AD9835: I want to realize a frequenzy agile FSK transmitter (5 to 21 MHz,
frequency hub of +-50 or 67 kHz). It seems that the AD9835 can do this. Is that
correct an can this FSK be simulated with the evaluation board?
In practice you can a DDS can be used to generate frequency up to approximately
one third of the input reference frequency. As you increase the output
frequency further, the filter requirements become prohibitive and
the SNR and THD start to degrade. As you are looking to generate an o/p freq of
21Mhz, this puts the AD9835 as an unsuitable choise for your application,
instead you should look at our AD9850 and AD9851.
In general, Direct Digital Synthesis (DDS) uses a crystal oscillator, Sine or
Cosine lookup table, some digital logic and a high AC performance DAC to
directly generate sine and cosine signals with of arbitrary frequency and
phase. All of the signal generation and processing is performed in the digital
domain and the high performance DAC does the conversion to the
analog domain. Because the signal is generated digitally, the phase and
frequency are very accurate (frequency resolution of 0.001Hz) and can be
changed very rapidly without the inherent lock-time required by a PLL solution.
On the AD9850, to change frequencies in your application, you need to load in a
new frequency tuning word and the DAC
starts to output the new frequency immediately after a delay of 18 clock
cycles. Programming the AD9850 is described on page 9 of the datasheet. The
AD9850 requires 5 byte writes to update the frequency register, this takes 5 x
7ns = 35ns. The latency between updating the frequency register and the output
frequency appearing at the output is 18 reference clock cycles. In theory you
can update the output frequency in 360ns + 35ns = 395ns. In practice, you will
limited by the speed of your digital interface.
Whether you are using serial or parallel programming, a complete 40 bit word
must be written to the AD9850 each time the frequency register is modified.
To change the the output frequency simply use the formula
Frequency Tuning Word = ( Output Frequency x 2^32 ) / Input Clock Frequency
To calculate the new word required.