We are using the AD725 in the following way:
RGB to composite video in NTSC mode.
Temperature: -40 degr.C to +85 degr.C
We need to know the actual bandwidth,because we got an requirement from our
customer witch are: The video bandwidth shall be a minimum of 4.5 Mhz at the 3
In the data sheet, the only spec for the AD725 Luma bandwidth for NTSC is
4.4MHz, typ. Please understand that the AD725 was designed as a low-cost
consumer part intended for use as the TV out feature of a PC or a video game.
These applications do not require tight video specs, so we only characterized
them to meet the market's need.
For proper operation the input RGB video signals must be filtered to accurate
limits. These filters are shown in the AD725 block diagram. For this case, the
filters in the Luma path are most appropriate.
The filters in the AD725 are active filters made with on-chip RC components and
op amps. As I'm sure that you are aware, the resistors and capacitors in an IC
process can vary quite a lot--as much as +/- 20% or more. This means that the
frequency response of the filters would vary significantly if this were all
that determined the bandwidth. Also, since the part needs to be low cost, it is
not possible to trim in production to get the required accuracy.
The approach taken to set the accuracy is called auto-tune. One of the features
of the AD725 is that it requires a very accurate frequency reference--the
subcarrier. Therefore, it is possible to measure the RC time constant on chip
and compare it to the period of the subcarrier frequency. Then, an on-chip
counter counts up or down to tune the filters. This comparison is made once for
every vertical period, or 60 times per second in NTSC mode.
The auto-tune RC elements are not the same ones that that control the filters.
They are just representative of the RC values on the particular device (IC).
Therefore, the accuracy of the filters on a chip depend on the matching
accuracy of the auto-tune RC elements and the RC components used in the
For this process, the matching spec for R and C elements on a given die is
+/-3%. There is also a +/-3% granularity in the step size of the auto-tune
circuit. This adds up to a total of +/-6%. Since the auto-tune circuit operates
all the time, it should be able to track the temperature changes, so there
should not be added tolerance variation due to temperature.
As a rough calculation, if the 4.4MHz spec is considered the average BW, then
with +/-6% tolerance, the minimum calculates to 41.3 MHz, and the maximum
calculates to 4.66 MHz.
Please understand that these are only guidelines and are not guaranteed. The
cost structure of this part and the target market are not compatible with
providing such accurate specs.