The separation of the clocks was because a lot of digital circuits back in the day used multiphase clocks (often generated by a Johnson counter). You could cascade the two stages with opposite clock phases to build a pair of 2-bit synchronous D flipflops (or put additional logic in between them). A lot of times, engineers back then wanted the not-Q outputs broken out, because they'd build all the logic with data selectors and AND-OR-INVERT networks, and needed both true and complement inputs. A 74151 could implement any Boolean function of 4 variables (and many more complex ones) as long as both true and complemented inputs were available. If they let you use a 24-pin package, a 74150 would do any 5-variable function (and lots of more complex ones). The later ones on the series had open-collector or tri-state outputs so you could make even wider functions by wire-ORing several together. I saw whole ALU's built mostly of 7415x and 7452/7461. There was a whole body of literature around doing things like balancing the workload among the clock transitions. All because the gates were so slow back then. I remember (circa 1979) trying to cram a 72-bit ECL barrel shifter (shift left, rotate left, shift right, shift right with sign extension) built out of JK flipflops and data selectors into one micropack. With the help of God and a shoehorn it all went in, but routing the thing was maddening!
Transparent latches are interesting in IC design because they need fewer transistors than edge-triggered flip-flops. When I took a computer design class in 1990, we used transparent latches instead of D flip-flops for this reason. We solved the transparency problem by using a 2-phase non-overlapping clock. Successive stages of latches were clocked from different phases, so they were never transparent at the same time. This reduced the transistor count by about 10% if I remember, which would have reduced the cost of silicon by about 20% if we were building real chips. When doing board-level design, D flip-flops are usually easier to use. Though oddly I've noticed Motorola 4000-series app notes from the 70's and 80's using transparent latches where I would expect flip-flops. I don't know why. Maybe in that old CMOS process latch chips were cheaper or faster than flip-flops, or maybe the app engineer just liked them. 🙂
Following up, I happened across an old Motorola CMOS data book that has schematics for the chips. A D flip-flop used 28 FETs and a D latch used 14. That's a bigger savings than I expected!
Wouldn't the Q and ~Q be perfectly out of phase when you applied some square wave to the D when C high? Just curious, this would be useful for creating differential signals, and C would be the "enable" bit.
Best part of my day. Settling into the couch with a drink and watching electronics videos
The separation of the clocks was because a lot of digital circuits back in the day used multiphase clocks (often generated by a Johnson counter). You could cascade the two stages with opposite clock phases to build a pair of 2-bit synchronous D flipflops (or put additional logic in between them).
A lot of times, engineers back then wanted the not-Q outputs broken out, because they'd build all the logic with data selectors and AND-OR-INVERT networks, and needed both true and complement inputs. A 74151 could implement any Boolean function of 4 variables (and many more complex ones) as long as both true and complemented inputs were available. If they let you use a 24-pin package, a 74150 would do any 5-variable function (and lots of more complex ones). The later ones on the series had open-collector or tri-state outputs so you could make even wider functions by wire-ORing several together. I saw whole ALU's built mostly of 7415x and 7452/7461.
There was a whole body of literature around doing things like balancing the workload among the clock transitions. All because the gates were so slow back then.
I remember (circa 1979) trying to cram a 72-bit ECL barrel shifter (shift left, rotate left, shift right, shift right with sign extension) built out of JK flipflops and data selectors into one micropack. With the help of God and a shoehorn it all went in, but routing the thing was maddening!
Thanks for sharing that! There was a lot of insightful engineering in the TTL days.
Happy Thanksgiving! Thanks for all your great videos.
Transparent latches are interesting in IC design because they need fewer transistors than edge-triggered flip-flops. When I took a computer design class in 1990, we used transparent latches instead of D flip-flops for this reason. We solved the transparency problem by using a 2-phase non-overlapping clock. Successive stages of latches were clocked from different phases, so they were never transparent at the same time. This reduced the transistor count by about 10% if I remember, which would have reduced the cost of silicon by about 20% if we were building real chips.
When doing board-level design, D flip-flops are usually easier to use. Though oddly I've noticed Motorola 4000-series app notes from the 70's and 80's using transparent latches where I would expect flip-flops. I don't know why. Maybe in that old CMOS process latch chips were cheaper or faster than flip-flops, or maybe the app engineer just liked them. 🙂
Following up, I happened across an old Motorola CMOS data book that has schematics for the chips. A D flip-flop used 28 FETs and a D latch used 14. That's a bigger savings than I expected!
I really like that SMD LED board you show there!
Bummer, I don't have the 7475 in my 74HC00s kit. That LED board is really cool.
I looked in probably the same set.
Whomp Whomp Whomp
SWPEET didn’t supply any in their set either.
😔
Thanks for the reminder of level-triggered vs. edge-triggered! Forgot the details of that. :)
Very Informative. Sounds like a chip used in a remote start for a car .
I applied 7475 to build DIY frequency counter along with 74160 and 7447, 30+ years ago.
Wouldn't the Q and ~Q be perfectly out of phase when you applied some square wave to the D when C high? Just curious, this would be useful for creating differential signals, and C would be the "enable" bit.
I did a lot of digital design back in the day with 74xx logic but never had a need for the 7475. I don't think that part had much use in real designs.
the 74hc174, 74LS273P and 74HC374 are also cool devices.