4 ms·

Great piece of history. Can someone explain the difference between TTL and CMOS chips?

kens 7y ago
I'm not sure how much explanation you want. The short answer is that TTL (transistor-transistor logic) is an older logic family, while CMOS (complementary metal oxide semiconductor) is what's used nowadays, such as in microprocessors. In more detail, first there were bipolar transistors (NPN and PNP) and later MOS transistor (NMOS and PMOS). Bipolar transistors are your basic semiconductor transistors with three layers of semiconductor. MOS (metal oxide semiconductor) transistors have an insulating oxide layer between the silicon and the metal (or polysilicon) on top. MOS transistors were developed later than bipolar, and started to become popular in the 1970s. One type of logic that you can make from bipolar transistors is TTL (transistor-transistor logic). TTL is better than earlier logic families such as resistor-transistor logic (RTL) or diode-transistor logic (DTL). TTL was cheap, reliable, fast, easy to use, and popular with minicomputer manufacturers. TTL has two main problems, though. It uses a fair bit of power, and you can't make it very dense. I.e. you can't put a lot on a chip. MOS, on the other hand, has the advantage that you can make very dense circuits from it. (I.e. Moore's law applies.) NMOS was used for early microprocessors such as the Z-80 and 6502. The problem with NMOS is that it uses resistors (sort of) in the logic gates, and these waste power. The solution was CMOS, complementary MOS. You use both NMOS and PMOS transistors (the complementary part), and you can get rid of the resistors, and your chip uses very low power. The problem with CMOS is it's more complicated because you need two types of transistors, and twice as many (sort of). However, the need for low power won out in the mid-1980s and microprocessors such as the 80386 started using CMOS. Use of CMOS has continued to the present. I'm oversimplifying the history somewhat. The books "To the digital age" and "History of semiconductor engineering" go into much more detail.
- gautamcgoel 7y ago
  This was great context, thanks for the explanation!
analog31 7y ago
A practical difference between TTL and CMOS is that a CMOS gate draws negligible current when it's sitting in one state (1 or 0). It only draws current when changing states. So CMOS circuits can run at much lower power, and this power consumption decreases as the gates get smaller. This in turn means you cram more gates onto a slab of silicon before heat dissipation starts to kill you. And it's kind of an invitation to miniaturize in the pursuit of both higher computing power and higher speed... nice. Now I'm dating myself, but there was a time period when CMOS gates were slower than TTL or even faster logic families like RTL. So a fast machine like the Cray was built with RTL, if I recall correctly. But eventually CMOS won this race.
- kens 7y ago
  The Cray-1 used bipolar ECL (emitter-coupled logic), which was very fast but power hungry. Thus, the Cray-1 needed to be immersed in Freon for cooling. You can think of ECL as differential amplifiers that steer current down one path or another. RTL (resistor-transistor logic) is entirely different. It was used in the Apollo Guidance Computer because that's all integrated circuits gave you at the time, but it's not very good. (The inputs affect each other because they are only separated by resistors. Even diode-transistor logic is much better.)
  - TheOtherHobbes 7y ago
    ECL is fast but hot because the transistors are never fully saturated, so current is always flowing. CMOS is more like the electronic equivalent of a stiff mechanical lever system. Logic voltages change the charge distribution inside the MOS transistors. There's almost no current flow outside of them. This is great for super-low power draw, but it's not particularly snappy, it can't drive LEDs directly, and it's not happy driving long wires without buffering. It's also very easily damaged by static electricity - which is why wrist grounding straps are a thing. Because the transistors are so tiny it's easy to scale it to VLSI. TTL is mid-way between the two. The transistors saturate cleanly, so switching is faster than CMOS. The source/sink currents are much smaller than with ECL, so power requirements are manageable, although switching glitches need good decoupling. It's a robust and reliable system, with low-power options. But it's not as dense as CMOS, which is why 74xx series TTL never got any more complex than a simple 4-bit ALU, while MOS soon expanded into full 8-bit VLSI microprocessors.

kens 7y ago

I'm not sure how much explanation you want. The short answer is that TTL (transistor-transistor logic) is an older logic family, while CMOS (complementary metal oxide semiconductor) is what's used nowadays, such as in microprocessors. In more detail, first there were bipolar transistors (NPN and PNP) and later MOS transistor (NMOS and PMOS). Bipolar transistors are your basic semiconductor transistors with three layers of semiconductor. MOS (metal oxide semiconductor) transistors have an insulating oxide layer between the silicon and the metal (or polysilicon) on top. MOS transistors were developed later than bipolar, and started to become popular in the 1970s. One type of logic that you can make from bipolar transistors is TTL (transistor-transistor logic). TTL is better than earlier logic families such as resistor-transistor logic (RTL) or diode-transistor logic (DTL). TTL was cheap, reliable, fast, easy to use, and popular with minicomputer manufacturers. TTL has two main problems, though. It uses a fair bit of power, and you can't make it very dense. I.e. you can't put a lot on a chip. MOS, on the other hand, has the advantage that you can make very dense circuits from it. (I.e. Moore's law applies.) NMOS was used for early microprocessors such as the Z-80 and 6502. The problem with NMOS is that it uses resistors (sort of) in the logic gates, and these waste power. The solution was CMOS, complementary MOS. You use both NMOS and PMOS transistors (the complementary part), and you can get rid of the resistors, and your chip uses very low power. The problem with CMOS is it's more complicated because you need two types of transistors, and twice as many (sort of). However, the need for low power won out in the mid-1980s and microprocessors such as the 80386 started using CMOS. Use of CMOS has continued to the present. I'm oversimplifying the history somewhat. The books "To the digital age" and "History of semiconductor engineering" go into much more detail.

gautamcgoel 7y ago
This was great context, thanks for the explanation!

analog31 7y ago

A practical difference between TTL and CMOS is that a CMOS gate draws negligible current when it's sitting in one state (1 or 0). It only draws current when changing states. So CMOS circuits can run at much lower power, and this power consumption decreases as the gates get smaller. This in turn means you cram more gates onto a slab of silicon before heat dissipation starts to kill you. And it's kind of an invitation to miniaturize in the pursuit of both higher computing power and higher speed... nice. Now I'm dating myself, but there was a time period when CMOS gates were slower than TTL or even faster logic families like RTL. So a fast machine like the Cray was built with RTL, if I recall correctly. But eventually CMOS won this race.

kens 7y ago
The Cray-1 used bipolar ECL (emitter-coupled logic), which was very fast but power hungry. Thus, the Cray-1 needed to be immersed in Freon for cooling. You can think of ECL as differential amplifiers that steer current down one path or another. RTL (resistor-transistor logic) is entirely different. It was used in the Apollo Guidance Computer because that's all integrated circuits gave you at the time, but it's not very good. (The inputs affect each other because they are only separated by resistors. Even diode-transistor logic is much better.)
- TheOtherHobbes 7y ago
  ECL is fast but hot because the transistors are never fully saturated, so current is always flowing. CMOS is more like the electronic equivalent of a stiff mechanical lever system. Logic voltages change the charge distribution inside the MOS transistors. There's almost no current flow outside of them. This is great for super-low power draw, but it's not particularly snappy, it can't drive LEDs directly, and it's not happy driving long wires without buffering. It's also very easily damaged by static electricity - which is why wrist grounding straps are a thing. Because the transistors are so tiny it's easy to scale it to VLSI. TTL is mid-way between the two. The transistors saturate cleanly, so switching is faster than CMOS. The source/sink currents are much smaller than with ECL, so power requirements are manageable, although switching glitches need good decoupling. It's a robust and reliable system, with low-power options. But it's not as dense as CMOS, which is why 74xx series TTL never got any more complex than a simple 4-bit ALU, while MOS soon expanded into full 8-bit VLSI microprocessors.