Antifuse OTP ROM

[jonsmirl]

I have an interesting application needing large amounts of ROM memory. Antifuse OTP ROM (Al or W) is an old technology from the 1990s. I'm wondering what the effect of Moore's law has been on it for the last thirty years. Is it possible to make OTP ROM like this at high density now? What kind of capacity is possible? The useful characteristic of OTP Antifuse is that it doesn't need a transistor for a cell.

 

[Fred Chen]

I have an interesting application needing large amounts of ROM memory. Antifuse OTP ROM (Al or W) is an old technology from the 1990s. I'm wondering what the effect of Moore's law has been on it for the last thirty years. Is it possible to make OTP ROM like this at high density now? What kind of capacity is possible? The useful characteristic of OTP Antifuse is that it doesn't need a transistor for a cell.

I thought mainstream OTP is precisely breaking down a transistor's gate oxide!

It can be made using advanced node CMOS.

A 28nm 2T cell ~ 0.05 um2: https://ieeexplore.ieee.org/document/10183939

 

[DerDrache]

Since the 130 nm node, foundry processes have generally moved to copper BEOL, so Al or W antifuse are no longer possible. Cu BEOL requires barrier or liner layers so the antifuse process becomes impractical. Instead, foundries typically offer eFuses instead. These are basically poly or metal resistors that can be blown away by the programming current. Note though that these are fuses (normally short) rather than antifuses (normally open).

As Fred points out though, mainstream OTP these days are gate dielectric based. They generally have better area density and lower power requirements for programming, at the cost of additional complexity in reading/sensing as you allude to. However, base OTP libraries are often offered as standard now by the foundry or EDA vendor, and there are also third party IP vendors if you have more stringent memory requirements.