Flash Memory Emulation

In addition to ASICs and microcontrollers, memories also become obsolete.  One of our projects is a replacement for an obsolete Intel 28F010 flash memory.  This is a 1 MB, 5 volt, flash memory, and does not have any sectors.

Unlike our other products which use standard processes, flash memories use a dedicated process.  And once that memory has become obsolete, the wafer fab is usually converted over to a newer process so that it can make newer flash memories.

This causes customer problems since the newer flash memories use lower voltages, and frequently have different programming algorithms.  As a result, the customer is faced with a system redesign because the older parts are not available.

We can’t recreate the old process.  What we can do is to use one of our ASICs as an interface between the old application and a newer flash memory.  We then buy flash die, and assemble both the flash memory and our ASIC in a single package that now works in the old application. 

This approach has several interesting engineering issues.  First, we have to design a voltage regulator so that the external 5 volt supply is stepped down to the 3.3 volts required by the memory.  And it has to have a fast response to changes in load current, because when a flash switches to write mode, the current can increase from 10ua to 50 mA in a few microseconds.

Another problem is the write algorithm.  Flash memories have a write algorithm that is used to prevent accidental writes to the memory.  For this chip, we have to detect when the write algorithm occurs, and replace it with the write algorithm required by our chip.  And since there is no clock available, we have to turn on a ring oscillator and make our own clock.

A third problem involves the use of high voltages.  The Intel part uses ~13 volts to indicate the programming and to enable a special device mode to read the manufacturer’s ID.  Our part will self-destruct if a voltage greater than 7 volts is applied to it.  To get around this, we have to put an internal resistor between the VPP pin and our internal detect circuit that will limit the voltage that the part actually sees.

A fourth problem occurs when using our device in some programmers.  They will immediately remove the power from our part after the write is finished.  Unfortunately, we require a little time to finish our own write algorithm, and so we will fail in some programmers.  I don’t think that we will be able to solve that problem, and so we will have to provide our own programmers, or we will have to ask the various programmer companies to alter their algorithms for us.

Date

2014-02-03