The victory of the losers, or the story of the creation of FLASH memory (part2)
This type of memory was made in the form of a two-dimensional matrix of conductors, at the intersection of which a thin jumper (20-30 μm) was created from metal or amorphous silicon. The programming of the microcircuits consisted of passing a large current pulse through an appropriate jumper that caused it to melt (burst). The likelihood of recovering torn jumpers was minimal. For greater reliability and eliminating the possibility of restoration of jumpers, the chip was subjected to thermal training (additional heating for a long time), and then it was checked and, if necessary, reprogrammed. The percentage of microcircuits that could not be correctly programmed by modern standards was quite large (for different memory PROM microcircuits. The coefficient of successful programming was in the range of 0.65-0.9).
Reference. Once user programmable memory chips with burnable jumpers are a matrix with a uniform array of conductive jumpers connecting rows and columns at all points of their intersections. The jumper in the chip acts as a memory element. The presence of a jumper encodes a logical “1”, and its absence – a logical “0”, or vice versa.
This memory (PROM) soon gained wide popularity among computer developers as a means to simplify loading, debugging and changing the “firmware” of microcodes in microprocessor programs. Microchips PROM memory were produced in the USSR. Figure 3 shows the EEPROMs that were most used in amateur radio and digital electronics.
They were produced in different versions and in different cases. For the military industry, they were produced in ceramic cases, using the technology of using precious metals (gold) (Fig. 4.). In this regard, they are now very difficult to find.
The advantage of PROM chips was their fast read time, which was about 35 ns, and the disadvantage was the inability to modify or destroy information stored in this memory. In addition, this memory had a small amount of data storage on the chip.
Part II. Flash predecessor
The next step in the development of non-volatile memory was a random discovery made by Intel engineer Dov Frohman in 1971, thanks to which EPROM memory was created – the predecessor of Flash. Frohman’s discovery marked the beginning of a new era of non-volatile memory.
Figure 5. Dov Frohman – creator of EPROM memory
From Dov Frohman’s memoirs: “Having come to Intel in 1969, I wanted to work on the MNOS project, but Intel was just created, and in that case, usually you don’t have to work on what you like.
Reference. MNOS was an alternative technology for manufacturing field effect transistors for integrated circuits in late 1968. Later it became the main one for creating solid-state non-volatile memory.
At that time, the company developed only two products using the old technology. These were the chips: INTEL 1101 256-bit NMOS RAM and 3101 64-bit bipolar memory.
Reference. NMOS (Metal-oxide-semiconductor with an n-type channel) is a type of field-effect transistor in which the control electrode is separated from the channel by a dielectric layer – silicon oxide;
Block diagram of a NMOS transistor. 1 – metal shutter; 2,3 – source and drain areas; 4 – silicon substrate.
Intel 1101 chip and its topology
At the time when I came to work for the company, there was an acute question of creating a new memory chip of 1 KB. The main task was to decide by which technology we will make this chip, should it be hybrid (flip chip) or monolithic.
Reference. Flip chip (Flip Chip) – one of the technologies for the production of hybrid microcircuits, in which a pair of identical devices that are connected in parallel are placed in one case.
It was proposed two options for creating this chip, the first is to combine four 256-bit microcircuits with an already developed topology into one, that is, to make a hybrid microcircuit, or to develop a new version of the topology of a monolithic microcircuit with a capacity of 1 Kb. Intel executives and Tom Innes, of the bipolar device development department, decided to put hybrid technology on the flip chips.
At that time, I did not know anything about flip chip technology. We worked for two months, after which an integrated circuit was created, which consisted of 4 256-bit NMOS memory chips. After all the tests and studies of the new integrated circuit, we came to a meeting with the leadership. At a meeting with Intel executives, we asked a question: “Well, we have a working prototype of a hybrid microcircuit, but I would like to know which way we will move on?” Gordon Moore, Andy Grove were at our meeting, and most likely there was also Robert Noah…