Close

Here's the transcript of what we talked about

A event log for HackChat topic: CMOS to TTL

Join us on the Hacker Channel to discuss all things 3.3V and 5V

sophi-kravitzSophi Kravitz 02/11/2016 at 17:161 Comment

Benchoff: Hey it's Bil! Bil Bil Bil Herd the computer nerd

Al Williams: Howdy Brian

Bil Herd: Just to be clear I dont get called that to my face.

Sophi Kravitz: :: laughing

Sophi Kravitz: alright: first Q: What are the differences between CMOS and TTL?

Bil Herd: Tobasco drinking contest! It will have to be based on the honor system tho...

jlbrian7: I had to switch to a pc, the tablet was killing me.

Benchoff: oh so peppercon...

Al Williams: Well truthfully, it isn't just CMOS and TTL -- there are tons of families with different characteristics

Sophi Kravitz: is there a chart?

Al Williams: And variations of even those

jlbrian7: The one I was working on is 75T. I think... i was more worried about where the wire were going.

Al Williams: Well there's two ways to look at it: inside and black box

Bil Herd: Some info in the video posts where I also showed the progression of the families as far as speed, etc.

Benchoff: https://cdn.hackaday.io/images/2325431455070140480.png

Al Williams: from the black box perspective what you want to know is what level is high and what level is low (on input) and how much you can drive on the output

Philip: The difference is the way the silicon is processed and how the chips are designed. Logically a 7404 (TTL) and a 74C04 (CMOS) do the same basic function

esot.eric: I found a nice graph some where, from TI I think, easier to visualize than a table... bbiaf

Benchoff: That picture above is very much incomplete...

Al Williams: So a TTL gate might take anything below 2V as a zero and anything above 3V as a 1. On output it might output 0.7V and 4.3V. However, usually the amount of current on a "1" at TTL is very low compared to the amount of current it can "sink" when doing a zero

Bil Herd: To me the family difference starts with wether its a current consuming device on its input or a voltage sensing... except that that is a simplification.

Sophi Kravitz: https://cdn.hackaday.io/images/7027181455070253379.png

Al Williams: CMOS is more fair... usually 1/2 threshold (e.g., 2.5V @5V) and usually pretty equal sink/source currents

Al Williams: You can dig into the internals to find out why, but for most people the black box approach is better.

Al Williams: The sink and source currents get important when you fan out (connect lots of inputs to an output) or try to drive an LED or something else directly

Bil Herd: https://cdn.hackaday.io/images/7425201455070364009.png

Al Williams: the threshold voltages get important when you have noise issues although they can play in other things too including interfacing between families

jlbrian7: What is the difference between CMOS and BiCMOS, and how do you identify the two families?

Bil Herd: https://cdn.hackaday.io/images/4435701455070439189.PNG

Al Williams: Now you are in the internals

Bil Herd: TTL Input sinks current from the transistors typically

Bil Herd: https://cdn.hackaday.io/images/7393871455070487081.png

Al Williams: The way I like to explain is... is think of this way... a logic gate output is like two switches... one to + and one to ground (except ECL... nevermind)... only one switch is on at one time.

Bil Herd: CMOS usually means MOSFET which has an insulated gate that senses voltage and doesn't really have something like a sharing of current

Al Williams: Now the families depend on what are the characteristics of those switches

Bil Herd: Love ECL

Sophi Kravitz: what kinds of characteristics are there

Sophi Kravitz: ?

Al Williams: Well look at some old logic... the switch to ground would be a transistor

Bil Herd: And over time the line blurred between CMOS and TTL and the Fast TTL families that CMOS can do almost everything TTL can

Al Williams: and the switch to + would be just a resistor

Al Williams: so the current through the resistor had to be low... but the transistor could sink a lot

Al Williams: Then as Bil says.... what turns the switch on

Al Williams: a bipolar switch looks like a diode from base to emitter (or you have the emitter inputs like in that schematic above)... but a CMOS (or NMOS or PMOS) input looks like a 10M resistor to nowhere usually

Al Williams: and those all make it different

Bil Herd: The output of TTL is a totem pols, however the highside can only pull so high because its an emitter follower

Al Williams: Yeah I was going back to RTL

Al Williams: I just said older... so for TTL the top and bottom switches are bipolar

Bil Herd: https://cdn.hackaday.io/images/6884821455070700066.PNG

Bil Herd: Here is RTL

Bil Herd: You talk, I will try and find images. :)

Al Williams: And there are others like NMOS and PMOS where one switch or the other is a FET and the other is a resistor (or effectively so)

Sophi Kravitz: how are these things tested when built?

Bil Herd: NMOS was more popular due to the process having some better characteristics, so it could pull down strong but had some issues pulling up

Al Williams: So think about a CMOS inverter which is two FETS tied together (up a schematic or two)... that input is a piece of polysilicon material going nowhere so the resistance is VERY high (like 10M)

Al Williams: In the RTL Bil is showing the input is going to see those resistors and not be very high

Philip: Characteristics: Voltage , Current, Time, Temperature, ... Typically within each family of parts (i.e 74ACTxxx) there is very easy to work with compatibility. The problem is that not all functions are available in all families. So you need to mix and match. That's when you have to know your way around the parameters. For each combination there are some relatively simple calculations that need to be made, if you want a robust design.

Bil Herd: Ready for DTL?

Al Williams: I talked about threshold voltage... if Q was hooked to a resistor equal to R2 and Q1 is off then the output is going to be 1/2 V+ which is probably too low to be a 1

Al Williams: You bet

Bil Herd: https://cdn.hackaday.io/images/4400011455070870098.PNG

Bil Herd: The current sharing aspect of resistors was swapped out for a diode that had an on or off state

Al Williams: Here the diodes are switches of a sort too. If I forward bias either diode the voltage at R1/R3 goes way down. If I don't the voltage is nearly V+ (assuming the resistors are scaled right). Then the output stage is still like the RTL

Al Williams: So from a practical standpoint

Al Williams: you have to look at what happens when you glue these together

jlbrian7: Why are there two diodes?

esot.eric: OR gate...?

Arsenijs: Yup

Al Williams: Jibrian7 -- one for each input... either input going to ground pulls the R1/R3 voltage down

Al Williams: yes

Al Williams: So say you have a TTL gate driving a CMOS gate

Bil Herd: And gate, when both are high the transistor will turn on.

Philip: Testing: for simple parts with 74xxxyyyyy or CD4xxxxx part numbers the chip testers apply an exhaustive list of input patterns, and check the output(s) to make sure the chip is doing the correct logic function, doing it within the required time, doing it both hot and cold, doing it with the correct logic levels. Testing challenge is to do all that in minimal time (measured in seconds)

Al Williams: Demorgan's theorem -- it is what you want it to be if you invert your inputs right and treat your outputs right ;-)

jlbrian7: ok.

Bil Herd: After RTL and DTL came TTL where all transistors were used. The main thing slowing down TTL was the transistors were in saturation

Al Williams: The diodes do make a kind of OR gate with neg inputs but the transistor inverts the output

Bil Herd: So they used schotky diodes to keep the transistor just barely on.

Al Williams: So say you have a TTL gate driving a CMOS gate... the CMOS won't draw much current so that's ok

Al Williams: and a 0.7V output will be < 2.5V so that's ok for a zero

Bil Herd: Agreed, I had typed the and part before seeing yours.

Al Williams: and a 4.3V output is >2.5 so that's ok for 1

Bil Herd: https://cdn.hackaday.io/images/2839741455071198177.png

Al Williams: I was getting ready to say, then you get into the ones where they aren't 5V

Al Williams: in the old days you often ran CMOS at >5V to get higher clock speeds. Now you see stuff running at lower voltages like 3.3V

Bil Herd: https://cdn.hackaday.io/images/8474821455071257120.PNG

Bil Herd: https://cdn.hackaday.io/images/8622351455071261669.PNG

esot.eric: (funny they don't show 3.3V CMOS...?)

Al Williams: This works because the 3.3V output is CMOS

Al Williams: so it will go "to the rail" (practically) and 3.3V is over the TTL "1" threshold (barely)

Bil Herd: LVC is low voltage cmos

Al Williams: it might not be OK if you have lots of noise

Al Williams: but generally ok

Al Williams: in the top picture it isn't ok because you are going >the supply voltage UNLESS

Al Williams: the part has some special feature that allows it to have 5V tolerant inputs

Al Williams: that is not unherd of (no pun Bil)

Bil Herd: https://cdn.hackaday.io/images/6986211455071399500.png

esot.eric: (74AHC?)

Bil Herd: Removing the protection diode can allow certain parts to be tolerant of 5V in spite of being 3.3 powered to Al's point.

Al Williams: The diode clamps the input, hopefully with some series resistance

Bil Herd: https://cdn.hackaday.io/images/27571455071468362.PNG

Al Williams: and there are other ways to do that

Bil Herd: Right, AHCT is missing a diode, most CMOS is not

Philip: 74AHC = Advanced High Speed CMOS

Al Williams: So let's go back for a second

Al Williams: Notice the two FETs?

esot.eric: Bil, wait, (A)HC vs (A)HCT, nother topic ;)

Al Williams: They are the switches I talked about before... the little circle means that one is on when the other one is off

Bil Herd: Yup, my bad, misread

Al Williams: So let's pause for questions

Philip: AHCT is CMOS, but with the middle of the transition region the same as TTL rather than VDD/2

Adam Fabio: did i make it?

Bil Herd: Yes, the part I got ahead on was when there is a "T" in the name

Philip: That's for the AHCT inputs. The outputs are rail to rail

Benchoff: YO ADAM

Bil Herd: Hi Adam

Al Williams: Hi Adam

Adam Fabio: Hey everyone :) Just walked in the door

Al Williams: http://www.ti.com/lit/sg/sdyu001aa/sdyu001aa.pdf -- not a bad document about many different families

Bil Herd: Yes. And CMOS can supply lots of current these days and very symetrical meaning same amount hi or lo

Al Williams: Oops I think thats the wrong link

Sophi Kravitz: hey everyone- quick pause

Adam Fabio: long day interfacing 1.2, 2.5, 3.3, and 5 V families

esot.eric: I mentioned AHC because when running at 3.3V it can accept 5V input, as you were discussing

Sophi Kravitz: does anyone care if this gets reposted?

Philip: Hi Adam.

Sophi Kravitz: hi Adam!

Adam Fabio: hey sophi!

Al Williams: I'm fine with reposting Sophi

Bil Herd: Also in the video posts I try and run through the families including F, etc.

Al Williams: So did we have any questions other than Sophis?

Adam Fabio: Did we cover ECL?

Bil Herd: Thats a great doc Al, my images are from there though I changed the colors

Al Williams: (oops I'm too cheap for an apostrophe)

Al Williams: Well Bil and I wistfully mentioned we miss doing ECL

Bil Herd: We acknowledged ECL exists

Sophi Kravitz: we had a few Qs from other people

Al Williams: More like analog... you want to talk about it?

Sophi Kravitz: Is static discharge really that dangerous for CMOS chips?

Bil Herd: Not as much as it used to be

jlbrian7: I am good. When A is High Not A drives power to Q, and when A is low Not A sinks Q.

Al Williams: That's a religious topic but two things

Al Williams: a) not as much as it used to be as Bil says

Bil Herd: Still a statistical gamble to not remove ESD in a production environment

Al Williams: b) not so much in Houston with 99% humidity all the time

Bil Herd: https://cdn.hackaday.io/images/3847501455071894472.png

Al Williams: That being said, NASA and everyone else even here still observes ESD discipline

Adam Fabio: I've never used it myself, but ECL was used in my first logic analyzer. the pods would heat up enough to soften their plastic cases... was enough to scare me away from it

Bil Herd: Used to be that static or a high voltage could also cause an event called a latch up

Al Williams: ECL operates the transistors in the linear region where they are FAST

Al Williams: but also where they draw lots of Ic

Bil Herd: Yeah the early CRAY's had freon circulating to each chip, but honestly that was to keep them the same temperature so they wouldn't get one faster than another

Al Williams: I have a Gould analyzer that is all ECL on the front end

Sophi Kravitz: I have to jump out- we'll post this somewhere tho, great information tonight.

Al Williams: and I have done a few ECL designs but not lately

Bil Herd: And a lot of Differential Signaling is ECL'ish under the covers

Adam Fabio: Al - I believe the one I used was a gould

Sophi Kravitz: Thanks @Bil Herd and @Al Williams!

Al Williams: True

Bil Herd: For you, anything.

esot.eric: I've an old SGI computer that seems to use ECL quite a bit, especially for clock-routing

Al Williams: Yes a K100D probably

Al Williams: Thank you @Sophi Kravitz

Adam Fabio: You Rock sophi!

Al Williams: More questions?

jlbrian7: @Sophi Kravitz, thanks for setting this up.

esot.eric: indeed

esot.eric: (agreed)

Bil Herd: https://cdn.hackaday.io/images/5222401455072178250.png

Al Williams: You are on the spot... I had mentioned earlier we used to run CMOS up over 5V to get the speed up

Al Williams: like the old RCA1802

Bil Herd: This is from the doc Al pointed at, shows relationship between speed and voltage and the oldest parts tend to be high and right

Bil Herd: My boss was going to use an an 1802 for a battey app until we showed him that a floating point calc took 250ms

Al Williams: Then you didn't do it right (fixed point is your friend Bil)

Al Williams: heh

Kevin: Ah, the 1802. Nice chip in its day. The first computer system I designed and built for myself used one.

Bil Herd: Heh, we were doing 3 pole laplacians

Al Williams: Me too Kevin

Bil Herd: Cosmac elf

Adam Fabio: So for the newbs who inherit piles of 74HC and 4000 series parts from friends and coworkers, should they use the old stuff, or order newer families?

Adam Fabio: (I say use it)

Al Williams: I can still hand assemble/dissassemble Hex for the 1802 and my first articles were in QuestData

esot.eric: LOL, piles of HC's? How about F's and LS's ;p

Kevin: Adam, they still have their uses.

Al Williams: If old stuff suits your purpose use old stuff

Bil Herd: Hard to find nixie drivers and BCD based stuff in new stuff

Philip: Floating point is for people who don't understand their algorithms.

Al Williams: there is nothing magic about newer stuff at this level... like Bil showed... lower prop delay, less power consumption, etc. but there's nothing you can't do with the old stuff

Al Williams: we went to the moon on less ;-)

Kevin: I still have two drawers full of old IC's. One is full of TTL and the other CMOS. Most of the parts are about 30 years old.

Bil Herd: We were doing weighing instrumentation with 20 bits of resolution in the late 70's. We need the precision to have a critically dampened low pass funtion

Al Williams: I don't like to talk about out how old my parts are lol

Adam Fabio: I've had people ask me that question in person a few times - as recently as last week, because we're tossing PILES of old stock at the day job.

Bil Herd: Are your pins tarnished?

esot.eric: (Random side-note, I found that LS's can run reasonably at 3.6V and output LVDS-compatible signals, weird discovery)

Adam Fabio: LOL, tarnished pins are an issue.

esot.eric: bah, that's why the sockets scrape 'em off ;)

Kevin: yea, some are, some aren't.

Al Williams: Well.. Bil... Adam... Brian? Anything else you want to add or have we flogged it to death?

Adam Fabio: You guys did great - I'm jealous. I only got like 1 question during my drone chat

Al Williams: Are you saying you droned on? ;-)

Bil Herd: I'm good. I do a series of posts and videos as listed at the top somewhere.

Adam Fabio: haha :) It was just a quiet night here :)

esot.eric: didn't see that listing

Bil Herd: On this topic. Including taking a picture of an old die with my kids microscope

Bil Herd: https://cdn.hackaday.io/images/3734461455072600877.png

Al Williams: https://hackaday.com/author/williamherd/

Adam Fabio: Hey Bil - remember that ROM robot i emailed you about?

Bil Herd: hang on I will get links

Bil Herd: https://hackaday.com/2015/06/30/gates-to-fpgas-ttl-electrical-properties/https://hackaday.com/2015/08/03/how-cmos-works/https://hackaday.com/2015/09/02/how-cmos-works-some-final-words-about-cmos/

Bil Herd: https://hackaday.com/2015/05/28/from-gates-to-fpgas-part-1-basic-logic/

Discussions

Eric Hertz wrote 02/17/2016 at 05:54 point

LOL I think you should've included the timestamps here... These folks must type 300WPM!

  Are you sure? yes | no