I have sitting around in case I need it one day.

Newer logic families like ALS, AS, or FAST will provide an improvement in speed or power consumption but the selection of logic functions is more limited than with LS.  The same applies when moving from HCT to AHCT or ACT except static CMOS power consumption is practically zero.

Quote
Lastly what are the names of the type of switches used on PDP an IMASI8080 front panels? If its possible I would like to pick up the "duck bill" style switches on digi key as well.

They are just another form of toggle switch but they will not be cheap.  They were not cheap in the past either.

I always feel some cognitive dissonance when people refer to 7[A]HC[T]... parts as TTL, but I have been a CMOS fan for the last 45 years. Could that misnaming be an issue?

Well strictly speaking CMOS also uses transistors, they don't have to be BJT. And TTL is used to distinguish from RTL, DTL, ECL (which really would have to be renamed if ever there could be a CMOS equivalent), etc. But I understand your feeling of dissonance.

It does not help that the later generations of TTL use DTL input structures in place of the multiple emitter inputs usually associated with TTL.  And then some replace the input diodes with PNP emitter followers.

CMOS ECL should be SCL (source coupled logic) and that term is sometimes used but more commonly it is referred to as CML (current mode logic).
 

Offline rwgast_lowlevellogicdesinTopic starter

  • Frequent Contributor
  • **
  • Posts: 659
  • Country: us
    • LowLevel-LogicDesign
So would I gain anything from leaving all the 70s 74LS chips in the drawerer and ordering all the parts in HC/HCT format? As far as moving to the 74Axx series I have no plans to do that at the moment, a lot of this is a learning project in hardware logic and cpu design, so while building and designing the thing i'm not looking to the problems 74Axx chips will bring with them when point to point wiring without a ground plane.

As far as the switches go maybe ill look around at some of the retro PDP kits and see if I can figure out where they source those kind of switches. I was under the impression the were SPDT "duckbill" switches but that term doesn't really seem to be helping me. I dont mind paying a bit for them, I was just hoping I could easily source switches like that on digikey or something.

Offline duak

  • Super Contributor
  • ***
  • Posts: 1048
  • Country: ca
Regarding the switches, I have a C & K sample kit and one switch looks like the one in the attached image.  I think these switches are called "Bat Handle".   Their website doesn't show it and my sample doesn't have a number so I think it's obsolete.  If you go to http://www.surplussales.com/switches/swtoggle-1.html you should see a few that migh be useable.

Having worked with a couple of PDP-11 minis I think I can say that their switches are a bit wider and longer than the switch I refer to above.  I also assembled and used an Altair 8800 micro for my high school in the 70s that used subminiature toggle switches.  I remember keying in the bootstrap loader and other small programs using these switches.  It would have been nicer to have slightly longer and wider toggles instead of the little cylindrical ones they used.  I see that some manufacturers have slip-on handles - look in Digikey under switch accessories.

Most logic families, even LSTTL, will require a ground plane (or at least a mesh) when you get to more than a few chips or the signal length gets over a few inches.  It's a function of edge rates, loop areas, and current spiking on signal transitions.  The old 74STTL was probably the worst and could require heroic layouts trying to create a ground plane and bypassing when using simple two sided boards.
 

Offline jfiresto

  • Frequent Contributor
  • **
  • Posts: 896
  • Country: de
Regarding the switches, I have a C & K sample kit and one switch looks like the one in the attached image.  I think these switches are called "Bat Handle".   Their website doesn't show it and my sample doesn't have a number so I think it's obsolete.  If you go to http://www.surplussales.com/switches/swtoggle-1.html you should see a few that might be usable.

I think in the U.S. that they are more commonly called paddle switches: good old American bats being round. I am sorry that they have gone out of style as I always liked their feel.
« Last Edit: June 04, 2020, 05:24:14 am by jfiresto »
-John
 

Offline jfiresto

  • Frequent Contributor
  • **
  • Posts: 896
  • Country: de
Most logic families, even LSTTL, will require a ground plane (or at least a mesh) when you get to more than a few chips or the signal length gets over a few inches.  It's a function of edge rates, loop areas, and current spiking on signal transitions.  The old 74STTL was probably the worst and could require heroic layouts trying to create a ground plane and bypassing when using simple two sided boards.

The old 4000-series CMOS was a lot easier to work with. I think I would use that if I were building a discrete logic computer. It would be a little slow, but you could probably power it off bits of paper towel, lemon juice and some spare change.
-John
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
The old 4000-series CMOS was a lot easier to work with. I think I would use that if I were building a discrete logic computer. It would be a little slow, but you could probably power it off bits of paper towel, lemon juice and some spare change.

Just about SFA at 5V, yup.  Plus you can overclock it to a massive 15, 18 volts even. Hold onto your hats we're cranking this baby up to TWO MEGAHERTZ! :-DD

Aside from power consumption (who cares?) and a little more freedom from signal quality issues (it's not THAT hard), I don't know that it'd be all that worthwhile; if it ends up clocked at 100kHz or something, there's simply fewer things you can use it for; being able to compute something at all is one thing, but being able to compute something in real time is a real challenge.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline jfiresto

  • Frequent Contributor
  • **
  • Posts: 896
  • Country: de
The old 4000-series CMOS was a lot easier to work with. I think I would use that if I were building a discrete logic computer. It would be a little slow, but you could probably power it off bits of paper towel, lemon juice and some spare change.

Just about SFA at 5V, yup.  Plus you can overclock it to a massive 15, 18 volts even. Hold onto your hats we're cranking this baby up to TWO MEGAHERTZ! :-DD

You could probably manage around 6 mhz at those voltages and be perhaps four times slower than the LSTTL equivalent. These days, I bet you could trounce a computer made with either, simulated in Python on a Raspberry Pi.
-John
 

Offline rwgast_lowlevellogicdesinTopic starter

  • Frequent Contributor
  • **
  • Posts: 659
  • Country: us
    • LowLevel-LogicDesign
@duak, Thanks a lot for that surplus link, the (SWT) 8V10128 model a quarter way down the page is exactly the kind of thing i was looking for and is WAY cheaper than a expected, but now looking at that page im wondering if the toggle switches with rollers on the end would be a cooler idea!

I have seen plenty of people breadboard pretty big 74LS projects with out any issues, most of the stuff I do is pretty low noise RF mixed with digital and hi power, so im pretty use to figuring out noise issue's, dealing with ringing etc. The thing is I don't want to deal with any of that crap while learning, hence why im not starting with the fastest logic chips I can. I would assume from some of the messy TTL projects I have seen that a good star grounding and decoupling rules of thumb like a .1uF on every chip along with a 1uf and 10uf for every 5 or 10 chips should be fine.

Still not sure if dumping the on hand 74LS and ordering everything in 74HC is worth the $50 bucks it will cost.

Offline jfiresto

  • Frequent Contributor
  • **
  • Posts: 896
  • Country: de
Quote
Still not sure if dumping the on hand 74LS and ordering everything in 74HC is worth the $50 bucks it will cost.

With 74LS, you might have to think about fanout, but beyond that, I bet you can tend to everything else with a little thought and little fuss.
-John
 

Offline tggzzz

  • Super Contributor
  • ***
  • Posts: 21227
  • Country: gb
  • Numbers, not adjectives
    • Having fun doing more, with less
I have seen plenty of people breadboard pretty big 74LS projects with out any issues, most of the stuff I do is pretty low noise RF mixed with digital and hi power, so im pretty use to figuring out noise issue's, dealing with ringing etc. The thing is I don't want to deal with any of that crap while learning, hence why im not starting with the fastest logic chips I can. I would assume from some of the messy TTL projects I have seen that a good star grounding and decoupling rules of thumb like a .1uF on every chip along with a 1uf and 10uf for every 5 or 10 chips should be fine.

"Ringing" isn't an appropriate concept, since that is associated with second (and higher) order filters.

If you are used to RF, you will know about transmission lines. Improper termination causes problems. The maximum wire length for which you can ignore transmission line effects depends solely on the transition time. For a quick intro, one place to start is https://www.edn.com/category/blog/bogatins-rules-of-thumb/

Power supply decoupling is required to deal with current transients due to changing voltage across capacitors and (with TTL, HTTL, STTL, LSTTL, FTTL) inputs consuming different current in the high and low state.

Decouplers need to have very short leads; you used to be able to buy flat capacitors that fitted under ICs and sockets with a built in capacitor under the IC.

Ideally you would have a solid ground plane and Vcc plane. On a two-sided board that would only be possible if you hand-wire all the signal connections.

For LSTTL an acceptable alternative is to have wide Vcc/GNd traces running under a row of ICs on the topside, and these should both be connected at both ends (don't have two interlocking fingers, one from each end). Put the 100nF decouplers on those traces by each IC. Add occasional "extra" grid grounds between the rows of ICs, to approximate a ground plane.

Wirewrap is OK for LSTTL but you have to be careful with faster families, and it fails for modern logic.

IDC is better than wirewrap, but I don't know if you can get the boards anymore.
There are lies, damned lies, statistics - and ADC/DAC specs.
Glider pilot's aphorism: "there is no substitute for span". Retort: "There is a substitute: skill+imagination. But you can buy span".
Having fun doing more, with less
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Honestly I'd stick with the TTL.  You have it, you can breadboard it today.  You may have more special functions on hand than are available in HC.  Do try to design some if you can, inventory what you have and make sure you have enough of various types to do it.  And you can always interface to HC and such with just a little buffering.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline tggzzz

  • Super Contributor
  • ***
  • Posts: 21227
  • Country: gb
  • Numbers, not adjectives
    • Having fun doing more, with less
Honestly I'd stick with the TTL.  You have it, you can breadboard it today.  You may have more special functions on hand than are available in HC.  Do try to design some if you can, inventory what you have and make sure you have enough of various types to do it.  And you can always interface to HC and such with just a little buffering.

Agreed. And the interfacing is even simpler with HCT, of course.
There are lies, damned lies, statistics - and ADC/DAC specs.
Glider pilot's aphorism: "there is no substitute for span". Retort: "There is a substitute: skill+imagination. But you can buy span".
Having fun doing more, with less
 

Offline duak

  • Super Contributor
  • ***
  • Posts: 1048
  • Country: ca
One of the most interesting problems I ever solved was in a manufactured product.  On occasion, an output bit on a register would change even though it wasn't supposed to.  It turned out that the data lines to the register were routed near the clock line to the register but because it was a two sided board, it wasn't all that close to the ground.  It was an unusually shaped board because it was a control interface to some part of the assembly.  When enough of the data lines changed at once, the clock line followed and caused the register contents to change.  The register was LS and the bus driver (on the same board) was LS.  One solution was to put a small cap on the register's clock input.  The other is to solder a metal ground plane made from a thin single sided piece of PCB  over the extension connecting all the ground points together and also laying close to the data lines.

Moral of the story - watch out for short glitches causing unintended clocking.  These can come from the most mundane sources.
 

Offline rstofer

  • Super Contributor
  • ***
  • Posts: 9964
  • Country: us
The PiDP-11 project uses a Raspberry Pi and the simh package to recreate a PDP-11/70.  Including the switches...

You might try to contact them for a lead on the switches:
https://obsolescence.wixsite.com/obsolescence/pidp-11

https://obsolescence.wixsite.com/obsolescence/contact
 

Offline rstofer

  • Super Contributor
  • ***
  • Posts: 9964
  • Country: us
Ben Eater's project is pretty nice!  I built most of it on breadboards until I just got bored with jumpers.  Some kind of prototype board and wire-wrap would be a lot better.

As to higher order machines, just about nobody is going to use TTL or any other chip family when they can build the whole thing in an FPGA.  I built an entire IBM1130 in a Spartan 3 Starter Board (now on a Nexys 2 Board) that runs the factory software, unchanged.  You can define the chips in terms of some HDL and wire them up with a keyboard instead of a jumper.

Software is the thing.  Without an assembler and some high level compiler, what good is the hardware?  I remember well when the Altair 8800 came out and it had just 256 bytes of RAM.  Every program had to be toggled in and, of course, they had to be tiny.  Today, you could grab the T80 core from OpenCores and implement a Z80 machine that would run at 50 MHz, maybe more.  You could even run the ROMs for PacMan.  I have that on a Nexys 2 Board as well.

I have wire-wrapped projects up to 100 chips of original TTL (predating LSTTL) and they worked well.  To do this, you really need a Gardner-Denver cut-strip-wrap gun.  Using the pencil type tool is too grim to contemplate.
 

Offline rwgast_lowlevellogicdesinTopic starter

  • Frequent Contributor
  • **
  • Posts: 659
  • Country: us
    • LowLevel-LogicDesign
I know this could be done in FPGA much easier, and at some point I may use 16v8gals or ATF1502 cpld's I have in the drawer. Really the point of this whole thing is to learn to efficiently use and think in logic gates (this way I can eventually move to an FPGA for some RF processing in other projects) and develop a software stack from the ground up and really understand implementing bare metal programming in assembler, which I feel would be more intuitive if im the one who built the hardware and understand its connections and mechanisms. The second goal is just to have a cool blinken lights piece of hardware sitting around doing simple tasks, which can easily be upgraded if I want to learn more advanced skills like predictive branching or graphics blitting.

You know I have a lot of copper clads I bought on the super cheap when RadioShack went out of businesses, not to mention the huge sheets of rogers I use for toner transfers when I need a quick way to build a comb-line filter or LNA for frequency's above VHF. I could actually dead bug the whole thing if i need too. As far as ringing, a square wave is full of harmonics and if the trace inductance is to high im sure a TTL chip can ring or spike on the rising edge, while this may not be an issue to logic threshold it will create a bunch of EMI, am I right about this? I mean a large enough harmonic spike somewhere could couple in to something causing a bit flip? At worse it will make a noisy ground!

Offline tggzzz

  • Super Contributor
  • ***
  • Posts: 21227
  • Country: gb
  • Numbers, not adjectives
    • Having fun doing more, with less
Signal trace inductance is less of a problem than ground lead inductance. "Ground bounce" can be a real problem, since the low level voltage margin is only 0.4V
There are lies, damned lies, statistics - and ADC/DAC specs.
Glider pilot's aphorism: "there is no substitute for span". Retort: "There is a substitute: skill+imagination. But you can buy span".
Having fun doing more, with less
 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 17428
  • Country: us
  • DavidH
So would I gain anything from leaving all the 70s 74LS chips in the drawerer and ordering all the parts in HC/HCT format?

Compared to LS, HC/HCT would draw less power for close to the same performance.  AHC/AHCT and AC/ACT would also give a performance improvement over LS.  Of course the same applies to the improved TTL families like ALS, AS, and FAST.  FAST is my favorite of the TTL families but unfortunately it has a limited selection of functions.

Unless you need more performance, then HCT would be my first choice and it has full compatibility with TTL signal levels so you can include some LS parts if needed and replace them later or not at all.

Signal trace inductance is less of a problem than ground lead inductance. "Ground bounce" can be a real problem, since the low level voltage margin is only 0.4V

AS and FAST never reached their intended performance because of lead inductance in DIP packages.  If anything, the fastest CMOS parts were even worse because of the difference in how CMOS and bipolar transistors switch.
 

Offline tggzzz

  • Super Contributor
  • ***
  • Posts: 21227
  • Country: gb
  • Numbers, not adjectives
    • Having fun doing more, with less
Signal trace inductance is less of a problem than ground lead inductance. "Ground bounce" can be a real problem, since the low level voltage margin is only 0.4V

AS and FAST never reached their intended performance because of lead inductance in DIP packages.  If anything, the fastest CMOS parts were even worse because of the difference in how CMOS and bipolar transistors switch.

That tallies with my recollection, but one of the claimed advantages of FAST was that controlled edge rates minimised the problem.

Nonetheless, excellent PSU planes were required - and that is difficult with a homebrew construction.

Putting the powe pins on the opposite corners always was a pessimal choice.
There are lies, damned lies, statistics - and ADC/DAC specs.
Glider pilot's aphorism: "there is no substitute for span". Retort: "There is a substitute: skill+imagination. But you can buy span".
Having fun doing more, with less
 

Offline rwgast_lowlevellogicdesinTopic starter

  • Frequent Contributor
  • **
  • Posts: 659
  • Country: us
    • LowLevel-LogicDesign
Just a curious question, when using the FAST or the A families cant you still reap there speed benefits whiles using appropriately sized resistors to slow the rising edge slow enough in order to avoid issues? People do this to control the rise time of a FETs gate all the time, and aren't the newer chip family's based on FETs is that part of what makes there layout issues so much more crucial? Seems like if you were going to use them without trying to slow there edges the key is low impeadence ground, and smashing there SMT versions as close together as possible to avoid trace length.

Offline CatalinaWOW

  • Super Contributor
  • ***
  • Posts: 5571
  • Country: us
I used several computers of the blinken light era.  Switches used for register entry and memory loads varied all over the map so you can use whatever you like and be "period correct".  The ones I liked best were on the HP 1000 series of of computers.  Lighted push buttons.  Press them and they latched down and turned on.  Press again and they popped up and turned off.  I am sure they weren't and aren't cheap. 
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf