There are a range of integrated circuits that most of us would regard as definitive examples of their type, devices which became the go-to for a particular function and which have entered our collective consciousness as electronics enthusiasts. They have been in production since the early days of consumer integrated circuits, remaining in use because of a comprehensive understanding of their characteristics among engineers, and the job they do well.
You can probably name the ones I’m going to rattle off here, the µA741 op-amp designed by David Fullagar for Fairchild in 1968, the NE555 timer from Hans Camenzind for Signetics in 1971, and a personal favourite, Bob Widlar’s µA723 linear regulator for Fairchild in 1967. There may be a few others that readers will name in the comments, but there’s one that until today it’s likely that few of you would have considered. Texas Instruments’ 5400 and 7400 TTL quad 2-input NAND gate has been in continuous production since 1964 and is the progenitor of what is probably the most numerous breed of integrated circuits, yet it doesn’t trip off the tongue when listing famous chips, and none of us can name its designer. So today we’re turning the spotlight on this neglected piece of silicon, and trying to bring it the adulation it deserves.
Can you name this anonymous IC designer?
![We could have featured a minicomputer such as a DEC PDP-11 as an example of a CPU built from 74 series logic. But the 74-driven Xerox Alto makes a greater point about 74 logic as the progenitor of modern computing devices. Joho345 [Public domain]](https://hackaday.com/wp-content/uploads/2018/08/768px-xerox_alto_mit_rechner.jpg)
![The die of a decapped National Semiconductor-branded 5400 from the 1980s Robert.Baruch [CC BY-SA 4.0]](https://hackaday.com/wp-content/uploads/2018/08/national_semiconductor_dm5400j_8838_die_image.jpg)
A 7400 today
Over the years there have been a dizzying array of 74-series compatible families that address the shortcomings of the originals with capabilities far exceeding them, and though many of the more esoteric devices are no longer made it remains available in a huge range of functions. It’s still perfectly normal to find a 74 derivative in a device manufactured in 2018, and no doubt it will continue to be so for the forseeable future. The original series may have been long-ago superseded, but they remain in production and you can still order a 7400 from all the usual sources. So we did just that, picking up a few TI SN7400s as part of a Mouser order. It’s interesting to note that at about a couple of dollars each in single quantities these are no longer a cheap part, while they would have cost tens of dollars at launch in the 1960s they have evidently passed their peak-production low price point and are now heading into the realm of small-production-run exotica. It seems that the HC series CMOS part 74HC(T)00 from the late 1980s is now the cheapest single-quantity equivalent coming in at about 30 cents.
A quad 2-input NAND gate is hardly the most exciting or exotic of components unless you have a pressing need for a bit of logic glue, so having secured a few we could hook up an LED or make an astable flip-flop if we wanted to. But the interesting story lies probably not in what can be done with a 7400, but in why we would now use a 74HC00 or any of the other families that superseded the original. Newer series almost all have either or both of higher speed and lower power consumption in their feature sets as you might expect, but they have also all addressed the original’s inherent flaw. The earliest logic gate families were digital circuits designed to represent logic 1 and logic 0 as a high voltage or a low voltage, but they also had a point during the voltage transition between logic states during which they functioned more as an analogue circuit than a digital one. The original 74 series shares this with its near-contemporary 4000 series CMOS.
In a digital circuit containing 74 series devices the moment during logic transition at which both output transistors were open caused a spike of high current to pass, which had the unintended side effects of making circuits extremely noisy, and requiring significant power supply decoupling efforts. It’s worth remarking at this point that unintended analogue properties in the 4000 series chips became popular in the world of experimental synthesisers as demonstrated by our own [Elliot Williams] in his Logic Noise series.
The 7400 then, a neglected survivor from a pivotal moment in computer design that has soldiered on unnoticed for over fifty years. You will no doubt still be able to buy 7400s for years to come, though we’re at a loss as to who would specify them in a new design, and for decades to come you’ll certainly be able to buy its derivatives. We hope this has shone a light upon it and accorded it the recognition it deserves, and in turn if we could put a name to its designer that would be a fascinating story in itself. Now, if you’ll excuse us we have a small pack of brand new 7400s to think of a project for.
7400 header image: Stefan506 [CC-BY-SA-3.0]
When I need an inverter and can use the 7400 in a circuit, I’ll use two of the unused inputs and one NAND becomes an inverter without using a 7404.
Both NAND and NOR are called the universal logic gates, because all other types of gates can be made with just one of these two.
One NAND can become a NOT, two can become an AND, and three can become an OR. So the quad IC always has your back.
And 4 NANDs become an XOR.
Or even a flip flop
AND, 3 rights make a left…
4 NAND can also become a 2 to 1 Multiplexer.
Sorry, I reported your comment with an accidental click (and you can’t undo it).
I have a tube of them next to my workbench. Inverters are about the only thing I use them for! lol
The local makerspace has a few hundred of them in the parts bin. I keep trying to come up with a simple, useful, educational project to use them for, but I haven’t got there yet. Usually I abandon an idea when I realize that it would take a breadboard the size of a full sheet of plywood, and that it could be done with a $1 uC instead.
Maybe there should be a 7400 contest, with a limit on additional components. Something to encourage designs that would make use of all the new-old stock of these things sitting around in bins.
It’ll be hard to top https://hackaday.com/2013/03/21/nandputer-is-mostly-wiring/
A solution to the Thompson Hack. Build everything from scratch including the software :)
I recently used one for the fist time to multiplex an SPI bus, plus turning one of the gates into an inverter
Yup, still pretty much the standard way to multiplex the SPI select line. In the board I’m turning on today I did it with single-gate NAND’s, simply because they’re smaller.
Great post. Reminds me why there no longer are Electronics/Hobbyist magazines.
Hackaday and it outstanding writers/contributers have replaced the magazines.
Sadly, I just cleaned up and threw away my 80s cheat sheets. They featured pinouts of the most common 74s, some eproms, transistors and connectors like shugart, apple2 slot, c64 game port, V.24 and many more.
All gone. Yesterdays‘ knowledge. Grampa tells stories about the great war.
We now get ourselves a Raspi (why?), install some linux (we need an OS?) and do bit banging using python scripts, occasionally written with scratch. Flashing LED, how cool. The rest of the day we use the Raspi to watch cute cat videos.
Well, the Gigatron is a new design based on the 7400-series…
This may have been the first logic IC I ever dealt with, in a Radio Shack hobby kit with those spring-based terminals. Looks like that one’s outlived Radio Shack.
Same here.
Tandy / Radio Shack Science Fair 200 in 1 Electronic Project Lab had a 7400 (quad 2 input nand and a dual j-k flip flop.
The 150:1 only had a little black sip that looks like an amplifier
The 100:1 had a one inch square white with clear covering thingy which included trace resistors and a discreet transistor
the 65:1 on didn’t have anything labled an integrated circuit ….
(garage sale finds!, except the 100:1 :)
I had the 100:1 as well. My ten (ish) year old self was slightly disappointed that the white module wasn’t a “real” IC.
74ttl is totally obsolete and there is no good reason to use it. 4000 CMOS however is still the same as it was 40? Years ago and still useful.
I wonder if I should hang on to my stash of MECL III chips?
Ditched all but a small museum amount of mine last year – I’m only a “recovering” packrat….
The PDP-11 came around a bit after TTL. The PDP8 started as kinda-rtl and DCD (diode-cap-diode) logic, but there was a newer version called the PDP8i you could move too once TTL came around – one of the few machines you could get in multiple logic families.
I disagree that ttl is totally obsolete – there are still uses and you can even buy single gates in small footprint SMD because of that (though those tend to be one of the other newer variants). I recently did as I just needed a couple of gates to make a fancy sample/hold peak detector for an X ray spectrometer and they were a good fit for that job (along with some much more modern comparators and fet switches). If you need to go medium fast, 4k cmos is the obsolete one – and despite the B versions with protection diodes, they’re still real easy to fry in a physics lab with a bit of “EMP” floating around, much easier than TTL. Main advantage of the CMOS is nil power consumption if they’re not doing anything, and maybe a super power wasting mode when you attempt to use them in a semi linear fashion (because there is shoot-through current in the output totem pole when you do that).
Use a bunch of 7400s and passives to create an easy-to-breadboard TTLECL interface! I still see ECL micros in parts bins, but I don’t have any other ECL components to test with so it would be an absolute PITA.
TTL is more robust against EMI and ESD than CMOS with a lower chance to bork your circuit just by touching it with a finger, and 5 V TTL is compatible with both 3.3 and 5 V logic input signals because the high level threshold is around 2.7 Volts vs 3.5 V for the CMOS part.
So the TTL part is more universal and handles more abuse at the cost of higher power consumption and lower circuit density.
For exactly this reason I have a bunch of new (hopefully, the’re from Ali) 74LS540 chips lying next to me.
Intended purpose is to interface between a Blue Pill (STM32F103C8T6) with Grbl and the optocouplers of some beefy stepper motor controllers. LS540 is inverting, but it can sink a lot of current to GND, and the opto’s are fed from a 5V rail which makes them invert it back to normal again.
LS also plays nice with the 3V3 outputs of the Blue Pill.
As this is a one-off personal project there will be no PCB, but old fashion vero board and enamalled wire.
CD4000 series is as slow as it was 40+ years ago, and doesn’t work well at low voltages. I do electronic engineering for a living, and we design in the 74LVC series stuff when we need popcorn glue logic, inverters, Schmitt triggers, etc.
The cool thing about 4000 series cmos is that you can easily put RC networks between them and create simple time delays. Years ago I won a bet with some other engineers I worked with. We needed a small, and simple turn on turn off circuit for our PA systems that we biamped. The popular amps of the day had a bit of a turn on and turn off thump that was not bad on speakers with passive crossovers but the HF horns in the bi amed setups did not like them one bit.
The runner up was predictable with a pair of 556 timers. I was a bit more off the wall, a bit more elaborate, and a lot simpler. My system did power on as a relay that clicked in and turned on all the low level stuff, one of the new opto 22 SSR’s that came on a few seconds later and tuned on the power amps, and a relay a few seconds later that connected the drivers to the amps. The first relay was initially an SSR, but early SSR’s leaked enough to light up the low level sturr that only took a few watts, so a relay was a better choice. On power down, it reversed the sequence. Drop the speakers, drop the power amps, drop the low level tuff. The entire thing was one hex inverter with simple RC networks, and a 2006 to drive the relays and SSD and status lights. I think I still have one of the “brains” from them kicking around someplace.
Contemporary 4000-series is slightly different to the original stuff.
For starters it has internal bleed resistors and won’t suffer ESD if you wave a charged hand over it (I did a thesis on this back in the day. The chunks that ESD could take out of IC internal traces were impressive when inspected under electron microscopes)
Secondly it has hysteresis on the inputs so you don’t get the analogue band that used to “grace” original 4000-series (and TTL and ECL) logic.
Who remembers having to decouple supply circuits with both electrolytic AND copious quantities of ceramic caps in order to ensure noise was properly shunted?
You can’t use the inverter for an amplifier anymore? That was promoted as a feature.
I recall a magazine article (April 1 edition) where Jim Williams built a class D amplifier out of a foot high stack of CMOS inverters soldered together pin-to-pin to get the output impedance down..
Am I the only one who thinks that chip pictured has two additional, unknown pins? (Just in case, 7400 comes in DIP14, not DIP16)
I count 14 pads. Vcc on the left, GND on the right. 2 gates above and 2 gates below, each with 3 pins.
The 16 pin chip at the very top is an artistic drawing, not a real chip :P
The de-capped 5400 that is real has the correct number of pads.
Those single gate 1G ‘little ligic’ chips are really useful.
I’ll generally use a SN74LVC1G00DBVR in favour of the 74HC00 because it makes the PCB layout easier/neater.
Yep, exactly what I was talking about in my comment above regarding the 74LVC series.
Between a letter to Rich Templeton, TI’s CEO, and these pages, the answer cannot be far away:
http://www.computerhistory.org/atchm/the-rise-of-ttl-how-fairchild-won-a-battle-but-lost-the-war/
http://corphist.computerhistory.org/corphist/documents/doc-47be05a3a124b.pdf
I recommend checking the following patents:
3,560,760
3,136,901
3,229,119
easiest way is to go to: http://www.pat2pdf.org which lets you input the numbers and it provides a downloadable Acrobat of the patent. The first above is the earliest TI I found on TTL type NAND gates. The second is an early RCA on NOR, OR and EX-OR. The last one is from Sylvania and is the only one showing the classic multi-emitter NPN transistors we think of with 5V TTL. R.E. Bohn is the inventor for the Sylvania patent.
patents must not have worked the same way back then – now days anyone inventing that type of stuff would patent it and kill of competitors for decades… Didn’t they know back then that patents were meant for killing innovation, not encouraging it?
Ian,
Think ‘cross licensing’. Users wanted second sources due to learning curve issues with chip fab. Manufacturers were kinda forced into cross licenses to serve the customer’s fear of committing to a single vendor that might fail to deliver. The company with the most patented parts rules the game, getting paid by customers and competitors. Or, something like that.
My first real job 40 years ago was with IBM designing VDUs. We were designing using one of IBMs technologies, and the prototypes were built with ‘sunburst cards’, where a 100 gate chip was exploded by the design system into TTL gates on a card. Of course, it was a card full of 7400 family logic. IBMs part number for a plain old 7400 was 2392700 – and you can tell I didn’t have to look that up; it’s burned into the memory.
NEDONAND is 8-bit homebrew computer entirely built out of many 74F00 chips (2-input NAND gates):
https://hackaday.io/project/9795-nedonand-homebrew-computer
Great article and I think highlights the work that many of us use but take for granted. BTW plenty of new designs using 74xx logic over at https://www.retrobrewcomputers.org/doku.php and many more other examples out there I am sure…
The 7400 didn’t have ESD protection diodes on the output (or input for that matter). This allowed the output voltage to be forced above VCC when VCC was at 0V without sinking much current (typically much less than 100nA at room temp) back into the output.
I actually used that feature in a CMOS RAM (remember the 6116?) battery backup circuit in a design from about 1982 or 1983.
This, of course, is not possible in the CMOS 7400 replacements which will clamp the output voltage to one diode drop above VCC under those conditions. It should be noted that a number of manufacturers came output with modified CMOS output stages in logic families from the ’90s and ’00s (e.g. LCX) that do allow the output voltage to go above VCC.
BTW, that was the last time I used 74xx series logic in a design.
Hi Allan, I’m currently dealing with the mentioned 6116 SRAM. Could you please elaborate on your backup circuit little bit more?
Nothing was more satisfying than coming up with an optimal solution using 74xx parts back in the 80’s. It was a giant puzzle and we all had the 74xx series memorized. People have it way too easy now, hardly challenging anymore.
I’m still sad that the 74154 can’t be had anymore, and never made it to 74LV.
Can’t you replace the 74154 with two 74xx138? The 74xx138 has 3 enable lines, in most cases you shouldn’t even need additional logic.
If you’re in Silicon Valley, get thee to HSC. They close on January 12th. They have a wall of TTL, many of which are NLA. Picked up some 154s and even a 74LS612 (Yes, an MMU in the form of a 40-pin DIP) a couple of weeks ago.
For audio folks they also had some NE566 VCOs (566, the VCO, not 556 our friendly timer).
No, this won’t help for anyone doing a new design, but if you’re looking to pick up some parts for your library that are no longer available at the major online suppliers, there’s no better time than now.
The intro to this got me thinking about the 555. In today’s age, Does it do anything that can not be done better and less expensively than what you can do with something else?
CMOS logic chips can be wired as RC or crystal oscillator, one shot time delays, amplifiers etc. I usually pick them over a 555 as the other gates can be used for logic too.
I have seen TTL version used, but they are ugly and not work as nicely as CMOS.
Great write-up.
Was thinking the reason NAND is used in so many digital-logic explanations (E.G. an SR Latch) was due to NANDs’ simplicity at the transistor-level… But the internal structure, shown above, uses 5 transistors! Hardly simple.
Am certain quite a bit of that circuitry is due to input/output requirements of TTL, plausibly also speed-related, and power…
But, from the looks of that, it might be, e.g. that an OR gate may require fewer transistors, and something like an SR-Latch would be much simpler to grasp using those, e.g. “if the input is high OR the output is high, then output High.” is much simpler than “If the upper input is high and the lower output is high, then the upper output goes low, forcing the lower NAND’s output high, which feeds back to the …” and “But, there’s no way to know the initial output state”…
So, 25+ years into digital designery, am drawing a blank as to why all those explanations are backed with NAND implementations. I understand *that* it *can* implement any other logic, but the question, I guess, is whether NANDs are really *used* at such a level, internal to more complex devices (again, e.g. SR-Latches) or whether there’s some other reason NAND-implementations are so ubiquitous in early logic training material.
“But the internal structure, shown above, uses 5 transistors”
4 actually, of which 1 is the actual NAND and the others form the output driver.
OK. But, again, surely the first-stage of the “output driver” varies depending on the logical operation’s implementation, and might even include part of the logic in it…
Looking at the internal circuits of the 7402 NOR gate shows the same kind of input transistor structure shown as V1 and V2 in the schematic above, with two sets of V1 and V2 transistors, one for each input. The V1s have a single emitter, and the V2s are connected in paralell, thus giving us the OR function. Looking at the AND-OR-INVERT gates such as 7451 shows this pattern in combination: Multiple emitters on the input makes AND, parallelled second transistors makes OR and the output stage gives the Inversion. http://www.ti.com/lit/ds/symlink/sn74ls51.pdf
Turns out, NAND (or AND-OR-INVERT) is the simplest circuit with TTL, and thus that is preferred. For CMOS the NAND and NOR are pretty much mirror images of each other (NAND has N-channel devices in series and P-channel devices in parallell, NOR has them the opposite way), and either one will be as good as the other for the basic gate.
Sometime around 1970, in between a couple RTL designs and a lot of TTL ones, I had a brief (and very unpleasant) encounter with Signetics Utilogic. Anybody remember that? Some types of gate could only source current, and others could only sink it. Some inputs needed to be fed current, while others needed to have it sucked out. The result was that a NAND could not drive another NAND, and the same for NORs. Oh, and, the devices ran really hot, too. Just before it was time to commit to PC layout, TI dropped the price on TTL from nearly 10 bucks a flop to under a buck (which mattered because our primary customers were radio and TV broadcasters, not the military). I tossed the Utilogic design and started over.
One sometimes useful 7400 trick is using 4 gates in one IC to make a short pulse generator operated by a spst switch. Each operation of the switch (usually a push button) produces one pulse of about 3 gate delays width. It is a combination of a switch “de-bouncer” and a “racy pulse generator”.
Handy but not repeatable from chip to chip
I used the 40H00 back in the 80’s as a selectable normal/turbo crystal oscillator. The 40H00 were wired as a MUX but with the 2 NAND wired as Pierce-Gate Crystal Oscillator. i.e. linearized with a 1M feedback between one of inputs and output while the other input is used as an enable pin.
Note: 40Hxx series were one of the earlier CMOS series long before the faster 74HC comes along.
The original NASA computers were made from NAND gates. Modules were Flat Pack design and had established reliability. I want to say MIT won the contract but not 100% on that fact.
Close – the Apollo Guidance Computer was made mostly of 3-port NOR gates.
This article made me start my 7400 collection (the original TTL 7400 quad 2-input NAND gate exclusively). So far I have chips from roughly 25 manufacturers all over the world :). Thank you, Jenny List.