                             What the Heck is A...

                   "Half-Flash Analog to Digital Converter"

                               by David L. Jones

A common method of converting an analog signal into an associated digital value
is to use a "flash" Analog to Digital Converter (ADC). As you may know, a flash
converter is very simple in principle. All it does is to simultaneously compare
the analog input with a large number of comparators that are set to linear
fractions of a reference voltage. Thus, only one comparator will be on at any
one time, which is digitally decoded to produce a corresponding binary value
that represents the analog voltage at that instant.

Flash converters have the great advantage of fast conversion speed, but one of
their major limitations is that they require a large number of comparators and
associated decoding circuitry. In fact, a flash convertor requires 2
comparators and resistors (where 'n' is the number of bits in the digital
sample), not to mention the decoding circuit!. So, an 8 bit flash ADC requires
256 comparators, resistors & decoding logic all on a single chip!.

The following information will take an 8 bit ADC as an example. However, the
half-flash technique is not limited to any particular number of bits.

An alternate method called "half-flash" conversion can drastically reduce the
number of circuit elements required by a full flash convertor, but will only
reduce the conversion time by about half. This can often be a much better option
than other conversion techniques such as sucessive approximation. For the
mathematically minded, the resultant number of comparators & resistors is in the
order of 2^(n/2).

As the name may suggest, the half-flash technique uses dual flash converters
that only require HALF the total number of bits. Therefore an 8 bit half-flash
ADC requires two 4 bit flash converters, a 4 bit DAC and an 8 bit latch, as
shown in the accompanying block diagram. This makes it a lot simpler to
manufacture, and gives it a speed capability somewhere between full flash and
successive approximation.

As a 4 bit flash only requires 16 comparators, the half-flash ADC only requires
32 comparators and resistors as opposed to 256 in a full 8 bit flash. This
lowers the chip complexity and cost significantly.

The basic principle of operation is as follows :

The most significant 4 bit ADC converts the voltage from the sample and hold,
and produces a corresponding 4 bit digital code. This value is put into the most
significant latch.

This 4 bit nibble, representing a "low resolution" sample of the input, is put
into the most significant latch. At the same time, it is converted back into an
analog voltage by the DAC.

This "approximate" voltage is then subtracted from the sampled input voltage,
which produces an "error" or difference voltage. This difference voltage is then
converted into another 4 bit digital value by the least significant ADC, which
is referenced to 1/16th (2^(n/2)) of the reference voltage used by the most
significant ADC. This value is then put into the least significant latch.

The 8 bit latch now contains an 8 bit digitised value of the input voltage. This
gives the same result as a full 8 bit flash, but it just takes a bit longer.

One advantage of this technique is that if you require a faster speed and only
need 4 bit resolution, then the most significant latch can be read out straight
away, as it acts just like a normal 4 bit flash.

So with a half-flash ADC, you have the option of selecting between a fast 4 bit
converter or a slower 8 bit version for higher resolution.

Some half-flash techniques use only one 4 bit flash ADC and "reuse" it for both
the most significant & least significant conversions. This is done by providing
additional sample & holds, and then switching these internal sample values
alternatively into the one ADC.

So as you can see, half-flash ADC's are great fun. Go out and build one today!.
