Control Engineering

Link This | Email this | Blog This | Comments (0)

How does Flash memory work?
April 7, 2008

Flash memory cells individually resemble metal-oxide-semiconductor field effect transistors (MOSFETs) with an extra electrode. N-channel MOSFETs consist of two highly doped N-type silicon spots (source and drain) in a lightly doped P-type substrate connected to ground. Electrical connections are made to these spots, and a thin non-conducting dielectric layer covers all. A metal or other conducting gate electrode overlays the dielectric between the source and drain. Applying a positive voltage to the gate sets up an electric field that drives the P-type charge carriers (holes) away from the region between the source and drain, creating a non-conductive depletion region there. Above a threshold voltage, the electric field becomes so strong it creates an inversion layer dominated by N-type charge carriers that pass current between the source and drain.

The additional electrode in a flash memory cell is a floating gate, completely surrounded by insulating dielectric. Any charge the gate picks up will be trapped there. In such case, the field of those trapped charges will modify the MOSFET threshold voltage. Trapping positive charge, for example, can significantly lower the threshold voltage needed to create the inversion layer and turn the transistor on (conducting state). A trapped negative charge makes it significantly harder to form the inversion layer.

Flash memory cells resemble MOSFETs with an extra electrode — the floating gate.

Suppose we say having the transistor on equals a 1, and having it off equals 0. With no charge on the floating gate and a control gate biased positively, but below threshold, we apply a small voltage between the source and drain. No current flows because the region between source and drain is completely depleted. The electronic circuit returns a 0.

Suppose, however, we have previously stored a one in the cell by applying a big negative voltage to the control gate — enough to set up an electric field strong enough to drive electrons out of the floating gate and across the thin insulating barrier into the P-type substrate below via, say, the tunnel effect. This leaves positively charged holes on the floating gate.

The field from those positively charged holes superposes with the field from the positively charge control gate to form a conductive inversion layer. The transistor conducts, and the circuit reports a 1.

All that remains is to address the particular cell. Setting the cells up in a rectangular grid with the sources connected together in rows, and the drains connected in columns. Each cell is then uniquely defined by its row and column numbers. Making row 357 slightly positive and column 492 negative uniquely addresses cell (357, 492). Applying a positive voltage just under threshold to all the control gates reads out its contents.

That is one scheme. Others are possible, such as grounding all of the sources except those in row 357, which you run into a current pickup, then apply a positive voltage to column 492. This will also uniquely read out the contents of cell (357, 492).

Flash memory gets its name from the erase scheme used. Simply applying a high negative voltage to all the gates in unison charges all of the floating gates positive, loading 1s into all the gates. To store data, the system selectively drives the charge off the floating gate to form 0s.

Driving charges across insulating barriers is a fairly violent process, which leads to flash memory’s most debilitating limitation: limited number of write/erase operations. Every time the system writes to or erases a cell, it slightly damages the insulating barrier. Eventually, the cell becomes useless.

On the positive side, the number of write/erase operations flash memory cells can endure is in the tens or hundreds of thousands, so Flash is quite useful for applications where the memory contents aren’t changed often once they’re laid down, such as memory sticks, archives, and firmware storage.

Control Engineering covers flash memory developments.

Posted by Charlie Masi on April 7, 2008 | Comments (0)