Detailed logic family differences? LS vs HC vs AH etc.

[gorplop]

Hello,

I'm trying to find a document that describes the uses and differences between the logic families (LS, HC, HCT, AC, ACT and so on). Ideally with some rough input/gate schematics, a short note on why it's done this or that way, advantages, tradeoffs etc.

First i found this: http://www.ti.com/lit/sg/sdyu001ab/sdyu001ab.pdf , which seems to be more of a marketing material summary about TI's products. I also found several links like http://www.quadibloc.com/comp/cp01.htm but these generally seem to stop at TTL vs CMOS differences.

After some more searching i gave up and decided to look at the datahseets for the same IC in different families. I chose 74xx04 since that is what I am experimenting with.

The 74LS04 datasheet has a nice single inverter schematic (even three!).

Then I looked at the datasheets for 74HC04 and to my surprise, neither TI nor other 4 vendors (NXP, Phillips, Fairchild, Diodes Inc.) whose datasheets I found decided to not include anything more than a logic symbol Same for HCT. At least TI included a nice comparison chart between HC and AHC, which is useful

Then I went onto looking at the 74AC04 datahseet and hit the same wall, there are no schematics, the only drawings are package dimensions (4 pages of package outlines...)

So, does anyone know an application note or other PDF that would outline the general behavior and usage of the various families and could link it? Or even point me in a good search direction?

Cheers

Footnote, I've found several threads on the internet that link TI literature but the links are all dead now, even these from 2018 (!) (way to go TI! )

[excitedbox]

I thought I had read somewhere (wikipedia?) that they just denote who manufactured them.not that there is an actual difference. Maybe that is why you can´t find anything about what the differences are?

[Zero999]

The 74HC series are buffered,  so the 74HC04 will really be three inverters in series.



Unbuffered parts are often available and are denoted by a U in the part number, i.e. 74HCU04.

[Gyro]

Unbuffered parts are often available and are denoted by a U in the part number, i.e. 74HCU04.

Useful for 'semi linear' applications like crystal oscillators.

[tggzzz]

"The Art of Electronics" chapter 10 discusses some of the more significant differences.

If you are starting to ask such questions which need detailed and subtle answers, then you are ready for TAoE and will benefit from it in many ways. See many other threads on this forum.

[Zero999]

Unbuffered parts are often available and are denoted by a U in the part number, i.e. 74HCU04.

Useful for 'semi linear' applications like crystal oscillators.

Yes, that's true. The problem with the buffered parts is three gates makes a ring oscillator, with an output frequency dependent on the propagation delays. A 74HC04 biased in its linear region can oscillate at a random high frequency.

[gorplop]

Thank you for your answers.

I thought I had read somewhere (wikipedia?) that they just denote who manufactured them.not that there is an actual difference. Maybe that is why you can´t find anything about what the differences are?

You are thinking about the prefix of the part no. Different manufacturers used different prefixes, especially the eastern block makers.

"The Art of Electronics" chapter 10 discusses some of the more significant differences.

If you are starting to ask such questions which need detailed and subtle answers, then you are ready for TAoE and will benefit from it in many ways. See many other threads on this forum.

Thank you, looks like I need to pick it up again. last time I finished reading it after the chapter about opamps. I have more free time now.

Yes, that's true. The problem with the buffered parts is three gates makes a ring oscillator, with an output frequency dependent on the propagation delays. A 74HC04 biased in its linear region can oscillate at a random high frequency.

That's a very valuable tip.
So would a 74HCU04 as a ring oscillator made from 3 inverters have more predictable frequency?

[SiliconWizard]

Unbuffered parts are often available and are denoted by a U in the part number, i.e. 74HCU04.

Useful for 'semi linear' applications like crystal oscillators.

Yes, that's true. The problem with the buffered parts is three gates makes a ring oscillator, with an output frequency dependent on the propagation delays. A 74HC04 biased in its linear region can oscillate at a random high frequency.

Yep!
And more "modern" variants, such as the "low-power" AUP series, suffers from the same potential issue.

[rstofer]

I haven't kept up with the changes since 74LS simply because I don't often use discrete logic.  However...

I might make a spreadsheet with rows for important parameters like Vdd range, logic thresholds, fan in, fan out, max sink current, max source current, switching speed, etc.  Then just add a column per device type.

In the end, I don't care how the chip works internally as long as I understand how to connect the pins.  OK, add a row for buffered versus unbuffered.  I haven't run into that just yet.

[excitedbox]

Here is a list of the sub families from wiki. There is also a table of at least some of their specs in the families section

https://en.wikipedia.org/wiki/7400-series_integrated_circuits#7400_series_derivative_families

I found some more info on the 7400 and the 4000 series

https://www.nutsvolts.com/magazine/article/understanding_digital_logic_ics_part_2

https://www.nutsvolts.com/magazine/article/understanding_digital_logic_ics_part_4

[Zero999]

So would a 74HCU04 as a ring oscillator made from 3 inverters have more predictable frequency?

Only if you put an RC circuit between each stage to slow the propagation delays down to a known time constant. If you just connect three gates together, the frequency would be no more predictable.

The problem with the 74HC04 is, in a design such as a crystal oscillator, when all that's required is an amplifier, the parasitic ring oscillator can select the frequency, rather than the crystal. The 74HCU04 only has one stage, thus less gain and no ring oscillator, so it should oscillate at the crystal's resonant frequency.

[David Hess]

A 2 or more input gate will be more representative of how input structures vary between logic families because multiple inputs are handled in different ways.

Older CMOS databooks include transistor level schematics.

[T3sl4co1l]

It's been my experience that inputs within a gate, tend to share ESD structures, so there's a small amount of hFE between inputs.  The hFE to one or the other supply is usually much more significant, though (enough to cause latchup above, say, 100mA in 74HC family -- coincidentally, these are rated "free of latchup below 100mA" ).

I don't happen to have any >2 input gates handy, so I don't know how it varies between adjacent or distant inputs.  Whole gates seem to be separate enough they have no measurable interaction.

This isn't useful for the most part, but a case where it might be of interest is a low power, gated RC oscillator: you might be using a schmitt NAND with an RC feedback, and the other input to enable/disable it.  If the /EN pin is subject to current injection on the order of 1mA, the other pin will experience leakage currents on the order of 30uA, skewing timing.

Another consequence is, injected current is actually shunted to the opposite rail, as much as it's clamped.  The ESD diodes are actually emitters of parasitic BJTs, and with hFE ~ 1 they effectively cascode about half the current to the opposite supply.  (Or was this more like 0.3, I forget; I've got the models on my website.)  So injected current isn't just shunted to the respective supply, it's actually increasing total chip dissipation!

Tim