What is the upper/lower frequency limit of the AWG2000s?

What is the upper/lower frequency limit of the AWG2000s?

Defining an upper frequency limit is a common question, but it is a difficult to answer. There is a hard upper limit of the Nyquist frequency, which is the maximum clock rate divided by 2. However this is not a very practical limit for a lot of applications due to lack of flexibility. Signal to noise ratio of a 2 or 3 point per cycle waveform can be poor. The waveform may be beyond the -3 dB point of the output amplifier's bandwidth, so the amplitude will be reduced > 30%. The practical upper limit which will work for most applications is 5 or more points per cycle or the maximum clock rate divided by 5. With some applications, waveforms of 2 to 4 points per cycle work just fine, but only the end user can determine that for sure.

A squarewave consists of the fundamental frequency and an infinite number of odd harmonics that rolls off in amplitude at a specific rate. This means that if you have a If you have a 50 MHz squarewave on a AWG2021 the 50 MHz squarewave is well within the 100 MHz bandwidth of the output amp. However, the first harmonic contained in the waveform is at 150 MHz, the next at 250, 350... These harmonics end up being filtered out of the waveform by the limited amplifier bandwidth. The result being that your squarewave looks more and more like a sine wave as you increase the frequency.

Another way to look at how amplifier bandwidth will impact your waveshape is risetime. Take for example a squarewave with 4 points per cycle (2 points high and 2 points low) running with a 250 MHz clock on the AWG2021. This would create a squarewave period of 16 ns with data that is transitioning between 0 and full scale instantly. It would be a perfect squarewave if you had an infinitely fast system to translate that data into a waveform. However in an AWG2021 the rise/fall time is 3.5 ns. This means that 7 of the 16 ns period will be spent between the 10% and 90% points of the waveform or, about 1/2 the time the waveform is transitioning from low to high and back again. While this would still look somewhat like a squarewave, it is certainly less than ideal.

The lower limit is much more cut and dried. You take the lowest available clock rate and divide it by the maximum record length. This method assumes that you create a waveform that is 1 cycle over the maximum record length.

Below is the lower limit and the practical/Nyquist upper limits for each AWG.