When is it appropriate to mix 74LSxx components with original TTL 74xx?

Question

I spend almost all my tinkering time with 1970s era logic circuits, which use almost all original 7400-series TTL. I do a fair bit of repair and am now starting to design additional logic to work with old circuits, and I am a little fuzzy on the risks/rewards of using (say) 74LS series parts in these designs, either as substitute/replacements or in new parts of a circuit.

The chart on this page indicates that the 74yy00 variants largely vary across propagation delay and power consumption:

My sense is that the LS variants are generally appropriate to substitute for the basic versions, with the general caveats that:

1. An existing circuit design might implicitly rely on certain propagation delays and that using parts with faster timing could alter the behavior undesirably.
2. I guess you could use more power than expected if you swap in just the S variant instead of LS?

Are there other, more functional or more concerning reasons to beware of intermixing, or specific other things to watch out for?

I want to not be foolish about these variants in old circuits but at the same time if a 74LS04 works just as well as a 7404, they're certainly cheaper and easier to find (and some of the logic chips never existed in the basic version, only in LS, etc.) AS and ALS (per the chart) look even better to me if I'm not sensitive to propagation delay, but again, I don't want to be blind about it.

Answer 1

In general, it's usually safe to mix 74xx logic families, with the following potential pitfalls:

• The TTL-based 74 series have logic levels concentrated around the low end of the supply voltage, while the CMOS ones (other than the xxT CMOS series) have logic levels spread more completely throughout the supply range. This means that a 74LS driving a 74HC, for instance, might not be able to drive the HC's input to a point that would read as high. The converse is not true; a 74HC output can easily drive a 74LS input with no issues. To compensate for this, the CMOS-based 74HCT (and other xxT designations) are specifically designed to have TTL input levels, and can be driven directly from TTL outputs.
• The output stage of the 74, 74S, 74AS, 74LS, and 74ALS series uses a resistive pull-up, meaning that they can safely drive an LED, BJT base, or similar load directly from the output without current-limiting. The output stage of the CMOS-based 74C, 74HC, 74AHC, 74HCT, 74AHCT, 74LVC, 74LVCT, and 74AUP series uses an active pull-up, and attempting to drive an LED-like load directly from the output will break either the load or the logic chip.
• Some of the more recent series might be too fast. By this I don't mean in terms of input to output delay (though that can also cause problems), but in terms of output slew rate. PCBs laid out with original 7400 chips in mind might encounter crosstalk or ringing issues from the very fast edges of 74HC or other more modern parts.
• Different logic series have different rated fanouts and different input loads. If one 74xx output is driving 10 inputs, that chip probably can't be replaced with a 74LS (which can't drive as much output load), though it would be fine to use a 74HC (which can drive significantly more output load). Back in the day, logic chips were rated in terms of how many "standard logic loads" (equivalent to one 7400 input) their outputs could drive, and how many "standard logic loads" their inputs counted as.
• Logic oscillators rely on IC delays to work. If you swap out a 7404 being used for a ring oscillator for a 74HC04, you might find that suddenly the entire board stops working because you've drastically increased the clock frequency. While oscillators like this are very unusual today, they were common in the heyday of 74-series logic.

Answer 2

One potential problem with substituting LS for regular TTL is the reduced fanout.

Regular TTL can typically drive 10 TTL loads. LS TTL can typically drive 5 TTL loads (8 mA). I use 'typically' here to refer to the guaranteed drive of a typical output (ie. excluding buffers and such like with higher output drive than normal).

Of course timing could be an issue if the circuit is pushing the limits (or is badly designed). Many older circuits used questionable asychronous techniques or even pressed the parts into some kind of linear-type operation such as oscillators. In that case, substituting types may well lead to interesting issues. Much faster parts may require better layout and bypassing than slower parts.

Answer 3

There are some designs where exact type of the chip may be required, as envisioned by the circuit designer:

• I have seen a 8 MHz oscillator consisting of three NAND gates connected in a closed circle, without any additional elements. It would obviously work at different frequency if the gates have more or less delay, or maybe not even at all.
• I have seen 5V voltage stabiliser where a single NAND gate that produces the reference voltage. I actually made such a stabiliser, and it worked fully OK at least at stable room temperature.
• Oscillators and Schmidt triggers consisting of the two gates connected sequentially (and indirectly in ring with positive feedback) were also quite common, almost standard. The required values of the additional discrete elements likely depend on the electronics inside the gates.
• Level indicators, flipping the gate from 0 to 1 at the threshold input voltage. Multiple gates were often connected sequentially to steepen the response.
• Also, some otherwise truly digital circuits are unusually loaded (like driving LEDs or powerful transistors). If the chip is rated for the lower maximal output current, it may not pull out.

Answer 4

One of the references I was thinking about in my comment is TI's Logic Guide, sdyu001ab from 2017. This is available on TI's web site, and provides a lot of good information on the compatibility of various logic families. Some of old S (AMD) and FAST (Fairchild) documents from the 80's have similar information.

Here are a couple of snippets from the TI document: