Not sure if better but you could probably "stack" a pair or two of '04 inverters for each 20ns range. A 74HCT04 has typical propagation of 8ns per gate, so two would add to 16 or so. The problem is, it might be too temperature dependant and only work properly if cold/warm. Worth a try though - HCT chips are very easy to obtain and cheap too. Keep in mind you need even number of inverters for each leg. 74HCT00 should be 10ns per spec but the caveat is this depends on the manufacturer of the chip. You can easily make an inverter out of NAND/NOR gate by connecting the inputs together.

These delay line chip actually seem 74LS14 with some pins snipped off, "repackaged" and sold for much more money.
74LS14 are Schmitt triggers which have a delay of typically ~15ns, 6 cascaded of them make the ~100ns you want here.
The small tolerances do not really matter for the creation of the DRAM timings (RAS, CAS etc), I used the LS14 myself for my DRAM cards back then in the 1980s.

Thank you!

For the 74HCT04 I think you need too many of them to get 100 ns at the end?

For the 74LS14 the delay seems not to be very specific? Its between 5 and 22 ns according to the datasheet. But if it doesnt matter it seems to be interesting.

How about SN74LS31? Think it should also be possible to get the required values? Think I have to take a look at the boards again and see which taps are really required.

But I seem to be right there is no drop-in replacement anymore...

I'd advise agains LS chip as delay lines. The H->L transition is much faster for TTL tech than L->H. HCT is much more uniform, better idea I think.

Deunan wrote:

I'd advise agains LS chip as delay lines. The H->L transition is much faster for TTL tech than L->H. HCT is much more uniform, better idea I think.

As the signal gets double inverted each stage, wouldn't this even out?

Predator99 wrote:

For the 74LS14 the delay seems not to be very specific? Its between 5 and 22 ns according to the datasheet. But if it doesnt matter it seems to be interesting.

Well this unspecificity made me wonder when I searched for a way to generate the delays I needed, and this was why I first used the LS14 in 1981. Instead of just cascading the inverters to get a roughly appropriate timing (as I did it back then) you might use a complicated contraption of wires and capacitors to get more precise timing, like described in the delay line's data sheet.

I don't think you need to take the effort of using the LS31, as the delay line package dimensions clearly indicates it was no '31 inside.

The LS14 was quite commonly used back then on ISA memory expansion cards, so I guess it might be worth a try.

retardware wrote:

As the signal gets double inverted each stage, wouldn't this even out?

Yes, it should - but inverters are not the only gates that can be used as a delay line.

Took again a look at it and LS14 seems to be a good choice. Only need to do 5 simple wire connections and the result is pin-compatible to the delay line.
I have ordered some and will try - thanks again! 😎

Dont have a soldering iron here, therefore I made a quick&dirty fix 😈

Result here...:

Suntac 80286 Mainboards

Thanks again!